textbook outcomes in gastric cancer surgery · 2019. 11. 21. · levy – textbook outcomes in...

T EXT BOO K O UT CO MES IN GA S T R IC CA NCER S UR GE R Y

BY

JORDAN LEVY , MD

A THESIS SUBMITTED I N CONFORMITY WITH THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE

IN CLINICAL EPIDEMIO LOGY AND HEALTH CARE RESEARCH

INSTITUTE OF HEALTH POLICY, MANAGEMENT AND EVALU ATION UNIVERSITY OF TORONTO

© COPYRIGHT BY JORDAN LEVY (2019)

Levy – Textbook Outcomes in Gastric Cancer Surgery ii

TEXTBOOK OUTCOMES IN GASTRIC CANCER SURGERY

JORDAN LEVY, MD

MASTER OF SCIENCE (CEHCR)

INSTITUTE OF HEALTH POLICY, MANAGEMENT AND EVALUATION

UNIVERSITY OF TORONTO

2019

EXECUTIVE SUMMARY

Textbook Outcomes, a composite of eight perioperative metrics, represent a comprehensive surgical quality

measure in gastric cancer surgery. The objectives of this thesis were to determine the long-term survival

implications of achieving Textbook Outcomes and to estimate the association between gastric cancer

surgery case-volume and Textbook Outcomes. A population-based cohort study of patients with non-

metastatic gastric adenocarcinoma treated with curative gastrectomy between 2004 and 2015 was

conducted. Multivariable regressions analyses were conducted in order to estimate the association

between (i) Textbook Outcomes and survival and (ii) gastrectomy volume and Textbook Outcomes.

Textbook Outcomes were achieved in merely 23% of the cohort and were associated with a 39% reduction

in mortality (aHR 0.61 [95%CI 0.50, 0.74], p<0.001). Neither surgeon nor hospital volumes were associated

with improved Textbook Outcome rates. Achieving Textbook Outcomes was strongly associated with

improved long-term survival in gastric cancer patients and future quality improvement strategies should

focus on more than increasing surgical volumes.

Levy – Textbook Outcomes in Gastric Cancer Surgery iii

ACKNOWLEDGEMENTS

There are a number of mentors, colleagues, friends and family I would like to thank and acknowledge

without which this thesis and my graduate studies would not have been possible.

First and foremost, to my supervisor and mentor Natalie Coburn, thank you for your unyielding support

throughout my graduate studies and research. You are an incredible role model for young researchers and

surgical residents alike. You gave me the time and freedom to consider numerous alternative projects until

we finally found the one I was passionate about. These two years have allowed me to grow as a scientist, a

resident and most importantly as a person and I can never express enough gratitude for your mentorship

and masterful guidance.

To Charles de Mestral, my thesis committee member, thank you for your encouragement and pragmatic

approach to research. Working with you as a medical student is what motivated me to pursue my General

Surgery residency and the Surgeon Scientist Program at the University of Toronto.

To Olli Saarela, my thesis committee member, thank you for taking the time to answer every esoteric

statistical question I could think of, without you my analyses would be nowhere near as robust.

To Alyson Mahar, my thesis committee member, thank you for pushing me to think more. By virtue of your

thoughtful questions, my methods are rigorous and defensible.

I wanted to acknowledge all those who supported me and this project financially: the Sherif and Mary Lou

Hanna Chair in Surgical Oncology at Sunnybrook Health Sciences Centre, the Division of General Surgery

and Department of Surgery at the University of Toronto and the Division of General Surgery at Sunnybrook

Health Sciences Center. I also wanted to thank my previous and current program directors Dr. Najma

Ahmed and Dr. Fred Brenneman, for coordinating the resources and planning necessary for me to pursue

this advanced degree.

To my fellow General Surgery residents, thank you for taking on the burden of all the SSTP residents’

absence in clinical duties. We and the Division owe you a debt of gratitude for this seemingly impossible

workload. I am very excited to be joining you on July 1st.

Levy – Textbook Outcomes in Gastric Cancer Surgery iv

To Vaibhav Gupta, my friend, colleague and research lab partner, thank you for your collaboration,

guidance and motivation. Good luck in the final six months of your PhD.

To Naheed Jivraj, my friend and CEHCR classmate, thank you for coding the things I couldn’t. I will cherish

our time in the ICES basement and in class as some of the most memorable experiences during my graduate

studies.

To Jared Toll, thank you for all the weekends we didn’t work.

To Fred Laliberte, thank you for all the weekends we did.

To Gab Kakon, Adrien Malka, Hillel Matthews and Nyv Segev, thank you for all the conversations and down-

time. I wouldn’t be where I am without you.

To Ari Zuckerbrot, thank you for reading my thesis when no one else would.

To Noah Abermad, thank you for standing by me during my long hours of writing and encouraging me

throughout the process.

To my twin sister Erika, thank you for always reminding me of what’s right, even when I don’t ask.

To my sister Jessica, thank you for your empathy.

To my grandfather, thank you for your laughter and wisdom.

To my mother, thank you for your love, compassion and support. You are the light that guides me.

And to my late grandmother who passed in April of this year, thank you for your infinite love and

kindness, you will always be in our hearts. I dedicate this thesis and the culmination of these past two

years of hard work to you.

Levy – Textbook Outcomes in Gastric Cancer Surgery v

TEXTBOOK OUTCOMES IN GASTRIC CANCER SURGERY II

Executive Summary ii

ACKNOWLEDGEMENTS III

CHAPTER 1 INTRODUCTION 1

1.1. GASTRIC CANCER 1 1.1.1. BIOLOGY AND PATHOLOGY 1 1.1.2. DESCRIPTIVE EPIDEMIOLOGY 2 1.1.3. DETECTION AND DIAGNOSIS 3 1.1.4. WORKUP AND STAGING 4 1.1.5. SURGICAL MANAGEMENT 5 1.2. GASTRECTOMY QUALITY IN ONTARIO 9 1.2.1. MORTALITY 9 1.2.2. LYMPHADENECTOMY 10 1.2.3. MARGIN STATUS 10 1.3. LONG TERM SURVIVAL IN ONTARIO 10 1.4. TEXTBOOK OUTCOMES: A BENCHMARK OF GASTRECTOMY QUALITY 11 1.5. VOLUME AND GASTRECTOMY OUTCOMES 12 1.6. RATIONALE AND OVERVIEW OF STUDY 14 1.6.1. RATIONALE AND PURPOSE 14 1.6.2. THESIS OBJECTIVES 14 1.6.3. THESIS FORMAT 14

CHAPTER 2 METHODS 15

2.1. STUDY DESIGN AND SETTING 15 FIGURE FOR 2.1. THESIS STUDIES TIMEFRAME 15 2.2. DATA SOURCES 16 2.2.1. DATA LINKAGE 16 2.2.2. ONTARIO CANCER REGISTRY 16 2.2.3. CANADIAN INSTITUTE FOR HEALTH INFORMATION – DISCHARGE ABSTRACT DATABASE 16 2.2.4. ONTARIO HEALTH INSURANCE PLAN 16 2.2.5. ESOPHAGOGASTRIC PATHOLOGY DATABASE 17 2.2.6. JOHNS HOPKINS ADJUSTED CLINICAL GROUPS® SYSTEM 17 2.2.7. ONTARIO MARGINALIZATION INDEX 18 2.2.8. CANCER ACTIVITY LEVEL REPORTING 18 2.2.9. REGISTERED PERSON DATABASE 18 2.3. PATIENT POPULATION 19 2.3.1. DEFINING PATIENTS WITH GASTRIC ADENOCARCINOMA 19 2.3.2. TIMEFRAME 19 2.3.3. DEFINING PATIENTS UNDERGOING CURATIVE-INTENT GASTRECTOMY 19 2.3.4. STUDY COHORT WITH PATHOLOGY REPORT 21

Levy – Textbook Outcomes in Gastric Cancer Surgery vi

2.4. GASTRECTOMY VOLUME 21 2.5. TEXTBOOK OUTCOMES 22 2.5.1. NEGATIVE RESECTION MARGINS 22 2.5.2. GREATER THAN 15 LYMPH NODES RESECTED 22 2.5.3. NO SEVERE COMPLICATIONS 23 2.5.4. NO REINTERVENTIONS 23 2.5.5. NO UNPLANNED ICU ADMISSION 23 2.5.6. LENGTH OF STAY 21 DAYS OR LESS 24 2.5.7. NO 30-DAY READMISSION 24 2.5.8. NO 30-DAY MORTALITY 24 2.6. CONFOUNDERS 26 Figure for 2.6. Causal Framework for Textbook Outcome and Survival Study 26 Figure for 2.6. Causal Framework for Volume and Textbook Outcome Study 27 2.6.1. PATIENT LEVEL CHARACTERISTICS 28 2.6.2. DISEASE LEVEL CHARACTERISTICS 28 2.7. DATA VALIDITY 29 2.8. MISSING DATA 32 2.9. POTENTIAL BIASES 34 2.9.1. SELECTION BIAS 34 2.9.2. MEASUREMENT BIAS 35 2.9.3. CONFOUNDING BIAS 36

CHAPTER 3 RESULTS 38

3.1. OUTLINE OF RESULTS 38 3.2. MANUSCRIPT #1: TEXTBOOK OUTCOMES AND SURVIVAL IN PATIENTS WITH GASTRIC CANCER 39

ABSTRACT 40

Background 40 Methods 40 Results 40 Conclusions 40

INTRODUCTION 41

METHODS 42

Study Design and Setting 42 Data Sources and Management 42 Study Cohort 43 Exposure 43 Outcomes 44 Covariates 44

Levy – Textbook Outcomes in Gastric Cancer Surgery vii

Statistical Analysis 45 Sensitivity Analysis 46 Missing Data 47 Ethical Approval 47

RESULTS 48

Study Cohort 48 Textbook Outcomes 48 Survival 49 Association Between Textbook Outcomes and Survival 49 Sensitivity Analyses 50

DISCUSSION 51

3.3. MANUSCRIPT #2: GASTRECTOMY CASE VOLUME AND TEXTBOOK OUTCOMES 67

ABSTRACT 68

Background 68 Methods 68 Results 68 Conclusions 68

INTRODUCTION 69

METHODS 70

Study Design and Setting 70 Data Sources and Management 70 Study Cohort 71 Exposure 71 Outcomes 72 Covariates 73 Statistical Analysis 74 Missing Data 75 Ethical Approval 75

RESULTS 76

Study Cohort 76 Textbook Outcomes 76 Gastrectomy Volumes 77

Levy – Textbook Outcomes in Gastric Cancer Surgery viii

Textbook Outcomes Across Gastrectomy Volumes 77 Textbook Outcome Metrics Across Gastrectomy Volumes 79

DISCUSSION 80

CHAPTER 4 DISCUSSION 100

4.1. REVIEW OF THESIS OBJECTIVES 100 4.2. SUMMARY OF MAIN FINDINGS 100 4.3. COMPARISON TO THE LITERATURE 101 4.4. LIMITATIONS AND METHODOLOGICAL ISSUES 103 4.4.1. MISSING DATA 103 4.4.2. INTERNAL AND EXTERNAL VALIDITY 105 4.5. IMPLICATIONS AND RECOMMENDATIONS 106 4.6. FUTURE RESEARCH 108 4.7. CONCLUSIONS 109

CHAPTER 5 TABLES 110

Table for 2.3.2. Morphology Codes and Histological Categories 110 Table for 2.3.3.A. Gastrectomy Intervention Codes 111 Table for 2.3.3.B. Esophagogastrectomy Intervention Codes 112 Table for 2.5.3. Severe Complication Categories – Diagnostic (ICD-10-CA) and Intervention (CCI) Codes

114 Table for 2.5.4. Reinterventions – CCI Codes 119 Table for 4.4.1.A. Comparison of Patients With and Without Pathology Report 126 Table for 4.4.1.B. Predictors of Missing Pathology Report 127

REFERENCES 128

Levy – Textbook Outcomes in Gastric Cancer Surgery 1

CHAPTER 1 INTRODUCTION

1.1. GASTRIC CANCER

1.1.1. BIOLOGY AND PATHOLOGY

The term gastric cancer represents all malignancies originating from the different cell lines of the

anatomical stomach. According to the most recent International Classification of Diseases for Oncology

(ICD-O-3) topography codes, the anatomical stomach begins at the gastric cardia and ends at the pylorus,

connecting the esophagus to the duodenum. The squamo-columnar junction, also known as the “Z-line,” is

where the esophageal squamous mucosa meets the gastric columnar mucosa. Tumors arising from this

gastroesophageal junction are variably classified as either esophageal or gastric in origin.

For the purposes of this thesis, and generally in the literature, gastric cancer refers to gastric

adenocarcinomas arising from the glandular cells lining the stomach wall [1]. Other primary malignant

tumors of the stomach include gastrointestinal stromal tumors (GIST) [2-4] arising from the interstitial cells

of Cajal [5]; endocrine tumors (carcinoids) arising from the enterochromaffin cells of the stomach [6];

mucosa-associated lymphoid tissue lymphomas (MALT lymphomas) arising from gastric lymphoid tissue,

typically in response to chronic Helicobacter pylori infection [7]; and leiomyosarcomas, which are soft

tissue sarcomas arising from smooth muscle [8, 9]. GIST account for 0.2% of all gastrointestinal tumors [10],

gastric endocrine tumors are on the rise and account for 1.8% of gastric cancers [11], and lymphomas and

leiomyosarcomas account for 5-10% of gastric cancers [8, 9, 12].

Adenocarcinomas account for greater than 90% of gastric cancers and are classified histologically according

to the Lauren classification [13] as intestinal (54%), diffuse (32%) or mixed subtypes (15%) [14]. Intestinal-

type is the most common variant, especially in sporadic cancers [15]. Alternatively, the 2010 World Health

Organization (WHO) classifies gastric adenocarcinomas into four major histologic patterns: tubular,

papillary, mucinous, signet ring and other poorly cohesive histologic variants[15]. Papillary and tubular

adenocarcinomas are the most common and usually fall under the Lauren intestinal type. Mucinous

adenocarcinomas represent about 10% of gastric cancers and can be intestinal or diffuse type. Finally

signet-ring cell carcinomas are of the diffuse type, however not all diffuse type gastric cancers are signet-

ring as other poorly cohesive carcinomas exist [16].


There are several known risk factors for gastric cancer. Helicobacter pylori infection is a declared Group 1

carcinogen [17] and the strongest known risk factor for gastric cancer [18]. It is known to preferentially

cause distal gastric cancers [19]. The attributable risk for non-cardia gastric cancer secondary to H.

pylori infection is 89% [20]. H. pylori initiates a multistep pathway [21] of longstanding inflammation,

mucosal atrophy, gastric gland intestinalization, and intraepithelial neoplasia with the risk of progression

to invasive adenocarcinoma [22]. Its exact role in this pathogenesis is hypothesized to include inflammation,

genetic instability in the host, and epigenetic alterations [23, 24]. However, some regions show high rates

of H. pylori infection without elevated risks of cancer, and thus H. pylori may represent a component, but

not sufficient cause of gastric cancer [25]. Epstein-Barr Virus is another infectious agent associated to

gastric cancer, with increased rates in male patients [26]. Other established risk factors include smoking

[27] and dietary factors [10, 28-34]. Family history has been associated with a relative risk increase of 1.5

to 10 depending on the geographic region, number of affected family members, and degree of relation to

them [35]. However, hereditary gastric cancer syndromes including Hereditary Diffuse Gastric Cancer [36,

37], Peutz-Jeghers [38, 39], Li-Fraumeni [40] and Familial Adenomatous Polyposis [41, 42] likely only

account for about 5% of diagnoses [10, 32, 43, 44].

1.1.2. DESCRIPTIVE EPIDEMIOLOGY

Globally, gastric cancer is the third most common cause of cancer-related mortality [45, 46]. Two thirds of

gastric cancers occurred in Eastern Asia, Eastern Europe, and South America [47]. According to the 2018

version of GLOBOCAN (Global Cancer Incidence, Mortality and Prevalence) [48], the world age-

standardized incidence rate in Eastern Asia as a whole is 32.1/100,000 men and 13.2/100,000 women,

however the rate in Korea was as high as 58/100,000 men and 24/100,000 women. These rates in North

America were 5.6/100,000 men and 2.8/100,000 women.

The Canadian age-standardized incidence rate is decreasing by 2.2% and 1.3% annually for males and

females, respectively. Meanwhile, the Canadian population age-standardized incidence rate (ASIR) has seen

a similar decline. Canada’s ASIR for males in 2017 was 11.8 per 100,000, a sharp decline from 23.1 per

100,000 in 1988. Still, gastric cancer represents the 14th most commonly diagnosed malignancy in Canada,

and the 11th most common cause of cancer-related mortality. One in 78 men (1.3%) and one in 133 women

(0.8%) will develop gastric cancer in their lifetime [49]. It is estimated that 3,500 Canadians will be

diagnosed with gastric cancer in 2018, and that 2,100 women and men will succumb to it [50]. In Ontario,

ASIRs are similar to the national average, with 850 new male and 540 new females cases annually [49].


1.1.3. DETECTION AND DIAGNOSIS

Given the capacity and deep anatomical location of the stomach, detection of a gastric tumor through

physical exam or mass-effect symptoms is rare. Early symptoms including vague discomfort, fullness,

heartburn, nausea or vomiting are non-specific and delay diagnosis. Thus, malignant progression often goes

unabated until “red-flags” such as obstruction, bleeding, perforation or significant weight loss are

recognized, at which time the cancer is often advanced [51]. In fact, in unscreened populations, 35 to 64%

of patients with gastric cancer have distant spread of their disease at the time of its first detection [50, 52,

53].

East Asia (China, Korea, Japan and Vietnam) is home to 60% of all stomach cancer patients [54], however

only Japan and Korea have established national gastric cancer screening programs [55]. In Japan, gastric

cancer screening was introduced as a national program in 1983. Prior to 2018, photofluorography - a

double-contrast barium upper gastrointestinal study - was the only modality recommended for either

population-based screening or opportunistic screening. This recommendation was based on the

conclusions of a meta-analysis demonstrating a 40% to 60% mortality reduction in patients screened using

photofluorography [56]. In 2018, an update to this meta-analysis was published with the addition of gastric

endoscopy (EGD) as a recommended tool for both population and opportunistic screening. Neither H. pylori

antibody, serum pepsinogen, or their combination had sufficient evidence for inclusion in the guidelines.

Population screening in Japan recommends EGD every 3 years beginning at the age of 50 [57]. Gastric

cancer screening rates in Korea are at least 47% and increasing by 4.3% annually [58]. Korea launched a

population-based screening program in 1999, and recommended both gastric endoscopy and

photofluorography for screening those 40 years of age and older every 2 years [59]. A 10-year nested-

control study using the Korean National Cancer Screening Program confirmed a 47% gastric cancer-specific

mortality reduction in those screened using gastric endoscopy, but found no significant benefit of

photofluorography [60].

The cost-effectiveness of screening programs depends largely on the prevalence of the disease screened,

and thus in Canada and other low-incidence countries, widespread screening programs represent a costly

and inefficient means of reducing gastric cancer-related mortality. A 2010 study [61] investigating the cost-

effectiveness of a one-time gastric endoscopy performed at the same time as a planned screening

colonoscopy (50 years old) demonstrated an incremental cost-effectiveness ratio (ICER) of 115,664USD per

quality adjusted life year (QALY) compared to no gastric cancer screening. A more recent study found that

despite a 21% relative reduction in lifetime risk of gastric cancer, endoscopic screening with endoscopic


mucosal resection did not meet the 100,000USD/QALY willingness to pay threshold. Serum pepsinogen

screening was more effective and cost-effective, but resulted in an ICER of 105,400USD/QALY [62]. These

studies have concluded that screening programs targeting high risk populations according to immigration

from high-prevalence areas [63, 64], race [65], and smoking status [62] may be more effective.

In Canada, gastric cancer is typically identified through symptomatic presentation related to an underlying

malignancy [51]. When gastric cancer is suspected, patients should undergo endoscopic biopsy for tissue

diagnosis, histological classification and molecular biomarking. It is recommended that this pathologic

examination be performed by a pathologist experienced in gastrointestinal malignancies [66]. Endoscopic

diagnosis consists of inspecting the gastric mucosa for any abnormalities, recording the location of any

tumor and taking multiple (six to eight) biopsies using standard-size endoscopy forceps in order to ensure

adequate material for histologic interpretation [67-69]. Endoscopic mucosal resection or submucosal

dissection of small lesions can provide a greater degree of information and may be therapeutic [70-73].

Cytologic brushings and gastric washings rarely provide a diagnosis but may be useful when initial biopsies

are not diagnostic [67].

1.1.4. WORKUP AND STAGING

Following diagnosis of gastric adenocarcinoma in Ontario, it is recommended [74] that all patients be

discussed at a multidisciplinary meeting in order to determine the optimal management of each patient

according to their personalized health status and extent of disease. It is thus imperative that patients

undergo initial risk assessment and tumor/lymph node/metastasis (TNM) staging according to the latest

edition of the Union for International Cancer Control (UICC) and American Joint Committee on Cancer

(AJCC) guidelines and staging manual [75, 76]. The TNM staging system classifies the depth of invasion into

the gastric wall or adjacent structures (T stage), the number of lymph nodes involved through lymphatic

spread (N stage) and metastatic or distant spread of primary gastric cancer (M stage).

Initially, patients are divided into two categories: curative-intent locoregional disease (AJCC stage I-III) and

non-curable systemic disease (AJCC stage IV). Oral and intravenous contrast enhanced computed

tomography of the chest, abdomen and pelvis is performed to evaluate for systemic and metastatic disease,

which is most commonly identified as peritoneal deposits, liver or lung metastatic lesions, or distant

lymphadenopathy and/or ascites [66, 67, 77, 78]. Solid lesions should undergo biopsy and ascites should

prompt peritoneal washing and cytology to confirm systemic disease [67, 78]. Non-curative gastrectomy in

the setting of low-volume metastatic disease is now inadvisable following the publication of the REGATTA


trial [79], which not only failed to show a survival advantage, but also highlighted the harm associated with

gastrectomies in patients with non-curable gastric cancer.

Subsequently, patients with locoregional disease can undergo further evaluation when there is a need to

assess for local invasion of other structures, nodal status and metastatic spread. Endoscopic ultrasound

provides accurate assessment of T stage as well as proximal and distal extent of the primary tumor [66, 67].

Computed tomography can identify tumor invasion into other organs or major vascular structures and aids

in operative planning [78]. Lymph node involvement (N stage) can be evaluated through computed

tomography and endoscopic ultrasound. On computed tomography, lymph nodes with short-axis diameter

greater than 6-8mm, round shape, central necrosis and heterogenous or high enhancement are suspicious

for lymph node involvement [66, 80-82]. Presently, computed tomography sensitivity for lymph node

involvement ranges between 63% and 92% [83]. Findings associated with lymph node metastases on

endoscopic ultrasound include hypoechogenicity, round shape, smooth, distinct margin and size > 1 cm

[66, 84, 85].

Computed tomography can miss lesions smaller than 5mm, and 20% to 30% of patients are found to have

occult intraperitoneal disease upon surgical exploration [78]. Therefore, it is recommended that all patients

with non-early gastric cancer (in-situ or stage IA) undergo diagnostic laparoscopy with peritoneal cytology

prior to planned primary tumor resection [66, 67, 77].

1.1.5. SURGICAL MANAGEMENT

Once patients are determined to have locoregional disease without evidence of distant spread, they should

undergo a preoperative assessment of their “medical fitness” and ability to withstand the physical stress of

surgery. Risk assessment includes a physical examination, full chemistry, blood count and differential, and

liver and renal function tests. Additionally, patients should undergo a nutritional assessment based on

physical exam, history of weight loss and nutritional intake, and albumin and prealbumin levels [66, 67, 78].

This assessment should focus on the individual patient’s risk for surgery, functional status and therapies

necessary for surgical optimization and pre-habilitation. These include nutritional counselling, smoking

cessation and iron supplementation [67]. Patients unfit for surgery without the potential for improvement

are provided best supportive care and should not undergo invasive surgery.

Locoregional gastric cancer is divided into two main subgroups: early and advanced gastric cancer [10].

Early gastric cancers are defined as tumors limited to the mucosa – those invading the lamina propria

and/or muscularis mucosa. These cancers have an extremely low incidence of lymph node metastases [86,


87] and as such may be amenable to localized endoscopic resection [67]. Given the high incidence of gastric

cancer in Japan, the classification of these lesions was further expounded [88, 89] and the criteria for

endoscopic resection was narrowed to tumors with the lowest risk of lymph node involvement [90-92].

However, patients with early gastric cancer undergoing endoscopic resection represent an exceedingly

small proportion of the Ontario population with gastric cancer and are not included in this study.

Patients with T1a lesions not amenable to endoscopic resection and all other patients with locoregional

gastric cancer should undergo a radical gastrectomy which includes removal of the gastric tissue, omentum

and associated lymph nodes [66, 67, 77, 93]. A basic tenet of radical gastrectomy is negative margins (R0

resection). In order to improve negative margin rates, it is recommended that the macroscopic distance

from the tumor to the resection line be at least 4 cm [67], with some groups suggesting the use of

intraoperative margin assessment (frozen section) when gross margins are under 5 cm and in particular

with tumours of diffuse histology where extent of invasion is heterogenous and occult [10]. The accuracy

of intraoperative frozen section margin evaluation has been reported to be 97% [94], but is of use only

when further resection can be safely performed. Patients with T4 tumours should undergo en-bloc multi-

visceral resection of involved structures [67]. For patients with distal gastric cancers, a subtotal gastrectomy

has been shown to have equivalent oncological outcomes [95, 96] but guidelines suggest a macroscopic

proximal margin of 4-5 cm between the tumour and the gastroesophageal junction in order to assure that

final pathology yields a negative margin. Due to a higher chance of submucosal infiltration, for diffuse

histology a margin of 8 cm is recommended [66].

This issue of lymph node dissection for gastric cancer has not been as straightforward as the acquisition of

negative margins. In 1995, the Japanese Research Society for the Study of Gastric Cancer (JRSGC)

standardized lymph node dissection and pathological evaluation for gastric cancer and consistently

recommended D2 lymph node dissection for the treatment of gastric cancer [97]. A D1 lymph node

dissection includes removal of the omentum, with perigastric nodes (stations 1 to 6) and lymph nodes along

the left gastric artery (station 7); whereas a D2 lymph node dissection includes the omentum, D1 nodes,

and nodes stations along all named gastric vessels--namely the hepatic artery (station 8a), celiac axis

(station 9), splenic artery (station 10 and 11) and hepatic artery proper (station 12a).

The need for an extended lymph node dissection inspired two major multicenter European randomized

trials that compared D1 with D2 dissection. The 5-year updates in the in Medical Research Council (MRC)

trial [98] and the Dutch trial [99] found no difference in long-term survival, both reporting 5-year survival

rates of 33% in the D2 group and 35% in the D1 group. However, some critics of these studies argue that


the researchers failed to properly train the surgeons participating, and that this significantly biased the

results of the trials. D2 lymphadenectomies are complex procedures, with a learning curve as high as 15-

100 cases, depending on the parameters measured [100, 101], which may take several years to achieve in

low-incident countries. The 32 participating surgeons in the MRC trial were provided a booklet and

videotapes as a means of D2 lymphadenectomy training, whereas eight surgeons in the Dutch trial trained

six months with a Japanese gastric cancer surgeon, and were then responsible for assisting local surgeons

participating in the study [102]. Surgical videos today, 3 decades later, remain suboptimal for training,

particularly for non-laparoscopic cases where recording lighting and focus are hindered by the surgeons

working in a dark and limited space. A subsequent quality control analysis of the Dutch trial found that non-

compliance (inadequate removal of lymph nodes from indicated stations) occurred in 26% of D2 patients,

and that contamination (lymph nodes detected in stations outside the intended level of dissection)

occurred in 23% of D1 patients. This was believed to pose a serious threat to the internal validity of the

trial.

Furthermore, critics often cite the exceedingly elevated post-operative mortality rates of the D2 groups in

the MRC trial [103] (D2 13% vs. D1 7%, p=0.04) and the Dutch trial [104] (D2 10% vs. D1 4%, p=0.004) as

evidence of poor surgeon training and the basis for the lack of survival benefit. These elevated mortality

rates are believed to be related to the high rate of distal pancreatectomy and splenectomy performed in

the D2 group, which was mandated by the operative protocol. In the Dutch trial [99], splenectomy was

performed in 37% of D2 patients and 11% of D1 patients, and pancreatectomy was performed in 30% of

D2 patients and 3% of D1 patients. The 15-year long term results of the Dutch trial [105] showed a higher

15-year survival rate in the D2 compared to the D1 group (29% vs. 21%, p=0.34) that was not statistically

significant, and several subgroup analyses favoured D2 lymphadenectomies, including female patients (D2

35% vs. D1 21%, p=0.03), stage II patients (D2 33% vs. D1 15%, p=0.03) and patients with N2 disease (D2

19% vs. D1 0%, p=0.07). Cancer control appeared superior in D2 lymphadenectomies: 48% of the D1 group

and 37% of the D2 group died secondary to their gastric cancers and regional recurrence occurred in 19%

of the D1 group and 13% of the D2 group.

This high rate of morbidity and mortality in patients undergoing D2 lymphadenectomies with associated

pancreatectomies and splenectomies became the main motivation for the creation of the Italian Gastric

Cancer Study Group. Their objective was to determine whether a strict quality-control system and pancreas

preservation could improve outcomes following D2 lymphadenectomies in Western patients. Their initial

one-arm phase I–II trial [106] included nine centers and 18 surgeons (two at each site). Prior to patient

enrolment, the study-supervising surgeon stayed at the National Cancer Center Hospital in Tokyo to learn


the D2 dissection from a specialist Japanese surgeon. The study-supervising surgeon then organized several

meetings with videos to train the other 17 surgeons, and at least one surgeon from each site attended his

first 10 D2 gastrectomies. The results of this study were quite encouraging. The mean number of nodes

removed was 39 and no patient had less than 22 nodes removed; the hospital mortality rate of 3.1% was

at least 3 times lower than in the Dutch and MRC trials; and the 5-year survival rate of 55% was significantly

greater, with the caveat that a large proportion (28%) of patients had early gastric cancer. In their

subsequent prospective multicenter randomised trial comparing D1 to D2 lymphadenectomy [107], five of

the nine centers which had performed more than 25 D2 gastrectomies in the phase I/II trial were included.

Pancreatectomies were only performed when the pancreas was invaded by the primary tumor, and

splenectomies were performed when the tumor was localized to the greater curvature of the stomach, of

T2 stage or greater, and in the proximal two thirds of the stomach. The spleen was removed in 4% of D1

and 10% of D2 patients, whereas distal pancreatectomy with splenectomy was performed in 1% of D1 and

3% of D2 patients. Contrary to the Dutch and MRC trials, this trial succeeded in establishing the safety of

D2 gastrectomy in experienced hands in their interim analysis, [107] where the hospital mortality rates

were 1% in the D1 group and 0% in the D2 group, and in their final analysis [108] where the hospital

mortality rates were 3% in the D1 group and 2% in the D2 group. However, similar to the two previous

European trials, they showed no difference in 5-year survival rates (D1 67% vs. D2 64%, p=0.70). This was

felt to be due to a number of methodological errors, including randomization failure, which was

demonstrated by the uneven distribution of early gastric cancers (31% in D1 group vs. 19% in D2 group)

and metastatic gastric cancer (7% in D1 group and 11% in D2 group); contamination (18% of D1 patients

underwent dissection of more than two LN stations that should not have been removed); and non-

compliance (34% of D2 specimens were missing more than two lymph node stations that should have been

excised). Consequently, a subgroup analysis of patients with pathologic T2-4 and positive lymph nodes and

no metastases found a 5-year survival rate of 38% in D1 patients compared to 59% in D2 patients, but this

was not statistically significant (p=0.055) [108].

All in all, the MRC and Dutch trials clearly showed that D2 lymphadenectomies, including pancreatectomies

and splenectomies, were associated with much greater morbidity and mortality than D1

lymphadenectomies. While a trial from Chile, comparing D2 gastrectomy with and without splenectomy

[109] found that, despite a relatively low post-operative mortality rate in the splenectomy group (4%), the

addition of splenectomy did not offer any survival benefit and should not be performed routinely, the long-

term results of the Dutch trial suggested improved cancer-control with D2 lymphadenectomy and potential

benefit in patients with more advanced disease. As well, the Italian Gastric Cancer Study Group showed


that in experienced hands and with strict quality control, pancreas preserving D2 lymphadenectomy could

be performed safely with a morbidity and mortality profile similar to D1 lymphadenectomy. The long-term

results of the Italian trial likewise suggested a benefit in patients with more advanced disease. Since then,

large population-based studies in the U.S. have reported improved survival with increasing number of

nodes resected for all stages [110, 111]. However, these findings may be due to stage migration rather than

the therapeutic effect of lymphadenectomy, and the true benefit of increased lymph node sampling may

be limited to patients with more advanced locoregional disease [112, 113].

Furthermore, with the dissemination of the 5th Edition of the AJCC staging manual in 1997, N-staging

depended on the number of lymph node metastases and proved a better estimate of prognosis than

previous staging systems [114]. A new category “N3” was defined as greater than 15 lymph node

metastases, which thus required that at least as many lymph nodes be sampled. The mean number of nodes

examined in a D1 lymphadenectomy has been reported to be 15 [115], which by definition would render

half of patients unstageable secondary to inadequate lymph nodes. This has been further corroborated in

a second study which determined that 62% of patients undergoing a D1 lymphadenectomy could not be

accurately staged [116].

In Ontario, provincial guidelines developed by Cancer Care Ontario recommend that all curative gastric

cancer resections be accompanied by a D2 lymphadenectomy [74], which is in line with the approaches of

the U.S. [67], Germany [117], the UK [118], Italy [119], and Europe in general (ESMO, ESSO) [120].

Exceptions to this rule include highly comorbid patients and those with T1N0 gastric cancers.

1.2. GASTRECTOMY QUALITY IN ONTARIO

Our research group has previously performed a population-based, retrospective cohort study of all gastric

cancer patients diagnosed between 2005 and 2008. A province-wide chart review was linked to

administrative data, and examined 30-day post-operative mortality rates, the number of lymph nodes

resected, and the status of the surgical margin, amongst other patient characteristics [121].

1.2.1. MORTALITY

In-hospital post-operative population-based mortality rates in Ontario have been estimated at 5%, and

range from 0% to 17% across 14 administrative health regions [121]. These results are consistent with both

an American College of Surgeons National Surgical Quality Improvement Program (NSQIP) database


analysis (4.7%) [122] and a population-based analysis of Dutch patients (5.5%) [123]. However, these post-

operative mortality rates are vastly greater than those reported in single and multi-institutional studies

from Korea (0.2% to 0.6%) [124-126] and national population studies from Japan (0.5% to 0.9%) [127, 128].

1.2.2. LYMPHADENECTOMY

The current UICC and AJCC TNM classification manuals [75, 76], as well as several guidelines [66, 77],

including Cancer Care Ontario [74], recommend sampling at least 16 lymph nodes when performing a

gastrectomy for adenocarcinoma. In Ontario, this was only achieved in 40% of patients (ranging from 26%

to 58% by region). These results are consistent with results in the American National Cancer Institute’s

Surveillance, Epidemiology, and End Results cancer registry data (29%-32%) [129, 130], and lower than the

Dutch Upper Gastrointestinal Cancer Audit data (57%) [123], which may speak to Holland’s regionalization

of gastrectomies since 2012.

1.2.3. MARGIN STATUS

As described above, positive surgical margins are associated with worse prognosis [131-133], and are

defined as tumor cells at the margin (College of American Pathologists) or within 1mm of it (Royal College

of Pathologists), depending on the guidelines. The positive resection margin rate in Ontario was 28% and

ranged between 15% and 50% [121]. Five-year survival rates were 51% and 27% in patients with negative

and positive margins, respectively. The positive margin rate in this study was much greater than the rate

found in two American National Cancer Database analyses (6% to 11%) [134, 135], and this may be due to

differences in margin positivity definitions or pathology assessment procedures.

1.3. LONG TERM SURVIVAL IN ONTARIO

Worldwide, gastric cancer is one of the top three causes of cancer-related deaths [45, 46]. However, there

is significant variation in survival across countries. Despite a relatively low incidence in North America,

gastric cancer represents a significant cause of cancer-related mortality, with only 25% to 31% of patients

surviving 5-years [50, 53]. In contrast, patients in high-incidence countries such as Japan and Korea have 5-

year survival rates of 65% to 75% [136, 137].

Factors contributing to the survival difference between high incident and low incident countries has long

been debated [138-141], and are generally thought to include patient characteristics, tumor biology, and


early detection through screening programs. Western patients tend to be diagnosed at an advanced age,

with associated increased comorbidity burden [126, 138]. There is also a higher incidence of Barret’s

esophagus, with a resulting increase in junctional tumors with lower survival [126, 142]. Furthermore,

patients in Japan and Korea have benefited from national screening programs since 1983 and 1999,

respectively, [55] and are twice as likely to be diagnosed with locoregional as opposed to metastatic disease

compared to North American patients [53, 142, 143].

Many of these differences in case-mix are, however, non-modifiable at the time of diagnosis. Efforts to

improve outcomes for gastric cancer patients in North America must therefore focus on optimizing their

peri-operative and oncologic care.

1.4. TEXTBOOK OUTCOMES: A BENCHMARK OF GASTRECTOMY QUALITY

Recently, healthcare research and quality improvement has seen a paradigm shift towards comprehensive

outcomes. Post-operative mortality has historically been the focus of many surgical quality studies because

it is clinically meaningful, available, and easily captured. However, death from surgery is only a crude

measure of surgical quality, and it represents merely one of many factors important to patients, their

families, and their health care professionals.

In 2017, a Dutch study combined 10 parameters to define high quality surgical care (henceforth entitled

“Textbook Outcomes”) [144]. This composite measure included intent of surgery, resection margin status,

lymph node harvest, intra-operative and post-operative complications, re-interventions, ICU admissions,

length of stay, readmissions, and mortality rates. They found that Textbook Outcomes occurred in only 32%

of patients. Importantly, there was high variation in the rate of Textbook Outcomes, ranging from 9% to

52% across hospitals.

