Predictors of Advanced Colorectal Neoplasia forColorectal Cancer Screening

Martin C.S. Wong, MD, MPH, Thomas Y.T. Lam, MSc, Kelvin K.F. Tsoi, PhD, Victor C.W. Chan, BSc,Hoyee W. Hirai, MSc, Jessica Y.L. Ching, MPH, Joseph J.Y. Sung, MD, PhD

From the InstChing, Sung)Faculty of MeHospital, Shat

Address coDigestive Dise7/F, Lui Che W32 Ngan Shingedu.hk

0749-3797/http://dx.do

& 2014 Publ

Background: The Asia-Pacific Colorectal Screening (APCS) score based on age, gender, familyhistory, and smoking is useful to predict advanced colorectal neoplasia (ACN) in asymptomaticAsian subjects.

Purpose: To evaluate the factors in addition to those of APCS associated with ACN colonoscopicfindings.

Methods: Data from 5,220 asymptomatic subjects aged between 50 and 70 years who underwentscreening colonoscopy in a community center between 2008 and 2012 were analyzed. One binarylogistic regression analysis was conducted in 2013 with the presence of ACN or cancer as theoutcome, controlling for APCS score, alcohol consumption, BMI, hypertension, and other chronicdiseases as independent variables.

Results: The average participant age was 57.7 years (SD¼4.9) and 47.5% were men. Advancedneoplasms or cancers were identified at colonoscopy in 5.6% of all screening participants. Frommultivariate regression analysis, APCS scoreZ4 (adjusted OR [AOR]¼1.74, 95% CI¼1.34, 2.25,po0.001); overweight (BMI¼23�24.9, AOR¼1.52, 95% CI¼1.12, 2.07, p¼0.007); obesity(BMIZ25, AOR¼1.56, 95% CI¼1.15, 2.10, p¼0.004); hypertension (AOR¼1.58, 95% CI¼1.21,2.06, p¼0.001); and alcohol consumption (AOR¼1.47, 95% CI¼1.05, 2.06, p¼0.025) wereassociated with ACN. The c-statistic of APCS score alone was 0.560 (95% CI¼0.524, 0.595,p¼0.001) and that of APCS score plus BMI, hypertension, and alcohol consumption was 0.613 (95%CI¼0.578, 0.648, po0.001).

Conclusions: Alcohol consumption, hypertension, and BMI are independent predictors of ACN,which could be incorporated into the APCS for prioritizing Asian asymptomatic subjects forcolorectal cancer screening.(Am J Prev Med 2014;46(5):433–439) & 2014 Published by Elsevier Inc. on behalf of American Journal ofPreventive Medicine

Introduction

Globally, colorectal cancer (CRC) is the thirdmost common malignancy in men, the secondmost common in women, and accounts for 10%

of all cancers.1 In recent decades, many Asian countries,

itute of Digestive Disease (Wong, Lam, Tsoi, Chan, Hirai,and School of Public Health and Primary Care (Wong),dicine, Chinese University of Hong Kong, Prince of Walesin NT, Hong Kongrrespondence to: Joseph J.Y. Sung, MD, PhD, Institute ofase, Faculty of Medicine, Chinese University of Hong Kong,oo Clinical Sciences Building, Prince of Wales Hospital, 30-Street, Shatin NT, Hong Kong SAR. E-mail: jjysung@cuhk.

$36.00i.org/10.1016/j.amepre.2013.12.008

ished by Elsevier Inc. on behalf of American Journal of P

namely, China, South Korea, Japan, Singapore, and HongKong, have experienced a rapid rise in CRC incidence.2

In future years, it will continue to impose a major publichealth burden in terms of mortality, morbidity, andmedical costs incurred.3