The concept of Textbook Outcomes advocates for a broad assessment of short-term patient outcomes

following a complex surgical procedure. Oncologically sound resections aim for cure in order to decrease

the risks of cancer recurrences and, ultimately, cancer-related mortality and morbidity. However, the risk-

benefit ratio of such an invasive and morbid procedure is negatively impacted when cancer is left behind

in cases of positive margins or inadequate lymph node sampling. Meanwhile, the occurrence of

complications and subsequent re-interventions, ICU admissions, prolonged length of stay, and

readmissions have important quality of life, clinical, and economic implications for the patient and the

health care system at large. Unlike standard surgical quality measures, including mortality, morbidity and,


more recently, readmission rates, Textbook Outcomes incorporate cancer-specific metrics (intent of

surgery, margins, and lymph nodes), and goes further by measuring all meaningful clinical events both

involved in and as a result of post-operative complications. As such, Textbook Outcomes may represent the

optimal benchmark in assessing gastrectomy quality for future clinical research and health care quality

improvement programs.

1.5. VOLUME AND GASTRECTOMY OUTCOMES

Cancer Care Ontario’s Staging and Surgical Approaches in Gastric Cancer guidelines recommend that

“patients should be referred to higher-volume centres for surgical resection,” but do not define the

minimum number of cases required. In the low incidence region of North America, a high-volume surgeon

has been previously defined as one performing greater than 3 resections annually [131, 145], whereas a

high volume center has been defined according to a cut-off between 15 and 35 annual cases [145-150]. In

East Asia, a high-incident region, definitions of high-volume surgeons and hospitals ranged from 30 to 35

[151, 152] and 56-386 [151, 153-155] cases annually. A recent international and multidisciplinary expert

panel [77] has provided consensus guidelines on the processes of care for gastric cancer, and has deemed

it necessary for practicing surgeons to be trained in gastric cancer surgery and appropriate for them to

perform greater than 6 cases annually at centers with case volumes greater than 15 annually.

The evidence for increased surgeon and hospital volume stems indirectly from the high morbidity and

mortality rates in the original Dutch and MRC trials and the advances made in the Italian trials, and is

substantiated directly by multiple North American studies that have shown decreased gastric surgery

operative mortality with increased center experience (hospital volume) [145, 146, 148-150] and surgeon

experience (surgeon volume) [145, 146]. However, these correlations have not been consistently shown at

the hospital [147, 156] or surgeon-level [131]. One U.S. state-wide analysis found that greater hospital-

volume was also associated with decreased length of stay and cost in gastric cancer surgery [156], and a

National Cancer Database analysis found that higher-volume hospitals examined more lymph nodes [157].

Several studies have investigated the effect of both hospital and surgeon-experience on gastric cancer

patient outcomes [131, 145, 146]. An earlier analysis of post-operative mortality in pancreas and

esophageal cancer resections found that the proportion of the effect of hospital volume attributable to

surgeon volume ranged between 46% and 54% [158], while procedures like aortic-valve replacements and

carotid endarterectomies were 100% associated with surgeon volumes and not at all with hospital volumes.

The authors hypothesized that mortality outcomes for high risk procedures with short lengths of stays were


likely completely associated with surgeon-technique and not the resources available at their centers.

Gastrectomies share many similarities with pancreatectomies and esophagectomies, and the effect of

volume on post-operative mortality is likely equally attributable to surgeon and hospital volume.


1.6. RATIONALE AND OVERVIEW OF STUDY

1.6.1. RATIONALE AND PURPOSE

Textbook Outcomes are a new concept in cancer surgery with oncology-specific and general operative and

post-operative metrics. We propose that Textbook Outcomes represent a comprehensive surgical quality

measure, and that by increasing their occurrence in Ontario, and in North America in general, we may begin

to close the gap between West and East patient prognosis. However, Textbook Outcomes’ long-term

survival implications remain unknown and must be examined. Furthermore, there exists some evidence for

a volume-outcome relationship in regard to Textbook Outcome components in gastric cancer, but the

majority of studies have focused on post-operative mortality alone. Further research is required to

determine whether surgeon experience, hospital experience, or both are associated with Textbook

Outcomes rates in Ontario.

1.6.2. THESIS OBJECTIVES

The objectives of this thesis are to describe the survival of patients with and without Textbook Outcomes;

to measure the independent association between Textbook Outcomes and long-term survival; and to

determine whether gastrectomy experience, in terms of surgeon and hospital-volumes, is associated with

achieving Textbook Outcomes and its metrics.

1.6.3. THESIS FORMAT

This thesis will be presented as two manuscripts, with expanded sections for introduction, methods and

discussion, with the results presented in manuscript-format.


CHAPTER 2 METHODS

Methodology descriptions in clinical manuscripts are intended to provide the essential information to judge

the validity of the study’s results based on the rationale for methodological choices and characteristics of

the study design. Unfortunately, given word count limitations, manuscript “methods” sections are

frequently too brief to allow for an in-depth exploration of all methodological considerations involved in

performing clinical studies, let alone replicate them. Given the important clinical implications of this thesis’

results and the ease of transition to knowledge translation and dissemination in such a format, this thesis

is presented in manuscript format. However, a robust “methods” chapter has been included in order to

complement and substantiate the manuscripts’ succinct methods sections included in Chapter 3.

2.1. STUDY DESIGN AND SETTING

The studies included in this thesis are population-based retrospective cohort analyses using routinely

collected administrative data and a prospectively collected pathology database. The time-frame for these

studies is included below (Figure for 2.1.). These studies were approved by the Research Ethics Board of

Sunnybrook Health Sciences Centre and the University of Toronto.

FIGURE FOR 2.1. THESIS STUDIES TIMEFRAME


2.2. DATA SOURCES

2.2.1. DATA LINKAGE

Several provincial and national administrative databases were used to derive the cohort and key variables

for both analyses. All datasets are held at ICES and linked using unique encoded identifiers. ICES is an

independent, non-profit research institute whose legal status under Ontario’s health information privacy

law allows it to collect and analyze health care and demographic data, for health system evaluation and

improvement. Secure access to these data is governed by policies and procedures that are approved by the

Information and Privacy Commissioner of Ontario.

2.2.2. ONTARIO CANCER REGISTRY

The Ontario Cancer Registry database is managed by Cancer Care Ontario. It is a passive, provincial registry

of all incident cancer diagnoses in Ontario and collects data from hospital discharge records, emergency

department visits, regional cancer centers, pathology reports from all public hospitals and community labs,

death certificates and interprovincial data sharing. As such, the Ontario Cancer Registry captures 96% of

cancer diagnoses in the province [159, 160].

2.2.3. CANADIAN INSTITUTE FOR HEALTH INFORMATION – DISCHARGE ABSTRACT DATABASE

The Canadian Institute for Health Information’s Discharge Abstract Database (CIHI-DAD) is a national data

repository of all inpatient hospital admission facility and patient-level data. Reporting of such data is

mandatory for all institutions within Canada. Trained medical records staff extract information from each

patient’s medical chart using standard diagnostic (International Classification of Diseases 10th Revision –

Canadian Version) and intervention (Canadian Classification of Health Intervention codes) codes [161].

Additional information includes administrative data (institution identifier, date of admission, date of

transfer, discharge date, discharge disposition) and clinical data (admission category, length of stay, special

care unit use and type).

2.2.4. ONTARIO HEALTH INSURANCE PLAN

In Ontario, 95% of physicians are financially reimbursed by the Ministry of Ontario Health and Long-Term

Care after submitting claims to the Ontario Health Insurance Plan for each covered service provided. This


database is a provincial dataset updated bimonthly, recording all such claims and was used to identify the

surgeon performing each gastrectomy [161]. A previous study has found that 95% of cancer surgeries in

the CIHI’s Discharge Abstract Database can be linked to an the Ontario Health Insurance Plan record [162].

Provincial coverage is available to all permanent residents of Ontario living within Ontario for at least 153

days in any 12-month period.

2.2.5. ESOPHAGOGASTRIC PATHOLOGY DATABASE

Our research group has generated a pathology database of all patients diagnosed with esophageal or gastric

cancer in Ontario since 2002. We began by obtaining electronic copies of these patients’ pathology reports

from the Ontario Cancer Registry. A committee with representation from gastrointestinal pathology,

surgical oncology, medical oncology, radiation oncology and clinical epidemiology developed a

standardized data collection instrument for pathology reports based on the American Joint Committee on

Cancer (AJCC) guidelines for variables to include in a cancer registry [163] and the College of American

Pathologists’ (CAP) published protocols for examining surgical specimens [164, 165]. A data dictionary was

developed as a reference standard for data abstraction and training. To ensure detailed and reliable

interpretation of the pathology reports, two medical doctors with graduate-level research training

(Catherine Allen-Ayodabo and Elmira Amirazodi) were recruited to perform the data abstraction. Both

abstractors were trained by the principal investigator and co-investigators (Natalie Coburn, Vaibhav Gupta,

Yunni Jeong, Jordan Levy) on esophagogastric cancer, relevant pathology data and key resources. Regular

meetings were held with the abstractors to review any questions, clarify abstraction protocols, and provide

advice or resources for unclear reports. Full committee data quality review was performed as part of data

cleaning. This included checks for plausibility, consistency, and accuracy, using clinical expertise, published

literature, and the abstraction manual. Pathology reports were obtained and queried for date of diagnosis,

histology, grade, molecular characteristics, lympho-vascular status, perineural status, tumor site, tumor

size, treatment response, margin status, lymph node involvement, extra-nodal extensions and stage.

Resection specimen pathology reports were further queried for quantification and location of lymph node

involvement, local invasion, evidence of metastatic disease and margin status.

2.2.6. JOHNS HOPKINS ADJUSTED CLINICAL GROUPS® SYSTEM

The Johns Hopkins Adjusted Clinical Groups® System (Version 10.0.1.), a validated predictor of 1-year

mortality [166] was used to derive a measure of baseline risk of death. The Johns Hopkins Adjusted Clinical

Groups concept stems from research examining the relationship between illness burden and service


utilization in pediatric populations, which validated that clustering of illnesses better predicted health

service resources than the presence of specific diseases. The now further refined ACG Case-Mix System

assigns all International Classification of Disease (ICD) codes to one of 32 diagnosis clusters known as

Aggregate Diagnosis Groups (ADG). Each disease or condition is placed into a single ADG based on five

clinical dimensions: (i) duration of the condition, (ii) severity of condition, (iii) diagnostic certainty, (iv)

etiology of the condition and (v) specialty care involvement. Diagnostic information is derived from

outpatient or ambulatory physician visit claims records, encounter records, inpatient hospital claims and

computerized discharge abstracts. ICD codes within the same ADG are similar in terms of clinical criteria

and expected healthcare utilization. Patients can have multiple ICD diagnostic codes and have up to 32

ADGs. Age, sex, presence of specific ADGs, number of major ADGs and total number of ADGs are then used

to form Adjusted Clinical Groups (ACGs). However, since patients within different ACGs may require the

same level of resources, and for parsimonious modelling, ACGs can be further combined into six classes of

Resource Utilization Bands: (0) No or Only Invalid Diagnosis, (1) Healthy Users, (2) Low Users, (3) Moderate

Users, (4) High Users and (5) Very High Users.

2.2.7. ONTARIO MARGINALIZATION INDEX

The Ontario Marginalization Index is a specialized database using census data to profile relative area-level

marginalization (socioeconomic information) at various geographic levels in Ontario [167]. It was used to

derive material deprivation as measure of poverty and is based on income, quality of housing, educational

attainment, and family structure characteristics. Material deprivation is scored from least marginalized

(quintile 1) to most marginalized (quintile 5).

2.2.8. CANCER ACTIVITY LEVEL REPORTING

The Cancer Activity Lever Reporting database is maintained by Cancer Care Ontario and mainly captures

information on systemic therapy delivery (chemotherapy and radiation therapy) and outpatient oncology

clinic visits occurring at one of the provincial Regional Cancer Centers since April 2005 [168].

2.2.9. REGISTERED PERSON DATABASE

The Registered Person Database is managed by the Ontario Ministry of Health and is updated bi-monthly.

It provides demographic data, vital status, and details such as date of last contact with the healthcare

system anywhere in Canada. It is enriched using other datasets at ICES collecting data on death [161].


2.3. PATIENT POPULATION

The thesis studies include adult Ontario residents diagnosed with gastric adenocarcinoma undergoing

curative-intent primary tumor resection between April 1, 2004 and March 31, 2015.

2.3.1. DEFINING PATIENTS WITH GASTRIC ADENOCARCINOMA

Patients with a diagnosis of malignant (International Classification of Disease for Oncology 3rd edition (ICD-

O-3) behaviour code = “3”) gastric tumor (ICD-O-3 topography code = “C16”) between October 1st, 2003

and March 31st, 2015 were initially identified from the OCR. Only those with adenocarcinoma according to

ICD-O-3 histology codes were included (Table for 2.3.2.). These ICD-O-3 codes were selected according to

expert opinion/previous studies [52, 64, 93, 121, 169], there have been no validation studies on this topic.

Patients were required to be Ontario Health Insurance Plan eligible adults and alive at the time of diagnosis.

2.3.2. TIMEFRAME

Patient cohorts were derived from the Ontario Cancer Registry database. It is important to note that the

index date for this study is date of surgery, and this date is not recorded within the Ontario Cancer Registry.

Given that this thesis pertains to patients undergoing curative-intent gastrectomy, according to expert

opinion and consensus (thesis committee and research group) we defined that patients undergoing

resection more than six months following diagnosis were unlikely to have curative gastrectomy. The Ontario

Cancer Registry collects date of diagnosis for all patients in Ontario, thus patients diagnosed with gastric

cancer between October 1st, 2003 and March 31st, 2015 were queried.

2.3.3. DEFINING PATIENTS UNDERGOING CURATIVE-INTENT GASTRECTOMY

We included patients deemed to have undergone a curative-intent gastrectomy, in other words a

gastrectomy performed as a means of curing the patient of their gastric cancer with the potential for

achieving Textbook Outcomes. All non-metastatic patients admitted as inpatients for their first elective

curative-intent gastrectomy 1 day to 6 months following diagnosis of gastric adenocarcinoma between April

1st, 2004 and March 31st, 2015 were included. Gastrectomy was defined according to the Canadian

Classification of Health Interventions (CCI) billing codes available in the CIHI dataset. The billing codes

selected for this study were developed by the CCO’s Gastric Cancer Surgery Quality Based Procedures (QBP)

methodology with the addition of similar historical codes and querying all gastric and gastroesophageal


resection codes recorded for the OCR cohort (Tables for 2.3.3.). Wedge gastric resections and all

endoscopic resections were excluded, as these are not adequate resections for gastric adenocarcinoma

except in rare cases of early gastric cancers. Patients with early gastric cancers undergoing endoscopic

resection cannot achieve Textbook Outcomes since their surgical specimens will never have greater than

15 lymph nodes, and they are more likely to have better long-term outcomes given their stage; this would

bias the results of the survival analysis.

Patients undergoing emergency surgery for bleeding, obstruction or perforation were excluded because

we could not confirm that the purpose of the surgery was cure and emergency surgeries are associated

with higher rates of morbidity and mortality. As such, patients with emergency surgeries are less likely to

achieve Textbook Outcomes and have shorter survival. In cases where patients were diagnosed following

their gastrectomy, responsible surgeons may not have known the underlying diagnosis and were less likely

to perform a proper oncologic resection with the necessary margins and lymph nodes.

Furthermore, patients undergoing gastrectomies greater than six months following diagnosis were felt to

be unlikely undergoing curative intent resection since time to resection impacts survival. Furthermore,

patients operated on prior to or on the same day of their diagnosis may not have undergone appropriate

curative gastrectomy in the context of unknown diagnosis. Thus, only patients undergoing resection 1 day

to 6 months following diagnosis were included. Gastrectomy date was available for 98% of patients as the

date of the CCI code, in the rest it was imputed from the date of the previous CCI code, since several CCI

codes are often coded together for one procedure with only the first assigned a date. Salvage gastrectomy

plays a limited role in patients with local recurrence and repeat gastrectomies are more likely for

complications than cure [170-172]. As such, only the first gastrectomy in patients with multiple resections

in the study period were included. The REGATTA trial [79], which demonstrated not only a lack of benefit,

but in fact the harmful effect of gastrectomy in the setting of a single non-curable finding, was published

in the same month as the end of this study’s inclusion period. Our research group has previously shown

that 37% of metastatic patients undergo gastrectomy [173] and that metastatic status has been shown to

be associated with more morbidity [174]. Furthermore, surgeons performing gastrectomies in metastatic

patients may prioritize function over cure and not adhere to standard oncologic targets, ultimately not

achieving Textbook Outcomes. Thus, we excluded all patients with metastatic disease. Patients with

metastatic disease were identified using an algorithm validated in an Ontario gastric cancer cohort [175]

and supplemented by the Esophagogastric Pathology Database stage variable when the pathologist

explicitly stated the presence of distant disease.


2.3.4. STUDY COHORT WITH PATHOLOGY REPORT

The cohort of patients with non-metastatic gastric adenocarcinoma who underwent curative-intent

gastrectomy was linked to the pathology report database using patient unique and anonymous ICES key

numbers (IKN). Given the potential for multiple reports per patient (biopsy, endoscopy, operative) rules

were set to identify the gastrectomy specimen pathology report. First, only pathology reports with a

“procedure date” within 30 days before or after the index gastrectomy date were kept. Next, only reports

with “report dates” 0 to 90 days following gastrectomy date were deemed eligible, as one would expect

that pathology review be reported within two to four weeks following surgery.

2.4. GASTRECTOMY VOLUME

Gastrectomy volumes, for both hospitals and surgeons, were defined as the average annual number of

institutionally performed gastrectomy procedures in the two years prior to the date of the index procedure.

Only gastrectomies performed non-urgently for in-patients 1 day to 6 months following diagnosis counted

towards gastrectomy volumes. The index procedure itself did not contribute to the numerator. This was

recorded for every patient included in this study. This definition has been used previously in an ICES-based

cancer volume-outcome study [176]. Compared to averaging the number of procedures over the years of

study, this definition accounts for year to year variation which may be present in our 11-year study period

and has more face-validity as procedures performed subsequent to the index procedure should not affect

the quality of care provided during the index procedure admission. Volumes were also considered as

categorical variables based on quintiles creating approximately equal number of patients (quintiles 1 to 5),

according to previously published studies [145, 150, 176, 177]. This method avoids using arbitrary numbers

from the literature which may not reflect the current state of care in Ontario for gastric cancer.

The gastrectomy cases included in the volume calculation are restricted to the rules applied to identify

patients undergoing gastrectomy. Most importantly, urgent gastrectomies and those performed outside of

the time-frame described (1 day to 6 months following diagnosis date) and those performed for non-gastric

adenocarcinomas likely represent a large portion of cases performed during the study period. As such, the

volumes reported in these studies are lower than what most surgeons would expect.

Patients included in this thesis were identified using CCI gastrectomy codes, which are inherently linked to

a discharge abstract and a hospital. Hospital volumes were calculated using the institution identifier in the

Discharge Abstract Database. Surgeons were identified by linking these discharge abstract records to the


Ontario Health Insurance Plan database when surgeons (general surgeons, thoracic surgeons,

cardiothoracic surgeons and vascular surgeons) billed gastrectomy and esophagogastrectomy codes (S089,

S090, S122, S123, S125, S128 and S129) within 30 days of the intervention date in the Discharge Abstract

Database.

2.5. TEXTBOOK OUTCOMES

Textbook Outcomes is based on the conceptual framework that patient-centered surgical care prioritizes

more than just post-operative mortality. This outcome is a composite of eight quality parameters related

to the oncologic resection, post-operative course and discharge of patients undergoing gastrectomy for

cancer.

1. Negative resection margins

2. Greater than 15 lymph nodes resected

3. No severe complications

4. No reinterventions

5. No unplanned ICU admission

6. Length of stay (LOS) 21 days or less

7. No readmission to hospital in 30 days following discharge

8. No death within 30 days of gastrectomy

2.5.1. NEGATIVE RESECTION MARGINS

Margin status was extracted from the Esophagogastric Pathology Database. Positive margin status was

defined by the pathology report. Three separate margins were queried for in each report: proximal, distal

and radial. Margin statuses were qualitatively defined: each margin was recorded as positive, negative or

unclear. In the case of multiple reports for the same patient, which may occur when a specimen is sent for

review, the qualitative value with the worse prognosis was favored and recorded. For example, if a patient’s

two reports stated that the proximal margin was unclear in one and positive in the other, a positive margin

was recorded. The Textbook Outcomes target was negative margins.

2.5.2. GREATER THAN 15 LYMPH NODES RESECTED


Lymph node counts were extracted from the Esophagogastric Pathology Database. In some reports it was

clearly stated how many nodes were found within the surgical specimen. In others, the number of nodes

from each station were enumerated; in these cases, the numbers were added up. If no nodes were

reported on, then 0 nodes were recorded. The Textbook Outcomes target was 16 or greater lymph nodes

sampled in the surgical specimen.

2.5.3. NO SEVERE COMPLICATIONS

Severe complications were defined according to ICD-10-CA diagnostic codes and CCI intervention codes.

Type 2 ICD-10-CA codes (post-admission comorbidities) were tabulated for the whole cohort. In order to

mitigate classification bias and data dredging, the frequency of each code was deleted from the

spreadsheet so that all that remained was the ICD-10-CA code and its description. Only codes determined

to represent an entity requiring pharmacological treatment (Clavien-Dindo Grade II), surgical, endoscopic

or radiological intervention (Clavien-Dindo Grade III) or ICU admission (Grade IV) were recorded [178].

These were reviewed and classified according to complication subtype. This process was repeated with CCI

intervention codes. These were stratified into two groups: those occurring on the same day as gastrectomy

and those occurring afterwards. All interventions occurring on the same day as gastrectomy which did not

fit into the standard gastrectomy procedure coding were deemed additional procedures and constituted

intra-operative complications. Interventions occurring after the day of surgery were split into those

representing complications and those representing re-interventions (described in 2.5.4.). A dichotomous

variable indicating whether any severe complication occurred was created and the Textbook Outcomes

target was no recorded severe complications during the post-operative period.

2.5.4. NO REINTERVENTIONS

The remaining “non-complication” CCI intervention codes representing procedures performed after the

day of gastrectomy were similarly tabulated and stratified into three groups: operative, endoscopic and

interventional radiology/percutaneous interventions. A dichotomous variable indicating whether any

reintervention occurred was created and the Textbook Outcomes target was no recorded reintervention

during the post-operative period.

2.5.5. NO UNPLANNED ICU ADMISSION


In CIHI, any admission to a medical, surgical, cardiac, coronary, trauma and burn intensive care unit or a

medical or surgical step-down unit is recorded as an admission to a special care unit (SCU) [179]. Patients

undergoing complex surgery in Ontario frequently undergo “observation” in one of these specialized, nurse

concentrated units, and admission to a unit on the day of gastrectomy does not imply a departure from

textbook care. Therefore, only patients admitted to an SCU for greater than 48 hours on the day of

gastrectomy, or those admitted on a subsequent day were recorded as having an unplanned ICU admission.

In some cases, patients are discharged from an SCU and admitted to a second SCU as part of the normal

post-operative progression, for example from a medical intensive care unit to a surgical step-down unit.

Thus, when an admission to an SCU occurred within 1 hour of a SCU discharge, this was recorded as a

“transfer” and the hours spent in each were added together; if these hours surpassed 48 hours, then this

would be recorded as an ICU admission. The Textbook Outcomes target was no unplanned ICU admission,

according to the definitions described.

2.5.6. LENGTH OF STAY 21 DAYS OR LESS

In CIHI, every admission to hospital is associated with an episode number and an episode visit number

[179]. Admissions with the same episode number are considered hospital transfers and represent a single

admission continuum. Length of stay (LOS) was measured from the day of gastrectomy until the last

discharge date for a given episode number. The Textbook Outcomes LOS target was 21 days and under as

defined by the original Dutch study [123]

2.5.7. NO 30-DAY READMISSION

Readmissions were defined as any admission to hospital within 30 days of the patient’s discharge date for

their gastrectomy admission. Only admissions with different episode numbers were recognized as separate

admissions. The Textbook Outcomes target was no readmission within 30 days of discharge, which is in line

with the original Dutch study [123] and the CIHI definition [180].

2.5.8. NO 30-DAY MORTALITY

The Registered Person Database was interrogated for date of death for all patients. If this date occurred

within 30 days of gastrectomy, then patients were recorded as having suffered a post-operative mortality.

The Textbook Outcomes mortality target was no death within 30 days of surgery, which is in line with the


original Dutch study [123] and the American College of Surgeons’ National Surgical Quality Improvement

Program definition [181].


2.6. CONFOUNDERS

This thesis pertains to two separate analyses of association: Textbook Outcomes with survival and

gastrectomy volume (surgeon and hospital) with Textbook Outcomes. The confounders for these

associations are depicted in Figures for 2.6. and the rationale for inclusion of each confounder is

subsequently described. Confounders are divided into patient and disease characteristics.

FIGURE FOR 2.6. CAUSAL FRAMEWORK FOR TEXTBOOK OUTCOME AND SURVIVAL STUDY

Disease CharacteristicsTumour Location, Histology Category, Lymphovascular

Invasion, Perineural Invasion, T and N Stage, Chemotherapy,

Radiation Therapy

Patient Characteristics

Age, Sex, Resource Utilization Band, Material Deprivation,

Year of Gastrectomy

Overall

Survival

Textbook

Outcomes

Confounders (Adjusted) – Rectangle

Mediators (Unadjusted) – Ellipse

Negative Margins>15 LN ResectedNo Complications

No Reinterventions

No ICU AdmissionLOS 21 Days or Less

No 30-Day ReadmissionNo 30-Day Mortality


FIGURE FOR 2.6. CAUSAL FRAMEWORK FOR VOLUME AND TEXTBOOK OUTCOME STUDY

Disease CharacteristicsTumour Location, Histology

Category, T and N Stage

Patient Characteristics

Age, Sex, Resource Utilization Band, Material Deprivation,

Year of Gastrectomy

Textbook

Outcomes

Gastrectomy

Volume

Confounders (Adjusted) – RectangleMediators (Unadjusted) – Ellipse

Surgery, Pathology,

Nursing, Radiology, Endoscopy and Allied

Health Expertise

Operating Room, Interventional Radiology Suite, Scanner and ICU

Availability


2.6.1. PATIENT LEVEL CHARACTERISTICS

Age (years) and sex (male, female) were available through the Discharge Abstract Database. Age is a

significant risk factor for both post-operative complications and mortality [182-184] and poor long term

survival [185] following gastrectomy. Whereas male sex has been identified as a poor prognostic indicator

[105, 186]. Patient comorbidity burden was measured according to Resource Utilization Bands (RUB 0 to

5). The ability to characterize and adjust for comorbidity burden is crucial in reducing bias when comparing

groups in observational studies. Material deprivation (quintile 1 to 5), derived from the Ontario

Marginalization Index [167], was used as measure of poverty and is based on income, quality of housing,

educational attainment, and family structure characteristics. Quintiles were based on 2011 dissemination

areas. Dissemination areas are small, relatively stable geographic units with a population of 400 to 700

persons; it is the smallest standard geographic area for which all census data are disseminated [187]. Year

of gastrectomy (2004 to 2015) was based on the date of the gastrectomy intervention code in CCI. We felt

that with regionalization of lung, esophageal, pancreas and liver surgery across Ontario, it was possible and

even likely that a natural centralization of gastrectomy would occur in parallel, given that surgeons

specializing in gastric cancer (thoracic surgeons, hepatopancreatobiliary surgeons and surgical oncologists)

would be affected by the regionalization of other cancer surgeries. Furthermore, it is our optimistic

expectation that surgical quality, multimodal therapy, ICU management and long-term survival would

improve over the 11-year study period.

2.6.2. DISEASE LEVEL CHARACTERISTICS

Tumor location (ICD-O-3 topography codes C16.0-16.9) was available through the Ontario Cancer Registry.

Proximal gastric cancers differ significantly from distal gastric cancers. Biologically, they are associated with

older patients with more advanced disease [188, 189] and anatomically they require more extensive

surgery (total gastrectomy vs. distal gastrectomy). Total gastrectomy is associated with more post-

operative complications and mortality and worse overall survival [188, 189]. Histology category (intestinal

vs. non-intestinal) was based on the ICD-O-3 morphological codes available through the Ontario Cancer

Registry database and supplemented by the Esophagogastric Pathology Database when the pathologist

reported on findings of linitis plastica, diffuse and/or signet ring cell histology. Non-intestinal histology, such

as signet ring cell, is associated with decreased rates of negative resection margins and median survival

[190]. Lymphovascular invasion (present or absent), perineural invasion (present or absent) and pathologic


T and N stages were available through the Esophagogastric Pathology Database. Lymphovascular and

perineural invasion are both associated with poor prognosis [191-194] and both may influence resection

margin status and the need for a more extensive resection, as such uneven distribution of these

characteristics could bias the results of the survival analysis. The very basis for staging tumors is

prognostication [186] and patients with larger tumors or more advanced T and N-stages have been found

to have lower rates of negative margins and worse overall survival [195, 196]. Chemotherapy treatment

(received, not received) was defined as at least three visits for treatment in the Activity Level Reporting

database or at least three billings for chemotherapy by a general practitioner, medical oncologist,

hematologist or internal medicine physician in the Ontario Health Insurance Plan physician billing database

using fee codes G281, G339, G345, G359, G381 and G388. Radiation treatment (received, not received)

was defined as at least three visits for treatment in the Activity Level Reporting database. Ontario Health

Insurance Plan billing was not used to identify radiation treatment because radiation oncology bills for

planning, thus patients can be billed for without ever undergoing treatment, furthermore the number of

treatments would be unquantifiable using this data source. Three visits or billings in the 183 days prior and

183 days subsequent to the date of gastrectomy were recorded as neoadjuvant and adjuvant, respectively.

Multimodal therapy with chemotherapy [197-200] and radiation [201, 202] have been associated with

improved long-term survival, and neoadjuvant treatment has been shown to improve negative margin rates

[203].

2.7. DATA VALIDITY

The data necessary to perform the studies included in this thesis are largely administrative collected at the

national and provincial level. Such administrative data allows for robust and efficient pragmatic analyses

but is not collected for specific research purposes and are thus vulnerable to issues with comprehensive

capture, coding and reporting. All of these pose a threat to the validity of studies using administartive data.

In order to facilitate identifying areas of data vulnerability, the steps from patient identification to analysis

are outlined in the following section.

The first step was identification of a patient with gastric cancer. This was done using the Ontario Cancer

Registry which captures the majority of incident cancer diagnoses in the province (96%) [159, 160]. The

diagnosis of gastric adenocarcinoma is recorded in this registry according to the International Classification

of Diseases for Oncology morphological codes identified by the pathologist. This is the first of several

occurrences where heterogeneity is introduced into the process secondary to subjectivity of the


pathologist. It is clear that the best method for creating a homogeneous cohort with homogeneously-

defined characteristics would be to have each included patient’s surgical specimen evaluated by several

trained, blinded and independent pathologists using standardized methods of evaluating each disease

characteristic. However, this is not economically, nor logistically, feasible for a retrospective and population

analysis of this size, and as such, there was no alternative to accepting the pathologist’s description as the

standard for all pathology-based characteristics.

Once a diagnosis of gastric adenocarcinoma was made, patients with a curative-intent gastrectomy were

selected. This was done using the CIHI-DAD CCI codes. CIHI has released several reports describing its

efforts to measure and improve the validity of its data holdings [204-206]. In 2012, CIHI released the results

of its 2009-2010 CIHI-DAD Data Quality re-abstraction study - a chart review of all patients in the country

(except territories due to small numbers) with an acute inpatients discharge date between April 1, 2009

and March 31, 2010 [206]. The proportion of interventions coded in the CIHI-DAD which were confirmed

on chart review was 96% in the country and 95% in Ontario. Specifically, therapeutic interventions on the

digestive and hepatobiliary tract were confirmed in 93% of records across the country. Whereas, examining

validity in the reverse direction, 88% of interventions identified in the chart review were recorded within

the CIHI-DAD across the country, and 86% in Ontario specifically. Therapeutic interventions on the digestive

and hepatobiliary tract fared worse of all and were recorded only 78% of the time. When an intervention

was available in both the CIHI-DAD and the chart review, coding consistency – whether the interventions

described matched on the type of health intervention, the anatomy site, the intervention performed, the

approach/technique, the device/method and the tissue involved in the procedure – was a perfect match in

91% and as high as 96% for a rubric match. A rubric match excludes describing approach, technique, device,

methods or tissue involved; these are not crucial in identifying gastrectomy interventions. These data

indicate that the capture rate of digestive interventions in the CIHI-DAD may be over- and under-recorded.

The risk of over-recoding is mitigated in this thesis by including only patients with gastric cancer specimen

pathology reports, and thus inclusion of non-gastrectomy patients is highly unlikely. The latter risk, that of

not capturing all patients undergoing gastrectomy is more problematic. The 2012 CIHI Data Quality Study

reported that interventions were identified in the chart review, but not recorded in the CIHI-DAD due to

“conflicting or vague documentation in the chart, which resulted in a difference in interpretation” in about

one-third of records, and “possible non-compliance with codebook directives/coding standards” in the

remaining two-thirds. It is possible that the CIHI-DAD underestimated the true number of patients

undergoing gastrectomy and influenced the number of patients included in this study. This could impact

the validity of this study if capture rates were systematically different across high-volume and low-volume


facilities, however the CIHI-DAD Data Quality Study did not report on inter-institution variation, and

moreover, one could not conclude in which direction such variation, if it existed, would bias the results of

this thesis. Given the potential for under-capturing patients undergoing gastric cancer, we considered

complementing the CIHI-DAD with the Ontario Health Insurance Plan database. However, this database is

populated according to physician billing fee codes with minimal scrutiny, whereas the CIHI-DAD is

populated by trained abstractors following stringent rules documented within the CIHI-DAD abstraction

manual. These analyses were limited to patients identified in the CIHI-DAD, a decision in-line with two

cancer atlases funded by ICES and Cancer Care Ontario and Cancer Care Ontario’s Quality Based Procedure

programs [168, 207, 208].

The CIHI-DAD was used to determine patient age, sex, RUB, year of gastrectomy, severe complications, re-

interventions, unplanned ICU admission, length of stay, readmissions. The validity of patient age, sex, year

of gastrectomy, length of stay and readmissions are outlined in a re-abstraction study led by ICES which

reported that in Ontario, sex was reliably recorded in the CIHI-DAD in 100% of cases, whereas age,

admission date and discharge date were reliably recorded in 99.9% of cases [209]. Reinterventions were

defined as interventions performed during the gastrectomy admission and the validity of these codes are

already covered in the preceding paragraph. RUB and severe complications were coded according to CIHI-

DAD diagnostic codes, with the addition of the Ontario Health Insurance Plan database fee codes for RUB.

The CIHI-DAD Data Quality Study determined that 84% of significant diagnoses reported to CIHI were

confirmed in the chart review and that 21% of significant diagnoses identified in the chart review were not

reported in the CIHI-DAD. As such, similar to interventions, it is possible that some complications were over

and under-recorded in the CIHI-DAD. It is unclear, whether variation in capture rates exists across hospital

volumes, and thus its impact on the validity of this study is unclear. Unplanned ICU admissions were

recorded according to the CIHI-DAD “special care unit” variable. A previous study [210] has investigated

the accuracy of administrative data in identifying patients admitted to the ICU in Ontario. The use of a

combination of CIHI SCU codes designating medical, surgical or combined medical-surgical intensive care

units accurately identified ICU admissions (sensitivity 92%, specificity 99%, positive predictive value 84%).

However, the gold-standard in this validation study did not include specialized ICU units (coronary, cardiac

surgery, burn and trauma), and thus it remains unclear whether inclusion of all SCU designations results in

the same accuracy as the proposed coding method.

The rest of the variables pertinent to these analyses were based on the Esophagogastric Pathology

Database and included tumor location, histology, lymphovascular invasion, perineural invasion, stage,

margin status and lymph node sampling. The missingness of such reports is discussed in Section 2.8. Up


until this point, data validity has been discussed in the context of the CIHI-DAD which is populated by

trained abstractors who follow a rigorous coding manual. Pathology reports in this study, originate from a

multitude of pathologists with different training backgrounds, clinical experience and focus, facility types

and availability of pathologist assistants [211]. There are guidelines on reporting for gastric cancer

specimens produced by the College of American Pathologists [212] and while these are not enforceable,

the introduction of synoptic reporting may have improved homogeneity and completeness of pathology

reports. A survey of Ontario pathologists in 2011 found that 92% to 98% of pathologists processed their

gastric specimens for vascular invasion, perineural invasion and signet ring cells, despite these elements

not being considered mandatory. However, only 67% of pathologists reported on using synoptic reporting,

despite 90% of pathologists identifying this as a useful process in gastric cancer specimen reporting [211].

2.8. MISSING DATA

Reliability and valid interpretation of the results of clinical studies are greatly influenced by the quality of

the data analyzed. Missing data are ubiquitous in all clinical research and pose a threat to causal inference.

Missing data can be classified as missing completely at random, missing at random and missing not at

random. Data are missing completely at random when missingness is not associated with any other

observed and unobserved variables in the dataset. Such that, the missing data have no relationship with

other known and unknown values in the data set and subjects with missing data represent a random sample

of the population [213]. When a missing completely at random mechanism is encountered, complete case

analysis (list-wise deletion of subjects missing data) or available-case analysis (variation in sample size for

different analyses) are appropriate since this should not result in biased estimates and will only reduce

statistical power [214]. These methods are also commonly used when a minimal amount of missing data is

encountered [214]. Data are missing at random when they can be completely explained by observed values

in the data set. When such a mechanism is encountered, multiple imputation can be used to determine the

missing values of data missing at random [215]. Data is missing not at random when it is associated with

variables not collected in the dataset and thus, data missing not at random cannot be estimated from

known values in the dataset. It is not routinely possible to distinguish between missingness mechanisms,

and unless conceptually explainable, the mechanism is typically assumed to be missing at random.