CRC screening using fecal tests4–6 and colonoscopy7

have been shown to reduce CRC mortality by up to 33%and 53%, respectively. Guidelines from authoritativesocieties in the U.S.8,9 and Europe,10 together with theAsia-Pacific Consensus statements,11 recommend CRCscreening for average-risk individuals aged 50 years orolder. However, the implementation of population-basedCRC screening programs could be hindered by resourcelimitations in many countries.12–14 These include con-cerns over colonoscopy capacity and the requirement ofestablishing screening infrastructure.14

reventive Medicine Am J Prev Med 2014;46(5):433–439 433

Wong et al / Am J Prev Med 2014;46(5):433–439434

In the light of these practical limitations, the Asia-Pacific Colorectal Screening (APCS) score was con-structed to prioritize asymptomatic Asian subjects forcolorectal screening.15 The scoring system was based onage, gender, family history, and smoking status to stratifythe risk of advanced colorectal neoplasia (ACN). TheAPCS is simple and can be used by general practitionersor nurse-educators. The tool is useful in clinical practiceto calculate risk, thus offering an option of prioritizinghigh-risk subjects for colonoscopy screening andaverage-risk subjects for fecal blood tests.15 The scoringsystem has been shown to be robust, with good matchbetween predicted and observed risk of ACN develop-ment. The APCS is currently used by physicians topredict ACN risk among asymptomatic Asian subjects,which helps to optimize the efficiency of screeningresources. Nevertheless, this study was limited by theabsence of BMI data. In addition, the sample size(n¼860) of the derivation cohort of this study is modest;thus, significant risk factors in the variable developmentphase of the scoring systemmight have been missed. Thisstudy recommends future research evaluating additionalrisk factors. A scoring system for predicting colorectalneoplasia was recently devised,16 but prediction of ACN,which bears greater malignant potential, is of greaterinterest.This study aimed to evaluate the variables associated

with ACN in addition to the APCS scores from a largercohort of asymptomatic subjects who received screeningcolonoscopy. The a priori hypothesis that these potentialrisk factors are independent predictors of ACN inaddition to the variables contained in the APCS systemwas tested.

MethodsSetting

A detailed description of the study setting has been describedelsewhere.17–19 Briefly, a bowel cancer screening center wasestablished in Hong Kong in 2008, providing free CRC screeningfor all eligible Hong Kong residents aged 50–70 years who wereasymptomatic for CRC. Data based on screening recruitmentbetween 2008 and 2012 were collected. This study was approved bythe Clinical Research Ethics Committee, the Chinese University ofHong Kong.

Study Design

The center prospectively recruited self-referred screening partic-ipants for CRC screening who were able to register via telephone,fax, e-mail, or walk-in. The eligibility criteria for this studyincluded (1) being aged between 50 and 70 years; (2) not havingexisting or previous symptoms suggestive of CRC such ashematochezia, melena, anorexia, or a change in bowel habits inthe past 4 weeks, or weight loss of greater than 5 kg in the past

6 months; and (3) not having undergone any CRC screening testsin the past 5 years. Exclusion criteria were a personal history ofCRC, colonic adenoma, diverticular disease, inflammatory boweldisease, a prosthetic heart valve, or vascular graft surgery.Participants with medical conditions that were contraindicationsto colonoscopy were also excluded.17–19

Registered participants were invited to complete a self-administered questionnaire, which included information on theirage, gender, family history of CRC, smoking status, drinkinghabits, previous medical history, and long-term medication use.Meanwhile, trained volunteers assisted survey completion forilliterate participants by reading the questions word by word. Allparticipants were then offered an educational session using astandard video followed by health seminars led by trainededucators. The video included information on the epidemiologyand natural history of CRC risk factors, clinical features of thiscondition, importance of regular screening, and procedures of thefecal immunochemical test (FIT) and colonoscopy. Hemosure(Manufacturer, W.H.P.M., Inc, El Monte CA) was used as the FITfor all subjects who chose fecal tests, which obviated the need fordietary restriction before testing. The seminars were delivered in astandardized manner, with both FIT and colonoscopy beingpresented nonpreferentially. Participants were given a choicebetween yearly FIT for up to 5 years or a direct colonoscopy forCRC screening. Our subjects consisted of participants who chosecolonoscopy, those who chose FIT initially but received a colono-scopy because of a positive FIT result, and those who receiveda colonoscopy after 3 consecutive years of negative FIT results.A total of 5,220 subjects were included in this study.