As mentioned above, the data necessary for the two planned studies can be categorized as data available

from the standard ICES datasets (the CIHI-DAD, the Ontario Health Insurance Plan database, the Johns

Hopkins Adjusted Clinical Groups® System, the Ontario Marginalization Index database, the Cancer Activity


Lever Reporting database, and the Registered Person Database) and data available from the

Esophagogastric Pathology Database.

Patient characteristics available directly from an ICES dataset, for which Data Quality Reports are available,

include age, sex and deprivation quintile. In the general population these are missing in 0%, 0% and 0.6%,

respectively [179]. Resource Utilization Bands are based on the CIHI-DAD diagnostic codes. Patients with

no previously recorded diagnoses would be assigned a RUB of 0, as such missingness is not possible but the

validity of a RUB of 0 is subject to the validity of diagnostic codes captured in the CIHI-DAD (discussed in

2.7.). Year of gastrectomy was computed from the Discharge Abstract Database “intervention date”

variables. There is a maximum of 20 interventions and their respective dates recorded per patient

admission, and missingness ranges from 49% (first intervention) to 100% (twentieth intervention).

However, the validity of interventions in the Discharge Abstract Database is as high as 95% in Ontario [206]

and the ranges reported by the Data Quality Reports likely represent a lack, rather than omission, of an

intervention. Furthermore, year of gastrectomy can be imputed from the admission date since all

gastrectomies are elective and likely to occur at the beginning of the admission.

Disease characteristics available directly from an ICES dataset, for which Data Quality Reports are available,

include tumor location, histology, chemotherapy treatment and radiation therapy. In the general

population these are missing in 0%, 0.4%, 0% and 0%, respectively. Importantly, chemotherapy and

radiation treatments recorded through the Activity Level Reporting database are only those performed at

Regional Cancer Centers and affiliated satellite sites [168]. This is appropriate for radiation therapy, which

is only provided at those respective sites, whereas chemotherapy, which is not restricted to those sites,

was further substantiated using the Ontario Health Insurance Plan database. Additionally, given that the

Activity Level Reporting database is only available as of April 2005 (outlined in 2.2.8.), sensitivity analyses

limited to relevant years are discussed further in Chapter 3. The primary exposures in the gastrectomy

volume and Textbook Outcomes analysis are hospital and surgeon-volumes. The ability to allocate a

hospital-level volume for each patient depends on determining the hospital where the gastrectomy was

performed. Patients are identified using CCI gastrectomy codes and are thus inherently linked to a

discharge abstract and a hospital. It is unlikely that any patient would be missing a hospital volume.

Identifying the surgeon performing the procedure is more complicated. In Ontario, 95% of physicians are

financially reimbursed by the Ministry of Ontario Health and Long-Term Care after submitting claims to the

Ontario Health Insurance Plan for each covered service provided. The Ontario Health Insurance Plan

database is a provincial dataset, recording all such claims and will be used to identify the surgeon

performing each gastrectomy. A previous study has found that 95% of cancer surgeries in CIHI can be linked


to the Ontario Health Insurance Plan database [162]. Given the relatively small proportion of patients

expected to be missing any of the variables, a complete case analysis was planned. Additionally, successful

linkage rates of patients to surgeons for surgeon volume was planned across included and excluded

patients for missing data.

The rest of variables pertinent to these analyses include Textbook Outcome metrics margin status and

adequate lymphadenectomy and pathology-based confounders including lymphovascular invasion,

perineural invasion and stage. All of these are obtained from the Esophagogastric Pathology Database. Our

research group has previously created a pathology database from OCR pathology reports for patients with

pancreas cancer and identified a large proportion of missing reports. Thus, we expected that a proportion

of eligible patients in this thesis will not be successfully linked to a pathology report. Given the sheer

number, complexity and variation of the variables required from the Esophagogastric Pathology Database,

it is highly unlikely that missing values in this dataset can be imputed from observed data. As such, we

planned a complete case analysis of patients with available pathology reports. Given the rare occurrence

of data missing completely at random, patients were stratified according to pathology report missingness

and statistical modelling was performed to determine whether study exposures and outcomes were

associated with pathology report missingness (Chapter 3).

2.9. POTENTIAL BIASES

2.9.1. SELECTION BIAS

No information is available on patient values, referral patterns and decision-making in regard to undergoing

surgery. This may lead to a self-selection bias where patients with higher health literacy are more likely to

undergo surgery. This would lead to differential inclusion favoring healthier patients which may hinder

generalizability of the study, briefly, the results of this study would only apply to patients referred for and

choosing to undergo surgery in the first place. Similarly, only patients with CCI gastrectomy intervention

codes are included in this study, as such those not captured are not represented, and it is not possible to

describe why uncaptured eligible patients were omitted.

The potential for differential loss to follow-up across volume groups is mitigated by measuring all metrics

making up Textbook Outcomes during the index admission (same admission as surgery) with the exception

of 30-day mortality and readmission. Death, as stated above, is nearly completely captured through the


Registered Person Database, whereas readmission could be differentially missed if patients from one of the

volume groups preferentially left the province for readmission, which is unlikely.

The most important selection bias present in this thesis is the exclusion of patients without pathology

reports. If these reports are missing not at random, then this selection bias may influence the results of the

studies in this thesis in an unpredictable way. It will be important to describe and compare the patients

excluded due to missing pathology reports, and most importantly determine whether pathology report

missingness is associated with Textbook Outcomes, surgeon/hospital gastrectomy volumes and/or survival.

For example, if patients in remote areas with limited access to health care, treated by the lowest volume

surgeons at the lowest volume hospitals were the most likely to be missing pathology reports, then the

results of the gastrectomy volume-Textbook Outcome study would be biased towards the null, because the

least favorable patient population would be excluded from the analysis.

2.9.2. MEASUREMENT BIAS

Volumes were systematically coded by a senior analyst for all patients prior to measuring Textbook

Outcomes for individual patients. CIHI abstractors are blinded to the study objective and CIHI-DAD

diagnostic and procedural codes have been validated [204-206, 216]. The Esophagogastric Pathology

Database was populated by two physician abstractors collecting pathology report data. They (i) did not

collect information on hospital volume and (ii) or collect institutional identifiers. However, hospital names

are available in the pathology reports, which may lead to an observer bias where results deviate away from

the null, if abstractors preferentially favored high volume centers. The bigger threat to these analyses lies

in the quality of the pathology reports themselves. As described previously, despite reporting

recommendations by the College of American Pathologists [212], not all pathology variables included in

these analyses were reported systematically. The first Textbook Outcome metric is margin status, which

according to the College of American Pathologists is determined to be positive if tumor is identified at the

margin (a distance of 0 millimetres). However, since the distances from the margins were not regularly

reported when a pathologist identified the margin as positive it is possible that a different definition was

used but not explicitly stated. It would have been more optimal if the distance was reported and our

physician abstractors would determine the margin status according to this distance, rather than what was

dictated by the pathologist. Similarly, determination of lymphovascular invasion, perineural invasion and

stage were left at the discretion of the pathologist which may have led to variability across institutions. If

pathologists at high volume institutions, with more experience, time and resources available to them were

more likely to identify patterns of tumor aggressiveness, positive margins and advanced stage, this would


lead to important measurement bias which could not be mitigated after the fact in these analyses. The only

solution to such heterogeneity would be to have an independent pathologist review the specimens of each

patient included in this study, however this represents an unfeasible undertaking.

2.9.3. CONFOUNDING BIAS

Gastrectomy volume and Textbook Outcomes and Textbook Outcomes and Survival share several common

causes which can be divided into patient and disease characteristics. The rationale for these confounders

is described in a section above.

The studies included in this thesis represent large, population-based analyses strengthened by the unique

opportunity to leverage administrative health data and pathology specimen data to measure surgical

quality outcomes and account for demographic, clinical and cancer-specific characteristics. As with all non-

randomized studies, confounding must be considered.

For the long-term survival analysis in Chapter 3, it was particularly important to adjust for differences in

baseline health status and tumor aggressiveness, since these impact patient survival. In terms of baseline

health status, we used age and RUB in order to control for potential and likely differences in the comparison

cohorts (those with and without Textbook Outcomes). As described further in the study, we expected that

these groups would differ significantly, and simply demonstrating an improved prognosis in patients

achieving Textbook Outcomes would not be enough to establish the importance of Textbook Outcomes in

this patient population. Rather, the aim was to demonstrate that in similar populations (those with the

same risk of death), those with Textbook Outcomes were more likely to survive longer than those without.

A cause-specific survival analysis, such as using recurrence as the outcome, would have been an

appropriate choice to mitigate any residual confounding based on baseline health status, since this would

account for cancer-related morbidity and mortality and not those attributable to other diseases. However,

recurrence is not currently captured within the administrative databases used. In terms of tumor

aggressiveness, if all the patients with smaller, less aggressive and less advanced cancers achieved Textbook

Outcomes, it could be assumed that it was the nature of their disease and not the quality of their peri-

operative care that had an impact on their prognosis. In order to mitigate this risk, we adjusted for tumor

location, histologic subtype, lymphovascular invasion, perineural invasion and tumor stage. We did not

identify additional confounders that were felt to be clinically meaningful for this association however the

risk of residual confounding remains, given the quality of the variables collected as described in 2.9.2.


For the Volume and Textbook Outcomes analysis, two important and likely scenarios were considered. It

was possible that high volume centers and surgeons attract patients with increased health literacy and, by

association, better health in general. This would bias our results away from the null, favoring high-volume

surgeons and institutions to achieve Textbook Outcomes. In order to mitigate this risk, we adjusted for

patient age, socioeconomic status and RUB. In this case, we did not have information about patient

preference and the reason why a patient underwent surgery at a particular institution or by a particular

surgeon. Patient values are not collected in administrative databases and represent an important missing

confounder in this analysis.

The other scenario considered was the possibility that high volume centers and surgeons perform more

difficult cases. This bias has been suggested when comparing outcomes across institutions or surgeons,

such as the National Surgical Quality Improvement Program [181] report cards. These report cards are

adjusted for the risk of unfavorable surgical outcomes which, similar to our analysis, are based on patient

age, sex, admission type (urgent vs. elective) baseline health status and comorbidities. However, NSQIP

also includes body mass index (BMI), which was not available for this analysis. BMI has been associated

with increased wound complications [217, 218], and if patients with higher BMI were more likely to be

operated on at a high volume institution or by a high volume surgeon this would bias the results of this

analysis towards the null and pose a threat to the validity of this analysis.


CHAPTER 3 RESULTS

3.1. OUTLINE OF RESULTS

The results of this thesis are presented in chapters 3.2. and 3.3. in manuscript format. Each of these

chapters includes its own set of tables and figures, numbered independently of the thesis itself. The

objectives of this thesis were two-fold, the first was to describe the association between Textbook

Outcomes and survival (Section 3.2.) and the second was to determine whether gastrectomy volume

predicts achieving Textbook Outcomes (Section 3.3.). Given their manuscript format, each chapter closes

with a discussion and conclusion, whereas this thesis has its own discussion and conclusion presented in

Chapter 4.


3.2. MANUSCRIPT #1: TEXTBOOK OUTCOMES AND SURVIVAL IN PATIENTS WITH

GASTRIC CANCER

Authors & Affiliations:

Jordan Levy, MD. Division of General Surgery, Department of Surgery and Institute of Health Policy,

Management, and Evaluation, University of Toronto, Canada.

Vaibhav Gupta, MD. Division of General Surgery, Department of Surgery and Institute of Health Policy,


Catherine Allen-Ayodabo, MD, MPH. Evaluative Clinical Sciences, Sunnybrook Research Institute, Toronto,

Canada.

Elmira Amirazodi, MD, MSc. Evaluative Clinical Sciences, Sunnybrook Research Institute, Toronto, Canada.

Naheed Jivraj, MD. Department of Anesthesia and Institute of Health Policy, Management, and

Evaluation, University of Toronto, Canada.

Alyson L Mahar, PhD. Manitoba Centre for Health Policy and Department of Community Health Sciences,

University of Manitoba, Canada.

Charles De Mestral, MD, PhD, Division of Vascular Surgery, Department of Surgery and Institute of Health

Policy, Management, and Evaluation, University of Toronto; Li Ka Shing Knowledge Institute and St.

Michael’s Hospital, Toronto, Canada.

Olli Saarela, PhD. Dalla Lana School of Public Health, University of Toronto, Canada.

Natalie G. Coburn, MD, MPH. Division of General Surgery, Department of Surgery and Institute of Health

Policy, Management, and Evaluation, University of Toronto; Sunnybrook Health Sciences Centre, Toronto,

Canada.

Conflicts of Interest: None

Funding: This study is supported by the Sherif and Mary Lou Hanna Chair in Surgical Oncology at

Sunnybrook Health Sciences Centre.

Acknowledgements: This study was supported by ICES, which is funded by an annual grant from the

Ontario Ministry of Health and Long-Term Care (MOHLTC). The opinions, results and conclusions reported

in this paper are those of the authors and are independent from the funding sources. No endorsement by

ICES or the Ontario MOHLTC is intended or should be inferred.


ABSTRACT

BACKGROUND

Gastric cancer represents a significant cause of cancer-related mortality in North America, where

population-based studies show that gastric cancer resections yield high rates of positive margins,

inadequate lymph node assessment and high peri-operative mortality. In order to assess surgical quality,

recently, a study from the Netherlands introduced the concept of Textbook Outcomes (TO): a composite

of surgical and post-operative metrics. However, the survival implications of TO remain unknown. The

objective of this study was to examine the association between Textbook Outcomes and long-term survival.

METHODS

This is a population-based retrospective cohort study of routinely collected administrative data and a

province-wide chart review of pathology reports. Adults with non-metastatic gastric adenocarcinoma

treated with gastrectomy between 2004 and 2015 were identified. Vital status information was available

until March 31, 2018. Post-operative outcomes were analyzed, and patients were assigned to TO vs. non-

TO groups. TO included (i) negative resection margins, (ii) greater than 15 lymph nodes sampled, (iii) no

severe complications, (iv) no re-interventions, (v) no unplanned ICU admission, (vi) length of stay (LOS) of

21 days or less, (vii) no readmission to hospital in 30 days following discharge and (viii) no mortality in the

30 days following surgery. Three-year survival was estimated using the Kaplan-Meier method. A marginal

Cox proportional hazards model accounting for clustering and regressed on patient confounders was used

to estimate the association between achieving TO metrics and long-term survival from gastrectomy date.

RESULTS

There were 1836 patients included in this study, of which 402 (22%) achieved all TO metrics. Patients

achieving TO were significantly younger, had fewer comorbidities, received multimodal therapy more

often, and had tumors that were less proximal and of lower T-stage. TO patients had a higher 3-year survival

rate compared to non-TO patients (75% vs. 55%, Log Rank p<0.001). Following adjustments for covariates

and clustering within hospitals, TO was associated with a 39% reduction in mortality (aHR 0.61 [95%CI 0.50,

0.74], p<0.001). These results were robust to exclusion of patients with gastroesophageal junctional tumors

and to potential residual confounding according to E-value methodology.

CONCLUSIONS

Achieving Textbook Outcomes is strongly associated with improved long-term survival in gastric cancer

patients and merits further focus in surgical quality improvement efforts.


INTRODUCTION

Worldwide, gastric cancer is one of the top three causes of cancer-related deaths [45, 46], however

prognosis varies significantly across countries. Despite a relatively low incidence in North America, gastric

cancer represents a significant cause of cancer-related mortality, with only 25 to 31% of patients surviving

5-years [50, 53]. In contrast, patients in high-incidence countries such as Japan and Korea have 5-year

survival rates of 65-75% [136, 137].

Factors contributing to the survival difference between high incidence countries and low incidence

countries have long been debated [138-141], and are thought to include patient characteristics, tumor

biology, early detection through screening programs, and differences in treatment. Western patients tend

to be diagnosed at an advanced age, with associated increased comorbidity burden and rates of obesity

[126, 138]. These patients have a higher incidence of Barret’s esophagus, with a resulting increase in

junctional tumors and decreased survival [126, 142]. Furthermore, patients in Japan and Korea have

benefited from national screening programs since 1983 and 1999, respectively, [55] and are twice as likely

to be diagnosed with locoregional as opposed to metastatic disease compared to North American patients

[53, 142, 143].

As many of these differences in case-mix are non-modifiable at the time of diagnosis, efforts to improve

outcomes for gastric cancer patients in North America must focus on optimizing their peri-operative and

oncologic care. Population studies in North America have reported perioperative mortality rates of 5% [121,

122], positive resection margins in 6 to 11% [134, 135] and inadequate lymph node sampling in 60 to 71%

[121, 129, 130]. These deficiencies may further contribute to the survival differential between East and

West.

Recently, a study from the Netherlands [144] introduced the concept of Textbook Outcomes (TO): a

composite of surgical quality metrics, including intent of surgery, resection margin status, adequate lymph

node sampling, intra-operative and post-operative complications, re-interventions, ICU admissions, length

of stay, readmissions and mortality rates. While Textbook Outcomes represent a composite of optimal

surgical, pathologic and clinical outcomes following gastric cancer resection, its long-term survival

implications remain unknown. The objective of this study was to examine the association between

Textbook Outcomes and long-term survival.


METHODS

STUDY DESIGN AND SETTING

This is a population-based retrospective cohort study using routinely collected administrative data and a

province-wide chart review of pathology reports. Adults who underwent gastrectomy for non-metastatic

gastric adenocarcinoma in Ontario were included. Ontario is Canada’s most populous province with over

14 million residents and a land mass greater than Texas and Arizona combined. In Ontario, all non-elective

health services are funded for all permanent residents through a universal and publicly administered health

care system. Reporting is in accordance with the REporting of studies Conducted using Observational

Routinely collected Data (RECORD) Statement [219] (Appendix 1).

DATA SOURCES AND MANAGEMENT

ICES is an independent, non-profit research institute whose legal status under Ontario’s health information

privacy law allows it to collect and analyze health care and demographic data, for health system evaluation

and improvement. Secure access to these data is governed by policies and procedures that are approved

by the Information and Privacy Commissioner of Ontario. These datasets were linked using unique encoded

identifiers and analyzed at ICES. These patients are identified through the Ontario Cancer Registry (OCR)

which captures 96% of incident cancer cases in Ontario [159, 160]. Patient demographics and vital status,

cancer center treatment information, inpatient and outpatient hospitalization records, physician billing

data, diagnostic codes and interventional codes are available at ICES for every resident covered by Ontario’s

public health insurance plan.

The Population Registry of Esophageal and Stomach Tumours in Ontario (PRESTO) is a population-based

clinical-pathological database of all adult patients with esophagogastric cancer (International Classification

of Disease for Oncology 3rd edition (ICD-O-3) topography codes “C15 and C16”) of any histology, diagnosed

and treated from 2002 onwards. It is derived from 18 datasets linked using unique encoded identifiers and

analyzed at ICES. Tumor-specific information is available through the Esophagogastric Pathology Database

which was populated by two physicians abstracting operative specimen pathology reports for all patients

within PRESTO. Surgical specimen pathology reports were abstracted for histology, tumor site, margin

status, lymph node count and stage.


STUDY COHORT

This study included Ontario residents diagnosed with gastric adenocarcinoma undergoing curative-intent

primary tumor resection between April 1, 2004 and March 31, 2015. Primary tumor site was defined

according to gastric ICD-O-3 topography code “C16”. ICD-O-3 morphological codes are included in Chapter

5 - Table for 2.3.2. Curative-intent gastrectomy definition was based on the following rules: (i) patients

must be electively admitted, (ii) zero days to six months following date of diagnosis for (iii) their first gastric

resection and (iv) have no evidence of metastatic disease at the time of their surgery. Gastrectomies were

identified through the hospital discharge abstract intervention codes (Chapter 5 - Table for 2.3.3.A/B).

Patients with metastatic disease at time of surgery were identified using an algorithm validated in an

Ontario gastric cancer cohort [175] and supplemented by the Esophagogastric Pathology Database stage

variable. Patients were excluded if their date of death erroneously preceded their diagnosis date, if they

were under the age of 18 years or if their resected tumor’s specimen pathology report was not available.

EXPOSURE

In this study, Textbook Outcome refers to a composite of eight quality metrics related to the oncologic

resection, post-operative course and discharge of patients undergoing gastrectomy for cancer. These

include (i) negative resection margins, (ii) greater than 15 lymph nodes sampled, (iii) no severe

complications, (iv) no re-interventions, (v) no unplanned ICU admission, (vi) length of stay (LOS) of 21 days

or less, (vii) no readmission to hospital in 30 days following discharge and (viii) no mortality in the 30 days

following surgery. Each of these metrics must be met in order to achieve a Textbook Outcome. This concept

was first proposed by the Dutch Upper Gastrointestinal Cancer Audit group [144] and we have modified it

to conform to this analysis and the data available through ICES. Intent of surgery was included in the original

study’s definition but was felt to represent a significant selection bias in this study given that metastatic

patients have less physiologic reserve, receive less extensive resections and are less likely to achieve TO.

Intra-operative complications could not be unambiguously differentiated from post-operative

complications within our datasets, and these were thus combined.

Resection margin status was defined according to the final pathology report (positive or negative). Lymph

node sampling was quantified in the pathology report. Severe Complications were defined according to the

hospital record’s post-admission comorbidity codes using the Canadian International Statistical

Classification of Diseases and Related Health Problems 10th Revision, and the Canadian Classification of


Health Intervention codes. Only codes determined to represent a complication requiring pharmacological

treatment (Clavien-Dindo Grade II), surgical, endoscopic or radiological intervention (Clavien-Dindo Grade

III) or ICU admission (Grade IV) were designated as severe complications (Chapter 5 - Table for 2.5.3.) [178].

Re-interventions were defined according to Canadian Classification of Health Intervention codes (Chapter

5 - Table for 2.5.4.). Patients admitted to a special care unit for longer than 48 hours from the day of their

gastrectomy, and those admitted on any subsequent day were considered to have required an unplanned

ICU admission; this definition was developed to avoid penalizing hospitals with protocols which require all

patients undergoing major surgery to be observed in a special care unit on the day of their surgery. Patient

LOS was measured from date of primary tumor resection until date of discharge. Thirty-day readmission

was defined as an admission to any hospital within Ontario in the 30 days following the discharge date.

Thirty-day mortality was defined as a death in the 30 days following gastrectomy and included both in-

hospital or home death.

OUTCOMES

The primary outcome of this study was overall survival, measured as the time between date of gastrectomy

and death. We reported on the 3-year survival rate in the TO and non-TO groups and the magnitude of the

association between TO and survival from date of gastrectomy. Maximum follow-up date was March 31,

2018.

COVARIATES

Age at diagnosis (years), sex (male, female) and tumor location (ICD-O-3 topography codes C16.0-16.9) were

available through the OCR. Resource utilization bands (RUB 0 to 5) were used in this study as a measure of

baseline risk of death; these are available through the Johns Hopkins Adjusted Clinical Groups® System

(Version 10.0.1) which is a validated predictor of 1-year mortality [166]. Material deprivation (quintile 1 to

5), available through the census-based, geographically derived Ontario Marginalization Index (ON-Marg)

[167] was used as measure of socioeconomic status and is based on income, quality of housing, educational

attainment, and family structure characteristics. Year of gastrectomy (2004 to 2015) was based on the date

of the gastrectomy intervention code. Histology category (intestinal vs. non-intestinal) was based on the

ICD-O-3 morphological codes available through OCR and supplemented by the Esophagogastric Pathology

Database when the pathologist reported on findings of linitis plastica, diffuse and/or signet ring cell


histology. Lymphovascular invasion (present or absent), perineural invasion (present or absent) and

pathologic T and N stages were available through the Esophagogastric Pathology Database. Chemotherapy

treatment (received, not received) was defined as at least three visits for treatment in the Activity Level

Reporting database (ALR) or at least three billings for chemotherapy by a general practitioner, medical

oncologist, hematologist or internal medicine physician in the Ontario Health Insurance Plan (OHIP)

physician billing database. Radiation treatment (received, not received) was defined as at least three visits

for treatment in the Activity Level Reporting database (ALR). Three visits or billings in the 183 days prior

and 183 days subsequent to the date of gastrectomy were recorded as neoadjuvant and adjuvant,

respectively.

STATISTICAL ANALYSIS

We first compared patients who did and did not achieve TO, as well as the frequency of each TO metric.

Next, we compared survival relative to achieving TO. In doing so, thirty-day mortality was removed from

the TO definition in order to prevent peri-operative mortality predicting overall survival. Three-year survival

rates were estimated using the Kaplan-Meier methods and the Log Rank Test was used to compare the

unadjusted survival functions in the TO and non-TO groups. Non-informative censoring on the maximum

date of follow-up was assumed.

A multivariable marginal Cox proportional hazards model was specified to characterize the association

between TO (binary exposure) and time to death. The proportional hazards assumption was verified by

testing the significance of a TO*time interaction term [220]. Furthermore, the survival analysis was

stratified by TO-status and the log-negative log survival and the cumulative hazards were plotted and

assessed graphically for parallel lines. The model accounted for clustering by hospital and adjusted for

patient-level confounders. In order to elucidate the independent associations of each TO metric (excluding

30-day mortality) with survival, the same model was run with all covariates and the seven TO metrics

replacing TO-status.

Covariates were selected a priori as potential confounders based on clinical grounds. These covariates were

assessed for statistical collinearity using a variation inflation factor cut-off of four. Influential outliers were

assessed for using “dfbeta” residuals. Model over-specification was assessed for according to an adopted

cut-off of 10 deaths for every degree of freedom. Statistical analyses were performed using SAS v9.4 (SAS


Institute Inc., Cary, NC, USA). All tests were two-sided with p-values of <0.05 and standardized differences

greater than 10% considered clinically meaningful.

SENSITIVITY ANALYSIS

We ran four sensitivity analyses. First, we excluded gastroesophageal junctional tumors, as these tumors

are often managed as esophageal tumors. These were defined as gastric cardia tumors (ICD-O-3

topography C16.0) with esophagogastrectomy in the Canadian Classification of Health Intervention codes.

Second, we set the 30th day following hospital discharge as “time zero” instead of gastrectomy date, and

excluded all patients dying prior to this date in order to prevent future events (TO metrics) from predicting

the outcome. Third, we restricted the analysis to patients who underwent gastrectomy as of October 1,

2005 as the ALR database used to partly define chemotherapy and completely define radiation treatment

was only available as of April 1, 2005. Patients who underwent surgery before October 1, 2005 could not

be accurately assessed for neoadjuvant therapy, while those who underwent surgery before April 1, 2005

could not be accurately assessed for adjuvant therapy. Fourth, we ran a fixed-effect multivariable cox

proportional hazards model regressed on identical covariates with the addition of institutional identifiers

in order to estimate identify institutional-level confounding on the association between TO and survival.

There are intrinsic differences between TO and non-TO patients; the resulting potential for residual

confounding may not be completely controlled for using available measured confounders. Sensitivity

analyses have often been used to assess the influence of unmeasured confounding on a study’s

conclusions. However, these make assumptions about prevalence, distribution and interactions of the

unmeasured confounder, and each assumption and decision introduces its own bias [221]. The “E-value”

[222] is an alternative approach to sensitivity analyses which avoids making assumptions. It represents the

minimum strength of association an unmeasured confounder would need to have with both the key

exposure (TO) and outcome (all-cause mortality), while controlling for other confounders, in order to

explain away the association between the key exposure and outcome. We computed an E-value for the

primary analysis using an online platform [222, 223].


MISSING DATA



A substantial amount of the data used in this study are available through validated databases [206, 209]

and missing data is quite rare (majority <1%) [179]. Unfortunately, this is not the case for the data held

within the Esophagogastric Pathology Database. A considerable proportion of patients were estimated to

be missing a pathology report, and secondarily, data to determine Textbook Outcome metrics. As such, a

mixed effects multivariable logistic regression accounting for clustering by hospital was used to determine

whether this study’s primary exposure (Textbook Outcomes) and/or outcome (survival) was associated with

pathology report missingness. All available covariates included in the primary analysis model were included

in the missing pathology data model.

ETHICAL APPROVAL

The use of data in this project was authorized under section 45 of Ontario’s Personal Health Information

Protection Act. This study was approved by the research ethics board of Sunnybrook Health Sciences

Centre and underwent privacy review at ICES.


RESULTS

STUDY COHORT

There were 2911 patients diagnosed with gastric adenocarcinoma who underwent an eligible resection

during the study period; all of these cases counted towards gastrectomy volumes. Of these, 2151 (74%)

had a pathology report available for extraction. Missingness of pathology reports was not significantly

associated with the non-pathology-based Textbook Outcome metrics (severe complications,

reinterventions, unplanned ICU admission, length of stay, 30-day readmission or 30-day mortality) or

surgeon volume, but was more likely at lower volume hospitals. As such, missingness of pathology reports

was not deemed a confounder in this analysis. Patients with metastatic disease (n=259) were excluded. Of

the remaining 1892 eligible patients, 1836 (97.0%) did not have any missing baseline demographic or tumor

characteristic data (Figure 1).

TEXTBOOK OUTCOMES

TO were achieved in 402 of 1836 patients (21.9%). The quality metrics that had the most negative impact

on the proportion achieving TO were adequate lymph nodes sampling and absence of severe complications,

which occurred in only 54.0% and 67.1% of cases, respectively. The distribution of each TO quality metric

for the non-TO group is presented in Figure 2. The baseline characteristics for the TO and non-TO groups

are presented in Table 1. TO patients were significantly younger (mean age 64.6 vs. 69.1, SD 37%), healthier

(RUB 5 14.9% vs. 19.5%, SD 12%), and received more neoadjuvant chemotherapy (15.7% vs. 10.5%, SD

15%), adjuvant chemotherapy (58.7% vs. 41.4%, SD 35%) and adjuvant radiation (40.5% vs. 26.6%, SD 30%).

Patients with TO had tumors that were less often localized to the GEJ (3.5% vs. 9.1%, 23%) and of lower T-

stage (T4a 20.1% vs 26.1%, SD 14%).


SURVIVAL

The 3-year survival functions of the TO and non-TO groups differed significantly according to the Log Rank

test (p<0.001, Figure 3). The overall survival functions show a clear separation in the unadjusted survival

curves from the onset, which widens for the remainder of the observation period. Indicating a long-term

effect of TO on patient survival. The 3-year survival proportions in the TO and non-TO differed significantly

(74.6% vs. 55.0%, LR p<0.001).

ASSOCIATION BETWEEN TEXTBOOK OUTCOMES AND SURVIVAL

The multivariable marginal Cox proportional hazards model was fitted to TO-status (excluding the 30-day

mortality metric) and the covariates described above. Removing 30-day mortality from the TO definition

had negligible effect on the number of patients achieving TO as there were less than six patients dying

within 30 days of their operation, who had no other TO metric excluding events. On univariate analysis, TO

was associated with a 47% decrease in the relative rate of death (HR 0.53 [95%CI 0.44, 0.62], p<0.001).

Following adjustments for covariates and clustering by hospital, TO was associated with a 39% decrease in

the relative rate of death (aHR 0.61 [95%CI 0.50, 0.74], p<0.001). Estimates for the multivariable marginal

model are reported in Table 2. Model variables were not collinear (all VIF < 1.2) and the proportional

hazards assumption was not violated according to a non-significant TO*time interaction term (p=1.00) and

by graphical inspection of TO-status stratified survival and cumulative hazard functions. There were no

influential outliers (all “dfbeta” centered around 0% and ranging from -2% to 0.5%) and the model was not

over-specified (30.2 deaths per degree of freedom).

The same model was run with all seven TO metrics (excluding 30-day mortality); none of these were

collinear (all VIF < 1.7). Negative margins, adequate lymph node sampling, no unplanned ICU admission and

no 30-day readmission were all found to be independently associated with survival (Table 3).


SENSITIVITY ANALYSES

The results of this analysis were robust to restricting to non-gastroesophageal junctional tumors (aHR 0.59

[95%CI 0.48, 0.72], p<0.001), measuring survival from 30-days following hospital discharge (aHR 0.64 [95%

CI 0.53, 0.79], p<0.001), restricting cohort to those with more reliable multimodal therapy assessment (aHR

0.61 [95% CI 0.51, 0.73], p<0.001), and including institutional identifiers in a fixed effect model instead of

adjusting for clustering by institution (aHR 0.60 [95% CI 0.50, 0.73], p<0.001).

Using the E-value methodology, we determined the strength of the unmeasured confounding necessary to

invalidate the results of the study. This simulated confounder would need to be associated with 1.8-fold

increased probability of being in the non-TO group and a 1.8-fold increased risk of death, while adjusting

for all other covariates outlined. None of the hazard ratio estimates presented in Table 2 reached such a

magnitude.


DISCUSSION

We conducted a population-based retrospective observational cohort study of 1836 patients undergoing

curative gastrectomy for adenocarcinoma. There are two major findings. First, 78% of patients did not

achieve TO and had high rates of positive margins (17%), inadequate lymphadenectomies (59%), intensive

care unit re-admissions (39%) and a high proportion required a readmission to hospital within 30-days of

discharge (19%). Second, patients with TO had significantly improved long-term survival, which persisted

following adjustment for all measured patient and tumor differences between both groups.

This study is the first to correlate TO, a new concept in cancer surgery, with survival, quantifying what most

clinicians know already--oncologically sound surgery void of complications and prolonged admissions result

in better short-term and long-term outcomes for patients. TO are an innovative composite metric of quality

markers that are feasible to measure and can be acted on in order to improve care. In terms of East and

West outcomes in gastric cancer, patients achieving TO had promising survival outcomes. These patients

also resembled patients in high incident countries and had better prognosis based on their younger age

[224], decreased comorbidities, and favorable tumor characteristics including smaller size, more distal

location and lower T-stage [186, 224]. As such, this study corroborates the implication of patient

characteristics and tumor biology in partially explaining the survival differential between East and West.

However, when controlling for all these factors, achieving TO was still strongly associated with improved

survival, implying that surgical quality moreover plays a critical role in patient survivorship. These results

were robust to exclusion of patients with GEJ tumors who were less likely to achieve TO and generally

present at later stages with worse overall survival [225, 226]. Given the surgico-pathologic implications of

TO, this study further supports the primary tenets of surgical oncology [66, 67, 227] – negative margins

[132, 133, 228] and adequate regional lymphadenectomy [224, 229, 230]. Both of these were independent

predictors of improved overall survival in the secondary analysis.

As described previously, the original TO definition in the Dutch study [144] was modified for this study in

two ways: curative intent was removed from the definition and intra-operative complications were

combined with post-operative severe complications. The latter was not felt to affect the results of this

study, but the former proved to be distinctly important. We should preface this description with the fact

that a survival analysis was not performed in the original paper and that its objective was to simply describe

the quality of the surgical care delivered to patients within the Netherlands. In this study, the curative intent

metric – namely whether the surgeon stated that the procedure was curative or palliative, was felt to

represent a selection bias that would persevere through any attempt at adjustment. The study period


ended one year prior to the release of REGATTA trial results [79]. This trial demonstrated not only a lack of

benefit, but in fact the harmful effect of gastrectomy in the setting of a single non-curable finding. As such,

we expected that some patients with metastatic disease would undergo a palliative gastrectomy without

adherence to resection margin and lymph node oncologic targets. These patients would be included in the

non-TO group and have worse overall survival secondary to their advanced disease. Thus, curative intent

was removed from the TO definition and we excluded all patients with metastatic disease identified in the

six months prior to surgery, or in the surgical specimen pathology report, those resected under emergent

circumstances, and those resected more than six months following diagnosis.

This study is a large, population-based analysis strengthened by the unique opportunity to leverage

administrative health data and pathology specimen data to measure surgical quality outcomes and account

for demographic, clinical and cancer-specific characteristics. As with all non-randomized studies,

confounding must be considered. In a long-term survival analysis such as this one, it is important to control

for comorbidities other than the primary malignancy as these may impact survival. A cause-specific survival

analysis, such as using recurrence as the outcome would have been an appropriate choice to mitigate this

confounding bias, however recurrence is not currently captured within most administrative databases. We

adjusted for age and RUB to assess factors that may impact upon clinical decision-making. While there may

be residual confounding, the necessary magnitude of such confounding (according to the E-value

computed) is larger than any other covariate in the model and TO-status itself, making such a scenario

highly unlikely, and the conclusions of this study robust to residual confounding. To put this into

perspective, in order for an unmeasured confounder to invalidate the results of this study, it would have to

be twice as common in the TO group and its effect estimate would have to have a magnitude superior to

comparing patients with T4b tumors to patients with T0-1 tumors. The most important limitation in this

study was the proportion of patients without pathology reports (26%). If missingness was informative, this

may affect the validity of this study. However, we used statistical modelling to determine that missingness

of a pathology report was not a confounder in the TO-survival relationship and as such the results of this

study remain valid and generalizable.

In conclusion, Textbook Outcomes are a new concept in gastric cancer surgery and represent a composite

of surgical quality metrics which is strongly associated with improved long-term survival. It is measurable,

reproducible and merits further focus in surgical quality improvement efforts in gastric cancer.