Outcome Variables and Covariates

Outcome variables included the proportion of screening partic-ipants who were found to have ACN; this was defined as CRC orany colorectal adenoma Z10 mm in diameter, high-gradedysplasia, villous or tubulovillous histologic characteristics, orany combination thereof. The major covariate included theAPCS score as a single variable, which was calculated for eachsubject on the basis of their age, gender, family history, andsmoking status. Additional variables tested for association withthe colonoscopic outcome of ACN included BMI (underweight,o18.5; normal, 18.5–22.3; overweight, 23–24.9; obese, Z 25);alcohol drinking (current drinkers with more than two drinksper week vs ex-drinkers or nondrinkers); self-reported history ofdiabetes mellitus, hypertension, cardiovascular diseases, livercirrhosis, stroke, gastroesophageal reflux disease, the use ofnonsteroidal anti-inflammatory drugs (NSAIDs) or aspirin; andself-reported current lifestyle habits as collected by validatedsurvey items (“frequent intake of meat,” “physical inactivity,”“frequent intake of barbecued food,” and “frequent intake offatty food”). The aforementioned BMI cut-off points were usedaccording to the recognized definition of obesity among Asiansubjects.20

Statistical Analyses

In 2013, all data were entered into a predesigned database withlogistic checking using Microsoft Access, and analyzed using SPSSsoftware, version 18.0. The proportion of participants who hadcolonoscopic findings of ACN was computed. A univariate

www.ajpmonline.org

Wong et al / Am J Prev Med 2014;46(5):433–439 435

analysis was conducted between ACN and each covariate consec-utively. All covariates were included into a binary logisticregression model if the initial p-value was o0.05 in the univariateanalysis. All variables selected in the multivariate regressionanalysis were detected for the presence of interactions.21 Eachvariable of interest was tested for interaction with other selectedvariables one by one. The variables with initial p-values o0.20 onunivariate analysis were included in another similar multivariateregression model. The adjusted ORs (AORs) and 95% CIs of thepotential independent predictors of ACN were estimated, and ap-value o0.05 in the multivariate regression analysis was consid-ered statistically significant. Furthermore, we compared the c-statistics of APCS score alone versus APCS score plus all theindependent predictors identified to evaluate whether the addi-tional predictors enhanced the performance of APCS in predictingACN risk factors. According to a generally accepted rule of thumb,a minimum of ten cases are required for each variable in theprediction rule.22 We assumed that the prevalence of ACN was 3%as evaluated in the validation cohort of the original APCS study.15

A total of 15 variables were tested in the prediction model; hence, atotal of at least 5,000 subjects needed to be recruited.

ResultsParticipant CharacteristicsAmong 10,728 screening participants, 5,220 received ascreening colonoscopy. The average participant agewas 57.7 (SD¼4.9) years, and 47.5% were men. Theaverage BMI was 23.5 (SD¼3.2); 8.3% were currentsmokers; and 10% were alcohol drinkers. More than14% had a first-degree relative with CRC. The mostfrequently reported chronic diseases included hyper-tension (22.9%); diabetes mellitus (7.4%); and gastro-esophageal reflux disease (5.3%). A few participantsshowed chronic use of NSAIDs (4.6%) or aspirin(2.5%). A significant number of participants reportedfrequent intake of meat (26.0%); physical inactivity(23.9%); frequent intake of barbecued food (21.4%);and frequent intake of fatty food (20.8%). The proportionof these screening participants found to have ACN orcancer was 5.6%, of which 265 subjects (5.1%) had ACN(Table 1).

Factors Associated with Advanced ColorectalNeoplasiaOn univariate analysis, APCS score (po0.001); BMI(p¼0.001); self-reported presence of diabetes (p¼0.047);hypertension (po0.001); and alcohol consumption(p¼0.001) were significantly associated with ACN. Inthe multivariate regression model, APCS score(AOR¼1.74, 95% CI¼1.34, 2.25, po0.001); overweight(BMI¼23�24.9, AOR¼1.52, 95% CI¼1.12, 2.07,p¼0.007); obesity (BMIZ25, AOR¼1.56, 95% CI¼1.15,2.10, p¼0.004); hypertension (AOR¼1.58, 95% CI¼1.21,2.06, p¼0.001); and alcohol consumption (AOR¼1.47,