FIGURE 1. PATIENT FLOW DIAGRAM


TABLE 1. BASELINE DEMOGRAPHICS AND CHARACTERISTICS OF PATIENTS WITH

TEXTBOOK VS. NON-TEXTBOOK OUTCOMES

Characteristic TO (n=402)

Non-TO (n=1434)

SD

Age (years), mean (STD) 64.6 (12.5) 69.1 (11.9) 37%

Age Groups, n (%)

<50 60 (14.9%) 102 (7.1%) 25% 50–59 72 (17.9%) 199 (13.9%) 11%

60–69 108 (26.9%) 346 (24.1%) 6%

70–79 121 (30.1%) 505 (35.2%) 11%

>80 41 (10.2%) 282 (19.7%) 27%

Sex, n (%)

Female 150 (37.3%) 498 (34.7%) 5% Male 252 (62.7%) 936 (65.3%) 5%

Resource Utilization Band, n (%)

0_2 21 (5.2%) 43 (3.0%) 11%

3 214 (53.2%) 683 (47.6%) 11%

4 107 (26.6%) 428 (29.8%) 7%

5 60 (14.9%) 280 (19.5%) 12% Material Deprivation Quintile, n (%)

1 68 (16.9%) 208 (14.5%) 7%

2 63 (15.7%) 248 (17.3%) 4%

3 77 (19.2%) 296 (20.6%) 4%

4 89 (22.1%) 337 (23.5%) 3% 5 105 (26.1%) 345 (24.1%) 5%

Year of Gastrectomy, median (IQR) 2010 (2008-2013) 2010 (2007-2012) 19%

Neoadjuvant Chemotherapy, n (%)

Received 63 (15.7%) 151 (10.5%) 15%

Not Received 339 (84.3%) 1,283 (89.5%) 15%

Adjuvant Chemotherapy, n (%) Received 236 (58.7%) 593 (41.4%) 35%

Not Received 166 (41.3%) 841 (58.6%) 35%

Adjuvant Radiation, n (%)

Received 163 (40.5%) 381 (26.6%) 30%

Not Received 223 (55.5%) 985 (68.7%) 27%

Missing 16 (4.0%) 68 (4.7%) 4% Tumor Type, n (%)

Gastric 388 (96.5%) 1,304 (90.9%) 23%

Gastroesophageal Junction 14 (3.5%) 130 (9.1%) 23%

Tumor Location, n (%)

Cardia, NOS 41 (10.2%) 277 (19.3%) 26%

Fundus of stomach 12 (3.0%) 67 (4.7%) 9% Body of stomach 46 (11.4%) 174 (12.1%) 2%

Gastric antrum 151 (37.6%) 476 (33.2%) 9%

Pylorus 16 (4.0%) 48 (3.3%) 3%

Lesser curvature of stomach, NOS 71 (17.7%) 183 (12.8%) 14%

Greater curvature of stomach, NOS 21 (5.2%) 58 (4.0%) 6%

Overlapping lesion of stomach 8 (2.0%) 37 (2.6%) 4% Stomach, NOS 36 (9.0%) 114 (7.9%) 4%

Histology Category, n (%)


Intestinal 246 (61.2%) 931 (64.9%) 8%

Non-Intestinal 156 (38.8%) 503 (35.1%) 8% Lymphovascular Invasion, n (%)

Absent 219 (54.5%) 788 (55.0%) 1%

Present 183 (45.5%) 646 (45.0%) 1%

Perineural Invasion, n (%)

Absent 279 (69.4%) 1,021 (71.2%) 4%

Present 123 (30.6%) 413 (28.8%) 4% T-Stage, n (%)

T0-1 105 (26.1%) 327 (22.8%) 8%

T2 79 (19.7%) 233 (16.2%) 9%

T3 120 (29.9%) 440 (30.7%) 2%

T4a 81 (20.1%) 374 (26.1%) 14%

T4b 17 (4.2%) 60 (4.2%) 0% N-Stage, n (%)

N0 152 (37.8%) 603 (42.1%) 9%

N+ 250 (62.2%) 831 (57.9%) 9%

SD: Standardized Differences, STD: Standard Deviation, IQR: Interquartile Range Standardized Difference of 10% or greater considered clinically meaningful


FIGURE 2. DISTRIBUTION OF TO METRICS IN NON-TO PATIENTS

6%

19%

16%

39%

15%

42%

59%

17%

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

No Death

No Readmission

No Prolonged LOS

No Unplanned ICU

No Reinterventions

No Complications

>15 Nodes Sampled

Negative Margins

Achieved Metric Did Not Achieve Metric


FIGURE 3. KAPLAN MEIER 3-YEAR SURVIVAL CURVES FOR TO (BLUE) VS. NON-TO (RED)

WITH 95% CONFIDENCE BANDS


TABLE 2. ASSOCIATION BETWEEN TEXTBOOK OUTCOMES AND SURVIVAL

Variable Adjusted HR (95% CI) p-value

Textbook Outcomes* (TO) TO 0.61 (0.50, 0.74) <0.001 Non-TO Reference

Age Groups <50 0.38 (0.29, 0.50) <0.001 50–59 0.48 (0.38, 0.61) <0.001 60–69 0.53 (0.45, 0.66) <0.001 70–79 0.72 (0.61, 0.86) <0.001 >80 Reference

Sex Female 0.81 (0.70, 0.94) 0.005 Male Reference

Resource Utilization Bands 0_2 0.67 (0.50, 0.89) 0.005 3 0.67 (0.57, 0.78) <0.001 4 0.91 (0.77, 1.06) 0.23 5 Reference

Material Deprivation Quintile 1 1.00 (0.82, 1.22) 0.98 2 1.00 (0.80, 1.25) 0.98 3 1.07 (0.88, 1.31) 0.49 4 1.06 (0.90, 1.24) 0.49 5 Reference

Year of Gastrectomy (2004-2015) 0.97 (0.95, 0.99) <0.001

Neoadjuvant Chemotherapy

Received 1.13 (0.91, 1.41) 0.26

Not Received Reference Adjuvant Chemotherapy

Received 0.84 (0.73, 0.96) 0.01

Not Received Reference

Adjuvant Radiation

Received 0.68 (0.58, 0.79) <0.001

Not Received Reference

Missing 0.92 (0.69, 1.21) 0.53

Tumor Location

Cardia, NOS Reference

Fundus of stomach 0.73 (0.49, 1.08) 0.11

Body of stomach 0.73 (0.58, 0.92) 0.009

Gastric antrum 0.66 (0.55, 0.79) <0.001

Pylorus 0.66 (0.47, 0.92) 0.01

Lesser curvature of stomach, NOS 0.66 (0.53, 0.82) 0.002

Greater curvature of stomach, NOS 0.70 (0.49, 1.00) 0.05

Overlapping lesion of stomach 1.10 (0.71, 1.71) 0.68

Stomach, NOS 0.84 (0.66, 1.06) 0.14

Histology

Intestinal Type 0.79 (0.70, 0.89) <0.001

Non-Intestinal Type Reference

Lymphovascular Invasion Absent 0.77 (0.67, 0.88) <0.001


Present Reference

Perineural Invasion Absent 0.78 (0.67, 0.90) <0.001 Present Reference

T-Stage

T0-1 0.30 (0.19, 0.47) <0.001

T2 0.43 (0.29, 0.64) <0.001

T3 0.51 (0.35, 0.74) <0.001 T4a 0.81 (0.54, 1.23) 0.33

T4b Reference

N-Stage

N0 0.55 (0.46, 0.66) <0.001

N+ Reference 95% CI: 95% Confidence Interval * Excluding 30-day mortality Multivariable Marginal Cox Proportional Hazards Model adjusted for Textbook Outcomes (excluding 30-day

mortality), covariates and clustering by hospital. Omnibus Wald Sandwich (2(df), p-value): 2060 (34), <0.001.


TABLE 3. ASSOCIATION BETWEEN TO METRICS AND SURVIVAL

TO Quality Metric Adjusted HR (95% CI) p-value

Negative Margins Achieved 0.62 (0.51, 0.76) <0.001 Not Achieved Reference

>15 Nodes Sampled Achieved 0.80 (0.70, 0.91) 0.001 Not Achieved Reference

No Severe Complications Achieved 0.88 (0.75, 1.04) 0.13 Not Achieved Reference

No Re-Interventions Achieved 0.87 (0.66, 1.14) 0.30 Not Achieved Reference

No Unplanned ICU Admission Achieved 0.72 (0.63, 0.82) <0.001 Not Achieved Reference

No Prolonged LOS (21 days or less) Achieved 0.83 (0.67, 1.04) 0.11 Not Achieved Reference

No 30-Day Readmission Achieved 0.77 (0.66, 0.89) <0.001

Not Achieved Reference 95% CI: 95% Confidence Interval Multivariable Marginal Cox Proportional Hazards Model adjusted for seven TO metrics (excluding 30-day

mortality), covariates and clustering by hospital. Omnibus Wald Sandwich (2(df), p-value): 2510 (40), <0.001.


APPENDIX 1. RECORD STATEMENT CHECKLIST

# STROBE items RECORD items Location in manuscript

1 (a) Indicate the study’s design with a commonly used term in the title or the abstract (b) Provide in the abstract an informative and balanced summary of what was done and what was found

RECORD 1.1: The type of data used should be specified in the title or abstract. When possible, the name of the databases used should be included. RECORD 1.2: If applicable, the geographic region and timeframe within which the study took place should be reported in the title or abstract. RECORD 1.3: If linkage between databases was conducted for the study, this should be clearly stated in the title or abstract.

Abstract

Background rationale

2 Explain the scientific background and rationale for the investigation being reported

Abstract

Objectives 3 State specific objectives, including any prespecified hypotheses

Abstract

Study Design 4 Present key elements of study design early in the paper

Study Design and Setting

Setting 5 Describe the setting, locations, and relevant dates, including periods of recruitment, exposure, follow-up, and data collection

Study Design and Setting, Study Cohort

Participants 6 (a) Cohort study - Give the eligibility criteria, and the sources and methods of selection of participants. Describe methods of follow-up

RECORD 6.1: The methods of study population selection (such as codes or algorithms used to identify subjects) should be listed in detail. If this is not possible, an explanation should be provided.

Study Cohort


(b) Cohort study - For matched studies, give matching criteria and number of exposed and unexposed

RECORD 6.2: Any validation studies of the codes or algorithms used to select the population should be referenced. If validation was conducted for this study and not published elsewhere, detailed methods and results should be provided. RECORD 6.3: If the study involved linkage of databases, consider use of a flow diagram or other graphical display to demonstrate the data linkage process, including the number of individuals with linked data at each stage.

Variables 7 Clearly define all outcomes, exposures, predictors, potential confounders, and effect modifiers. Give diagnostic criteria, if applicable.

RECORD 7.1: A complete list of codes and algorithms used to classify exposures, outcomes, confounders, and effect modifiers should be provided. If these cannot be reported, an explanation should be provided.

Exposure, Outcomes, Covariates

Data sources/ measurement

8 For each variable of interest, give sources of data and details of methods of assessment (measurement). Describe comparability of assessment methods if there is more than one group

Data Sources and Management, Exposure, Outcomes, Covariates

Bias 9 Describe any efforts to address potential sources of bias

Data Sources and Management, Exposure, Outcomes, Covariates, Statistical Analysis, Sensitivity Analysis, Missing Data


Study size 10 Explain how the study size was arrived at

Study Cohort

Quantitative variables

11 Explain how quantitative variables were handled in the analyses. If applicable, describe which groupings were chosen, and why

Statistical Analysis

Statistical methods

12 (a) Describe all statistical methods, including those used to control for confounding (b) Describe any methods used to examine subgroups and interactions (c) Explain how missing data were addressed (d) Cohort study - If applicable, explain how loss to follow-up was addressed (e) Describe any sensitivity analyses

Statistical Analysis, Sensitivity Analysis, Missing Data

Data access and cleaning methods

.. RECORD 12.1: Authors should describe the extent to which the investigators had access to the database population used to create the study population. RECORD 12.2: Authors should provide information on the data cleaning methods used in the study.

Data Sources and Management, Study Cohort, Exposure, Outcome, Covariates

Linkage .. RECORD 12.3: State whether the study included person-level, institutional-level, or other data linkage across two or more databases. The methods of linkage and methods of linkage quality evaluation should be provided.



Participants 13 (a) Report the numbers of individuals at each stage of the study (e.g., numbers potentially eligible, examined for eligibility, confirmed eligible, included in the study, completing follow-up, and analysed) (b) Give reasons for non-participation at each stage. (c) Consider use of a flow diagram

RECORD 13.1: Describe in detail the selection of the persons included in the study (i.e., study population selection) including filtering based on data quality, data availability and linkage. The selection of included persons can be described in the text and/or by means of the study flow diagram.

Results: a. Study Cohort b. Figure 1

Descriptive data 14 (a) Give characteristics of study participants (e.g., demographic, clinical, social) and information on exposures and potential confounders (b) Indicate the number of participants with missing data for each variable of interest (c) Cohort study - summarise follow-up time (e.g., average and total amount)

Results: a. Gastrectomy Volumes, Tables 1-4 b. Figure 1 c. Study Cohort

Outcome data 15 Cohort study - Report numbers of outcome events or summary measures over time

Results: Textbook Outcomes, Figure 2

Main results 16 (a) Give unadjusted estimates and, if applicable, confounder-adjusted estimates and their precision (e.g., 95% confidence interval). Make clear which confounders were

Results: a. Survival, Figure 3, Association Between Textbook Outcomes and Survival, Table 2, Tables 3, Sensitivity Analyses b. Table 1 c. N/A


adjusted for and why they were included (b) Report category boundaries when continuous variables were categorized (c) If relevant, consider translating estimates of relative risk into absolute risk for a meaningful time period

Other analyses 17 Report other analyses done—e.g., analyses of subgroups and interactions, and sensitivity analyses

Results: Sensitivity Analyses

Key results 18 Summarise key results with reference to study objectives

Discussion: Paragraphs 1-2

Limitations 19 Discuss limitations of the study, taking into account sources of potential bias or imprecision. Discuss both direction and magnitude of any potential bias

RECORD 19.1: Discuss the implications of using data that were not created or collected to answer the specific research question(s). Include discussion of misclassification bias, unmeasured confounding, missing data, and changing eligibility over time, as they pertain to the study being reported.

Discussion: Paragraph 4

Interpretation 20 Give a cautious overall interpretation of results considering objectives, limitations, multiplicity of analyses, results from similar studies, and other relevant evidence

Discussion: Paragraph 1, 2, 5

Generalisability 21 Discuss the generalisability (external validity) of the study results



Funding 22 Give the source of funding and the role of the funders for the present study and, if applicable, for the original study on which the present article is based

Title Page “Funding”

Accessibility of protocol, raw data, and programming code

.. RECORD 22.1: Authors should provide information on how to access any supplemental information such as the study protocol, raw data, or programming code.

Appendices


3.3. MANUSCRIPT #2: GASTRECTOMY CASE VOLUME AND TEXTBOOK OUTCOMES

Authors & Affiliations:

Jordan Levy, MD. Division of General Surgery, Department of Surgery and Institute of Health Policy,


Vaibhav Gupta, MD. Division of General Surgery, Department of Surgery and Institute of Health Policy,


Elmira Amirazodi, MD, MSc. Evaluative Clinical Sciences, Sunnybrook Research Institute, Toronto, Canada.

Catherine Allen-Ayodabo, MD, MPH. Evaluative Clinical Sciences, Sunnybrook Research Institute, Toronto,

Canada.

Naheed Jivraj, MD. Department of Anesthesia and Institute of Health Policy, Management, and

Evaluation, University of Toronto, Canada.

Alyson L Mahar, PhD. Manitoba Centre for Health Policy and Department of Community Health Sciences,

University of Manitoba, Canada.

Charles De Mestral, MD, PhD, Division of Vascular Surgery, Department of Surgery and Institute of Health

Policy, Management, and Evaluation, University of Toronto; Li Ka Shing Knowledge Institute and St.

Michael’s Hospital, Toronto, Canada.

Olli Saarela, PhD. Dalla Lana School of Public Health, University of Toronto, Canada.

Natalie G. Coburn, MD, MPH. Division of General Surgery, Department of Surgery and Institute of Health

Policy, Management, and Evaluation, University of Toronto; Sunnybrook Health Sciences Centre, Toronto,

Canada.

Conflicts of Interest: None

Funding: This study is supported by the Sherif and Mary Lou Hanna Chair in Surgical Oncology at

Sunnybrook Health Sciences Centre.

Acknowledgements: This study was supported by ICES, which is funded by an annual grant from the

Ontario Ministry of Health and Long-Term Care (MOHLTC). The opinions, results and conclusions reported

in this paper are those of the authors and are independent from the funding sources. No endorsement by

ICES or the Ontario MOHLTC is intended or should be inferred.


ABSTRACT

BACKGROUND

Many gastric cancer resections in North America fall short of the standard of care. Increased experience

(volume) with gastric cancer surgery has been associated with decreased operative mortality. However,

death following surgery is a crude measure of surgical quality and there are several other important factors

pertaining to oncologic and post-operative outcomes. A recent Dutch study combined 10 surgical and post-

operative metrics to define high quality surgical care, entitled Textbook Outcomes (TO), and found that

rates of TO varied significantly across institutions. The objective of this study was to determine the

association between gastric cancer surgery case-volume at the surgeon and hospital level and Textbook

Outcomes.

METHODS

This is a population-based cohort study using routinely collected administrative data and a province-wide

chart review of pathology reports. Adults with non-metastatic gastric adenocarcinoma and treated with

gastrectomy between 2004 and 2015 were identified from a population-based cancer registry. TO included

(i) negative resection margins, (ii) greater than 15 lymph nodes sampled, (iii) no severe complications, (iv)

no re-interventions, (v) no unplanned ICU admission, (vi) length of stay (LOS) of 21 days or less, (vii) no

readmission to hospital in 30 days following discharge and (viii) no mortality in the 30 days following

surgery. Patients meeting each of these metrics were identified as achieving TO. We used multivariable

generalized estimating equation (GEE) logistic regression modelling to estimate the association between

surgeon and hospital gastrectomy volumes and Textbook Outcomes. Volumes were considered as

continuous variables and quintiles.

RESULTS



which occurred in only 53.9% and 68.3% of patients, respectively. TO were achieved in a similar proportion

of the patients treated by surgeons in the highest volume quintile (median volume: 4.5 annual

gastrectomies) and lowest volume quintile (median volume: 0 gastrectomies annually) (24.0% vs. 20.8%,

standardized difference (SD) 8%), whereas it was achieved in a higher proportion of the patients treated

at the hospitals in the highest volume quintile (median volume: 14.5 gastrectomies annually) compared to

the lowest volume quintile (median volume: 1 gastrectomy annually) (23.5% vs. 16.2%, SD 18%). However,

following adjustment for covariates and clustering, neither surgeon volume, nor hospital volume was

significantly associated with TO in Ontario.

CONCLUSIONS

In conclusion, improving gastric cancer surgery quality may require more than increasing volumes through

centralization. Future centralization policies should consider using case-mix adjusted Textbook Outcome

rates as a means of identifying high-quality surgeons and providing feedback in order to systematically

improve gastric cancer surgical quality at a population level.


INTRODUCTION

Many gastric cancer resections in North America fall short of the standard of care. Current guidelines for

oncologically sound resections recommend negative margins and adequate lymph node sampling (greater

than 15) [66, 67, 77]. However, population-based studies in the United States and Canada have reported

positive resection margins in 6 to 11% of patients [134, 135], inadequate lymph node sampling in 60 to 71%

[121, 129, 130, 231], and post-operative mortality rates of 5% [121, 122].

Increased experience (volume) with gastric cancer surgery at the hospital- [145, 146, 148, 150] and

surgeon-level [131, 145, 146] have both been associated with decreased operative mortality in multiple

North American studies. Several studies have investigated the effect of both hospital and surgeon-

experience on gastric cancer patient outcomes [131, 145, 146]. One such study found that surgeon volume

accounted for 46% to 54% of the apparent effect of hospital volume on post-operative mortality in pancreas

and esophageal resections [158]. Gastrectomies share many similarities with pancreatectomies and

esophagectomies, and the effect of volume on post-operative mortality is likely equally attributable to

surgeon and hospital volume.

However, death following surgery is a crude measure of surgical quality and represents just one of many

factors important to patients, their families and their health care professionals. One U.S. state-wide analysis

found that greater hospital-volume was also associated with decreased length of stay and cost in gastric

cancer surgery [156], and a National Cancer Database analysis found that higher-volume hospitals

examined more lymph nodes [157]. In 2017, a Dutch study combined 10 metrics to define high quality

surgical care, henceforth entitled Textbook Outcomes (TO) [144]. This composite measure included intent

of surgery, resection margin status, lymph nodes sampled, intra-operative and post-operative

complications, re-interventions, ICU admissions, length of stay, readmissions and mortality rates.

Achieving Textbook Outcomes likely depends on expert decision-making, surgical technique and the

processes of care available at an institution. We hypothesized that centers and surgeons with greater

experience in gastric cancer surgery are more likely to achieve a Textbook Outcome. The present study

leverages the clinical and pathologic data in the Population Registry of Esophageal and Stomach Tumours

of Ontario (PRESTO) to determine the association between gastric cancer surgery case-volume and

Textbook Outcomes.


METHODS

STUDY DESIGN AND SETTING

This is a population-based retrospective cohort study using routinely collected administrative data and a

province-wide chart review of pathology reports. Adults who underwent curative gastrectomy for non-

metastatic gastric adenocarcinoma in Ontario were included. Ontario is Canada’s most populous province

with over 14 million residents and a land mass greater than Texas and Arizona combined. In Ontario, all

non-elective health services are funded for all permanent residents through a universal and publicly

administered health care system. Reporting is in accordance with the REporting of studies Conducted using

Observational Routinely collected Data (RECORD) Statement [219] (Appendix 3).

DATA SOURCES AND MANAGEMENT

The Population Registry of Esophageal and Stomach Tumours in Ontario (PRESTO) is a population-based

clinical-pathological database of all adult patients with esophagogastric cancer (International Classification

of Disease for Oncology 3rd edition (ICD-O-3) topography codes “C15 and C16”) of any histology, diagnosed

and treated from 2002 onwards. It is derived from 18 datasets linked using unique encoded identifiers and

analyzed at ICES. ICES is an independent, non-profit research institute whose legal status under Ontario’s

health information privacy law allows it to collect and analyze health care and demographic data, for health

system evaluation and improvement. Secure access to these data is governed by policies and procedures

that are approved by the Information and Privacy Commissioner of Ontario. These datasets were linked

using unique encoded identifiers and analyzed at ICES. These patients are identified through the Ontario

Cancer Registry (OCR) which captures 96% of incident cancer cases in Ontario [159, 160]. Patient

demographics and vital status, cancer center treatment information, inpatient and outpatient

hospitalization records, physician billing data, diagnostic codes and interventional codes are available at

ICES for every resident covered by Ontario’s public health insurance plan. Tumor-specific information is

available through the Esophagogastric Pathology Database which was populated by two physicians

abstracting operative specimen pathology reports for all patients within PRESTO (EA, CA). Surgical specimen

pathology reports were abstracted for histology, tumor site, margin status, lymph node count and stage.


STUDY COHORT

This study included Ontario residents diagnosed with gastric adenocarcinoma undergoing curative-intent

primary tumor resection between April 1, 2004 and March 31, 2015. Primary tumor site was defined

according to gastric ICD-O-3 topography code “C16”. ICD-O-3 morphological codes are included in Chapter

5 - Table for 2.3.2. Curative-intent gastrectomy definition was based on the following rules: (i) patients

must be electively admitted, (ii) one day to six months following date of diagnosis for (iii) their first gastric

resection and (iv) have no evidence of metastatic disease at the time of their surgery. Gastrectomies were

identified through the hospital discharge abstract intervention codes (Chapter 5 - Table for 2.3.3.A/B).

Patients with metastatic disease at time of surgery were identified using an algorithm validated in an

Ontario gastric cancer cohort [175] and supplemented by the Esophagogastric Pathology Database stage

variable. Patients were excluded if their date of death preceded their diagnosis date, if they were under

the age of 18 years or if their resected tumor’s specimen pathology report was not available. Tumors at the

gastroesophageal junction are often managed as esophageal cancers and esophageal surgery has been

regionalized in Ontario since 2005. As such, we excluded patients with gastroesophageal junctional tumors

undergoing esophagectomies as these would be overrepresented in high-volume hospitals.

Esophagectomy was defined according to ICD-O-3 topography codes (C16.0) and esophagectomy

Intervention codes (Chapter 5 - Table for 2.3.3.A/B).

EXPOSURE

The main exposures in this study are hospital- and surgeon-volumes. Patients undergoing gastrectomies

were identified through the hospital discharge abstract database intervention codes (Chapter 5 - Table for

2.3.3.A/B) and were thus inherently linked to a hospital. As such, hospitals and their respective volume

were identified for each patient. Surgeons were identified through linkage to the provincial physician billing

database using fee codes included in Appendix 1, in order to reduce any potential bias in penalizing low

volume surgeons, we excluded physicians not identified as general surgeons, thoracic surgeons,

cardiothoracic surgeons or vascular surgeons. We estimated that 95% [162] of patients would be

successfully linked and planned a complete case analysis (excluding all patients without a surgeon) for the

analysis.

Only gastrectomies performed non-urgently for in-patients one day to six months following diagnosis

counted towards gastrectomy volumes. The index procedure itself did not contribute to the numerator.


Further, only gastrectomies performed for resection of adenocarcinoma were considered. Given these

restrictions, the volumes reported in this study are lower than expected. Volumes were also considered as

categorical variables by ranking surgeons and hospitals in order of increasing annual volumes and selecting

cut-points that most closely sorted patients into five evenly sized groups (quintiles 1 to 5), according to

previously published studies [145, 150, 176, 177]. This method avoids using arbitrary numbers from the

literature which may not reflect the current state of care in Ontario for gastric cancer.

Volume was defined using the number of gastrectomy procedures performed in the two years prior to the

date of the index procedure for each patient’s surgeon and hospital, divided by two years for an average

annual volume [176]. In order to measure volumes for patients in the first two years of the study,

gastrectomies performed two years prior to the start date of this study were included to calculate volumes.

This definition has been used previously in an ICES-based cancer volume-outcome study [176]. Compared

to averaging the number of procedures over the years of study, this definition accounts for year to year

variation which may be present in our 11-year study period and has more face-validity as procedures

performed subsequent to the index procedure should not affect the quality of care provided during the

index procedure admission. The unit of analysis in this study is the patient. In this regard, surgeon and

hospital volumes are recorded for every patient, and specific surgeons and hospitals may have several

different volumes (and volume quintiles) as these changed throughout the study period.

OUTCOMES

In this study, Textbook Outcome refers to a composite of eight quality metrics related to the oncologic

resection, post-operative course and discharge of patients undergoing gastrectomy for cancer. These

include (i) negative resection margins, (ii) greater than 15 lymph nodes sampled, (iii) no severe

complications, (iv) no re-interventions, (v) no unplanned ICU admission, (vi) length of stay (LOS) of 21 days

or less, (vii) no readmission to hospital in 30 days following discharge and (viii) no mortality in the 30 days

following surgery. Each of these metrics must be met in order to achieve a Textbook Outcome. This concept

was first proposed by the Dutch Upper Gastrointestinal Cancer Audit group [144] and we have modified it

to conform to this analysis and the data available through ICES. Intent of surgery was included in the original

study’s definition but was felt to represent a significant selection bias in this study given that metastatic

patients have less physiologic reserve, receive less extensive resections and are less likely to achieve TO.

Intra-operative complications could not be unambiguously differentiated from post-operative

complications and these were thus combined.


Resection margin status was defined according to the final pathology report (positive or negative). Lymph

node sampling was quantified in the pathology report. Severe Complications were defined according to the

hospital record’s post-admission comorbidity codes using the Canadian International Statistical

Classification of Diseases and Related Health Problems 10th Revision, and the Canadian Classification of

Health Intervention codes. Only codes determined to represent a complication requiring pharmacological

treatment (Clavien-Dindo Grade II), surgical, endoscopic or radiological intervention (Clavien-Dindo Grade

III) or ICU admission (Grade IV) were designated as severe complications (Chapter 5 - Table for 2.5.3.) [178].

Re-interventions were defined according to Canadian Classification of Health Intervention codes (Chapter

5 - Table for 2.5.4.). Patients admitted to a special care unit for longer than 48 hours from the day of their

gastrectomy, and those admitted on any subsequent day were considered to have required an unplanned

ICU admission; this definition was developed to avoid penalizing hospitals with protocols which require all

patients undergoing major surgery to be observed in a special care unit on the day of their surgery. Patient

LOS was measured from date of primary tumor resection until date of discharge. Thirty-day readmission

was defined as an admission to any hospital within Ontario in the 30 days following the discharge date.

Thirty-day mortality was defined as a death in the 30 days following gastrectomy and included both in-

hospital or home death.

COVARIATES

Age at diagnosis (years), sex (male, female) and tumor location (ICD-O-3 topography codes C16.0-16.9) were

available through the OCR. Resource utilization bands (RUB 0 to 5) were used in this study as a measure of

baseline risk of death; these are available through the Johns Hopkins Adjusted Clinical Groups® System

(Version 10.0.1) which is a validated predictor of 1-year mortality [166]. Material deprivation (quintile 1 to

5), available through the census-based, geographically derived Ontario Marginalization Index (ON-Marg)

[167] was used as measure of socioeconomic status and is based on income, quality of housing, educational

attainment, and family structure characteristics. Year of gastrectomy (2004 to 2015) was based on the date

of the gastrectomy intervention code in the Canadian Classification of Health Intervention codes. Histology

category (intestinal vs. non-intestinal) was based on the ICD-O-3 morphological codes available through

OCR and supplemented by the Esophagogastric Pathology Database when the pathologist reported on

findings of linitis plastica, diffuse and/or signet ring cell histology. Pathologic T and N stages were available

through the Esophagogastric Pathology Database.


STATISTICAL ANALYSIS

Patient baseline demographics and tumor characteristics of the highest surgeon quintile (Q5) and the

lowest surgeon volume quintile (Q1) were compared using standardized differences (SD). This was

repeated using hospital volume quintiles.

A multivariable generalized estimating equation (GEE) logistic model was specified to characterize the

association between surgeon volume (continuous distribution) and Textbook Outcomes as the odd ratio of

achieving Textbook Outcomes with increasing surgeon volume. The model accounted for clustering by

surgeon and adjusted for patient-level confounders. Covariates were selected a priori as potential

confounders based on clinical grounds. These covariates were assessed for statistical collinearity using a

variation inflation factor cut-off of four. This was repeated with hospital volume (continuous distribution).

In order to estimate the association between surgeon and hospital volumes and Textbook Outcomes as the

odd ratio of achieving Textbook Outcomes with increasing surgeon and hospital volumes, both volume

variables were subsequently included in the model. Naïve model (non-GEE) calibration and discrimination

were assessed using the Hosmer and Lemeshow Goodness-of-Fit test (satisfactory fit if p>0.05) and the “c”

statistic, respectively [232]. Influential observations were detected using “dfbeta” residuals. Violations in

log-linearity of surgeon and hospital volumes as continuous variables were tested for using deviance

residuals. Model over-specification was assessed, according to an adopted cut-off of 10 events (Textbook

Outcomes) for every degree of freedom.

For ease of interpretation, this was repeated using surgeon volume quintiles (Surgeon Q5 vs. Q1) and

hospital volume quintiles (Hospital Q5 vs. Q1). Correlation between surgeon and hospital volumes as

continuous variables was determined using the Spearman Correlation Coefficient (significant if p<0.05).

Correlation between surgeon and hospital volumes as categorical quintiles was determined using the

Polychoric Correlation Coefficient (significant if p<0.05). Multicollinearity was determined according to a

VIF cut-off of four. In order to elucidate the associations between surgeon volume quintiles, hospital

volume quintiles and TO metrics, identical models were fitted with both volume quintile variables and

covariates for each TO metric independently.

Statistical analyses were performed using SAS v9.4 (SAS Institute Inc., Cary, NC, USA). All tests were two-

sided with p-values of <0.05 considered statistically significant and standardized differences greater than

10% considered clinically meaningful.


MISSING DATA



A substantial amount of the data used in this study are available through validated databases [206, 209]

and missing data is quite rare (majority <1%) [179]. Unfortunately, this is not the case for the data held

within the Esophagogastric Pathology Database. A considerable proportion of patients were estimated to

be missing a pathology report and secondarily, data to determine Textbook Outcome metrics. As such, a

mixed effects multivariable logistic regression accounting for clustering by hospital was used to determine

whether this study’s primary exposure (gastrectomy volume) and/or outcome (Textbook Outcomes) was

associated with pathology report missingness. All available covariates included in the primary analysis

model were included in the missing pathology data model.

ETHICAL APPROVAL

The use of data in this project was authorized under section 45 of Ontario’s Personal Health Information

Protection Act. This study was approved by the research ethics board of Sunnybrook Health Sciences

Centre and underwent privacy review at ICES.


RESULTS

STUDY COHORT

There were 2911 patients diagnosed with gastric adenocarcinoma who underwent an eligible resection

during the study period; all of these cases counted towards gastrectomy volumes. Of these, 2151 (74%)

had a pathology report available for extraction. Missingness of pathology reports was not significantly

associated with the non-pathology based Textbook Outcome metrics (severe complications,

reinterventions, unplanned ICU admission, length of stay, 30-day readmission or 30-day mortality) or

surgeon volume but was more likely at lower volume hospitals. As such missingness of pathology reports

was not deemed a confounder in this analysis. Patients with metastatic disease (n=259) and those with

gastroesophageal junctional tumors (n=150) were excluded. Of the remaining 1742 eligible patients, 1707

(98%) were successfully linked to a surgeon using the physician billing database. Finally, following exclusion

of patients with missing baseline demographic or tumor characteristic data (3%), a study cohort of 1660

patients was available for analysis (Figure 1).

TEXTBOOK OUTCOMES



which occurred in only 53.9% and 68.3% of patients, respectively. The distribution of each quality metric

and their effect on the TO rate are presented in Figure 2. The proportion of TO varied significantly by year

of surgery and displayed a significant and positive trend (Cochran-Armitage Trend test, p<0.001); it was

20.3% in 2004 and 29.3% in 2015 (Figure 3). On adjusted GEE logistic regression, every increase in year of

gastrectomy (2004 to 2015) was independently associated with an 8% increase in the odds of achieving TO

(aOR 1.08 [1.03-1.13], p=0.001); patients undergoing gastrectomy in 2015 were 88% more likely to achieve

TO than patients in 2004.


GASTRECTOMY VOLUMES

The 1660 patients included in the study cohort were divided into five approximately equally-sized patient

groups based on (i) their surgeon’s annual volume (surgeon volume quintiles Q1 to Q5) and (ii) their

hospital’s annual volume (hospital volume quintiles Q1 to Q5). The number of patients, surgeons, hospitals,

volume cut points and median volumes for each volume-patient group are presented in Tables 1 and 2.

Volumes across Ontario were low. There were 355 surgeons operating at 69 hospitals. The highest and

lowest volume quintile surgeons performed 3.5 to 9.5 and 0 annual gastrectomies, respectively. Whereas

the highest and lowest volume quintiles hospitals performed 12 to 22 and 0 to 2 annual gastrectomies,

respectively. Surgeons performed a median of a single case a year, and hospitals performed a median of

6.5 cases annually. The highest volume surgeons, or those in the top quintile, performed a median of 4.5

cases annually. The highest volume hospitals performed a median of 14.5 cases annually. Patients were

much more likely to undergo surgery by the lowest volume surgeons if they went to the lowest volume

hospitals than if they went to the highest volume hospitals (38.5% vs. 12.4%). Baseline characteristics and

statistical comparisons using standardized differences between patients treated by the highest compared

to the lowest volume surgeons (surgeon Q5 vs. Q1) and between those treated at the highest compared to

the lowest volumes hospitals (hospital Q5 vs. Q1) are presented in Tables 3 and 4. Patients in the highest

surgeon and hospital quintiles were more contemporary (had their gastrectomy more recently) and were

more likely to have proximal tumors (gastric cardia or fundus). Patients treated at the highest volume

hospitals were more comorbid (elevated RUB) and of lower socioeconomic status (elevated material

deprivation quintile).

TEXTBOOK OUTCOMES ACROSS GASTRECTOMY VOLUMES

Achieving TO was rare across all surgeons (median rate 0%, interquartile range (IQR) 0-33%) and all

hospitals (median rate 17%, IQR 0-28%). TO were achieved in a similar proportion of the patients treated

by the highest volume surgeons compared to the lowest volume surgeons (Surgeon Q5 24.0% vs. Q1 20.8%,

SD 8%) (Table 5). Adequate lymph node sampling and absence of severe complications metrics were

superior in the highest volume surgeon group, but this was counteracted by the higher reintervention rate

in this group. TO were achieved in a higher proportion of the patients treated at the highest volume

hospitals compared to the lowest volume hospitals (Hospital Q5 23.5% vs. Q1 16.2%, SD 18%) (Table 6).

This difference was driven by the adequate lymph node sampling, lower rate of unplanned ICU admissions


and lower rates of 30-day mortality. The reintervention rate remained higher in this highest volume hospital

group.

On unadjusted logistic regression, surgeon volume as a continuous variable (p=0.20) and hospital volume

as a continuous variable (p=0.09) were not associated with TO. When surgeon (p=0.46) and hospital

(p=0.19) volumes were both included together as continuous variables in the model, neither was associated

with TO. Given the marked heterogeneity in the surgeon and hospital volume quintile groups and potential

for clustering by surgeon, a multivariable generalized estimating equation logistic model was fitted to

patient age, sex, RUB, deprivation quintile, year of gastrectomy, tumor location, histology category and T

and N stage. Continuous surgeon volumes (p=0.41) and continuous hospital volumes (p=0.24) were not

significantly associated with TO. Continuous volumes were moderately correlated (Spearman Correlation

Coefficient 41%, p<0.001) and not collinear (VIF=1.19). When both were included together, neither surgeon

(p=0.67) nor hospital (p=0.33) volume as continuous variables, was associated with TO. Model variables

were not collinear (all VIF < 1.2). There were no influential outliers (all “dfbeta” centered around 0% and

ranging from -10% to 10%) nor were there violations in log-linearity of surgeon and hospital volumes (nearly

all deviance residuals within 1 standard deviation). The model was not over-specified (12.2 TO per degree

of freedom).

Quintile volumes were moderately correlated (Polychoric Correlation Coefficient 45%, p<0.001) and not

collinear (VIF=1.26). Compared to the lowest volume surgeons, patients treated by the highest volume

surgeons has no difference in TO (Surgeon Q5 vs. Q1 – aOR 1.18 [0.78-1.78], p=0.42). Compared to the

lowest volume hospitals, patients treated at the highest volume hospitals were 45% more likely to achieve

TO, however this difference was likewise not statistically significant (Hospital Q5 vs. Q1 – aOR 1.45 [0.93-

2.29], p=0.10). When both quintile volumes were included in the model neither surgeon volume quintile

(Surgeon Q5 vs. Q1 – aOR 1.05 [0.68-1.62], p=0.82) nor hospital volume quintile (Hospital Q5 vs. Q1 – aOR

1.37 [0.84-2.22], p=0.21) was associated with TO (Appendix 2).


TEXTBOOK OUTCOME METRICS ACROSS GASTRECTOMY VOLUMES

Using GEE logistic regression adjusted for covariates and clustering by surgeon, the association of both

surgeon and hospital volume quintiles with each TO metric was estimated (Table 7). Compared to the

lowest volume hospitals, patients treated at the highest volume hospitals were a 92% more likely to have

adequate lymphadenectomies (Hospital Q5 vs. Q1 – aOR 1.92 [1.24-2.97], p=0.003) and 323% more likely

to avoid an unplanned ICU admission (Hospital Q5 vs. Q1 – aOR 3.23 [1.97-5.31], p<0.001). Surgeon and

hospital volume quintiles were not significantly associated to any of the other TO.


DISCUSSION

This study first and foremost provides evidence that measuring TO is feasible and informative. Cancer

surgery quality is multidimensional and generating population data on TO provides a comprehensive

overview of the current state of surgical management of gastric cancer. Gastrectomies are an invasive

procedure with important rates of post-operative complications and mortality. Optimally, patients benefit

most when the procedure is effective at reducing the risk of local recurrence and is safe.

TO were achieved in 22.8% of the study cohort, this was lower than the proportion of 32.1% reported in

the original Dutch study [144]. This was not completely unexpected given the regionalization of gastric

surgery in the Netherlands since 2012. The quality metric that had the most negative impact on achieving

TO in both studies was the adequacy of lymph node harvest which occurred in only 53.9% of Canadian and

57.1% of Dutch patients.

Where the results of this study differed most from the Dutch study was the proportion of patients with

severe complications (31.7% vs. 11.7%) and unplanned admissions to the ICU (27.8% vs. 9%). It is important

to note that contrary to the Dutch study, intra-operative and post-operative complication were combined

because complication diagnostic codes lacked an associated date and thus were indistinctively reported for

the whole admission. Nevertheless, the intraoperative complication rate in the Dutch study was only 5%

and even its addition to the complication rate would result in half as many patients with severe

complications than the present study. Furthermore, the majority of patients with severe complications in

both studies had unplanned ICU admissions, which supports the severity of the complications identified in

these studies. Patients undergoing gastrectomies at the highest volume hospitals were more than three

times more likely to avoid an unplanned ICU admission than those treated at the lowest volume hospitals.

Despite decreased ICU admissions, complication rates were the same across these hospital groups. These

results substantiate a previous volume study, which found that hospitals in the highest volume quintile (>11

annual gastrectomy cases) had decreased failure to rescue rates, but similar complication rates [233].

TO rates increased significantly during the study period; patients undergoing gastrectomy in 2015 were

88% more likely to achieve TO than patients in 2004. Despite identifying a natural regionalization of

gastrectomies over the study period – the median years of gastrectomies performed by the highest and

lowest volume surgeons were 2010 and 2009, and the median years of gastrectomies performed at the

highest and lowest volume hospitals were 2010 and 2008 – this time-dependent improvement in TO rates

was independent of gastrectomy volume. In Ontario, there has been government-backed regionalization


policies for thoracic surgery since 2005 and hepatopancreatobiliary surgery since 2006 [234, 235]. The

association between regionalization of these related complex cancer surgeries and the improvement of TO

in gastric cancer over time is speculative, but there is evidence of non-specific surgical volumes and

improved outcomes in lung, esophageal, colorectal resections and repair of aortic aneurysms in Ontario

[236].

The gastrectomy cases included in the volume calculation are restricted to gastrectomies performed in the

elective setting, subsequent to a diagnosis of gastric adenocarcinoma and within six months of this

diagnosis. As such, the volumes reported in these studies are lower than what most surgeons would expect.

In the low incidence region of North America, a high-volume surgeon has been previously defined as one

performing greater than 3 resections annually [131, 145], whereas a high volume center has been defined

according to a cut-off between 15 and 35 annual cases [145-150, 237]. In East Asia, a high-incident region,

definitions of high-volume surgeons ranged from 30 to 35 [151, 152] and high volume hospital were

defined according to cut-offs ranging from 56 to 386 cases annually [151, 153-155, 238], although very high

volume hospitals performed greater than 881 annual cases. A recent international and multidisciplinary

expert panel [77] has provided consensus guidelines on the processes of care for gastric cancer, and has

deemed it necessary for practicing surgeons to be trained in gastric cancer surgery and appropriate for

them to perform greater than 6 cases annually at centers with case volumes greater than 15 annually. As

such, the volumes of the highest volume surgeons and hospitals reported in this study are in line with

previous studies performed in North America and recent consensus guidelines but, as expected, fall short

of the volumes expected in high-incident countries in East Asia. Furthermore, D2 lymphadenectomies are

complex procedures, with a learning curve as high as 15 to 100 cases before a plateau is reached and varies

according to what parameters are measured [100, 101]. Given the current volume patterns in Ontario, it

would take the highest volume surgeons many years to achieve competency, whereas surgeons performing

less than two annual cases (Surgeon Volume Quintile 1-3) may never attain this goal.

These findings have important clinical and economic implication and should be considered for future policy-

making. Centralization of a rare procedure is important in order for specialists to achieve sufficient

experience to develop and maintain their expertise, training of future specialists and advancement of

medical knowledge through research. Centralization of complex procedures allows for human resources

and physical structures necessary for achieving TO [239], for example through standardized clinical

pathways and availability of ICU and interventional radiology. It is our opinion that centralization of gastric

cancer surgery should happen, but that simply concentrating the number of cases per hospital or surgeon

is not enough. What is of equal importance is the need to improve care when regionalizing. The volume


outcome-relationship is akin to “practice makes perfect”, when in fact “perfect practice makes perfect”

may be appropriate [239]. Perhaps future policies should be more aligned to meeting quality metrics than

an absolute volume. TO and outcome metrics could be used to identify surgeons with excellent gastric

cancer surgery outcomes to serve as mentors, to participate in the centralization of gastric cancer care, or

as a means to identify areas requiring improvements for all surgeons and the processes of care available to

them. Surgicopathologic quality (margins and lymph nodes) in this study was poor and represents a

combined responsibility between surgeons and pathologists [240, 241]. Intraoperative frozen section

margin is highly accurate [94], but requires the presence of a timely and efficient pathology review if it is

to be used. Achieving adequate lymph node numbers depend on adequate resection in addition to

specimen processing and evaluation. D2 lymphadenectomies [66, 67, 77] are recommended by several

guidelines, but true uptake in this and several North American population studies remains low [121, 129,

130, 231]. In low volume countries, specialized fellowship training, D2 gastrectomy courses [242], or

international observerships could be utilized to improve outcomes. Following centralization to fewer gastric

cancer surgeons, using TO and its metrics as a means of quality feedback to continue improving outcomes

after centralization. Regrettably, providing metrics to surgeons and hospitals will only be meaningful to

those with an adequate volume, thus reinforcing the need to centralize rare procedures [243].

This study is a large, population-based analysis which attempted to capture all patients who underwent a

potentially curative gastrectomy for gastric adenocarcinoma. The strengths of this study are based on its

stringent methods, expansive data linkage and granular pathology data. Compared to institutional-series

and clinical trials, the findings in this study are more widely generalizable and represent Ontario’s real-

world experience with gastric cancer surgery. Extensive collaboration with our provincial cancer registry

and a significant investment in pathology report extraction has allowed for a unique opportunity to leverage

pathology specimen data to measure surgical quality outcomes and account for cancer-specific

confounders. Unfortunately, 26% of eligible patients were lacking a pathology report, despite significant

attempts to acquire them. We used statistical modelling to determine that missingness of a pathology

report was not a confounder in the gastrectomy volume-TO relationship. A second limitation to this study

was the inability to conclusively determine the intent of surgery for the study cohort. The study period

ended one year prior to the release of REGATTA trial results [79]. This trial failed to show a survival

advantage, and more-so highlighted the negative impact of gastrectomy in the setting of non-curable

gastric cancer. Thus, it is possible, that some gastrectomies performed during the study period were non-

curative, in which case function would be prioritized over oncologic targets including margin status and

extent of lymphadenectomy. We attempted to mitigate this risk in several ways; we excluded all patients


with emergency admissions, previous gastrectomies, gastrectomies occurring more than six months

following diagnosis and those identified with metastatic disease in their surgical pathology reports or by

validated metastatic algorithm.

In conclusion, neither surgeon volume, hospital volume nor their combination was associated with

improved rates of TO. However, patients treated at hospital with the highest gastrectomy volumes were

more likely to undergo adequate lymphadenectomy and avoid ICU admissions without an associated

increase in post-operative morbidity or mortality. Improving gastric cancer surgery quality requires more

than increasing the number of cases an individual surgeon or a hospital completes. Future centralization

policies should consider using case-mix adjusted Textbook Outcome rates as a means of identifying high-

quality surgeons and hospitals and providing them feedback in order to systematically improve gastric

cancer surgical quality at a population level.


FIGURE 1. PATIENT FLOW DIAGRAM

Jordan Levy Textbook Outcomes in Patients with Gastric Cancer

85

FIGURE 2. PROPORTION AND CUMULATIVE INCIDENCE OF PATIENTS ACHIEVING EACH QUALITY

METRIC AND TEXTBOOK OUTCOMES


86

FIGURE 3. PROPORTION OF TEXTBOOK OUTCOMES BY YEAR OF GASTRECTOMY

20%14%

22%13% 16%

28% 24% 28% 26% 23%28% 29%

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015

TO Proportion Non-TO Proportion


87

TABLE 1. SURGEON VOLUME QUINTILES

Surgeon Volume Quintiles

Q1 Q2 Q3 Q4 Q5

Number of Patients 356 296 368 319 321

Number of Surgeons 277 194 154 85 36

Annual Volume Cut Points 0 0.5 1.0-1.5 2.0-3.0 3.5-9.5

Annual Volume, Median (IQR) 0 (0) 0.5 (0.5) 1 (1.0-1.5) 2.5 (2.0-3.0) 4.5 (4.0-6.0)

Number of Hospitals 67 52 47 39 24

Hospital Volume Quintiles, n (%)

Q1 137 (38.5%) 68 (23.0%) 50 (13.6%) 2 (0.6%) 2 (0.6%)

Q2 82 (23.0%) 85 (28.7%) 107 (29.1%) 76 (23.8%) 26 (8.1%)

Q3 41 (11.5%) 48 (16.2%) 81 (22.0%) 67 (21.0%) 71 (22.1%) Q4 52 (14.6%) 49 (16.6%) 81 (22.0%) 100 (31.4%) 111 (34.6%)

Q5 44 (12.4%) 46 (15.5%) 49 (13.3%) 74 (23.2%) 111 (34.6%)

IQR: Interquartile Range Caption: Patients included in the study cohort were divided into five approximately equally-sized patient groups based on their surgeon’s annual volume (surgeon volume quintiles Q1 to Q5) and their hospital’s annual volume (hospital volume quintiles Q1 to Q5).

TABLE 2. HOSPITAL VOLUME QUINTILES

Hospital Volume Quintiles

Q1 Q2 Q3 Q4 Q5

Number of Patients 259 376 308 393 324

Number of Hospitals 64 42 31 22 12 Annual Volume Cut Points 0-2.0 2.5-5.0 5.5-7.5 8.0-11.5 12.0-22.0

Annual Volume, Median (IQR) 1.0 (0.5-1.5) 4.0 (3.0-4.5) 6.5 (6.0-7.0) 9.5 (8.5-10.5) 14.5 (12.5-16.5)

Number of Surgeons 147 164 122 110 79

IQR: Interquartile Range Caption: Patients included in the study cohort were divided into five approximately equally-sized patient groups based on their hospital’s annual volume (hospital volume quintiles Q1 to Q5) and their surgeon’s annual volume (surgeon volume quintiles Q1 to Q5).


88

TABLE 3. COMPARISON OF BASELINE CHARACTERISTICS FOR PATIENTS TREATED BY THE

HIGHEST VS. LOWEST VOLUME SURGEONS (SURGEON VOLUME QUINTILE 5 VS. 1)

Characteristic Surgeon Volume Q1 Surgeon Volume Q5 SD

Age Groups, n (%)

<50 32 (9.0%) 29 (9.0%) 0%

50–59 44 (12.4%) 54 (16.8%) 13%

60–69 94 (26.4%) 67 (20.9%) 13%

70–79 121 (34.0%) 114 (35.5%) 3%

>80 65 (18.3%) 57 (17.8%) 1%

Sex, n (%)

Female 123 (34.6%) 107 (33.3%) 3%

Male 233 (65.4%) 214 (66.7%) 3%


0-2 17 (4.8%) 11 (3.4%) 7%

3 166 (46.6%) 156 (48.6%) 4%

4 109 (30.6%) 88 (27.4%) 7%

5 64 (18.0%) 66 (20.6%) 7%

Material Deprivation Quintile, n (%)

1 54 (15.2%) 49 (15.3%) 0%

2 65 (18.3%) 54 (16.8%) 4%

3 70 (19.7%) 61 (19.0%) 2%

4 81 (22.8%) 74 (23.1%) 1%

5 86 (24.2%) 83 (25.9%) 4%



Cardia, NOS 24 (6.7%) 41 (12.8%) 20%

Fundus 8 (2.2%) 20 (6.2%) 20%

Body 50 (14.0%) 47 (14.6%) 2%

Antrum 155 (43.5%) 106 (33.0%) 22%

Pylorus 18 (5.1%) 9 (2.8%) 12%

Lesser Curvature, NOS 53 (14.9%) 45 (14.0%) 2%

Greater Curvature, NOS 14 (3.9%) 10 (3.1%) 4%

Overlapping Lesion 7 (2.0%) 8 (2.5%) 4%

NOS 27 (7.6%) 35 (10.9%) 11%


Intestinal 227 (63.8%) 203 (63.2%) 1%

Non-Intestinal 129 (36.2%) 118 (36.8%) 1%

T-Stage, n (%)

T0-1 86 (24.2%) 90 (28.0%) 9%

T2 66 (18.5%) 47 (14.6%) 10% T3 101 (28.4%) 94 (29.3%) 2%

T4a 92 (25.8%) 78 (24.3%) 4%

T4b 11 (3.1%) 12 (3.7%) 4%

Node Involvement, n (%)

N0 143 (40.2%) 129 (40.2%) 0%

N+ 213 (59.8%) 192 (59.8%) 0%

SD: Standardized Differences, IQR: Interquartile Range Standardized Difference of 10% or greater considered clinically meaningful


89

TABLE 4. COMPARISON OF BASELINE CHARACTERISTICS FOR PATIENTS TREATED AT THE HIGHEST VS. LOWEST VOLUME HOSPITALS (HOSPITAL VOLUME QUINTILE 5 VS. 1)

Characteristic Hospital Volume Q1 Hospital Volume Q5 SD

Age Groups, n (%)

<50 16 (6.2%) 31 (9.6%) 13%

50–59 35 (13.5%) 53 (16.4%) 8%

60–69 70 (27.0%) 78 (24.1%) 7%

70–79 86 (33.2%) 104 (32.1%) 2%

>80 52 (20.1%) 58 (17.9%) 6%

Sex, n (%)

Female 99 (38.2%) 101 (31.2%) 15%

Male 160 (61.8%) 223 (68.8%) 15%


0-2 8 (3.1%) 17 (5.2%) 11%

3 125 (48.3%) 152 (46.9%) 3%

4 87 (33.6%) 84 (25.9%) 17%

5 39 (15.1%) 71 (21.9%) 18%


1 49 (18.9%) 46 (14.2%) 13%

2 54 (20.8%) 50 (15.4%) 14%

3 53 (20.5%) 64 (19.8%) 2%

4 50 (19.3%) 76 (23.5%) 10%

5 53 (20.5%) 88 (27.2%) 16%



Cardia/Fundus 14 (7.3%) 41 (20.1%) 25%

Body 32 (12.4%) 40 (12.3%) 0%

Antrum 104 (40.2%) 115 (35.5%) 10%

Pylorus 12 (4.6%) 9 (2.8%) 10%



Overlapping Lesion 10 (3.9%) 9 (2.8%) 6%

NOS 27 (10.4%) 18 (5.6%) 18%


Intestinal 161 (62.2%) 199 (61.4%) 2%

Non-Intestinal 98 (37.8%) 125 (38.6%) 2%

T-Stage, n (%)

T0-1 65 (25.1%) 71 (21.9%) 8%

T2 44 (17.0%) 56 (17.3%) 1%

T3 71 (27.4%) 107 (33.0%) 12% T4a 62 (23.9%) 79 (24.4%) 1%

T4b 17 (6.6%) 11 (3.4%) 15%

Node Involvement, n (%)

N0 143 (55.2%) 188 (58.0%) 6%

N+ 116 (44.8%) 136 (42.0%) 6%



90

TABLE 5. COMPARISON OF TEXTBOOK OUTCOMES AND METRICS FOR PATIENTS TREATED BY THE HIGHEST VS. LOWEST VOLUME SURGEONS (SURGEON VOLUME QUINTILE 5 VS. 1)

Characteristic Surgeon Volume Q1 Surgeon Volume Q5 SD

Textbook Outcomes 20.8% 24.0% 8%

Negative Margins 89.6% 88.5% 4%

Greater than 15 LN Sampled 45.5% 62.0% 34%

Number of Nodes Sampled, Median (IQR) 14 (8-20) 19 (13-29) 53%

No Severe Complication 69.1% 74.1% 11%

No Reintervention 93.3% 88.5% 17%

No Unplanned ICU Admission 75.0% 71.0% 9%

Length of Stay 21 Days or Less 88.5% 89.7% 4%

Length of Stay (Days), Median (IQR) 9 (7-13) 9 (7-13) 9%

No 30-Day Readmission 83.4% 82.6% 2%

No 30-Day Mortality 96.6% 95.3% 7%



91

TABLE 6. COMPARISON OF TEXTBOOK OUTCOMES AND METRICS FOR PATIENTS TREATED AT THE HIGHEST VS. LOWEST VOLUME HOSPITALS (HOSPITAL VOLUME QUINTILE 5 VS. 1)

Characteristic Hospital Volume Q1 Hospital Volume Q5 SD

Textbook Outcomes 16.2% 23.5% 18%

Negative Margins 84.9% 84.9% 0%

Greater than 15 LN Sampled 38.2% 62.0% 49%

Number of Nodes Sampled, Median (IQR) 13 (8-20) 18 (13-25) 59%

No Severe Complication 67.2% 69.4% 5%

No Reintervention 91.5% 86.7% 15%

No Unplanned ICU Admission 66.8% 80.2% 31%

Length of Stay 21 Days or Less 88.0% 88.0% 0%

Length of Stay (Days), Median (IQR) 9 (7-13) 9 (7-12) 11%

No 30-Day Readmission 83.8% 80.2% 9%

No 30-Day Mortality 93.4% 96.9% 16%



92

TABLE 7. ASSOCIATION BETWEEN SURGEON AND HOSPITAL VOLUME QUINTILES AND TEXTBOOK OUTCOME METRICS

(MULTIVARIABLE GEE LOGISTIC REGRESSION ADJUSTED ODDS RATIOS)

Characteristic Surgeon

Volume Q1 Surgeon

Volume Q5 p-value

Hospital Volume Q1

Hospital Volume Q5

p-value

Textbook Outcomes Reference 1.05 (0.68-1.62)

0.82 Reference 1.37 (0.84-2.22)

0.21

Negative Margins Reference 0.96 (0.48-1.93)

0.92 Reference 0.98 (0.55-1.73)

0.94

Greater than 15 LN Sampled Reference 1.33 (0.84-2.10)

0.22 Reference 1.92 (1.24-2.97)

0.003

No Severe Complication Reference 1.38 (0.82-2.32)

0.22 Reference 1.11 (0.70-1.76)

0.65

No Reintervention Reference 0.72 (0.37-1.40)

0.33 Reference 0.71 (0.36-1.42)

0.33

No Unplanned ICU Admission Reference 0.66 (0.39-1.12)

0.13 Reference 3.23 (1.97-5.31)

<0.001

Length of Stay 21 Days or Less Reference 1.18 (0.69-2.03)

0.54 Reference 0.91 (0.50-1.67)

0.76

No 30-Day Readmission Reference 0.94 (0.58-1.52)

0.79 Reference 0.85 (0.52-1.37)

0.50

No 30-Day Mortality Reference 0.96 (0.64-1.45)

0.86 Reference 0.73 (0.46-1.17)

0.19

Multivariable Generalized Estimating Equation Logistic Model adjusted for surgeon and hospital volume quintiles, covariates and clustering by surgeon.


93

APPENDIX 1. PHYSICIAN LINKING ALGORITHM

Eligible Physicians General Surgeons, Thoracic Surgeons, Cardiothoracic Surgeons, Vascular Surgeons

Eligible Physician Billing Fee Codes† S089, S090, S122, S123, S125, S128, S129

Time Period Within 30 days of gastrectomy date

Eligible Setting Inpatient Admission

† The Ontario Health Insurance Plan (OHIP) Schedule of Benefits Fee Codes

APPENDIX 2. ASSOCIATION BETWEEN SURGEON AND HOSPITAL VOLUME QUINTILES AND TEXTBOOK OUTCOMES

(MULTIVARIABLE GEE LOGISTIC REGRESSION ADJUSTED ODDS RATIOS)

Variable Adjusted HR (95% CI) p-value

Surgeon Volume Quintile 1 Reference 2 0.93 (0.65, 1.33) 0.70 3 1.09 (0.74, 1.60) 0.68 4 1.21 (0.83, 1.77) 0.33 5 1.05 (0.68, 1.62) 0.82

Hospital Volume Quintile 1 Reference 2 1.39 (0.90, 2.14) 0.14 3 1.48 (0.94, 2.33) 0.09 4 1.45 (0.91, 2.33) 0.12 5 1.37 (0.84, 2.23) 0.21

Age Groups <50 4.72 (2.79, 7.98) <0.001 50–59 2.75 (1.71, 4.42) <0.001 60–69 2.20 (1.42, 3.39) <0.001 70–79 1.57 (1.03, 2.40) 0.04 >80 Reference

Sex Female 1.04 (0.82, 1.31) 0.77 Male Reference

Resource Utilization Bands 0_2 2.04 (1.04, 4.02) 0.04 3 1.33 (0.90, 1.96) 0.15 4 1.28 (0.82, 2.01) 0.27 5 Reference

Material Deprivation Quintile 1 1.18 (0.82, 1.69) 0.38 2 0.87 (0.59, 1.27) 0.47 3 0.97 (0.69, 1.36) 0.85 4 0.92 (0.68, 1.25) 0.61 5 Reference


94

Year of Gastrectomy (2004-2015) 1.08 (1.03, 1.13) 0.001

Tumor Location

Cardia, NOS Reference

Fundus of stomach 0.95 (0.43, 2.11) 0.91

Body of stomach 1.49 (0.82, 2.70) 0.19 Gastric antrum 1.84 (1.11, 3.07) 0.02

Pylorus 2.13 (1.02, 4.48) 0.05

Lesser curvature of stomach, NOS 2.11 (1.27, 3.50) 0.004

Greater curvature of stomach, NOS 2.30 (1.10, 4.80) 0.027

Overlapping lesion of stomach 1.15 (0.43, 3.08) 0.78

Stomach, NOS 2.30 (1.27, 4.15) 0.006 Histology

Intestinal Type 1.16 (0.90, 1.50) 0.26

Non-Intestinal Type Reference

T-Stage

T0-1 1.40 (0.73, 2.67) 0.31

T2 1.47 (0.75, 2.88) 0.27 T3 1.08 (0.61, 1.93) 0.78

T4a 0.79 (0.44, 1.41) 0.42

T4b Reference

N-Stage

N0 0.69 (0.52, 0.91) 0.008

N+ Reference Multivariable Generalized Estimating Equation Logistic Model adjusted for surgeon and hospital volume quintiles, covariates and clustering by surgeon.

APPENDIX 3. RECORD STATEMENT CHECKLIST

# STROBE items RECORD items Location in manuscript

1 (a) Indicate the study’s design with a commonly used term in the title or the abstract (b) Provide in the abstract an informative and balanced summary of what was done and what was found

RECORD 1.1: The type of data used should be specified in the title or abstract. When possible, the name of the databases used should be included. RECORD 1.2: If applicable, the geographic region and timeframe within which the study took place should be reported in the title or abstract. RECORD 1.3: If linkage between databases was conducted for the study, this should be clearly stated in the title or abstract.

Abstract


95

Background rationale

2 Explain the scientific background and rationale for the investigation being reported

Abstract

Objectives 3 State specific objectives, including any prespecified hypotheses

Abstract

Study Design 4 Present key elements of study design early in the paper

Study Design and Setting

Setting 5 Describe the setting, locations, and relevant dates, including periods of recruitment, exposure, follow-up, and data collection

Study Design and Setting, Study Cohort

Participants 6 (a) Cohort study - Give the eligibility criteria, and the sources and methods of selection of participants. Describe methods of follow-up (b) Cohort study - For matched studies, give matching criteria and number of exposed and unexposed

RECORD 6.1: The methods of study population selection (such as codes or algorithms used to identify subjects) should be listed in detail. If this is not possible, an explanation should be provided. RECORD 6.2: Any validation studies of the codes or algorithms used to select the population should be referenced. If validation was conducted for this study and not published elsewhere, detailed methods and results should be provided. RECORD 6.3: If the study involved linkage of databases, consider use of a flow diagram or other graphical display to demonstrate the data linkage process, including the number of individuals with linked data at each stage.

Study Cohort


96

Variables 7 Clearly define all outcomes, exposures, predictors, potential confounders, and effect modifiers. Give diagnostic criteria, if applicable.

RECORD 7.1: A complete list of codes and algorithms used to classify exposures, outcomes, confounders, and effect modifiers should be provided. If these cannot be reported, an explanation should be provided.

Exposure, Outcomes, Covariates

Data sources/ measurement

8 For each variable of interest, give sources of data and details of methods of assessment (measurement). Describe comparability of assessment methods if there is more than one group

Data Sources and Management, Exposure, Outcomes, Covariates

Bias 9 Describe any efforts to address potential sources of bias

Data Sources and Management, Exposure, Outcomes, Covariates, Statistical Analysis, Missing Data

Study size 10 Explain how the study size was arrived at

Study Cohort

Quantitative variables

11 Explain how quantitative variables were handled in the analyses. If applicable, describe which groupings were chosen, and why

Statistical Analysis

Statistical methods

12 (a) Describe all statistical methods, including those used to control for confounding (b) Describe any methods used to examine subgroups and interactions

Statistical Analysis and Missing Data


97

(c) Explain how missing data were addressed (d) Cohort study - If applicable, explain how loss to follow-up was addressed (e) Describe any sensitivity analyses

Data access and cleaning methods

.. RECORD 12.1: Authors should describe the extent to which the investigators had access to the database population used to create the study population. RECORD 12.2: Authors should provide information on the data cleaning methods used in the study.


Linkage .. RECORD 12.3: State whether the study included person-level, institutional-level, or other data linkage across two or more databases. The methods of linkage and methods of linkage quality evaluation should be provided.

Data Sources and Management, Study Cohort, Exposure, Outcome, Covariates, Appendix 1

Participants 13 (a) Report the numbers of individuals at each stage of the study (e.g., numbers potentially eligible, examined for eligibility, confirmed eligible, included in the study, completing follow-up, and analysed) (b) Give reasons for non-participation at each stage. (c) Consider use of a flow diagram

RECORD 13.1: Describe in detail the selection of the persons included in the study (i.e., study population selection) including filtering based on data quality, data availability and linkage. The selection of included persons can be described in the text and/or by means of the study flow diagram.

Results: a. Study Cohort b. Figure 1

Descriptive data 14 (a) Give characteristics of

Results: a. Gastrectomy


98

study participants (e.g., demographic, clinical, social) and information on exposures and potential confounders (b) Indicate the number of participants with missing data for each variable of interest (c) Cohort study - summarise follow-up time (e.g., average and total amount)

Volumes, Tables 1-4 b. Figure 1 c. Study Cohort

Outcome data 15 Cohort study - Report numbers of outcome events or summary measures over time

Results: Textbook Outcomes, Figure 2, Figure 3

Main results 16 (a) Give unadjusted estimates and, if applicable, confounder-adjusted estimates and their precision (e.g., 95% confidence interval). Make clear which confounders were adjusted for and why they were included (b) Report category boundaries when continuous variables were categorized (c) If relevant, consider translating estimates of relative risk into absolute risk for a meaningful time period

Results: a. Textbook Outcomes Across Gastrectomy Volumes, Textbook Outcome Metrics Across Gastrectomy Volumes, Tables 5-6, Appendix 2 b. Tables 1-4 c. N/A

Other analyses 17 Report other analyses done—e.g., analyses of subgroups and interactions, and sensitivity analyses

Textbook Outcome Metrics Across Gastrectomy Volumes


99

Key results 18 Summarise key results with reference to study objectives

Discussion: Paragraphs 1-4

Limitations 19 Discuss limitations of the study, taking into account sources of potential bias or imprecision. Discuss both direction and magnitude of any potential bias

RECORD 19.1: Discuss the implications of using data that were not created or collected to answer the specific research question(s). Include discussion of misclassification bias, unmeasured confounding, missing data, and changing eligibility over time, as they pertain to the study being reported.


Interpretation 20 Give a cautious overall interpretation of results considering objectives, limitations, multiplicity of analyses, results from similar studies, and other relevant evidence

Discussion: Paragraph 1, 2, 5, 6, 7, 9

Generalisability 21 Discuss the generalisability (external validity) of the study results


Funding 22 Give the source of funding and the role of the funders for the present study and, if applicable, for the original study on which the present article is based

Title Page “Funding”

Accessibility of protocol, raw data, and programming code

.. RECORD 22.1: Authors should provide information on how to access any supplemental information such as the study protocol, raw data, or programming code.

Appendices


100

CHAPTER 4 DISCUSSION

4.1. REVIEW OF THESIS OBJECTIVES

The objectives of this thesis were to describe the survival of patients with and without Textbook Outcomes;

to measure the independent association between Textbook Outcomes and long-term survival; and to

determine whether gastrectomy experience, in terms of surgeon and hospital-volumes, is associated with

achieving Textbook Outcomes and its metrics.

4.2. SUMMARY OF MAIN FINDINGS

The studies included in this thesis provide evidence that measuring Textbook Outcomes is feasible and

informative. Cancer surgery quality is multidimensional and generating population data on TO provides a

comprehensive overview of the current state of surgical management of gastric cancer. Gastrectomies are

an invasive procedure with important rates of post-operative complications and mortality. Optimally,

patients benefit most when the procedure is effective at reducing the risk of cancer recurrence and safe.

Textbook Outcomes were achieved in 23% of patients across Ontario (22% when including

gastroesophageal junctional tumors). The quality metric that had the most negative impact on achieving

Textbook Outcomes was the adequacy of lymph node harvest which occurred in 54% and a post-operative

course void of severe complications which occurred in 68%.

The first study (Section 3.2.) laid the foundation for Textbook Outcomes as an important measure of surgical

quality with real long-term implications. Patients with Textbook Outcomes had 3-year survival rates of 75%

compared to 55% for patients without Textbook Outcomes. Following adjustments for potential

confounders, achieving Textbook Outcomes was still associated with a 39% reduction in mortality

throughout the observation period. This survival benefit was found to be driven by negative margins,

adequate lymph node sampling and fewer unplanned ICU admissions or re-admissions.

Given the strong survival benefit of Textbook Outcomes, the second study (Section 3.3.) intended to identify

a process to improve Textbook Outcome rates in Ontario: surgical experience measured by gastrectomy

volumes across surgeons and hospitals. Volumes across Ontario were exceedingly low. Surgeons performed

an average of a single case a year, and hospitals performed a median of 6.5 cases annually. The highest


101

volume surgeons, or those in the top quintile, performed a median of 4.5 cases annually. The highest

volume hospitals performed a median of 14.5 cases annually. Perhaps due to persistently low volumes,

neither surgeons, nor hospitals in the top volume quintile achieved Textbook Outcomes in significantly

more patients than those in the lowest quintile. However, patients treated at hospitals with the highest

gastrectomy volumes were more likely to undergo adequate lymphadenectomy and avoid ICU admissions

without an associated increase in post-operative morbidity or mortality.

4.3. COMPARISON TO THE LITERATURE

In the first study presented in this thesis (Section 3.2.) we aimed to establish the significance of Textbook

Outcomes in patient long-term survival. Survival outcomes in patients with gastric cancer have been

extensively compared between high incident East and low incident West countries [138-141]. This survival

differential has even been used to substantiate the necessity of D2 gastrectomies, which has been inspired

and driven by Japanese experience and excellent outcomes [97-99, 103, 104]. However, many critics of this

more extensive procedure have capitalized on systematic differences in Eastern and Western patients.

Western patients tend to be diagnosed at an advanced age, with associated increased comorbidity burden

and rates of obesity [126, 138] and have a higher incidence of Barret’s esophagus, with a resulting increase

in junctional tumors and decreased survival [126, 142]. Furthermore, patients in Japan and Korea have

benefited from national screening programs since 1983 and 1999, respectively [55] and are twice as likely

to be diagnosed with locoregional as opposed to metastatic disease compared to North American patients

[53, 142, 143]. As such, it has been debated whether survival differences were due to differences in surgical

quality and case volumes or simply differences in baseline health and tumor prognostic factors. Through

the rich data available in this study, we were able to account for patient age, sex, comorbidity burden,

socioeconomic status and tumor location, histology, invasiveness patterns and stage. We concluded that

patients with Textbook Outcomes very much resembled Eastern patients. They were younger, healthier

and had tumors that were more distal, and of lower stage. This finding corroborated what was expected

and validated our choice of confounders and their prognostic ability. Most importantly, on adjusted analysis

Textbook Outcomes remained a significant predictor of long-term survival, and more specifically, adequate

lymph node sampling was significantly associated with improved survival, independent of margin status

and post-operative complications, reinterventions, ICU admissions, length of stay and readmissions. This

study further establishes the implication of sampling at least 16 lymph nodes during all. While baseline


102

patient and tumor characteristics have prognostic implications in their own right, surgical quality remains

an undermeasured and undervalued factor in the gastric cancer outcome discussion.

Textbook Outcome rates were appallingly low during the study period (22% to 23%), and only as high as

24% at the hands of the highest volume surgeons. The data did offer a glimmer of hope for patients with

gastric cancer in Ontario. Patients in 2015 had a 33% reduction in their risk of mortality compared to

patients in 2004. It was interesting to note that in the second thesis study, we similarly identified that

patients undergoing gastrectomy in 2015 were 88% more likely to achieve TO than patients in 2004.

Unfortunately, a longitudinal analysis of temporal trends in gastric cancer processes of care in Ontario was

not under the mandate of this thesis, but makes for an attractive follow-up study.

According to current guidelines [66, 67, 77, 120, 243] and clinical trial evidence [197-202, 244], all patients

in this study cohort should have undergone some form of adjuvant chemotherapy. However, we found that

only 45% of patients received post-operative chemotherapy, a proportion similar to the proportion

reported to have completed peri-operative chemotherapy in the MAGIC trial and similar to the proportion

of patients completing at least one pre-operative and post-operative cycle of chemotherapy in the French

Trial [199, 244]. Our definition of receiving chemotherapy was much more sensitive than these trials; any

patient with at least three visits for systemic therapy at a Regionalized Cancer Center and/or three physician

billings for chemotherapy in the 183 days following gastrectomy was recognized as having undergone

adjuvant chemotherapy; depending on the regimen this may represent less than a single full cycle. This is

not the first study to identify underutilization of adjuvant chemotherapy in Ontario, it has been previously

reported for primary and metastatic colorectal resections, pancreatic resections and bladder resections

[245-248]. Textbook Outcomes are specific to the peri-operative period, however future gastric cancer

outcome studies should focus on determining why chemotherapy is underused and establish whether

there are temporal or regional variations in its use across Ontario.

In the second study presented in this thesis (Section 3.3.) we aimed to establish the significant association

of gastrectomy experience in terms of volumes and achieving Textbook Outcomes. While Textbook

Outcomes were not associated with either surgeon or hospital volumes, we did find that in addition to

decreasing unplanned ICU admissions, the highest volume hospitals had improved adequate

lymphadenectomy rates. When this was accounted for in the model, surgeon volume was no longer

associated with adequate lymphadenectomy rates. Highest volume hospitals resected a median of 18

nodes, and at least 16 nodes in 62%. Lowest volume hospitals resected a median of 13 nodes, and at least


103

16 nodes in 38%. This may speak to the multidisciplinary nature of lymph node harvest that is not uniquely

dependent on surgeon ability, but a collaboration between surgeons and pathologists. This is a very

important finding given the evidence from the colorectal cancer literature that specimen processing [241]

and collaborative educational-programs for surgeons and pathologists [240] can significantly increase

lymph node harvest in oncologic resections. These types of programs may represent yet another

institutional policy that may greatly benefit not only patients with gastric cancer, but nearly all solid tumor

malignancies with staging dependent on lymph node status.

During the study period, we did note the seemingly natural regionalization in gastric cancer surgery, such

that patients cared for by the highest volume surgeons or in the highest volume hospitals had undergone

their gastrectomy more recently than those cared for by the lowest volume surgeons or in the lowest

volume hospitals. It is unclear whether this was secondary to patient health literacy, referral patterns or

regionalization of hepatopancreatobiliary and thoracic surgery during the same time period. In parallel, this

study found that, compared to the lowest volume hospitals, the highest volume hospitals tended to care

for more comorbid patients (RUB 5, 21.9% vs. 15.1%), those of the lowest SES (material deprivation quintile

5, 27.2% vs. 20.5%) and those with more proximal tumors despite excluding patients with GEJ tumors

(cardia tumors, 12.7% vs. 5.4%). Therefore, if Textbook Outcome rates, or for that matter any surgical

quality measures, are to be performed across hospitals at the provincial-level, these must be case-mix

adjusted in order not to penalize higher volume hospitals.