May 2014

95% CI¼1.05, 2.06, p¼0.025) were associated with thecolonoscopic findings of ACN (Table 2). Diabetes,cardiovascular diseases, chronic obstructive pulmonarydisease, stroke, cirrhosis, gastroesophageal reflux disease,use of aspirin or other NSAIDs, and all self-reportedlifestyle factors were not significantly associatedwith ACN.No interaction exists between covariates in the binary

logistic regression model, thereby implying that theregression analysis is robust. Sensitivity analyses usingpo0.05 and o0.20 as the cut-off points for variableselection into the multivariate regression analyses,respectively, did not change the independent covariatesidentified. The area under the curve (AUC) of thereceiver operating characteristic (ROC) curve of APCSscore alone was 0.560 (95% CI¼0.524, 0.595, p¼0.001)and that of APCS score plus BMI; hypertension; andalcohol consumption was 0.613 (95% CI¼0.578, 0.648,po0.001). For the comparison of these two ROC curves,the revised model with additional risk factors performedsignificantly better than the APCS score alone in terms ofprediction of advanced neoplasia (z-statistic¼2.082,p¼0.037).

DiscussionIn this large cohort of CRC screening participants, theprevalence of ACN was 5.6%. Apart from the APCSscoring system, overweight or obesity, hypertension, andalcohol consumption were independently associated withthe findings of ACN. Adding these three additional riskfactors to the APCS increased its performance to predictACN risk.

Relationship with Existing LiteratureThe prevalence of ACN as reported in other Asianstudies is between 3% and 12%.23–25 In the derivationand validation cohort of the APCS study,15 the preva-lence of ACN was 4.5% and 3.0%, respectively. Theprevalence of ACN in this study (5.1%) was thereforehigher than the validation cohort (po0.001) but similarto the derivation cohort (p¼0.499) and the figures fromother studies. One possible explanation for this differenceis the younger mean age of the validation cohort ascompared to our subjects (51.0 vs 57.7 years)—hence ahigher proportion of subjects were found to have ACNin our study.Regarding the relationship between BMI and color-

ectal neoplasia, a recent pooled analysis of sevenprospective studies involving 8,213 screening partici-pants confirmed that BMI is associated with the risk ofmetachronous colorectal lesions, but is confined to menand lesions of the proximal colon.26 In a prospective

Table 1. Participant characteristics (N¼5,220), n (%) unlessotherwise noted

Characteristics M (SD), n (%)

Age (years, M [SD]) 57.72 (4.89)

BMI (M [SD]) 23.51 (3.17)

Gender, male 2,181 (47.5)

Current smoking 431 (8.3)

Alcohol consumption 522 (10.0)

Diabetes mellitus 386 (7.4)

Family history present for a first-degreerelative

735 (14.1)

Hypertension 1,196 (22.9)

IHD/heart disease 91 (1.7)

COPD 39 (0.7)

Stroke 37 (0.7)

Cirrhosis 8 (0.2)

GERD 277 (5.3)

Use of NSAIDs 240 (4.6)

Use of aspirin 133 (2.5)

Frequent intake of red meat 1,358 (26.0)

Physical inactivity 1,245 (23.9)

High intake of barbecued food 1,115 (21.4)

Frequent intake of fatty food 1,087 (20.8)

Advanced neoplasia and colorectal cancer 289 (5.6)

Colorectal cancer 24 (0.5)

Advanced neoplasia 265 (5.1)

COPD, chronic obstructive pulmonary disease; GERD, gastroesophagealreflux disease; IHD, ischemic heart disease; NSAIDs, nonsteroidal anti-inflammatory drugs

Wong et al / Am J Prev Med 2014;46(5):433–439436

cohort study of 33,403 African-American women agedZ30 years with no prior cancer or polyps, an increasedrisk of colon polyps was demonstrated among thosewith BMIZ35 compared to subjects with BMIo25.27 Across-sectional study conducted in Korea demonstrateda weak association between BMI and the risk ofadenoma among men at the highest BMI levels, butthis association was abolished after adjustment forwaist circumference.28 However, these studies wereperformed either on Western or African subjects, orsuffered from a small sample size. The relatively largenumber of asymptomatic screening participantsincluded in the present study reported the new findingthat obese subjects had a 29% increased risk of ACN. Inaddition, this study also showed a dose-response

relationship between BMI and ACN, which impliesthat a potential cause–effect relationship might exist.The best established biochemical mediator betweenobesity and colorectal neoplasia is the pro-carcinogeninsulin-like growth factor 1 and, to a lesser extent,leptin.29 From a clinical utilization perspective, thesefindings imply that Asian subjects who are overweight(BMI¼23–24.9) have statistically similar odds of devel-oping ACN when compared with obese subjects(BMIZ25) (1.52-fold, 95% CI¼1.12, 2.07, vs 1.56-fold,95% CI¼1.15, 2.10). Therefore, overweight patientsshould also be considered as having a higher risk ofACN and receive higher priority for screening thanaverage-risk subjects.Turning to the association between alcohol consump-

tion and ACN, a cross-sectional study conducted in theU.S. found that patients who consumed more than eightdrinks of spirits or eight servings of beer per week weremore likely to have significant colorectal neoplasia butthat the consumption of one to eight glasses of wine perweek was associated with a lower risk.30 A pooledanalysis of primary data frommore than 489,000 personsin eight cohort studies in North American and Europeancountries has shown that alcohol drinkers of more than2–2.5 drinks daily had a 1.16-fold increased risk ofdeveloping CRC.31 A large prospective cohort studyconducted in Japan found a positive association betweenalcohol drinking and CRC,32 yet a smaller-scale case–control study in the same country reported only modestlyincreased risks of distal colon and rectal adenomas, butnot large adenomas.33 However, subjects in both analyseswere not CRC screening participants, and their inclusioncriteria did not include the absence of CRC symptoms.From the existing literature, the relationship betweenalcohol drinking and colorectal neoplasia among asymp-tomatic subjects in Asia has been inconclusive. Ethanolhas been shown to promote, rather than initiate, color-ectal carcinogenesis.33 Thus, the present finding thatalcohol drinkers had a 50% higher risk of ACN than non-drinkers implies that alcohol consumption should beincluded as one of the important predictive factors for thepresence of ACN.The positive association between hypertension and

ACN has not been reported in previous literature andthus could represent a novel finding. Moreover, it isinteresting that diabetes was only found to be significantlyassociated with ACN in univariate analysis but not inmultivariate regression models. There are a few explan-ations for the absence of this association. First, diabetesand ACN share some common risk factors, such asalcohol consumption and obesity. This may compromisethe independent predictive ability of diabetes for ACNwhen these other factors are controlled for. Also, a weaker

www.ajpmonline.org

Table 2. Univariate and multivariate predictors of advanced colorectal neoplasia and colorectal cancer

Unadjusted (univariate analysis) Adjusted (multivariate regression analysis)

Risk factors OR (95% CI) p-value B coefficient SE OR (95% CI) p-value

APCS High risk versus low ormoderate risk

1.858 (1.439,2.398)

o0.001 0.553 0.133 1.738 (1.339,2.254)

o0.001

BMI o0.001 0.009

o18.5 (underweight) 0.812 (0.371,1.776)

0.601 �0.157 0.400 0.854 (0.390,1.873)

0.694

Z18.5 to o23 (normal) ref ref

Z23 to o25 (overweight) 1.610 (1.188,2.184)

0.002 0.420 0.157 1.522 (1.119,2.069)

0.007

Z25 (obesity) 1.830 (1.371,2.445)

o0.001 0.442 0.153 1.556 (1.154,2.099)

0.004

Diabetes 1.489 (1.005,2.205)

0.047 0.160 0.209 1.173 (0.779,1.767)

0.444

Hypertension 1.821 (1.415,2.343)

o0.001 0.456 0.137 1.578 (1.206,2.064)

0.001

IHD/heart disease 1.663 (0.797,3.470)

0.175 — — — —

Alcohol 1.772 (1.275,2.462)