4.4. LIMITATIONS AND METHODOLOGICAL ISSUES

4.4.1. MISSING DATA



A substantial amount of the data used in the studies were available through validated databases [206, 209]

with rare missing data [179]. The first study (Section 3.2.) excluded only 3% of eligible patients for missing

covariate, exposure or outcome data, whereas the second study (Section 3.3.) excluded 5% secondary to

the addition of a surgeon-linking exclusion, such that 2% of patients could not be linked to a surgeon and

were thus missing a surgeon volume, which was not required in the first study. As such, these missing data

were felt to represent a very minor threat to the validity of the thesis studies.


104

Unfortunately, prior to the exclusion of these data, 26% of patients were excluded due to missing pathology

reports. Given the sheer number, complexity and variation of the variables required from the

Esophagogastric Pathology Database, missing values in this dataset could not be imputed from observed

and measured data. In order to better characterize the implications of these missing pathology reports,

patients were stratified according to pathology report missingness and statistical comparisons of observed

patient characteristics, disease characteristics, hospital and surgeon volumes, non-pathology based

Textbook Outcome metrics and survival across both patient groups were performed (Table for 4.4.1.A.).

Secondly, given the potential correlation between these observed variables, statistical modelling using a a

mixed effects multivariable logistic regression accounting for clustering by hospital to determine whether

study exposures and outcomes were independently associated with pathology report missingness (Table

for 4.4.1.B.).

Compared to the those excluded based on missing pathology report, those with pathology reports were

slightly older (mean age 68.1 vs. 66.1, Standardized Difference (SD) 15% - age <50, 9.4% vs. 13.0%, SD 11%)

and were more contemporary (median year of gastrectomy 2010 vs. 2008, SD 36%) but were otherwise

similar in sex, socioeconomic status (deprivation quintile), baseline health status (resource utilization band),

neoadjuvant and adjuvant chemotherapy treatment, adjuvant radiation treatment and tumor location,

with the exception of tumor location “Gastric NOS” which was less common in the non-excluded patients,

potentially due to improved tumor location capture in more recent years. Most importantly, patients

missing pathology reports were no different in the proportion achieving non-pathology-based Textbook

Outcome metrics or 3-year survival rates. Given the exclusion of numerous patients, it was possible that

some hospitals would not be represented in the included patients. This may have biased our results towards

the null in the volume-Textbook Outcome study if lower hospital volumes were systematically excluded due

to missing pathology reports. However, of the 12 of 82 hospitals that were not represented in the non-

excluded patients, only two of these were missing in a meaningful way (SD > 10%) owing to the fact that

the other 10 had performed a maximum of 3 gastrectomies during the 11-year study period. The two

excluded hospitals with SD>10% had performed 4 and 5 cases. If these very low volume hospitals’ exclusion

was associated with hospital-volume, this would result in a more conservative estimate. Patients missing

pathology reports were more likely to have undergone gastrectomy by the lowest volume surgeons

(Surgeon Volume Q1, 30.1% vs. 21.4%, SD 20%) and at the lowest volume hospitals (Hospital Volume Q1,

29.6% vs. 16.5%, SD 32%). In the second thesis study (section 3.3.) we did identify a natural regionalization

of gastric resections, such that more recent gastrectomies were more likely to be performed by higher


105

volume surgeons and at higher volume hospitals. As such, it was unclear whether the association between

pathology report missingness was associated with gastrectomy-volumes independent of year of

gastrectomy. In order to address this concern, a mixed effect multivariable logistic regression accounting

for clustering by hospital was modelled. Similar to the unadjusted comparisons described above and

reported in Table for 4.4.1.A, hospital volume, age and year of gastrectomy were all independently

associated with pathology report missingness. This missingness was not associated with the Textbook

Outcome metrics included in the model or survival and is unlikely to represent an unmeasured confounder

in the relationships of interest in the first and second study.

4.4.2. INTERNAL AND EXTERNAL VALIDITY

Pathology reports missingness should be considered when evaluating the validity of these studies’ results.

Younger patients, especially those under 50 years old were underrepresented in these studies; as were

patients treated earlier on in the study period. Younger patients were more likely to be cared for by the

highest volume surgeons and hospitals, achieve Textbook Outcomes, and have superior survival, whereas

patients treated earlier on in the study period were more often cared for by the lowest volume surgeons

and hospitals, were less likely to achieve Textbook Outcomes and had inferior survival. Omitting a

proportion of younger patients and those treated earlier on in the study period may have biased the results

of both studies towards the null and resulted in more conservative estimates in the associations studied.

As such, it is possible that in the first study, the impact of Textbook Outcomes on long-term survival was

underestimated. In the second study, the exclusion of such patients may have biased the results towards

the null or reduced the study’s statistical power. Whether this would reverse accepting the null hypothesis

that gastrectomy volumes are not associated with Textbook Outcomes cannot be determined since post-

hoc power analyses in this context systematically underestimate study power [249-252]. Given the

selection bias present, the results of these studies may be less applicable to the youngest patients and

those treated at by the lowest volume hospitals. However, given that the majority of patients undergoing

gastrectomy for gastric cancer are older than 50 years old (90%) and are not cared for by the lowest volume

hospitals (80%) the results of these studies remain widely applicable to the Ontario population.

The patient population and volumes reported in these studies are comparable to other low-incident

population-based analyses [131, 145-150, 237, 253-255] and as such, our findings should be considered for

future policy-making in all low-incident Western countries. As described in Section 3.3., centralization of a


106

rare procedure allows for specialists to achieve sufficient experience to develop and maintain their

expertise, training of future specialists, advancement of medical knowledge through research and makes it

more feasible to mobilize human resources and physical structures necessary for achieving Textbook

Outcomes [239]. However, increased volumes alone were not associated with improved perioperative care,

as measured by the Textbook Outcome metrics. Whereas high volume hospitals achieved adequate lymph

node sampling in more patients with less ICU admissions, these findings could improve staging and

prognosis for patient undergoing gastric cancer resections and reduce health care expenditures. For

regions who have not yet implemented centralization policies, and even for those that have, Textbook

Outcomes are highly prognostic and should be utilized in two strategies. First, they can be used to identify

high quality surgeons and hospitals for centralization policies, however a sufficient number of cases is

required in order to estimate accurate Textbook Outcome rates and as such, volumes will play a role in

selecting appropriate surgeons and hospitals for future gastric cancer patient care. Second, they should be

used to identify clinical care areas in need of improvement. Improvement strategies will depend on specific

deficiencies, but specialized fellowship training, D2 gastrectomy courses [242], international observerships

and multidisciplinary education programs could be utilized to improve surgeon-specific outcomes, while

standardized pre-operative and post-operative clinical pathways can improve clinical peri-operative

outcomes.

4.5. IMPLICATIONS AND RECOMMENDATIONS

The results of these studies have wide-ranging implications at the surgeon, hospital and provincial level.

Overall, outcomes for gastric cancer resections in Ontario are worse than expected and fall short of what

was reported in the Netherlands, another low-incident Western country. Most importantly, resection

margins were positive in 12% of patients, inadequate lymph node sampling occurred in 46% of patients,

32% of patients experienced a severe medical complication and 28% required an ICU admission.

These results have been presented at several meetings attended by a high number of surgeons performing

gastric cancer resections. The reaction has been somewhat consistent, all surgeons are shocked by the

results and no surgeon can believe these results represent their own practice. It is important to note that

these represent population-level distributions and it is likely that some surgeons performed much better

than others, but even the 36 surgeons at the peak of their volumes achieved Textbook Outcomes in merely

24% of their patients and had positive resection margins in 11%, inadequate lymph node sampling in 38%,


107

severe complications in 26% and unplanned ICU admissions in 29% of their patients, respectively. High

volume centres and surgeons may take more complex cases than their lower volume counterparts, and we

have attempted to adjust for this through the ADG, tumor location, and cancer stage.

These studies do not aim at blaming surgeons, Textbook Outcome metrics are multifaceted and the

responsibility of surgeons, anesthesiologists, pathologists, intensivists, nurses, radiologists,

physiotherapists, occupational therapists, nutritionists, social workers, hospital administrators, and most

importantly patients buying-in to pre-operative and post-operative functional optimization regimens

including smoking cessation, nutrition and exercise [66, 67, 74, 77, 78]. However, there are no

gastrectomies without surgeons and much of what occurs in the operating room predicts a patient’s post-

operative course. As such, post-operative outcomes will undeniably be linked back to surgeons. High quality

gastrectomies are complex and morbid procedures, and surgeons cannot accept the status-quo when faced

with the harsh reality of the state of patients undergoing gastric cancer resections in Ontario. A strong

sense of optimism is crucial when caring for complex patients with poor prognoses, health care

professionals must believe they can help their patients attain superior outcomes, but without measuring

these outcomes in a practical way, a false sense of optimism or even an illusion of superiority impedes

efforts to genuinely improve outcomes on a large scale. Surgeons require and should demand some sort of

feedback, in order to have objective measures of their current care. The Textbook Outcomes presented in

these studies are measurable, reproducible and clinically meaningful case-mix adjusted measures of peri-

operative quality and can be used by surgeons to identify areas of weakness in their own operating room

and/or hospital, which would allow them to champion processes necessary to improve outcomes for their

own patients. Further, utilization of common outcomes between countries and regions can allow for

external bench-marking, and assessment of impact of programs to improve patient care across

jurisdictions.

In Ontario, hospitals function on a strict budget with limited resources. In these studies, we have confirmed

the resource intensive nature of caring for patients with resectable gastric cancer; a large proportion of

patients have severe complications, require reinterventions, are admitted to the ICU, have long hospital

admissions and are often readmitted. Standardized pathways have proven beneficial in gastric, esophageal,

colorectal, pancreatic, lung, liver, knee and spine surgery amongst others [250-258], but these cannot

feasibly be implemented for a handful of patients a year at low volume hospitals. Perhaps a more granular

understanding of patients’ perioperative course will inspire a genuine conversation about whether lowest


108

volume hospitals should continue offering surgical care for gastric cancer patients in light of the human,

structural and financial resources required to do so, in addition to their detrimental effect on patients.

The implications of these studies are critical for future policy-making in gastric cancer. While regionalization

of gastric cancer surgery may ensure that specialists achieve sufficient experience to develop and maintain

their expertise, it also simplifies the logistics and feasibility of a provincial quality improvement program.

While gastrectomy volumes were not significantly associated with Textbook Outcomes, we can see from

the Dutch experience (who achieved Textbook Outcomes in 50% more patients) that achieving higher rates

of Textbook Outcomes is possible. This thesis studied the current volume-outcome relationship in Ontario,

and with it determined that even the highest volume surgeons performed a median of 4.5 curative

gastrectomy cases a year. A single gastrectomy case every three months. Practically, it is impossible to build

experience with such numbers, especially for a multidisciplinary team such as the one required in gastric

cancer. It is possible that the volumes in Ontario at present were systematically too similar across low and

high-volume providers and hospitals to identify the association at hand. Unfortunately, the results of these

studies emphasize the fact that the majority of surgeons and hospitals have poor outcomes, and most

importantly, given the survival consequences of Textbook Outcomes, patients are left dealing with the

consequences. Despite low volumes across the province, the highest volume hospitals have already been

found to perform more extensive lymph node sampling without increased morbidity or mortality and

decreased use of ICUs; these findings alone have meaningful survival and financial implications. Future

centralization policies should use Textbook Outcomes to identify a small group of motivated surgeons with

the proper hospital structures to support them. Textbook Outcome scores can be used to as a feedback

mechanism, with surgeons championing the changes necessary to improve the care delivered to gastric

cancer patients at their hospitals. Ontario can do better, and the first step was describing the current state

of care for patients with gastric cancer in this province.

4.6. FUTURE RESEARCH

Illusory superiority is a cognitive bias whereby people overestimate their own abilities, and health care

professionals are no exception. The results of these studies establish the current state of care for patients

with gastric cancer in Ontario as whole and in part according to volume quintiles. However, more is needed

in order illicit rapid change. The next step should entail personalized Textbook Outcome report cards for

hospitals and the multidisciplinary teams caring for patient with curative gastric cancer. In this way, each


109

team will receive a personal objective measure of their actual care quality and have the data necessary to

implement quality improvement initiatives. The results of these report cards may dissuade very low volume

providers from continuing to manage these patients and may inspire others advocate for these patients in

an evidence-based approach.

Textbook Outcome rates and patient survival improved during the study period, the reasons for which were

not explored in this thesis. A longitudinal analysis of temporal trends in gastric cancer processes of care in

Ontario would allow us to describe the patterns of care throughout the study period and potentially identify

a strategy for further improving care in Ontario and in other similar jurisdictions.

Textbook Outcomes have proven to be measurable and informative, and several studies have investigated

them in non-gastric surgery patients, including hepatic, pancreatic, biliary, colorectal, lung and even

bariatric surgery patients [265-269]. It will be interesting to describe the quality of care delivered to patients

with resectable esophageal cancer in Ontario, and the survival implications of Textbook Outcomes in this

related group of patients. This will further validate and emphasize the results of this thesis and hopefully

enforce the need for an oncologic surgery quality feedback mechanism in the province that is usable for

the majority of resectable solid tumors.

In the area of gastric cancer, these studies have established a strong association between Textbook

Outcomes and long-term survival, and as such, future work exploring additional processes to improve

Textbook Outcomes is advisable. These may include identifying surgeon and hospital-level characteristics

which predict Textbook Outcomes and investigating the efficacy of focused training programs for surgeons

and pathologists.

4.7. CONCLUSIONS

In conclusion, Textbook Outcomes are a new concept in cancer surgery and represent a composite of

perioperative quality metrics which is strongly associated with survival in gastric cancer patients. Increased

gastrectomy volume was not associated with improved Textbook Outcome rates and future policies for

improving gastric cancer surgery quality should focus on a combination of centralization and surgical quality

feedback systems using Textbook Outcome metrics.

Jordan Levy Gastric Cancer Surgery Textbook Outcomes in Ontario

110

CHAPTER 5 TABLES

TABLE FOR 2.3.2. MORPHOLOGY CODES AND HISTOLOGICAL CATEGORIES

Morphology Code† Description Histological Category

8140 Adenocarcinoma, NOS Intestinal*

8142 Linitis plastica Non-Intestinal

8144 Adenocarcinoma, intestinal type Intestinal*

8145 Carcinoma, diffuse type Non-Intestinal

8210 Adenocarcinoma in adenomatous polyp Intestinal

8211 Tubular adenocarcinoma Intestinal

8214 Parietal cell carcinoma Intestinal

8250 Bronchiolo-alveolar adenocarcinoma, NOS Intestinal

8255 Adenocarcinoma with mixed subtypes Intestinal*

8260 Papillary adenocarcinoma, NOS Intestinal

8261 Adenocarcinoma in villous adenoma Intestinal

8262 Villous adenocarcinoma Intestinal

8310 Clear cell adenocarcinoma, NOS Intestinal

8323 Mixed cell adenocarcinoma Intestinal

8480 Mucinous adenocarcinoma Non-Intestinal

8481 Mucin-producing adenocarcinoma Non-Intestinal

8490 Signet ring cell carcinoma Non-Intestinal

8574 Adenocarcinoma with neuroendocrine differentiation Intestinal

† International Classification of Diseases of Oncology 3rd Edition (ICD-O-3) Morphological Codes * Some tumors with these morphology codes were considered as non-intestinal based on pathologist reported diffuse or linitis plastica histological findings


111

TABLE FOR 2.3.3.A. GASTRECTOMY INTERVENTION CODES

Intervention

Codes† Description

1NF87DG Excision partial, stomach without vagotomy endoscopic [laparoscopic] approach

esophagogastric anastomosis

1NF87DH Excision partial, stomach without vagotomy endoscopic [laparoscopic] approach

gastroduodenal anastomosis

1NF87DJ Excision partial, stomach with vagotomy endoscopic [laparoscopic] approach gastroduodenal

anastomosis

1NF87DL Excision partial, stomach with vagotomy endoscopic [laparoscopic] approach gastrojejunal [or

gastroenteral NEC] anastomosis

1NF87DQ Excision partial, stomach without vagotomy endoscopic [laparoscopic] approach gastrojejunal

[or gastroenteral NEC] anastomosis

1NF87GX Excision partial, stomach with vagotomy endoscopic [laparoscopic] approach esophagogastric

anastomosis

1NF87RG Excision partial, stomach without vagotomy open approach gastroduodenal anastomosis

1NF87RH Excision partial, stomach with vagotomy open approach gastroduodenal anastomosis

1NF87RJ Excision partial, stomach without vagotomy open approach gastrojejunal [or gastroenteral NEC]

anastomosis

1NF87RK Excision partial, stomach with vagotomy open approach gastrojejunal [or gastroenteral NEC]

anastomosis

1NF87RP Excision partial, stomach without vagotomy open approach esophagogastric anastomosis

1NF87SH Excision partial, stomach with vagotomy open approach esophagogastric anastomosis

1NF89DAXXF Excision total, stomach using endoscopic [laparoscopic] approach with esophagojejunal

anastomosis (interposition pouch formation with or without Roux en Y)

1NF89DZ Excision total, stomach using esophagojejunal anastomosis, laparoscopic

1NF89GW Excision total, stomach using esophagoduodenal anastomosis, laproscopic

1NF89LAXXF Excision total, stomach using open approach with esophagojejunal anastomosis (interposition

pouch formation with or without Roux en Y)

1NF89SG Excision total, stomach using esophagoduodenal anastomosis, open

1NF89TH Excision total, stomach using esophagojejunal anastomosis, open

1NF90LAXXG Excision total with reconstruction, stomach using an open approach with jejunum

1NF91LAXXF Excision radical, stomach using open approach with esophagojejunal anastomosis (pouch

formation with or without Roux en Y)

1NF91RG Excision radical, stomach using open approach with gastroduodenal anastomosis

1NF91RJ Excision radical, stomach with gastrojejunal anastomosis

1NF91SG Excision radical, stomach with esophagoduodenal anastomosis

1NF91RJXXF Excision radical, stomach using open approach with gastrojejunal anastomosis (pouch

formation with or without Roux en Y)

1NF91RP Excision radical, stomach using open approach with esophagogastric anastomosis

1NF92LAXXG Excision radical with reconstruction, stomach using open approach with jejunum

† Canadian Classification of Health Interventions (CCI) codes


112

TABLE FOR 2.3.3.B. ESOPHAGOGASTRECTOMY INTERVENTION CODES

Intervention

Codes† Description

1NA87DB Excision partial, esophagus with anastomosis using endoscopic abdominal approach [e.g. open

cervical with laparoscopic approach]

1NA87DC Excision partial, esophagus abdominal [level] anastomosis (esophago-esophagostomy,

esophago-gastrostomy) using endoscopic abdominal approach [includes: open cervical with

laparoscopic approach]

1NA87DBXXF Excision partial, esophagus with interpositional (intestine) flap using endoscopic abdominal

approach [e.g. open cervical with laparoscopic approach]

1NA87DBXXG Excision partial, esophagus with gastric pull up using endoscopic

abdominal approach [e.g. open cervical with laparoscopic

approach]

1NA87EZ Excision partial, esophagus with anastomosis using endoscopic thoracic approach [e.g. open

cervical with thoracoscopic approach]

1NA87FA Excision partial, esophagus with anastomosis using combined endoscopic thoracoabdominal

approach

1NA87FAXXF Excision partial, esophagus with interpositional (intestine) flap using combined endoscopic

thoracoabdominal approach

1NA87FAXXG Excision partial, esophagus with gastric pull up using combined endoscopic thoracoabdominal

approach

1NA87FB Excision partial, esophagus abdominal [level] anastomosis (esophago-esophagostomy,

esophago-gastrostomy) using combined endoscopic thoraco-abdominal approach

1NA87LB Excision partial, esophagus with anastomosis using open cervical with abdominal [transhiatal]

approach

1NA87LBXXF Excision partial, esophagus with interpositional (intestine) flap using open cervical with

abdominal [transhiatal] approach

1NA87LBXXG Excision partial, esophagus with gastric pull up using open cervical with abdominal [transhiatal]

approach

1NA87LD Excision partial, esophagus abdominal [level] anastomosis (esophago-esophagostomy,

esophago-gastrostomy) using open abdominal approach [includes cervical with abdominal

approach, transhiatal approach]

1NA87LP Excision partial, esophagus with anastomosis using open cervical approach

1NA87QB Excision partial, esophagus with anastomosis using open thoracic approach [e.g. cervicothoracic

approach]

1NA87QBXXF Excision partial, esophagus with interpositional (intestine) flap using open thoracic approach

[e.g. cervicothoracic approach]

1NA87QBXXG Excision partial, esophagus with gastric pull up using open thoracic approach [e.g.

cervicothoracic approach]

1NA87QF Excision partial, esophagus with anastomosis using combined open (cervico)thoracoabdominal

approach

1NA87QFXXF Excision partial, esophagus with interpositional (intestine) flap using combined open

(cervico)thoracoabdominal approach

1NA87QFXXG Excision partial, esophagus with gastric pull up using combined open

(cervico)thoracoabdominal approach

1NA87QG Excision partial, esophagus abdominal [level] anastomosis (esophago-esophagostomy,

esophago-gastrostomy) using combined open thoraco-abdominal approach [includes: open

cervical with thoraco-abdominal approach]

1NA89DB Excision total, esophagus using endoscopic abdominal approach with anastomosis


113

1NA89DBXXF Excision total, esophagus using endoscopic abdominal approach [e.g. open cervical with

laparoscopic approach] with interpositional (intestine) flap

1NA89DBXXG Excision total, esophagus using endoscopic abdominal approach [e.g. open cervical with

laparoscopic approach] with gastric pull up

1NA89FA Excision total, esophagus using combined endoscopic thoracoabdominal approach with

anastomosis

1NA89FAXXF Excision total, esophagus using combined endoscopic thoracoabdominal approach with

interpositional (intestine) flap

1NA89FAXXG Excision total, esophagus using combined endoscopic thoracoabdominal approach with gastric

pull up

1NA89LB Excision total, esophagus using open cervical with abdominal [transhiatal] approach with

anastomosis

1NA89LBXXF Excision total, esophagus using open cervical with abdominal [transhiatal] approach with


1NA89LBXXG Excision total, esophagus using open cervical with abdominal [transhiatal] approach with gastric

pull up

1NA89QF Excision total, esophagus using combined open (cervico)thoracoabdominal approach with

anastomosis

1NA89QFXXF Excision total, esophagus using combined open (cervico)thoracoabdominal approach with


1NA89QFXXG Excision total, esophagus using combined open (cervico)thoracoabdominal approach with

gastric pull up

1NA91DB Excision radical, esophagus using endoscopic abdominal approach [e.g. open cervical with

laparoscopic approach] with anastomosis

1NA91DBXXF Excision radical, esophagus using endoscopic abdominal approach [e.g. open cervical with

laparoscopic approach] with interpositional (intestine) flap

1NA91DBXXG Excision radical, esophagus using endoscopic abdominal approach [e.g. open cervical with

laparoscopic approach] with gastric pull up

1NA91FA Excision radical, esophagus using combined endoscopic thoracoabdominal approach with

anastomosis

1NA91FAXXF Excision radical, esophagus using combined endoscopic thoracoabdominal approach with


1NA91FAXXG Excision radical, esophagus using combined endoscopic thoracoabdominal approach with

gastric pull up

1NA91LB Excision radical, esophagus using open cervical with abdominal [transhiatal] approach with

anastomosis

1NA91LBXXF Excision radical, esophagus using open cervical with abdominal [transhiatal] approach with


1NA91LBXXG Excision radical, esophagus using open cervical with abdominal [transhiatal] approach with

gastric pull up

1NA91QF Excision radical, esophagus using combined open (cervico)thoracoabdominal approach with

anastomosis

1NA91QFXXF Excision radical, esophagus using combined open (cervico)thoracoabdominal approach with


1NA91QFXXG Excision radical, esophagus using combined open (cervico)thoracoabdominal approach with

gastric pull up

† Canadian Classification of Health Interventions (CCI) codes


114

TABLE FOR 2.5.3. SEVERE COMPLICATION CATEGORIES – DIAGNOSTIC (ICD-10-CA) AND

INTERVENTION (CCI) CODES

Bleeding or Vascular Injury Complication

ICD-10-CA

D62 Acute posthaemorrhagic anaemia

K661 Haemoperitoneum

K920 Haematemesis

K921 Melaena

K922 Gastrointestinal haemorrhage, unspecified

R58 Haemorrhage, not elsewhere classified

T810 Haemorrhage and haematoma complicating a procedure, not elsewhere classified

S352 Injury of coeliac or mesenteric artery

T817 Vascular complications following a procedure, not elsewhere classified

CCI

1LZ37LAFP Installation of external appliance, circulatory system NEC extracorporeal blood

salvage device (cell saver) (intraoperative)

1OA13DAW3 Control of bleeding, liver endoscopic [laparoscopic] approach and fibrin glue

1OB13LA Control of bleeding, spleen open [abdominal] approach Using apposition

technique [e.g. suturing or NOS]

1OB13LAGX Control of bleeding, spleen open [abdominal] approach using device NEC

1OB13LAX7 Control of bleeding, spleen open [abdominal] approach Using chemical cautery

[e.g. topical thrombin]

1OB13LAXXN Control of bleeding, spleen open [abdominal] approach Using synthetic tissue

[e.g. mesh, teflon pledgets]

1OJ13LA Control of bleeding, pancreas using open approach and apposition technique (e.g.

suturing)

1OT13LA Control of bleeding, Abdominal cavity using open approach

1OT13LANP Control of bleeding, Abdominal cavity using open approach and leaving packing in

situ

1GY13LA Control of bleeding, thoracic cavity NEC using open approach

1NF13LA Control of bleeding, stomach using open approach (for devascularization)

1IN80LA Repair, pulmonary vein using open approach without tissue

1KE51LA Occlusion, abdominal arteries NEC open approach(e.g. arteriotomy) using direct

suture

1KE51LAFF Occlusion, abdominal arteries NEC open approach(e.g. arteriotomy) using band or

clip

1KE57LAGX Extraction, abdominal arteries NEC open approach no tissue used using device

NEC

1KE80LA Repair, abdominal arteries NEC using open approach

1KQ51LA Occlusion, abdominal veins NEC open approach(e.g. venotomy) using direct

suture

1KQ80LA Repair, abdominal veins NEC using open approach without tissue [suture]

1KT51DAFF Occlusion, vessels of the pelvis, perineum and gluteal region endoscopic

[laparoscopic] approach using clips

1KV80LA Repair, artery NEC using open approach

1LZ19HHU1J Transfusion, circulatory system NEC using homologous transfusion of red cell

concentrates

1LZ35HAC5 Pharmacotherapy (local), circulatory system NEC percutaneous injection

approach of blood related products

1LZ35HHC5 Pharmacotherapy (local), circulatory system NEC percutaneous infusion approach

of blood related products


115

Cardiac Complication

ICD-10-CA

I200 Unstable angina

I210 Acute transmural myocardial infarction of anterior wall

I211 Acute transmural myocardial infarction of inferior wall

I214 Acute subendocardial myocardial infarction

I2140 Acute subendocardial myocardial infarction of anterior wall

I2141 Acute subendocardial myocardial infarction of inferior wall

I2149 Acute subendocardial myocardial infarction, unspecified site

I219 Acute myocardial infarction, unspecified

I249 Acute ischaemic heart disease, unspecified

I308 Other forms of acute pericarditis

I309 Acute pericarditis, unspecified

I313 Pericardial effusion (noninflammatory)

I319 Disease of pericardium, unspecified

I339 Acute endocarditis, unspecified

I420 Dilated cardiomyopathy

I441 Atrioventricular block, second degree

I442 Atrioventricular block, complete

I447 Left bundle-branch block, unspecified

I454 Nonspecific intraventricular block

I460 Cardiac arrest with successful resuscitation

I461 Sudden cardiac death, so described

I469 Cardiac arrest, unspecified

I472 Ventricular tachycardia

I4900 Ventricular fibrillation

I500 Congestive heart failure

I509 Heart failure, unspecified

R570 Cardiogenic shock

CCI

1HA52QB Drainage, pericardium using thoracic approach (e.g. sternotomy, thoracotomy)

1HZ09JAFS Stimulation, heart NEC external approach using electrode converter/defibrillator

1HZ09JAJF Stimulation, heart NEC external approach using electrode with synchronized DC

shock

1HZ09LAFS Stimulation, heart NEC open approach using electrode converter/defibrillator

1HZ30JN Resuscitation, heart NEC by external manual compression with or without

concomitant ventilation

1HZ30JY Resuscitation, heart NEC with external manual compression and ventilation

1HZ37JANN Installation of external appliance, heart NEC of temporary (external) cardiac

pacemaker

1HZ53GRFS Implantation of internal device, heart NEC percutaneous transluminal approach

[transvenous] or approach NOS cardioverter/defibrillator [AICD, CRT-D, BiV-ICD]

1HZ53GRNK Implantation of internal device, heart NEC percutaneous transluminal approach

[transvenous] or approach NOS dual chamber programmable rate responsive

pacemaker (DVI,DDD, DDDR modes)

1HZ53GRNM Implantation of internal device, heart NEC percutaneous transluminal approach

[transvenous] or approach NOS Single chamber, programmable, rate responsive

pacemaker (VVD,VVI, AAI, VVIR, AAIR modes)

1HZ53GRNN Implantation of internal device, heart NEC percutaneous transluminal approach

[transvenous] or approach NOS Temporary pacemaker


116

1IC53GQQL Implantation of internal device, thoracic [descending] aorta of intra aortic balloon

using percutaneous transluminal approach [e.g. through femoral artery]

1IJ76LAXXQ Bypass, coronary arteries open approach [sternotomy] using combined sources of

tissue (e.g. graft/pedicled flap)

1IL16HTT9 Perfusion, vessels of heart percutaneous intermittent flow technique of

pharmacological agent NEC

1HA52HA Drainage, pericardium using percutaneous (needle) approach

1HA52HATS Drainage, pericardium using percutaneous (needle) approach leaving drainage

tube [catheter] in situ

Delirium, Cerebral Ischemia and Encephalopathic Complication

ICD-10-CA

F050 Delirium not superimposed on dementia, so described

F051 Delirium superimposed on dementia

F058 Other delirium

F059 Delirium, unspecified

G8199 Hemiplegia of unspecified type of unspecified [unilateral] side

G931 Anoxic brain damage, not elsewhere classified

I633 Cerebral infarction due to thrombosis of cerebral arteries

G934 Encephalopathy, unspecified

I635 Cerebral infarction due to unspecified occlusion or stenosis of cerebral arteries

I638 Other cerebral infarction

I639 Cerebral infarction, unspecified

I64 Stroke, not specified as haemorrhage or infarction

Thromboembolic Complication

ICD-10-CA

I269 Pulmonary embolism without mention of acute cor pulmonale

I800 Phlebitis and thrombophlebitis of superficial vessels of lower extremities

I802 Phlebitis and thrombophlebitis of other deep vessels of lower extremities

Hollow Organ Complication

ICD-10-CA

K222 Oesophageal obstruction

K223 Perforation of oesophagus

K251 Gastric ulcer, acute with perforation

K253 Gastric ulcer, acute without haemorrhage or perforation

K254 Gastric ulcer, chronic or unspecified with haemorrhage

K280 Gastrojejunal ulcer, acute with haemorrhage

K316 Fistula of stomach and duodenum

K430 Ventral hernia with obstruction, without gangrene

K436 Other and unspecified ventral hernia with obstruction, without gangrene

K440 Diaphragmatic hernia with obstruction, without gangrene

K529 Noninfective gastroenteritis and colitis, unspecified

K550 Acute vascular disorders of intestine

K562 Volvulus

K565 Intestinal adhesions [bands] with obstruction

K631 Perforation of intestine (nontraumatic)

K632 Fistula of intestine

K660 Peritoneal adhesions

K822 Perforation of gallbladder

K831 Obstruction of bile duct

K833 Fistula of bile duct

K913 Postoperative intestinal obstruction

N321 Vesicointestinal fistula


117

S36460 Laceration of small intestine, excluding duodenum without open wound into

cavity

S36510 Laceration of colon without open wound into cavity

T8152 Perforation due to foreign body accidentally left in body cavity or operation

wound following a procedure

T8182 Persistent postoperative fistula

T8183 Anastomotic breakdown/dehiscence

T855 Mechanical complication of gastrointestinal prosthetic devices, implants and

grafts

Infectious Complication – NOS

ICD-10-CA

A490 Staphylococcal infection, unspecified site

A498 Other bacterial infections of unspecified site

A499 Bacterial infection, unspecified

T814 Infection following a procedure, not elsewhere classified

Infectious Complication – Organ/Deep Space

ICD-10-CA

A047 Enterocolitis due to Clostridium difficile

A090 Other and unspecified gastroenteritis and colitis of infectious origin

A099 Gastroenteritis and colitis of unspecified origin

J853 Abscess of mediastinum

K521 Toxic gastroenteritis and colitis

K650 Acute peritonitis

K658 Other peritonitis

K659 Peritonitis, unspecified

R100 Acute abdomen

Pulmonary Complication

ICD-10-CA

J14 Pneumonia due to Haemophilus influenzae

J150 Pneumonia due to Klebsiella pneumoniae

J151 Pneumonia due to Pseudomonas

J152 Pneumonia due to Staphylococcus

J154 Pneumonia due to other streptococci

J155 Pneumonia due to Escherichia coli

J156 Pneumonia due to other aerobic Gram-negative bacteria

J180 Bronchopneumonia, unspecified

J181 Lobar pneumonia, unspecified

J189 Pneumonia, unspecified

J219 Acute bronchiolitis, unspecified

J690 Pneumonitis due to food and vomit

J698 Pneumonitis due to other solids and liquids

J80 Adult respiratory distress syndrome

J81 Pulmonary oedema

J860 Pyothorax with fistula

J869 Pyothorax without fistula

J938 Other pneumothorax

J939 Pneumothorax, unspecified

J940 Chylous effusion

J941 Fibrothorax

J942 Haemothorax

J9580 Postprocedural pneumothorax

J9588 Other postprocedural respiratory disorders


118

J959 Postprocedural respiratory disorder, unspecified

CCI

1GV35HH1C Pharmacotherapy (local), pleura transcatheter [chest tube] instillation or

insufflation approach using thrombolytic agent

1GV35HHC1 Pharmacotherapy (local), pleura transcatheter [chest tube] instillation or

insufflation approach using antithrombotic agent

Renal Injury Complication

ICD-10-CA

N144 Toxic nephropathy, not elsewhere classified

N19 Unspecified kidney failure

N170 Acute renal failure with tubular necrosis

N179 Acute renal failure, unspecified

N990 Postprocedural renal failure

CCI 1PZ21HQBR Dialysis, urinary system NEC hemodialysis

1PZ21HQBS Dialysis, urinary system NEC continuous venovenous hemodialysis

Respiratory Complication

ICD-10-CA

J951 Acute pulmonary insufficiency following thoracic surgery

J952 Acute pulmonary insufficiency following nonthoracic surgery

J960 Acute respiratory failure

J9600 Acute respiratory failure, type 1 [hypoxic]

J9601 Acute respiratory failure, type II [hypercapnic]

J9609 Acute respiratory failure, type unspecified

J969 Respiratory failure, unspecified

J9690 Respiratory failure, unspecified, type I [hypoxic]

J9691 Respiratory failure, unspecified, type II [hypercapnic]

J9699 Respiratory failure, unspecified, type unspecified

J9818 Other pulmonary collapse

R090 Asphyxia, unspecified

R092 Respiratory arrest

CCI

1GZ31CAEP Ventilation, respiratory system NEC using invasive per orifice approach by

endotracheal intubation manual hand assisted (e.g. ambu bag)

1GZ31CAND Ventilation, respiratory system NEC invasive per orifice approach by endotracheal

intubation and positive pressure

1GZ31CBND Ventilation, respiratory system NEC non-invasive approach and positive pressure

ventilation (e.g. CPAP, BIPAP)

1GZ31CRND Ventilation, respiratory system NEC invasive per orifice with incision approach for

intubation through tracheostomy positive pressure

1GZ31GPND Ventilation, respiratory system NEC invasive percutaneous transluminal approach

(e.g. transtracheal jet) through needle and positive pressure

Sepsis or Shock Complication

ICD-10-CA

A408 Other streptococcal sepsis

A410 Sepsis due to Staphylococcus aureus

A411 Sepsis due to other specified staphylococcus

A413 Sepsis due to Haemophilus influenzae

A4150 Sepsis due to Escherichia coli [E.coli]

A4152 Sepsis due to Serratia

A4158 Sepsis due to other Gram-negative organisms

A4180 Sepsis due to Enterococcus

A4188 Other specified sepsis

A419 Sepsis, unspecified


119

B377 Candidal sepsis

R571 Hypovolaemic shock

R572 Septic shock

R579 Shock, unspecified

R651 Systemic inflammatory response syndrome of infectious origin with acute organ

failure

T811 Shock during or resulting from a procedure, not elsewhere classified

Solid Organ Complication

ICD-10-CA

D735 Infarction of spleen

K720 Acute and subacute hepatic failure

K729 Hepatic failure, unspecified

K746 Other and unspecified cirrhosis of liver

K750 Abscess of liver

K763 Infarction of liver

K85 Acute pancreatitis

K850 Idiopathic acute pancreatitis

K858 Other acute pancreatitis

K859 Acute pancreatitis, unspecified

K863 Pseudocyst of pancreas

K868 Other specified diseases of pancreas

S36150 Liver haematoma NOS, laceration NOS, injury to liver NOS without open wound

into cavity

T812 Accidental puncture and laceration during a procedure, not elsewhere classified

Lymphatic or Thoracic Duct Complication

CCI

1MM51DA Occlusion, thoracic duct using endoscopic approach

1MM51LB Occlusion, thoracic duct using open abdominal approach

1MM51LP Occlusion, thoracic duct using open cervical [thoracic] approach

1MM80LP Repair, thoracic duct no tissue used using open cervical [thoracic] approach