0.001 0.386 0.172 1.471 (1.050,2.061)

0.025

COPD 1.963 (0.693,5.562)

0.204 — — — —

Stroke 0.472 (0.065,3.456)

0.460 — — — —

Cirrhosisa N/A 1.000 — — — —

GERD 1.198 (0.732,1.962)

0.473 — — — —

Use of NSAIDs 0.581 (0.317,1.064)

0.078 — — — —

Use of aspirin 1.365 (0.717,2.601)

0.344 — — — —

Frequent intake of red meat, n (%) 0.907 (0.688,1.197)

0.907 — — — —

Physical inactivity, n (%) 0.777 (0.577,1.045)

0.096 — — — —

Low intake of fruit or vegetables, n (%) 1.010 (0.756,1.348)

0.948 — — — —

Frequent intake of fatty food, n (%) 1.210 (0.915,1.601)

0.180 — — — —

Note: Boldface indicates po0.05.aNone of the eight subjects with cirrhosis had advanced neoplasia.APCS, Asia Pacific Colorectal Scoring System; COPD, chronic obstructive pulmonary disease; GERD, gastroesophageal reflux disease; IHD, ischemicheart disease; NSAIDs, nonsteroidal anti-inflammatory drugs

Wong et al / Am J Prev Med 2014;46(5):433–439 437

association exists between diabetes and ACN amongAsian subjects compared to Western populations. Forinstance, a systematic review of 16 cohort studies andeight case-control studies involving more than 3.6 millionindividuals showed that the presence of diabetes was

May 2014

associated with a 1.26-fold increased risk of CRC,34 whichwas lower among Asian populations (relative risk¼1.19,95% CI¼1.11, 1.28) than European ones (relativerisk¼1.39, 95% CI¼1.26, 1.53). Lastly, most previousstudies have found an association between type 2 diabetes

Wong et al / Am J Prev Med 2014;46(5):433–439438

and ACN, but not type 1 diabetes.35,36 The associ-ation between ACN and type 1 diabetes remains to beexplored.

Strengths and LimitationsThis is a large-scale study that evaluated asymptomaticsubjects who were self-referred for CRC screening. It isunique because the analysis was based on the well-validated APCS scoring system, which allows the explora-tion of additional risk factors for ACN. However, somelimitations should be acknowledged. First, consecutivesampling was adopted, which might have limited itsgeneralizability to the whole population. However, giventhe screening infrastructure of the community, it was notfeasible to conduct a study using population-based, simplerandom-sampling methodology owing to the absence ofan accurate sampling frame and high refusal rates. Second,the characteristics of Chinese subjects aged between 50and 70 years might be different from those in the originalAPCS study, which included asymptomatic subjects with awider age range recruited from different centers in theAsia-Pacific region. In addition, although our recruitmentexcluded subjects with some well-known risk factors, suchas a past history of CRC and inflammatory bowel disease,not all confounders could be controlled in this analysis.Lifestyle factors such as a high intake of red meat, fat, andbarbecued food37,38 and physical inactivity39 were onlymeasured in a simplified manner. Additionally, the surveydid not collect information on the type of alcoholicbeverages and their total consumption. Nevertheless, thevariables pertaining to dietary habits require food fre-quency questionnaires or the accurate keeping of a 24-hour food diary for assessment; this process is time-consuming and therefore unlikely to be practical forinclusion as a component in the risk scoring system.Finally, the presence of chronic diseases was self-reportedand may be subject to ascertainment bias, andtheir duration and treatment statuses have not beenanalyzed.

Implications for Clinical Practice and FutureResearchBMI, alcohol consumption, and self-reported hyperten-sion should be considered as additional risk factors apartfrom the variables of the APCS scoring system.15 Theseadditional factors could be easily obtained or measuredin clinical or community settings. Patients with these riskfactors should receive additional counseling efforts forCRC screening, as the colonoscopic yield of ACN is hig-her when compared with the average-risk general pop-ulation. In resource-limited countries, physicians mayuse these risk factors to prioritize screening services, and

in low-prevalence countries, risk stratification could beadopted to selectively offer screening to high-risk indi-viduals. Built on the APCS system, these additionalvariables could potentially inform policy makers andphysicians when allocating CRC screening resources inthe community. It is anticipated that these factors couldrisk-stratify CRC screening participants in a more cost-effective manner. Because the additional variables foundto be significant may have different associations withACN in other patient groups, their generalizability toother populations such as Chinese immigrants livingoutside of China should be studied further. Futurestudies should also focus on developing and validatinga modified system based on the present findings yetmaintaining the simplicity of the predictive tool.