ICD-10-CA: International Classification of Diseases 10th Revision – Canadian Version, CCI: Canadian Classification of Health Interventions

TABLE FOR 2.5.4. REINTERVENTIONS – CCI CODES

CCI Code Intervention Description

Operative Interventions

1EQ52LA Drainage, soft tissue of head and neck using open (incisional) approach 1GJ77HA Bypass with exteriorization, trachea using percutaneous needle approach (e.g.

percutaneous dilational tracheostomy - PDT) 1GJ77LA Bypass with exteriorization, trachea using open approach (e.g. collar incision)

1GJ77LALG Bypass with exteriorization, trachea using open approach and temporary implant 1GJ77QB Bypass with exteriorization, trachea using mediastinal approach

1GJ50CANG Dilation, trachea per orifice approach using endotracheal tube (without ventilator connection)

1GM86MEXXE Closure of fistula, bronchus NEC for fistula terminating in abdominal cavity [e.g. bronchoesophageal, bronchogastric] using local flap

1GM86MEXXF Closure of fistula, bronchus NEC for fistula terminating in abdominal cavity [e.g. bronchoesophageal, bronchogastric] using free flap

1GR87DA Excision partial, lobe of lung using endoscopic approach [VATS] 1GR87QB Excision partial, lobe of lung using open thoracic approach 1GR89QB Excision total, lobe of lung using open thoracic approach


120

1GV52HA Drainage, pleura using percutaneous (needle) approach 1GV52HAHE Drainage, pleura using percutaneous catheter (intracostal) with underwater seal drainage

system 1GV52HATK Drainage, pleura using percutaneous catheter with suction pump, (under water seal or

negative pressure) 1GV52LA Drainage, pleura using open approach

1GV52LATS Drainage, pleura using open approach and leaving drainage tube in situ 1GV59DAGX Destruction, pleura using endoscopic approach [VATS] and device NEC 1GV59HAZ9 Destruction, pleura using percutaneous instillation of chemical agent NEC (e.g. talc)

1GV87DA Excision partial, pleura using endoscopic approach [VATS] 1GV87LA Excision partial, pleura using open approach 1GV89LA Excision total, pleura using open approach

1GW52LATS Drainage, mediastinum using open approach and leaving drainage tube in situ 1GX80LB Repair, diaphragm with simple closure open abdominal approach

1GX80QB Repair, diaphragm with simple closure open thoracic approach 1GY72DA Release, thoracic cavity NEC using endoscopic approach 1GY72LA Release, thoracic cavity NEC using open approach

1KQ51LAW3 Occlusion, abdominal veins NEC open approach (e.g. venotomy) using fibrin glue 1KQ80LA Repair, abdominal veins NEC using open approach without tissue [suture] 1KT80LA Repair, vessels of the pelvis, perineum and gluteal region using open approach without

tissue [anastomosis] 1MM51DA Occlusion, thoracic duct using endoscopic approach 1MM51LB Occlusion, thoracic duct using open abdominal approach 1MM51LP Occlusion, thoracic duct using open cervical [thoracic] approach 1MM80LP Repair, thoracic duct no tissue used using open cervical [thoracic] approach

1NA52LPTS Drainage, esophagus using open (cervical) approach and leaving indwelling drainage tube (esophagostomy tube)

1NA77SQ Bypass with exteriorization, esophagus using chest wall [subcutaneous] tunnel exteriorization technique

1NA77TC Bypass with exteriorization, esophagus using end cervical exteriorization technique 1NA80DB Repair, esophagus using apposition technique [e.g. suturing] for closure using endoscopic

abdominal approach [includes: open cervical with laparoscopic approach] 1NA80FA Repair, esophagus using apposition technique [e.g. suturing] for closure using combined

endoscopic thoraco-abdominal approach 1NA80LB Repair, esophagus using apposition technique [e.g. suturing] for closure using open

abdominal approach [includes: cervical with abdominal approach, transhiatal approach] 1NA80LBXXE Repair, esophagus using local transposition flap [e.g. gastric fundus wrap, fundoplication]

for closure using open abdominal approach [includes: cervical with abdominal approach, transhiatal approach]

1NA80QB Repair, esophagus using apposition technique [e.g. suturing] for closure using open thoracic approach [includes: open cervicothoracic approach]

1NA80QG Repair, esophagus using apposition technique [e.g. suturing] for closure using combined open thoraco-abdominal approach (includes: open cervical with thoraco-abdominal approach)

1NA82TH Reattachment, esophagus of esophagoenterostomy 1NA86MT Closure of fistula, esophagus with simple excision [with or without closure] for fistula

travelling through multiple cavities and terminating in any organ(s) or at skin 1NA87DC Excision partial, esophagus abdominal [level] anastomosis (esophago-esophagostomy,

esophago-gastrostomy) using endoscopic abdominal approach [includes: open cervical with laparoscopic approach]

1NA87QC Excision partial, esophagus cervical [level] anastomosis (esophago-esophagostomy) using open thoracic approach [includes: open cervicothoracic approach]


121

1NA87QD Excision partial, esophagus thoracic [level] anastomosis (esophago-esophagostomy) using open thoracic approach [includes: open cervicothoracic approach]

1NA88QFXXG Excision partial with reconstruction, esophagus open thoraco-abdominal approach [includes: cervical with thoraco-abdominal approach] with gastric pull-up

1NA89LB Excision total, esophagus using open cervical with abdominal approach (with anastomosis) 1NA90LBXXG Excision total with reconstruction, esophagus open cervical with abdominal approach

[includes: transhiatal approach] with gastric pull-up 1NA90QFXXG Excision total with reconstruction, esophagus open thoraco-abdominal approach [includes:

cervical with thoraco-abdominal approach] with gastric pull-up 1NA91FA Excision radical, esophagus using endoscopic thoraco abdominal approach with gastric

anastomosis 1NE80LA Repair, pylorus open approach without concomitant vagotomy 1NE80XN Repair, pylorus open approach with truncal vagotomy [includes: vagotomy NOS] 1NF13LA Control of bleeding, stomach using open approach (for devascularization)

1NF53HATS Implantation of internal device, stomach of (gastric) tube using percutaneous approach 1NF53LATS Implantation of internal device, stomach of (gastric) tube using open (laparotomy)

approach 1NF76GW Bypass, stomach endoscopic [laparoscopic] approach Esophagoduodenostomy 1NF76SG Bypass, stomach open approach Esophagoduodenostomy 1NF76TH Bypass, stomach open approach Esophagoenterostomy NEC 1NF78SH Repair by decreasing size, stomach open approach using gastric bypass technique with

gastroenterostomy [e.g. Roux-en-Y] 1NF80LA Repair, stomach open approach using apposition technique [e.g. sutures] 1NF82RJ Reattachment, stomach of gastroenterostomy using open approach

1NF87DA Excision partial, stomach without vagotomy using endoscopic [laparoscopic] approach simple apposition technique or no closure needed for tissue regeneration [e.g. for polypectomy]

1NF87LA Excision partial, stomach open approach and simple apposition technique or no closure needed for tissue regeneration [e.g. for polypectomy]

1NF87RG Excision partial, stomach without vagotomy open approach gastroduodenal anastomosis 1NF87RJ Excision partial, stomach without vagotomy open approach gastrojejunal [or gastroenteral

NEC] anastomosis 1NF87RP Excision partial, stomach without vagotomy open approach esophagogastric anastomosis

1NF89LAXXF 1NF89LAXXF 1NF89TH Excision total, stomach open approach Esophagojejunal [or esophagoenteral NEC]

anastomosis 1NF90LAXXG Excision total with reconstruction, stomach using open approach with jejunum (to construct

pouch) 1NK52HATS Drainage, small intestine percutaneous (needle) approach leaving drainage/decompression

tube in situ 1NK52LA Drainage, small intestine open approach aspiration [suction] technique

1NK53BTTS Implantation of internal device, small intestine of feeding tube [jejunal] using endoscopic per orifice approach with percutaneous incision

1NK53CATS Implantation of internal device, small intestine of feeding tube [jejunal] using per orifice approach [e.g. naso intestinal]

1NK53DATS Implantation of internal device, small intestine of feeding tube [jejunal] using endoscopic [laparoscopic] approach

1NK53HATS Implantation of internal device, small intestine of feeding tube (jejunal) using percutaneous approach

1NK53LATS Implantation of internal device, small intestine of feeding tube [jejunal] using open approach

1NK53TGTS Implantation of internal device, small intestine of feeding tube [jejunal] using open approach and formation of mucous fistula


122

1NK76RF Bypass, small intestine open approach Enteroenterostomy bypass technique 1NK76RJ Bypass, small intestine open approach Gastroenterostomy bypass technique

1NK77EM Bypass with exteriorization, small intestine endoscopic [laparoscopic] approach feeding enterostomy (e.g. jejunostomy)

1NK77RQ Bypass with exteriorization, small intestine open approach feeding enterostomy (e.g. jejunostomy)

1NK77RR Bypass with exteriorization, small intestine open approach end enterostomy (e.g. terminal, end or loop ileostomy)

1NK80DA Repair, small intestine endoscopic [laparoscopic] approach using apposition technique [e.g. suturing, stapling]

1NK80DAXXE Repair, small intestine endoscopic [laparoscopic] approach using local transposition flap [e.g. omental patch]

1NK80LA Repair, small intestine open approach using apposition technique [e.g. suturing, stapling] 1NK80LAXXE Repair, small intestine open approach using local transposition flap [e.g. omental patch]

1NK82RQ Reattachment, small intestine open approach of feeding jejunostomy 1NK87DP Excision partial, small intestine endoscopic [laparoscopic] approach Enteroenterostomy

anastomosis technique 1NK87LA Excision partial, small intestine open approach Simple excisional technique 1NK87RE Excision partial, small intestine open approach Enterocolostomy anastomosis technique 1NK87RF Excision partial, small intestine open approach Enteroenterostomy anastomosis technique 1NK87TF Excision partial, small intestine open approach Stoma formation with distal closure

1NM77RS Bypass with exteriorization, large intestine colostomy using open approach 1NM80LA Repair, large intestine open approach using apposition technique [e.g. suturing, stapling] 1NM87BA Excision partial, large intestine endoscopic per orifice approach Simple excisional technique 1NM87DF Excision partial, large intestine endoscopic [laparoscopic, laparoscopic-assisted, hand-

assisted] approach colocolostomy anastomosis technique 1NM87DX Excision partial, large intestine stoma formation and distal closure 1NM87RE Excision partial, large intestine open approach Enterocolostomy anastomosis technique 1NM87RN Excision partial, large intestine open approach Colocolostomy anastomosis technique 1NM87TF Excision partial, large intestine open approach Stoma formation with distal closure 1NM87TG Excision partial, large intestine open approach Stoma formation with creation of mucous

fistula 1NP72DA Release, small with large intestine using endoscopic [laparoscopic] approach 1NP72LA Release, small with large intestine using open approach 1NP73LA Reduction, small with large intestine using open approach

1NP86ME Closure of fistula, small with large intestine with simple excision [with or without closure] for fistula terminating in abdominal cavity [any organ of digestive or biliary tract]

1NP86MQXXG Closure of fistula, small with large intestine using pedicled flap [e.g. advancement muscle flap] for fistula terminating in thoracic cavity

1NP86MT Closure of fistula, small with large intestine with simple excision [with or without closure] for fistula travelling through multiple cavities and terminating in any organ(s) or at skin

1NQ87BA Excision partial, rectum endoscopic per orifice approach closure by apposition technique [e.g. suturing, stapling] or no closure required (for tissue regeneration)

1NV89LA Excision total, appendix using open approach 1OA13LA Control of bleeding, liver using open approach (e.g. manual pressure, suturing or packing)

1OA13LAW3 Control of bleeding, liver open approach and fibrin glue 1OA13LAX7 Control of bleeding, liver open approach and chemical cautery agent

1OA52DA Drainage, liver using endoscopic [laparoscopic] approach 1OA52LA Drainage, liver using open [abdominal] approach 1OA87LA Excision partial, liver using open approach 1OB89LA Excision total, spleen using open [abdominal] approach 1OD89LA Excision total, gallbladder open approach without extraction of calculi cholecystectomy

alone


123

1OE52LATS Drainage, bile ducts using open approach leaving catheter (tube) in situ 1OE80LA Repair, bile ducts open approach using apposition technique [e.g. suturing]

1OJ52LATS Drainage, pancreas leaving drainage tube in situ using open (abdominal) approach 1OJ76VK Bypass, pancreas open approach using pancreaticoenterostomy diversion 1OT13LA Control of bleeding, Abdominal cavity using open approach 1OT52DA Drainage, abdominal cavity using endoscopic (laparoscopic) approach

1OT52DATS Drainage, abdominal cavity using endoscopic (laparoscopic) approach and leaving drainage tube in situ

1OT52LA Drainage, abdominal cavity using open approach 1OT52LATS Drainage, abdominal cavity using open (incisional) approach and leaving drainage tube in

situ 1OT72LA Release, abdominal cavity open approach using device NEC 1OT80LA Repair, abdominal cavity using open approach without tissue

1OT80LAXXE Repair, abdominal cavity using open approach with local transposition flap [e.g. omental, mesenteric patch]

1OT87LA Excision partial, abdominal cavity using open approach 1OW80LA Repair, surgically constructed sites in digestive & biliary tract using open approach 1OW87LA Excision partial, surgically constructed sites in digestive & biliary tract using open approach 1OW89LA Excision total, surgically constructed sites in digestive & biliary tract using open approach 1SY80DA Repair, muscles of the chest and abdomen endoscopic [laparoscopic] approach without

tissue [e.g. suturing or stapling] 1SY80LA Repair, muscles of the chest and abdomen open approach without tissue [e.g. suturing or

stapling] 1SY80LAFF Repair, muscles of the chest and abdomen open approach and using temporary abdominal

closure device [e.g. Bogota bag] 1SY80LAXXN Repair, muscles of the chest and abdomen open approach using synthetic tissue [e.g. mesh,

sponge] 1SZ52LA Drainage, soft tissue of the chest and abdomen using open (incisional) approach 1SZ59LA Destruction, soft tissue of the chest and abdomen using open approach

1SZ87LAXXN Excision partial, soft tissue of the chest and abdomen using open approach and synthetic tissue [e.g. mesh] (to close surgical defect)

1YG52LA Drainage, skin of neck using incisional approach 1YS52LA Drainage, skin of abdomen and trunk using incisional approach

1YS80LAXXB Repair, skin of abdomen and trunk using split-thickness autograft

Endoscopic Interventions

1NF13BA Control of bleeding, stomach using endoscopic per orifice approach 1NF13BAAG Control of bleeding, stomach using endoscopic per orifice approach and laser 1NF13BAGX Control of bleeding, stomach using endoscopic per orifice approach and device NEC [e.g.

electrocautery, endoclips] 1NF13BAKK Control of bleeding, stomach using endoscopic per orifice approach and special electrical

heat device [e.g. argon beam coagulator, gold probe] 1NF13BAW4 Control of bleeding, stomach using endoscopic per orifice approach and glue [e.g.

superglue, Histocryl] 1FX52BA Drainage, oropharynx using endoscopic per orifice approach

1GJ50BANR Dilation, trachea endoscopic approach using stent 1GM35BAD1 Pharmacotherapy (local), bronchus NEC using endoscopic per orifice approach and

antiinfective irrigating solution 1GM35BAD2 Pharmacotherapy (local), bronchus NEC using endoscopic per orifice approach and salt

irrigating solution 1GM50BANQ Dilation, bronchus NEC using endoscopic per orifice and insertion of stent

1GM52BATJ Drainage, bronchus NEC using endoscopic per orifice approach (bronchoscope) with suction device


124

1GM52CATJ Drainage, bronchus NEC using per orifice approach with suction device 1NA13BA Control of bleeding, esophagus using endoscopic per orifice approach

1NA50BABD Dilation, esophagus using endoscopic per orifice approach and balloon dilator 1NA50BABJ Dilation, esophagus using endoscopic per orifice approach and flexible dilator 1NA50BABP Dilation, esophagus using endoscopic per orifice approach and rigid dilator 1NA50BANR Dilation, esophagus using endoscopic per orifice approach and stent 1NA50CABJ Dilation, esophagus using per orifice approach and (unguided) flexible dilator

1NA52BA Drainage, esophagus using endoscopic per orifice approach and aspiration 1NA52CANR Drainage, esophagus using per orifice approach and stent (for dilation and drainage) 1NA52CATS Drainage, esophagus using per orifice approach and leaving indwelling tube 1NE35BAL7 Pharmacotherapy (local), pylorus of bacterial toxin [e.g. botulinum toxin or Botox] using

endoscopic per orifice approach 1NE50BA Dilation, pylorus endoscopic per orifice approach without concomitant vagotomy 1NF13BA Control of bleeding, stomach using endoscopic per orifice approach

1NF13BAAG Control of bleeding, stomach using endoscopic per orifice approach and laser 1NF13BAGX Control of bleeding, stomach using endoscopic per orifice approach and device NEC [e.g.

electrocautery, endoclips] 1NF13BAKK Control of bleeding, stomach using endoscopic per orifice approach and special electrical

heat device [e.g. argon beam coagulator, gold probe] 1NF13BAW4 Control of bleeding, stomach using endoscopic per orifice approach and glue [e.g.

superglue, Histocryl] 1NF50BABL Dilation, stomach using endoscopic per orifice approach and balloon (hydrostatic) 1NF53BTQB Implantation of internal device, stomach of (gastric) valved tube using per orifice

endoscopic approach with percutaneous incision 1NF53BTTS Implantation of internal device, stomach of (gastric) tube using per orifice endoscopic

approach with percutaneous incision 1NF53DATS Implantation of internal device, stomach of (gastric) tube using endoscopic (laparoscopic)

approach 1NK50BABD Dilation, small intestine using endoscopic per orifice approach with balloon dilator 1NK52CATS Drainage, small intestine per orifice approach leaving drainage/decompression tube in situ

1OW13BAGX Control of bleeding, surgically constructed sites in digestive and biliary tract using endoscopic per orifice [or via stoma] approach and device NEC (e.g. bicap electrocautery)

1OW50BABD Dilation, surgically constructed sites in digestive & biliary tract using endoscopic per orifice approach and balloon dilator

1OW52BATS Drainage, surgically constructed sites in digestive and biliary tract using endoscopic per orifice approach and leaving drainage tube in situ

1OW80BA Repair, surgically constructed sites in digestive and biliary tract using endoscopic per orifice approach (or via stoma)

1PL50BAGX Dilation, bladder neck using endoscopic per orifice (transurethral) approach and device NEC 1PL59BAGX Destruction, bladder neck endoscopic per orifice approach using device NEC 1PM52BATS Drainage, bladder using endoscopic per orifice approach and drainage catheter

Interventional Radiology and Percutaneous Interventions

1GV52DATS Drainage, pleura using endoscopic approach and leaving drainage tube in situ 1GW52HATS Drainage, mediastinum using percutaneous approach and drainage tube 1KA57GQFV Extraction, abdominal aorta percutaneous transluminal approach no tissue used using

atherectomy device (e.g. transluminal extractor catheter, rotoablator) 1KE51GQGE Occlusion, abdominal arteries NEC percutaneous transluminal approach using [detachable]

coils 1KE51GQW0 Occlusion, abdominal arteries NEC percutaneous transluminal approach using synthetic

agent [e.g. gelfoam, silicone, microspheres, polystyrene, polyvinyl alcohol, contour particles]


125

1KG57GQFV Extraction, arteries of leg NEC percutaneous transluminal approach no tissue used using atherectomy device (e.g. transluminal extractor catheter, rotoablator)

1KT51GQGE Occlusion, vessels of the pelvis, perineum and gluteal region percutaneous transluminal approach using (detachable) coil

1IS51GRKA Occlusion, vena cava (superior and inferior) using filtering device percutaneous transluminal approach

1OA52HATS Drainage, liver using percutaneous [trans abdominal] approach leaving drainage tube in situ 1OB13GQW0 Control of bleeding, spleen percutaneous transluminal approach using synthetic agent [e.g.

gelfoam, silicone, microspheres, polystyrene, polyvinyl alcohol, contour particles]] 1OD52HA Drainage, gallbladder using percutaneous (needle) approach

1OD52HATS Drainage, gallbladder using percutaneous (needle) approach and leaving drainage tube in situ

1OE50HANR Dilation, bile ducts percutaneous [transhepatic] transluminal approach using rigid dilator [e.g. stent]

1OE52GPTS Drainage, bile ducts using percutaneous transluminal approach [e.g. transhepatic] leaving catheter (tube) in situ

1OT52HA Drainage, abdominal cavity using percutaneous (needle) approach 1OT52HATS Drainage, abdominal cavity using percutaneous (needle) approach and leaving drainage

tube in situ 1OT52HHD1 Drainage, abdominal cavity using percutaneous transcatheter approach and anti-infective

irrigating solution 1OT52HHD3 Drainage, abdominal cavity using percutaneous transcatheter approach and other irrigating

solution 1PE52HH Drainage, renal pelvis using percutaneous approach with insertion of tube (e.g.

nephrostomy, pyelostomy) 1PM52HHTS Drainage, bladder using percutaneous transcatheter approach and drainage catheter

1SQ52HA Drainage, pelvis using percutaneous (needle) approach 1SZ52HA Drainage, soft tissue of the chest and abdomen using percutaneous (needle) approach

1SZ52HATS Drainage, soft tissue of the chest and abdomen using percutaneous approach with drainage tube left in situ

1OW35HAD3 Pharmacotherapy, surgically constructed sites in digestive & biliary tract using percutaneous needle approach [injection] and other irrigating solution

1YS52HA Drainage, skin of abdomen and trunk using needle aspiration CCI: Canadian Classification of Health Interventions


126

TABLE FOR 4.4.1.A. COMPARISON OF PATIENTS WITH AND WITHOUT PATHOLOGY REPORT

Characteristic Report Unavailable (n=682)

Report Available (n=1,963)

SD

Surgeon Volume Quintile

1 205 (30.1%) 421 (21.4%) 20%

2 120 (17.6%) 344 (17.5%) 0% 3 119 (17.4%) 431 (22.0%) 11%

4 98 (14.4%) 367 (18.7%) 12%

5 101 (14.8%) 357 (18.2%) 9%

Missing 39 (5.7%) 43 (2.2%) 18%

Hospital Volume Quintile

1 202 (29.6%) 324 (16.5%) 32% 2 171 (25.1%) 445 (22.7%) 6%

3 101 (14.8%) 361 (18.4%) 10%

4 106 (15.5%) 450 (22.9%) 19%

5 102 (15.0%) 383 (19.5%) 12%

No Severe Complications 462 (67.7%) 1,333 (67.9%) 0% No Reinterventions 597 (87.5%) 1,740 (88.6%) 3%

No Unplanned ICU Admission 475 (69.6%) 1,419 (72.3%) 6%

Length of Stay 21 Days or Less 588 (86.2%) 1,720 (87.6%) 4%

No 30-Day Readmission 593 (87.0%) 1,675 (85.3%) 5%

No 30-Day Mortality 665 (97.5%) 1,883 (95.9%) 9%

3-Year Survival Rate 54.7% 56.1% LR p=0.70 Age, mean (SDV) 66.1 (13.0) 68.1 (12.3) 15%

Age Group, n (%)

<50 89 (13.0%) 185 (9.4%) 11%

50-59 99 (14.5%) 284 (14.5%) 0%

60-69 181 (26.5%) 472 (24.0%) 6%

70-79 217 (31.8%) 665 (33.9%) 4% >80 96 (14.1%) 357 (18.2%) 11%

Sex, n (%)

Female 262 (38.4%) 717 (36.5%) 4%

Male 420 (61.6%) 1,246 (63.5%) 4%


0-2 35 (5.1%) 90 (4.6%) 3% 3 313 (45.9%) 971 (49.5%) 7%

4 205 (30.1%) 553 (28.2%) 4%

5 129 (18.9%) 349 (17.8%) 3%


1 94 (13.8%) 288 (14.7%) 3%

2 127 (18.6%) 331 (16.9%) 5% 3 136 (19.9%) 399 (20.3%) 1%

4 152 (22.3%) 443 (22.6%) 1%

5 157 (23.0%) 477 (24.3%) 3%

Missing 16 (2.3%) 25 (1.3%) 8%


Neoadjuvant Chemotherapy, n (%) 78 (11.4%) 220 (11.2%) 1%

Adjuvant Chemotherapy, n (%) 330 (48.4%) 902 (46.0%) 5%

Adjuvant Radiation, n (%) 187 (27.4%) 558 (28.4%) 2%


127


Cardia, NOS 71 (10.4%) 208 (10.6%) 1%

Fundus 40 (5.9%) 100 (5.1%) 3%

Body 81 (11.9%) 244 (12.4%) 2%

Antrum 234 (34.3%) 718 (36.6%) 5%

Pylorus 26 (3.8%) 77 (3.9%) 1%



Overlapping Lesion 17 (2.5%) 56 (2.9%) 2% NOS 88 (12.9%) 175 (8.9%) 13%

SD: Standardized Differences, SDV: Standard Deviation, IQR: Interquartile Range

Standardized Difference of 10% or greater considered clinically meaningful

TABLE FOR 4.4.1.B. PREDICTORS OF MISSING PATHOLOGY REPORT

Variable p-value

Surgeon Volume 0.30

Hospital Volume <0.001

Age Groups <0.001

Sex 0.65

Resource Utilization Bands 0.25 Material Deprivation Quintile 0.77

Year of Gastrectomy (2004-2015) <0.001

Tumor Location 0.34

No Severe Complications 0.27 No Re-Interventions 0.55

No Unplanned ICU Admission 0.61

No Prolonged LOS (21 days or less) 0.96

No 30-Day Readmission 0.37

No 30-Day Mortality 0.02 Mixed Effect Multivariable Logistic Model adjusted for six non-pathology-

based TO metrics (excluding margin status and adequate

lymphadenectomy), covariates and clustering by hospital.

p-value of <0.004 significant according to Bonferroni adjustment 0.05/14


128

REFERENCES

1. Kumar, V., et al., Cotran pathologic basis of disease. 2010. Saunders Elsevier, 8th ed. P, 2010. 784. 2. Connolly, E., E. Gaffney, and J. Reynolds, Gastrointestinal stromal tumours. British Journal of

Surgery, 2003. 90(10): p. 1178-1186. 3. Blackstein, M.E., et al., Gastrointestinal stromal tumours: consensus statement on diagnosis and

treatment. Canadian Journal of Gastroenterology and Hepatology, 2006. 20(3): p. 157-163. 4. Kitamura, Y., S. Hirota, and T. Nishida, Gastrointestinal stromal tumors (GIST): a model for

molecule‐based diagnosis and treatment of solid tumors. Cancer science, 2003. 94(4): p. 315-320. 5. Kindblom, L.-G., et al., Gastrointestinal pacemaker cell tumor (GIPACT): gastrointestinal stromal

tumors show phenotypic characteristics of the interstitial cells of Cajal. The American journal of pathology, 1998. 152(5): p. 1259.

6. Borch, K., et al., Gastric carcinoids: biologic behavior and prognosis after differentiated treatment in relation to type. Annals of surgery, 2005. 242(1): p. 64.

7. Isaacson, P., Gastric MALT lymphoma: from concept to cure. Annals of Oncology, 1999. 10(6): p. 637-645.

8. Davis, G.B., et al., Tumors of the stomach. World journal of surgery, 2000. 24(4): p. 412-420. 9. Shiu, M., et al., Myosarcomas of the stomach: natural history, prognostic factors and

management. Cancer, 1982. 49(1): p. 177-187. 10. Wright, F.C., et al., Surgical Oncology Manual. Second Edition ed. 2016: Springer. 11. Modlin, I.M., K.D. Lye, and M. Kidd, A 50-year analysis of 562 gastric carcinoids: small tumor or

larger problem? The American journal of gastroenterology, 2004. 99(1): p. 23. 12. Kelley, J.R. and J.M. Duggan, Gastric cancer epidemiology and risk factors. Journal of clinical

epidemiology, 2003. 56(1): p. 1-9. 13. Lauren, P., The two histological main types of gastric carcinoma: diffuse and so‐called intestinal‐

type carcinoma: an attempt at a histo‐clinical classification. Acta Pathologica Microbiologica Scandinavica, 1965. 64(1): p. 31-49.

14. Polkowski, W., et al., Prognostic value of Lauren classification and c-erbB-2 oncogene overexpression in adenocarcinoma of the esophagus and gastroesophageal junction. Annals of surgical oncology, 1999. 6(3): p. 290-297.

15. Bosman, F.T., et al., WHO classification of tumours of the digestive system. 2010: World Health Organization.

16. Hu, B., et al., Gastric cancer: Classification, histology and application of molecular pathology. Journal of gastrointestinal oncology, 2012. 3(3): p. 251.

17. International Agency for Research on Cancer, Infection with Helicobacter pylori. In: IARC monographs on the Evaluation of Carcinogenic Risks to Humans. Schisto-somiasis, Liver Flukes and Helicobacter pylori, 1994. 61: p. 177-241.

18. Herrera, V. and J. Parsonnet, Helicobacter pylori and gastric adenocarcinoma. Clinical Microbiology and Infection, 2009. 15(11): p. 971-976.

19. Kamangar, F., et al., Opposing risks of gastric cardia and noncardia gastric adenocarcinomas associated with Helicobacter pylori seropositivity. Journal of the National Cancer Institute, 2006. 98(20): p. 1445-1452.

20. Plummer, M., et al., Global burden of gastric cancer attributable to Helicobacter pylori. International journal of cancer, 2015. 136(2): p. 487-490.

21. Correa, P., et al., A model for gastric cancer epidemiology. The Lancet, 1975. 306(7924): p. 58-60. 22. Rugge, M., M. Fassan, and D.Y. Graham, Epidemiology of gastric cancer, in Gastric Cancer. 2015,

Springer. p. 23-34.


129

23. Delahay, R.M. and M. Rugge, Pathogenesis of Helicobacter pylori infection. Helicobacter, 2012. 17: p. 9-15.

24. Shiotani, A., P. Cen, and D.Y. Graham. Eradication of gastric cancer is now both possible and practical. in Seminars in cancer biology. 2013. Elsevier.

25. Holcombe, C., Helicobacter pylori: the African enigma. Gut, 1992. 33(4): p. 429. 26. Murphy, G., et al., Meta-analysis shows that prevalence of Epstein–Barr virus-positive gastric

cancer differs based on sex and anatomic location. Gastroenterology, 2009. 137(3): p. 824-833. 27. González, C.A., et al., Smoking and the risk of gastric cancer in the European Prospective

Investigation Into Cancer and Nutrition (EPIC). International journal of cancer, 2003. 107(4): p. 629-634.

28. Correa, P., Gastric cancer: overview. Gastroenterology Clinics of North America, 2013. 42(2): p. 211.

29. Tsugane, S., Salt, salted food intake, and risk of gastric cancer: epidemiologic evidence. Cancer science, 2005. 96(1): p. 1-6.

30. Jakszyn, P. and C.A. González, Nitrosamine and related food intake and gastric and oesophageal cancer risk: a systematic review of the epidemiological evidence. World journal of gastroenterology: WJG, 2006. 12(27): p. 4296.

31. González, C.A. and A. Agudo, Carcinogenesis, prevention and early detection of gastric cancer: where we are and where we should go. International journal of cancer, 2012. 130(4): p. 745-753.

32. Karimi, P., et al., Gastric cancer: descriptive epidemiology, risk factors, screening, and prevention. Cancer Epidemiology and Prevention Biomarkers, 2014. 23(5): p. 700-713.

33. Plummer, M., S. Franceschi, and N. Muñoz, Epidemiology of gastric cancer. IARC scientific publications, 2004(157): p. 311-326.

34. Brenner, H., D. Rothenbacher, and V. Arndt, Epidemiology of stomach cancer, in Cancer Epidemiology. 2009, Springer. p. 467-477.

35. Yaghoobi, M., R. Bijarchi, and S. Narod, Family history and the risk of gastric cancer. British journal of cancer, 2010. 102(2): p. 237.

36. Pharoah, P.D., et al., Incidence of gastric cancer and breast cancer in CDH1 (E-cadherin) mutation carriers from hereditary diffuse gastric cancer families. Gastroenterology, 2001. 121(6): p. 1348-1353.

37. Fitzgerald, R.C., et al., Hereditary diffuse gastric cancer: updated consensus guidelines for clinical management and directions for future research. Journal of medical genetics, 2010. 47(7): p. 436-444.

38. Giardiello, F.M., et al., Very high risk of cancer in familial Peutz–Jeghers syndrome. Gastroenterology, 2000. 119(6): p. 1447-1453.

39. Van Lier, M., et al., High cancer risk in Peutz–Jeghers syndrome: a systematic review and surveillance recommendations. The American journal of gastroenterology, 2010. 105(6): p. 1258.

40. Masciari, S., et al., Gastric cancer in individuals with Li-Fraumeni syndrome. Genetics in Medicine, 2011. 13(7): p. 651.

41. Jagelman, D., et al., Upper gastrointestinal cancer in familial adenomatous polyposis. The Lancet, 1988. 331(8595): p. 1149-1151.

42. Horii, A., et al., The APC gene, responsible for familial adenomatous polyposis, is mutated in human gastric cancer. Cancer research, 1992. 52(11): p. 3231-3233.

43. Fassan, M., R. Baffa, and A. Kiss, Advanced precancerous lesions within the GI tract: the molecular background. Best Practice & Research Clinical Gastroenterology, 2013. 27(2): p. 159-169.

44. Oliveira, C., R. Seruca, and F. Carneiro, Genetics, pathology, and clinics of familial gastric cancer. International journal of surgical pathology, 2006. 14(1): p. 21-33.


130

45. Parkin, D.M., The global health burden of infection‐associated cancers in the year 2002. International journal of cancer, 2006. 118(12): p. 3030-3044.

46. Siegel, R., et al., Cancer statistics, 2014. CA: a cancer journal for clinicians, 2014. 64(1): p. 9-29. 47. Guggenheim, D.E. and M.A. Shah, Gastric cancer epidemiology and risk factors. Journal of surgical

oncology, 2013. 107(3): p. 230-236. 48. Bray, F., et al., Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality

worldwide for 36 cancers in 185 countries. CA: a cancer journal for clinicians, 2018. 68(6): p. 394-424.

49. Canadian Cancer Statistics Advisory Committee, Canadian Cancer Statistics 2017, Canadian Cancer Society, Editor. 2017: Toronto, ON.

50. Canadian Cancer Statistics Advisory Committee, Canadian Cancer Statistics 2018, Canadian Cancer Society, Editor. 2018: Toronto, ON.

51. Maconi, G., G. Manes, and G.B. Porro, Role of symptoms in diagnosis and outcome of gastric cancer. World journal of gastroenterology: WJG, 2008. 14(8): p. 1149.

52. Mahar, A.L., et al., Evaluating TNM stage prognostic ability in a population-based cohort of gastric adenocarcinoma patients in a low-incidence country. Canadian Journal of Public Health, 2018. 109(4): p. 480-488.

53. National Cancer Institute. Cancer Stat Facts: Stomach Cancer. 2017; Available from: https://seer.cancer.gov/statfacts/html/stomach.html.

54. Bray, F., et al., Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin, 2018. 68(6): p. 394-424.

55. Leung, W.K., et al., Screening for gastric cancer in Asia: current evidence and practice. The lancet oncology, 2008. 9(3): p. 279-287.

56. Hamashima, C., et al., The Japanese guidelines for gastric cancer screening. Jpn J Clin Oncol, 2008. 38(4): p. 259-67.

57. Hamashima, C., G. Systematic Review, and G. Guideline Development Group for Gastric Cancer Screening, Update version of the Japanese Guidelines for Gastric Cancer Screening. Jpn J Clin Oncol, 2018. 48(7): p. 673-683.

58. Suh, M., et al., Trends in Participation Rates for the National Cancer Screening Program in Korea, 2002-2012. Cancer Res Treat, 2017. 49(3): p. 798-806.

59. Kim, Y., et al., Overview of the National Cancer screening programme and the cancer screening status in Korea. Asian Pac J Cancer Prev, 2011. 12(3): p. 725-30.

60. Jun, J.K., et al., Effectiveness of the Korean National Cancer Screening Program in Reducing Gastric Cancer Mortality. Gastroenterology, 2017. 152(6): p. 1319-1328 e7.

61. Gupta, N., et al., Endoscopy for upper GI cancer screening in the general population: a cost-utility analysis. Gastrointestinal endoscopy, 2011. 74(3): p. 610-624. e2.

62. Yeh, J.M., et al., Gastric adenocarcinoma screening and prevention in the era of new biomarker and endoscopic technologies: a cost-effectiveness analysis. Gut, 2016. 65(4): p. 563-574.

63. Dan, Y.Y., J. So, and K.G. Yeoh, Endoscopic screening for gastric cancer. Clinical gastroenterology and hepatology, 2006. 4(6): p. 709-716.

64. Sutradhar, R., et al., Higher risk of gastric cancer among immigrants to Ontario: a population-based matched cohort study with over 2 million individuals. Gastric Cancer, 2018. 21(4): p. 588-597.

65. Saumoy, M., et al., Cost Effectiveness of Gastric Cancer Screening According to Race and Ethnicity. Gastroenterology, 2018.

66. Smyth, E.C., et al., Gastric cancer: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up. Ann Oncol, 2016. 27(suppl 5): p. v38-v49.

https://seer.cancer.gov/statfacts/html/stomach.html


131

67. Ajani, J.A., et al., Gastric Cancer, Version 3.2016, NCCN Clinical Practice Guidelines in Oncology. J Natl Compr Canc Netw, 2016. 14(10): p. 1286-1312.

68. Graham, D.Y., et al., Prospective evaluation of biopsy number in the diagnosis of esophageal and gastric carcinoma. Gastroenterology, 1982. 82(2): p. 228-31.

69. Hatfield, A.R., et al., Importance of the site of endoscopic gastric biopsy in ulcerating lesions of the stomach. Gut, 1975. 16(11): p. 884-6.

70. Akiyama, M., et al., Endoscopic mucosal resection of gastric neoplasms using a ligating device. Gastrointestinal endoscopy, 1997. 45(2): p. 182-186.

71. Ono, H., et al., Endoscopic mucosal resection for treatment of early gastric cancer. Gut, 2001. 48(2): p. 225-229.

72. Gotoda, T., H. Yamamoto, and R.M. Soetikno, Endoscopic submucosal dissection of early gastric cancer. Journal of gastroenterology, 2006. 41(10): p. 929-942.

73. Isomoto, H., et al., Endoscopic submucosal dissection for early gastric cancer: a large-scale feasibility study. Gut, 2009. 58(3): p. 331-336.

74. Coburn, N., et al., Staging and Surgical Approaches in Gastric Cancer. A Quality Initiative of the Program in Evidence-Based Care (PEBC), Cancer Care Ontario (CCO), 2017.

75. Gospodarowicz, M.K., J.D. Brierley, and C. Wittekind, TNM classification of malignant tumours. 2017: John Wiley & Sons.

76. American Joint Committee on Cancer, AJCC Cancer Staging Manual. 8th ed. ed. 2017, New York: Springer.

77. Brar, S.S., et al., Processes of care in the multidisciplinary treatment of gastric cancer: results of a RAND/UCLA expert panel. JAMA Surg, 2014. 149(1): p. 18-25.

78. Strong, V.E., Gastric cancer: Principles and practice. 2015: Springer. 79. Fujitani, K., et al., Gastrectomy plus chemotherapy versus chemotherapy alone for advanced

gastric cancer with a single non-curable factor (REGATTA): a phase 3, randomised controlled trial. The Lancet Oncology, 2016. 17(3): p. 309-318.

80. Kim, Y.N., et al., Gastric cancer staging at isotropic MDCT including coronal and sagittal MPR images: endoscopically diagnosed early vs. advanced gastric cancer. Abdominal imaging, 2009. 34(1): p. 26-34.

81. Chen, C.Y., et al., Gastric cancer: preoperative local staging with 3D multi-detector row CT--correlation with surgical and histopathologic results. Radiology, 2007. 242(2): p. 472-82.

82. Ba-Ssalamah, A., et al., Dedicated multidetector CT of the stomach: spectrum of diseases. Radiographics, 2003. 23(3): p. 625-44.

83. Kwee, R.M. and T.C. Kwee, Imaging in assessing lymph node status in gastric cancer. Gastric Cancer, 2009. 12(1): p. 6-22.

84. Bhutani, M.S., R.H. Hawes, and B.J. Hoffman, A comparison of the accuracy of echo features during endoscopic ultrasound (EUS) and EUS-guided fine-needle aspiration for diagnosis of malignant lymph node invasion. Gastrointest Endosc, 1997. 45(6): p. 474-9.

85. Catalano, M.F., et al., Endosonographic features predictive of lymph node metastasis. Gastrointest Endosc, 1994. 40(4): p. 442-6.

86. Sano, T., O. Kobori, and T. Muto, Lymph node metastasis from early gastric cancer: endoscopic resection of tumour. Br J Surg, 1992. 79(3): p. 241-4.

87. Gotoda, T., et al., Incidence of lymph node metastasis from early gastric cancer: estimation with a large number of cases at two large centers. Gastric Cancer, 2000. 3(4): p. 219-225.

88. Japanese Gastric Cancer Association, Japanese classification of gastric carcinoma: 3rd English edition. Gastric Cancer, 2011. 14(2): p. 101-12.

89. Japanese Gastric Cancer Association, Japanese Classification of Gastric Carcinoma - 2nd English Edition. Gastric Cancer, 1998. 1(1): p. 10-24.


132

90. The Paris endoscopic classification of superficial neoplastic lesions: esophagus, stomach, and colon: November 30 to December 1, 2002. Gastrointest Endosc, 2003. 58(6 Suppl): p. S3-43.

91. Pimentel-Nunes, P., et al., Endoscopic submucosal dissection: European Society of Gastrointestinal Endoscopy (ESGE) Guideline. Endoscopy, 2015. 47(9): p. 829-54.

92. Gotoda, T., Endoscopic resection of early gastric cancer: the Japanese perspective. Curr Opin Gastroenterol, 2006. 22(5): p. 561-9.

93. Coburn, N.G., et al., Management of gastric cancer in Ontario. Journal of surgical oncology, 2010. 102(1): p. 54-63.

94. Shen, J.G., et al., Intraoperative frozen section margin evaluation in gastric cancer of the cardia surgery. Hepatogastroenterology, 2006. 53(72): p. 976-8.

95. Bozzetti, F., et al., Subtotal versus total gastrectomy for gastric cancer: five-year survival rates in a multicenter randomized Italian trial. Italian Gastrointestinal Tumor Study Group. Ann Surg, 1999. 230(2): p. 170-8.

96. Gouzi, J.L., et al., Total versus subtotal gastrectomy for adenocarcinoma of the gastric antrum. A French prospective controlled study. Ann Surg, 1989. 209(2): p. 162-6.

97. Nishi, M., Y. Omori, and K. Miwa, Japanese Classification of Gastric Carcinoma. Japanese Research Society for Gastric Cancer (JRSGC, 1995. 1st English Edition(Tokyo: Kanehara & Co.).

98. Cuschieri, A., et al., Patient survival after D 1 and D 2 resections for gastric cancer: long-term results of the MRC randomized surgical trial. British journal of cancer, 1999. 79(9): p. 1522.

99. Bonenkamp, J., et al., Extended lymph-node dissection for gastric cancer. New England Journal of Medicine, 1999. 340(12): p. 908-914.

100. Parikh, D., et al., D2 gastrectomy: lessons from a prospective audit of the learning curve. British journal of surgery, 1996. 83(11): p. 1595-1599.

101. Kim, C.Y., et al., Learning curve for gastric cancer surgery based on actual survival. Gastric Cancer, 2016. 19(2): p. 631-638.

102. Coburn, N.G., Lymph nodes and gastric cancer. Journal of surgical oncology, 2009. 99(4): p. 199-206.

103. Cuschieri, A., et al., Postoperative morbidity and mortality after D1 and D2 resections for gastric cancer: preliminary results of the MRC randomised controlled surgical trial. The Lancet, 1996. 347(9007): p. 995-999.

104. Bonenkamp, J., et al., Randomised comparison of morbidity after D1 and D2 dissection for gastric cancer in 996 Dutch patients. The Lancet, 1995. 345(8952): p. 745-748.

105. Songun, I., et al., Surgical treatment of gastric cancer: 15-year follow-up results of the randomised nationwide Dutch D1D2 trial. The lancet oncology, 2010. 11(5): p. 439-449.

106. Degiuli, M., et al., Morbidity and mortality after D2 gastrectomy for gastric cancer: results of the Italian Gastric Cancer Study Group prospective multicenter surgical study. J Clin Oncol, 1998. 16(4): p. 1490-3.

107. Degiuli, M., et al., Morbidity and mortality after D1 and D2 gastrectomy for cancer: interim analysis of the Italian Gastric Cancer Study Group (IGCSG) randomised surgical trial. Eur J Surg Oncol, 2004. 30(3): p. 303-8.

108. Degiuli, M., et al., Randomized clinical trial comparing survival after D1 or D2 gastrectomy for gastric cancer. Br J Surg, 2014. 101(2): p. 23-31.

109. Csendes, A., et al., A prospective randomized study comparing D2 total gastrectomy versus D2 total gastrectomy plus splenectomy in 187 patients with gastric carcinoma. Surgery, 2002. 131(4): p. 401-407.

110. Smith, D.D., R.R. Schwarz, and R.E. Schwarz, Impact of total lymph node count on staging and survival after gastrectomy for gastric cancer: data from a large US-population database. J Clin Oncol, 2005. 23(28): p. 7114-24.


133

111. Samples, J.E., K.B. Stitzenberg, and M.O. Meyers, Lymph node yield and survival in gastric carcinoma. 2014, American Society of Clinical Oncology.

112. Siewert, J.R., et al., Relevant prognostic factors in gastric cancer: ten-year results of the German Gastric Cancer Study. Ann Surg, 1998. 228(4): p. 449-61.

113. Macalindong, S.S., et al., Effect of total number of harvested lymph nodes on survival outcomes after curative resection for gastric adenocarcinoma: findings from an eastern high-volume gastric cancer center. BMC Cancer, 2018. 18(1): p. 73.

114. Klein Kranenbarg, E., et al., Evaluation of the 5th edition of the TNM classification for gastric cancer: improved prognostic value. Br J Cancer, 2001. 84(1): p. 64-71.

115. Wagner, P., et al., Lymph node counts in the upper abdomen: anatomical basis for lymphadenectomy in gastric cancer. British journal of surgery, 1991. 78(7): p. 825-827.

116. de Manzoni, G., et al., The new TNM classification of lymph node metastasis minimises stage migration problems in gastric cancer patients. Br J Cancer, 2002. 87(2): p. 171-4.

117. Meyer, H.J., et al., [Current S3 guidelines on surgical treatment of gastric carcinoma]. Chirurg, 2012. 83(1): p. 31-7.

118. Allum, W.H., et al., Guidelines for the management of oesophageal and gastric cancer. Gut, 2011. 60(11): p. 1449-72.

119. De Manzoni, G., et al., The Italian Research Group for Gastric Cancer (GIRCG) guidelines for gastric cancer staging and treatment: 2015. Gastric Cancer, 2017. 20(1): p. 20-30.

120. Waddell, T., et al., Gastric cancer: ESMO-ESSO-ESTRO Clinical Practice Guidelines for diagnosis, treatment and follow-up. Ann Oncol, 2013. 24 Suppl 6: p. vi57-63.

121. Mahar, A., et al., Geographic variation in surgical practice patterns and outcomes for resected, non-metastatic gastric cancer across Ontario. Current Oncology, 2018.

122. Bartlett, E.K., et al., Morbidity and mortality after total gastrectomy for gastric malignancy using the American College of Surgeons National Surgical Quality Improvement Program database. Surgery, 2014. 156(2): p. 298-304.

123. Busweiler, L.A., et al., Textbook outcome as a composite measure in oesophagogastric cancer surgery. Br J Surg, 2017. 104(6): p. 742-750.

124. Park, C., K. Song, and S. Kim, Treatment results for gastric cancer surgery: 12 years’ experience at a single institute in Korea. European Journal of Surgical Oncology (EJSO), 2008. 34(1): p. 36-41.

125. Kim, H.-H., et al., Long-Term Results of Laparoscopic Gastrectomy for Gastric Cancer: A Large-Scale Case-Control and Case-Matched Korean Multicenter Study. Journal of Clinical Oncology, 2014. 32(7): p. 627-633.

126. Strong, V.E., et al., Comparison of gastric cancer survival following R0 resection in the United States and Korea using an internationally validated nomogram. Ann Surg, 2010. 251(4): p. 640-6.

127. Watanabe, M., et al., Total gastrectomy risk model: data from 20,011 Japanese patients in a nationwide internet-based database. Annals of surgery, 2014. 260(6): p. 1034-1039.

128. Nashimoto, A., et al., Gastric cancer treated in 2002 in Japan: 2009 annual report of the JGCA nationwide registry. Gastric Cancer, 2013. 16(1): p. 1-27.

129. Baxter, N.N. and T.M. Tuttle, Inadequacy of lymph node staging in gastric cancer patients: a population-based study. Annals of surgical oncology, 2005. 12(12): p. 981.

130. Coburn, N.G., et al., Significant regional variation in adequacy of lymph node assessment and survival in gastric cancer. Cancer, 2006. 107(9): p. 2143-2151.

131. de Gara, C.J., J. Hanson, and S. Hamilton, A population-based study of tumor-node relationship, resection margins, and surgeon volume on gastric cancer survival. Am J Surg, 2003. 186(1): p. 23-7.

132. Kim, S.H., et al., Effect of microscopic resection line disease on gastric cancer survival. J Gastrointest Surg, 1999. 3(1): p. 24-33.


134

133. Wang, S.Y., et al., Clinical impact of positive surgical margin status on gastric cancer patients undergoing gastrectomy. Ann Surg Oncol, 2009. 16(10): p. 2738-43.

134. Ecker, B.L., et al., Minimally invasive gastrectomy for gastric adenocarcinoma in the United States: Utilization and short-term oncologic outcomes. J Surg Oncol, 2015. 112(6): p. 616-21.

135. Rhome, R.M., et al., Predictors of Positive Margins After Definitive Resection for Gastric Adenocarcinoma and Impact of Adjuvant Therapies. Int J Radiat Oncol Biol Phys, 2017. 98(5): p. 1106-1115.

136. Center for Cancer Control and Information Services, Monitoring of Cancer Incidence in Japan - Survival 2006-2008 Report National Cancer Center, 2016.

137. Jung, K.-W., et al., Cancer Statistics in Korea: Incidence, Mortality, Survival, and Prevalence in 2014. Cancer research and treatment : official journal of Korean Cancer Association, 2017. 49(2): p. 292-305.

138. Bickenbach, K. and V.E. Strong, Comparisons of gastric cancer treatments: east vs. west. Journal of gastric cancer, 2012. 12(2): p. 55-62.

139. Markar, S.R., et al., Long-Term Survival After Gastrectomy for Cancer in Randomized, Controlled Oncological Trials: Comparison between West and East. Annals of Surgical Oncology, 2013. 20(7): p. 2328-2338.

140. Siewert, J.R., Gastric cancer: the dispute between East and West. Gastric Cancer, 2005. 8(2): p. 59-61.

141. Strong, V.E., et al., Differences in gastric cancer survival between the US and China. Journal of surgical oncology, 2015. 112(1): p. 31-37.

142. Noguchi, Y., et al., Is gastric carcinoma different between Japan and the United States? Cancer, 2000. 89(11): p. 2237-46.

143. Eom, B.W., et al., Trends in gastric cancer incidence according to the clinicopathological characteristics in Korea, 1999-2014. Cancer research and treatment: official journal of Korean Cancer Association, 2018. 50(4): p. 1343.

144. Busweiler, L., et al., Textbook outcome as a composite measure in oesophagogastric cancer surgery. British Journal of Surgery, 2017. 104(6): p. 742-750.

145. Hannan, E.L., et al., The influence of hospital and surgeon volume on in-hospital mortality for colectomy, gastrectomy, and lung lobectomy in patients with cancer. Surgery, 2002. 131(1): p. 6-15.

146. Callahan, M.A., et al., Influence of surgical subspecialty training on in-hospital mortality for gastrectomy and colectomy patients. Annals of surgery, 2003. 238(4): p. 629.

147. Finlayson, E.V., P.P. Goodney, and J.D. Birkmeyer, Hospital volume and operative mortality in cancer surgery: a national study. Archives of Surgery, 2003. 138(7): p. 721-725.

148. Smith, D.L., et al., Factors influencing the volume-outcome relationship in gastrectomies: a population-based study. Ann Surg Oncol, 2007. 14(6): p. 1846-52.

149. Birkmeyer, J.D., et al., Hospital volume and late survival after cancer surgery. Annals of surgery, 2007. 245(5): p. 777.

150. Birkmeyer, J.D., et al., Hospital volume and surgical mortality in the United States. N Engl J Med, 2002. 346(15): p. 1128-37.

151. Xirasagar, S., et al., Procedure volume of gastric cancer resections versus 5-year survival. European Journal of Surgical Oncology (EJSO), 2008. 34(1): p. 23-29.

152. Fujita, T. and Y. Yamazaki, Influence of surgeon's volume on early outcome after total gastrectomy. European Journal of Surgery, 2002. 168(10): p. 535-538.

153. Nomura, E., et al., Population‐based study of relationship between hospital surgical volume and 5‐year survival of stomach cancer patients in Osaka, Japan. Cancer science, 2003. 94(11): p. 998-1002.


135

154. Yun, Y., et al., The influence of hospital volume and surgical treatment delay on long-term survival after cancer surgery. Annals of oncology, 2012. 23(10): p. 2731-2737.

155. Ichikawa, D., et al., Effect of hospital volume on long-term outcomes of laparoscopic gastrectomy for clinical stage I gastric cancer. Anticancer research, 2013. 33(11): p. 5165-5170.

156. Gordon, T.A., et al., Complex gastrointestinal surgery: impact of provider experience on clinical and economic outcomes1. Journal of the American College of Surgeons, 1999. 189(1): p. 46-56.

157. Bilimoria, K.Y., et al., Effect of hospital type and volume on lymph node evaluation for gastric and pancreatic cancer. Archives of surgery, 2008. 143(7): p. 671-678.

158. Birkmeyer, J.D., et al., Surgeon volume and operative mortality in the United States. New England Journal of Medicine, 2003. 349(22): p. 2117-2127.

159. Robles, S.C., et al., An application of capture-recapture methods to the estimation of completeness of cancer registration. J Clin Epidemiol, 1988. 41(5): p. 495-501.

160. Clarke EA, Marrett LD, and Kreiger N, Cancer registration in ontario: a computer approach, in Cancer registration principles and methods. Pub No. 95 ed, Jensen OM, et al., Editors. 1991, IARC: Lyon, France. p. 246–57.

161. Williams, J. and W. Young, A summary of studies on the quality of health care administrative databases in Canada. Patterns of health care in Ontario: the ICES practice atlas. 2nd ed. Ottawa: Canadian Medical Association, 1996. 339: p. 45.

162. Simunovic, M., et al., Using administrative databases to measure waiting times for patients undergoing major cancer surgery in Ontario, 1993-2000. Can J Surg, 2005. 48(2): p. 137-42.

163. Rice, T.W., D.T. Patil, and E.H. Blackstone, 8th edition AJCC/UICC staging of cancers of the esophagus and esophagogastric junction: application to clinical practice. Ann Cardiothorac Surg, 2017. 6(2): p. 119-130.

164. Shi, C., et al., Protocol for the Examination of Specimens From Patients With Carcinoma of the Esophagus, Version 4.0.0.0. 2017, College of American Pathologists: Northfield, IL.

165. Shi, C., et al., Protocol for the Examination of Specimens From Patients With Carcinoma of the Stomach, Version 4.0.0.0. 2017, College of American Pathologists: Northfield, IL.

166. Austin, P.C., et al., Using the Johns Hopkins Aggregated Diagnosis Groups (ADGs) to predict mortality in a general adult population cohort in Ontario, Canada. Medical care, 2011. 49(10): p. 932-939.

167. Matheson, F.I., et al., Development of the Canadian Marginalization Index: a new tool for the study of inequality. Canadian Journal of Public Health/Revue Canadienne De Sante'e Publique, 2012: p. S12-S16.

168. Cancer Care Ontario. Cancer Care Ontario Data Book 2018-2019. April 27, 2019]. 169. Mahar, A.L., et al., Predictors of hospital stay and home care services use: A population-based,

retrospective cohort study in stage IV gastric cancer. Palliative medicine, 2015. 29(2): p. 147-156. 170. Badgwell, B., et al., Attempted Salvage Resection for Recurrent Gastric or Gastroesophageal

Cancer. Annals of Surgical Oncology, 2009. 16(1): p. 42-50. 171. Carboni, F., et al., Treatment for isolated loco-regional recurrence of gastric adenocarcinoma:

does surgery play a role? World journal of gastroenterology, 2005. 11(44): p. 7014. 172. Bohner, H., et al., Detection and prognosis of recurrent gastric cancer--is routine follow-up after

gastrectomy worthwhile? Hepatogastroenterology, 2000. 47(35): p. 1489-94. 173. Dixon, M., et al., Prognostic factors in metastatic gastric cancer: results of a population-based,

retrospective cohort study in Ontario. Gastric Cancer, 2016. 19(1): p. 150-159. 174. Jeong, Y., et al., Outcomes of Non-curative Gastrectomy for Gastric Cancer: An Analysis of the

American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP). Ann Surg Oncol, 2018. 25(13): p. 3943-3949.


136

175. Mahar, A.L., et al., Validating an algorithm to identify metastatic gastric cancer in the absence of routinely collected TNM staging data. BMC Health Serv Res, 2018. 18(1): p. 309.

176. Eskander, A., et al., Volume-outcome relationships for head and neck cancer surgery in a universal health care system. Laryngoscope, 2014. 124(9): p. 2081-8.

177. Finlayson, E.V., P.P. Goodney, and J.D. Birkmeyer, Hospital volume and operative mortality in cancer surgery: a national study. Arch Surg, 2003. 138(7): p. 721-5; discussion 726.

178. Dindo, D., N. Demartines, and P.A. Clavien, Classification of surgical complications: a new proposal with evaluation in a cohort of 6336 patients and results of a survey. Ann Surg, 2004. 240(2): p. 205-13.

179. ICES. ICES Data Dictionary. April 29, 2019]; Available from: https://www.cancercareontario.ca/en/data-book-reporting-standards.

180. Canadian Institute for Health Information All-Cause Readmission to Acute Care and Return to the Emergency Department.

181. Program;, A.C.o.S.N.S.Q.I. User Guide for the 2017 ACS NSQIP Participant Use Data File (PUF). October 2018 April 29 2019]; Available from: https://www.facs.org/~/media/files/quality%20programs/nsqip/nsqip_puf_userguide_2017.ashx.

182. Yu, J., et al., The impact of age and comorbidity on postoperative complications in patients with advanced gastric cancer after laparoscopic D2 gastrectomy: results from the Chinese laparoscropic gastrointestinal surgery study (CLASS) group. European Journal of Surgical Oncology (EJSO), 2013. 39(10): p. 1144-1149.

183. Pacelli, F., et al., Risk factors in relation to postoperative complications and mortality after total gastrectomy in aged patients. The American surgeon, 1991. 57(6): p. 341-345.

184. Bittner, R., et al., Total gastrectomy. Updated operative mortality and long-term survival with particular reference to patients older than 70 years of age. Annals of surgery, 1996. 224(1): p. 37.

185. Lepage, C., et al., Operative mortality after gastric cancer resection and long-term survival differences across Europe. Br J Surg, 2010. 97(2): p. 235-9.

186. Hundahl, S.A., J.L. Phillips, and H.R. Menck, The National Cancer Data Base Report on poor survival of US gastric carcinoma patients treated with gastrectomy: American Joint Committee on Cancer staging, proximal disease, and the “different disease” hypothesis. Cancer, 2000. 88(4): p. 921-932.

187. Statistics Canada, 2011 Census Dictionary, in Statistics Canada Catalogue. 2012/2018: Ottawa, Ontario.

188. Piso, P., et al., Proximal versus distal gastric carcinoma—what are the differences? Annals of surgical oncology, 2000. 7(7): p. 520-525.

189. Pacelli, F., et al., Proximal compared with distal gastric cancer: multivariate analysis of prognostic factors. The American surgeon, 2001. 67(7): p. 697.

190. Piessen, G., et al., Signet ring cell histology is an independent predictor of poor prognosis in gastric adenocarcinoma regardless of tumoral clinical presentation. Annals of surgery, 2009. 250(6): p. 878-887.

191. Dicken, B.J., et al., Lymphovascular invasion is associated with poor survival in gastric cancer: an application of gene-expression and tissue array techniques. Annals of surgery, 2006. 243(1): p. 64.

192. Kim, H., et al., Lymphovascular invasion is an important predictor of lymph node metastasis in endoscopically resected early gastric cancers. Oncology reports, 2011. 25(6): p. 1589-1595.

193. Tanaka, A., et al., Perineural invasion as a predictor of recurrence of gastric cancer. Cancer, 1994. 73(3): p. 550-555.

194. Bilici, A., et al., Prognostic significance of perineural invasion in patients with gastric cancer who underwent curative resection. Annals of surgical oncology, 2010. 17(8): p. 2037-2044.

https://www.cancercareontario.ca/en/data-book-reporting-standards

https://www.facs.org/~/media/files/quality%20programs/nsqip/nsqip_puf_userguide_2017.ashx


137

195. Wang, S.-Y., et al., Clinical impact of positive surgical margin status on gastric cancer patients undergoing gastrectomy. Annals of surgical oncology, 2009. 16(10): p. 2738-2743.

196. Songun, I., et al., Prognostic value of resection-line involvement in patients undergoing curative resections for gastric cancer. European Journal of Cancer, 1996. 32(3): p. 433-437.

197. Paoletti, X., et al., Benefit of adjuvant chemotherapy for resectable gastric cancer a meta-analysis. 2010.

198. Bang, Y.J., et al., Adjuvant capecitabine and oxaliplatin for gastric cancer after D2 gastrectomy (CLASSIC): a phase 3 open-label, randomised controlled trial. Lancet, 2012. 379(9813): p. 315-21.

199. Cunningham, D., et al., Perioperative chemotherapy versus surgery alone for resectable gastroesophageal cancer. N Engl J Med, 2006. 355(1): p. 11-20.

200. Sasako, M., et al., Five-year outcomes of a randomized phase III trial comparing adjuvant chemotherapy with S-1 versus surgery alone in stage II or III gastric cancer. J Clin Oncol, 2011. 29(33): p. 4387-93.

201. Smalley, S.R., et al., Updated analysis of SWOG-directed intergroup study 0116: a phase III trial of adjuvant radiochemotherapy versus observation after curative gastric cancer resection. Journal of Clinical Oncology, 2012. 30(19): p. 2327.

202. Macdonald, J.S., et al., Chemoradiotherapy after surgery compared with surgery alone for adenocarcinoma of the stomach or gastroesophageal junction. New England Journal of Medicine, 2001. 345(10): p. 725-730.

203. Schuhmacher, C., et al., Neoadjuvant chemotherapy compared with surgery alone for locally advanced cancer of the stomach and cardia: European Organisation for Research and Treatment of Cancer randomized trial 40954. Journal of clinical oncology, 2010. 28(35): p. 5210.

204. Canadian Institute for Health Information, Data Quality Study of Ontario Emergency Department Visits for 2004–2005—Executive Summary. 2007: p. 1-22.

205. Canadian Institute for Health Information, Data quality study of the 2008-2009 discharge abstract database. 2010.

206. Canadian Institute for Health Information, CIHI Data Quality Study of the 2009–2010 Discharge Abstract Database. 2012.

207. Eskander, A., et al., Head and Neck Cancer Surgery in Ontario, 2003-2010: An ICES Atlas. 2016: Institute for Clinical Evaluative Sciences.

208. Urbach, D., M. Simunovic, and S. Schultz, Cancer Surgery in Ontario: ICES Atlas. 2008. Institute for Clinical Evaluative Sciences, Toronto.

209. Juurlink, D., et al., Canadian Institute for HealthInformation Discharge Abstract Database: A Validation Study. Institute for Clinical Evaluative Sciences, 2006.

210. Scales, D.C., et al., Administrative data accurately identified intensive care unit admissions in Ontario. Journal of clinical epidemiology, 2006. 59(8): p. 802-807.

211. Mahar, A.L., et al., A descriptive analysis of gastric cancer specimen processing techniques. Journal of surgical oncology, 2011. 103(3): p. 248-256.

212. C, S., et al. Protocol for the Examination of Specimens From Patients With Carcinoma of the Stomach. 2017. 4.0.0.0.

213. Donders, A.R.T., et al., A gentle introduction to imputation of missing values. Journal of clinical epidemiology, 2006. 59(10): p. 1087-1091.

214. Haukoos, J.S. and C.D. Newgard, Advanced statistics: missing data in clinical research—part 1: an introduction and conceptual framework. Academic Emergency Medicine, 2007. 14(7): p. 662-668.

215. Newgard, C.D. and J.S. Haukoos, Advanced statistics: missing data in clinical research—part 2: multiple imputation. Academic Emergency Medicine, 2007. 14(7): p. 669-678.


138

216. Juurlink, D., et al., Canadian institute for health information discharge abstract database: a validation study. ICES investigative report. Institute for Clinical Evaluative Sciences, Toronto, 2006.

217. Choban, P.S., et al., Increased incidence of nosocomial infections in obese surgical patients. The American surgeon, 1995. 61(11): p. 1001-1005.

218. Mullen, J.T., et al., Impact of body mass index on perioperative outcomes in patients undergoing major intra-abdominal cancer surgery. Annals of surgical oncology, 2008. 15(8): p. 2164.

219. Benchimol, E.I., et al., The REporting of studies Conducted using Observational Routinely-collected health Data (RECORD) statement. PLoS Med, 2015. 12(10): p. e1001885.

220. Kleinbaum, D.G., Epidemiologic methods: the “art” in the state of the art. Journal of clinical epidemiology, 2002. 55(12): p. 1196-1200.

221. Haneuse, S., T.J. VanderWeele, and D. Arterburn, Using the E-value to assess the potential effect of unmeasured confounding in observational studies. Jama, 2019. 321(6): p. 602-603.

222. VanderWeele, T.J. and P. Ding, Sensitivity analysis in observational research: introducing the E-value. Annals of internal medicine, 2017. 167(4): p. 268-274.

223. Mathur, M.B., et al., Web site and R package for computing E-values. Epidemiology, 2018. 29(5): p. e45-e47.

224. Han, D.-S., et al., Nomogram predicting long-term survival after d2 gastrectomy for gastric cancer. Journal of clinical oncology, 2012. 30(31): p. 3834-3840.

225. Harrison, L.E., M.S. Karpeh, and M.F. Brennan, Proximal gastric cancers resected via a transabdominal-only approach. Results and comparisons to distal adenocarcinoma of the stomach. Ann Surg, 1997. 225(6): p. 678-83; discussion 683-5.

226. Piso, P., et al., Proximal versus distal gastric carcinoma--what are the differences? Ann Surg Oncol, 2000. 7(7): p. 520-5.

227. Coburn, N., et al., Optimal management of gastric cancer: results from an international RAND/UCLA expert panel. Ann Surg, 2014. 259(1): p. 102-8.

228. de Gara, C.J., J. Hanson, and S. Hamilton, A population-based study of tumor-node relationship, resection margins, and surgeon volume on gastric cancer survival. The American journal of surgery, 2003. 186(1): p. 23-27.

229. Degiuli, M., et al., Morbidity and mortality in the Italian Gastric Cancer Study Group randomized clinical trial of D1 versus D2 resection for gastric cancer. Br J Surg, 2010. 97(5): p. 643-9.

230. Gholami, S., et al., Number of lymph nodes removed and survival after gastric cancer resection: an analysis from the US Gastric Cancer Collaborative. Journal of the American College of Surgeons, 2015. 221(2): p. 291-299.

231. Karna Sura, H.Y., et al., How many lymph nodes are enough?—defining the extent of lymph node dissection in stage I–III gastric cancer using the National Cancer Database. Journal of gastrointestinal oncology, 2018. 9(6): p. 1168.

232. Hosmer Jr, D.W., S. Lemeshow, and R.X. Sturdivant, Applied logistic regression. Vol. 398. 2013: John Wiley & Sons.

233. Ghaferi, A.A., J.D. Birkmeyer, and J.B. Dimick, Hospital volume and failure to rescue with high-risk surgery. Medical care, 2011: p. 1076-1081.

234. Marcaccio, M., Hepatic, pancreatic, and biliary tract (HPB) surgical oncology standards. 2006. 235. Sundaresan, S., Thoracic surgical oncology standards. 2005. 236. Urbach, D.R. and N.N. Baxter, Does it matter what a hospital is “high volume” for? Specificity of

hospital volume-outcome associations for surgical procedures: analysis of administrative data. Bmj, 2004. 328(7442): p. 737-740.

237. Ghaferi, A.A., J.D. Birkmeyer, and J.B. Dimick, Complications, Failure to Rescue, and Mortality With Major Inpatient Surgery in Medicare Patients. Annals of Surgery, 2009. 250(6): p. 1029-1034.


139

238. Kim, C.G., E.K. Kwak, and S.i. Lee, The relationship between hospital volume and outcome of gastrointestinal cancer surgery in Korea. Journal of surgical oncology, 2011. 104(2): p. 116-123.

239. Langer, B., Role of volume outcome data in assuring quality in HPB surgery. HPB, 2007. 9(5): p. 330-334.

240. Smith, A.J., et al., Multimodal CME for surgeons and pathologists improves colon cancer staging. Journal of Cancer Education, 2003. 18(2): p. 81-85.

241. Scott, K. and R. Grace, Detection of lymph node metastases in colorectal carcinoma before and after fat clearance. British Journal of Surgery, 1989. 76(11): p. 1165-1167.

242. Hardwick, R.H., D2 gastrectomy course, The Royal College of Surgeons of England, February 13–15, 2002. Gastric Cancer, 2002. 5(4): p. 0244-0245.

243. Coburn, N., et al., Staging and surgical approaches in gastric cancer: a clinical practice guideline. Current Oncology, 2017. 24(5): p. 324.

244. Ychou, M., et al., Perioperative chemotherapy compared with surgery alone for resectable gastroesophageal adenocarcinoma: an FNCLCC and FFCD multicenter phase III trial. J Clin Oncol, 2011. 29(13): p. 1715-1721.

245. Booth, C.M., et al., Perioperative chemotherapy for muscle‐invasive bladder cancer: A population‐based outcomes study. Cancer, 2014. 120(11): p. 1630-1638.

246. Krishnamurthy, A., et al., Chemotherapy delivery for resected colorectal cancer liver metastases: Management and outcomes in routine clinical practice. European Journal of Surgical Oncology (EJSO), 2017. 43(2): p. 364-371.

247. Booth, C.M., et al., Use and effectiveness of adjuvant chemotherapy for stage III colon cancer: a population-based study. Journal of the National Comprehensive Cancer Network, 2016. 14(1): p. 47-56.

248. Latchana, N., et al., Population‐based study of the impact of surgical and adjuvant therapy at the same or a different institution on survival of patients with pancreatic adenocarcinoma. BJS open, 2019. 3(1): p. 85-94.

249. Goodman, S.N. and J.A. Berlin, The use of predicted confidence intervals when planning experiments and the misuse of power when interpreting results. Annals of internal medicine, 1994. 121(3): p. 200-206.

250. Hoenig, J.M. and D.M. Heisey, The abuse of power: the pervasive fallacy of power calculations for data analysis. The American Statistician, 2001. 55(1): p. 19-24.

251. Walters, S.J., Consultants' forum: should post hoc sample size calculations be done? Pharmaceutical Statistics: The Journal of Applied Statistics in the Pharmaceutical Industry, 2009. 8(2): p. 163-169.

252. Levine, M. and M.H. Ensom, Post hoc power analysis: an idea whose time has passed? Pharmacotherapy: The Journal of Human Pharmacology and Drug Therapy, 2001. 21(4): p. 405-409.

253. Coupland, V.H., et al., Hospital volume, proportion resected and mortality from oesophageal and gastric cancer: a population-based study in England, 2004–2008. Gut, 2013. 62(7): p. 961-966.

254. Jensen, L.S., A. Bendixen, and H. Kehlet, Organisation and early outcomes of major upper gastrointestinal cancer surgery in Denmark 1996–2004. Scandinavian journal of surgery, 2007. 96(1): p. 41-45.

255. Wenner, J., et al., The influence of surgical volume on hospital mortality and 5-year survival for carcinoma of the oesophagus and gastric cardia. Anticancer research, 2005. 25(1B): p. 419-424.

256. Auyong, D.B., et al., Reduced length of hospitalization in primary total knee arthroplasty patients using an updated enhanced recovery after orthopedic surgery (ERAS) pathway. The Journal of arthroplasty, 2015. 30(10): p. 1705-1709.


140

257. Low, D.E., et al., Esophagectomy—it’s not just about mortality anymore: standardized perioperative clinical pathways improve outcomes in patients with esophageal cancer. Journal of Gastrointestinal Surgery, 2007. 11(11): p. 1395-1402.

258. Page, A., et al., Patient outcomes and provider perceptions following implementation of a standardized perioperative care pathway for open liver resection. British Journal of Surgery, 2016. 103(5): p. 564-571.

259. Pędziwiatr, M., et al., Short hospital stays after laparoscopic gastric surgery under an Enhanced Recovery After Surgery (ERAS) pathway: experience at a single center. European Surgery, 2014. 46(3): p. 128-132.

260. Porter, G.A., et al., Cost and utilization impact of a clinical pathway for patients undergoing pancreaticoduodenectomy. Annals of Surgical Oncology, 2000. 7(7): p. 484-489.

261. Rogers, L.J., et al., The impact of enhanced recovery after surgery (ERAS) protocol compliance on morbidity from resection for primary lung cancer. The Journal of thoracic and cardiovascular surgery, 2018. 155(4): p. 1843-1852.

262. Varadhan, K.K., et al., The enhanced recovery after surgery (ERAS) pathway for patients undergoing major elective open colorectal surgery: a meta-analysis of randomized controlled trials. Clinical nutrition, 2010. 29(4): p. 434-440.

263. Wainwright, T.W., T. Immins, and R.G. Middleton, Enhanced recovery after surgery (ERAS) and its applicability for major spine surgery. Best practice & research Clinical anaesthesiology, 2016. 30(1): p. 91-102.

264. Zehr, K.J., et al., Standardized clinical care pathways for major thoracic cases reduce hospital costs. The Annals of thoracic surgery, 1998. 66(3): p. 914-919.

265. Karthaus, E.G., et al., Textbook outcome: a composite measure for quality of elective aneurysm surgery. Annals of surgery, 2017. 266(5): p. 898-904.

266. Merath, K., et al., A multi-institutional international analysis of textbook outcomes among patients undergoing curative-intent resection of intrahepatic cholangiocarcinoma. JAMA surgery, 2019: p. e190571-e190571.

267. Merath, K., et al., Textbook Outcomes Among Medicare Patients Undergoing Hepatopancreatic Surgery. Annals of surgery, 2018.

268. Numan, R.C., et al., How to evaluate quality for Thoracic Lung Cancer Surgery: A ‘’textbook outcome’’. Medical Research Archives, 2015(3).

269. Poelemeijer, Y.Q., et al., Textbook Outcome: an Ordered Composite Measure for Quality of Bariatric Surgery. Obesity surgery, 2019. 29(4): p. 1287-1294.

textbook outcomes in gastric cancer surgery · 2019. 11. 21. · levy – textbook outcomes in...

Documents