We acknowledge the full funding support of the Hong KongJockey Club Charities Trust for this study. We are grateful toDr. David Wilmshurst, the Chinese University of Hong Kong’sAcademic Editor, for editing the first draft of this paper.No financial disclosures were reported by the authors of

this paper.

References1. Ferlay J, Shin HR, Bray F, et al. Estimates of worldwide burden of

cancer in 2008: GLOBOCAN 2008. Int J Cancer 2010;127(12):2893–917.

2. Sung JJY, Lau JYW, Goh KL, et al. for the Asia Pacific Working Groupon Colorectal Cancer. Increasing incidence of colorectal cancer in Asia:implications for screening. Lancet Oncol 2005;6(11):871–6.

3. Wong CK, Lam CL, Poon JT, et al. Direct medical costs of care forChinese patients with colorectal neoplasia: a health care serviceprovider perspective. J Eval Clin Pract 2012;18(6):1203–10.

4. Hardcastle JD, Chamberlain JO, Robinson MH, et al. Randomisedcontrolled trial of faecal-occult-blood screening for colorectal cancer.Lancet 1996;348(9040):1472–7.

5. Jorgensen OD, Kronborg O, Fenger C. A randomised study ofscreening for colorectal cancer using faecal occult blood testing: resultsafter 13 years and seven biennial screening rounds. Gut 2002;50(1):29–32.

6. Mandel JS, Church TR, Ederer F, et al. Colorectal cancer mortality:effectiveness of biennial screening for fecal occult blood. J Natl CancerInst 1999;91(5):434–7.

7. Zauber AG, Winawer SJ, O’Brien MJ, et al. Colonoscopic polypectomyand long-term prevention of colorectal-cancer deaths. N Engl J Med2012;366(8):687–96.

8. Pignone M, Rich M, Teutsch SM, et al. Screening for colorectal cancerin adults at average risk: a summary of the evidence for the U.S.Preventive Services Task Force. Ann Intern Med 2002;137(2):132–41.

9. Levin B, Lieberman AD, McFarland B, et al. Screening and surveillancefor the early detection of colorectal cancer and adenomatous polyps,2008: a joint guideline from the American Cancer Society, the U.S.Multi-Society Task Force on Colorectal Cancer, and the AmericanCollege of Radiology. CA Cancer J Clin 2008;58(3):130–60.

10. Advisory Committee on Cancer Prevention. Recommendations on cancerscreening in the European Union. Eur J Cancer 2000;36(12):1473–8.

11. Sung JJ, Lau JY, Young GP, et al. Asia Pacific consensus recommen-dations for colorectal cancer screening. Gut 2008;57(8):1166–76.

www.ajpmonline.org

Wong et al / Am J Prev Med 2014;46(5):433–439 439

12. Brown ML, Klabunde CN, Mysliwiec P. Current capacity for endo-scopic colorectal cancer screening in the U.S.: data from the NationalCancer Institute Survey of Colorectal Cancer Screening Practices. Am JMed 2003;115(2):129–33.

13. Levin TR. Colonoscopy capacity: can we build it? Will they come?Gastroenterology 2004;127(6):1841–4.

14. Rex DK, Lieberman DA. Feasibility of colonoscopy screening: dis-cussion of issues and recommendations regarding implementation.Gatrointest Endosc 2001;54(5):662–7.

15. Yeoh KG, Ho KY, Chiu HM, et al. for the Asia Pacific Working Groupon Colorectal Cancer. The Asia-Pacific Colorectal Screening score: avalidated tool that stratifies risk for colorectal advanced neoplasia inasymptomatic Asian subjects. Gut 2011;60(9):1236–41.

16. Wong MCS, Lam TYT, Tsoi KKF, et al. A validated tool to predictcolorectal neoplasia and inform screening choice for asymptomaticsubjects. Gut 2013. http://dx.doi.org/10.1136/gutjnl-2013-305639.

17. Wong MCS, Tsoi KKF, Ng SSM, et al. A comparison of the acceptanceof FIT and colonoscopy in colorectal cancer screening: a prospectivestudy among Chinese. Aliment Pharmacol Ther 2010;32(1):74–82.

18. WongMCS, JohnGK, Hirai HW, et al. Changes in the choice of colorectalcancer screening tests in primary care settings from 7,845 prospectivelycollected surveys. Cancer Causes Control 2012;23(9):1541–8.

19. Wong MC, Hirai HW, Luk AK, et al. The knowledge of colorectalcancer symptoms and risk factors among 10,078 screening partici-pants: are high risk individuals more knowledgeable? PLoS One 2013;8(4):e60366.

20. WHO/IASO/IOTF. The Asia-Pacific perspective: redefining obesityand its treatment. Sydney Australia: Health Communication AustraliaPty Ltd, 2000.

21. Katz MH, ed. Multivariable analysis: a practical guide for clinicians.Cambridge MA: Cambridge University Press, 2006.

22. Peduzzi P, Concato J, Kemper E, et al. A simulation study of thenumber of events per variable in logistic regression analysis. J ClinEpidemiol 1996;49(12):1373–9.

23. Sung JJ, Chan FK, Leung WK, et al. Screening for colorectal cancer inChinese: comparison of fecal occult blood test, flexible sigmoidoscopy,and colonoscopy. Gastroenterology 2003;124(3):608–14.

24. Soon MS, Kozarek RA, Ayub K, et al. Screening colonoscopy inChinese and Western patients: a comparative study. Am J Gastro-enterol 2005;100(12):2749–55.

25. Chiu HM, Wang HP, Lee YC, et al. A prospective study of thefrequency and the topographical distribution of colon neoplasia in

May 2014

asymptomatic average-risk Chinese adults as determined by colono-scopic screening. Gastrointest Endosc 2005;61(4):547–53.

26. Jacobs ET, Ahnen DJ, Ashbeck EL, et al. Association between bodymass index and colorectal neoplasia at follow-up colonoscopy: apooling study. Am J Epidemiol 2009;169(6):657–66.

27. Wise LA, Rosenberg L, Palmer JR, et al. Anthropometric risk factors forcolorectal polyps in African-American women. Obesity (Silver Spring)2008;16(4):859–68.

28. Kim Y, Kim Y, Lee S. An association between colonic adenoma andabdominal obesity: a cross-sectional study. BMC Gastroenterol 2009;9:4.

29. Frezza EE, Wachtel MS, Chiriva-Internati M. Influence of obesity onthe risk of developing colon cancer. Gut 2006;55(2):285–91.

30. Anderson JC, Alpern Z, Sethi G, et al. Prevalence and risk of colorectalneoplasia in consumers of alcohol in a screening population. Am JGastroenterol 2005;100(9):2049–55.

31. Cho E, Smith-Warner SA, Ritz J, et al. Alcohol intake and colorectalcancer: a pooled analysis of 8 cohort studies. Ann Intern Med 2004;140(8):603–13.

32. Otani T, Iwasaki M, Yamamoto S. Alcohol consumption, smoking, andsubsequent risk of colorectal cancer in middle-aged and elderlyJapanese men and women: Japan Public Health Center-based pro-spective study. Cancer Epidemiol Biomarkers Prev 2003;12(12):1492–500.

33. Toyomura K, Yamaguchi K, Kawamoto H, et al. Relation of cigarettesmoking and alcohol use to colorectal adenomas by subsite: the self-defense forces health study. Cancer Sci 2004;95(1):72–6.

34. Deng L, Gui Z, Zhao L, et al. Diabetes mellitus and the incidence ofcolorectal cancer: an updated systematic review and meta-analysis. DigDis Sci 2012;57(6):1576–85.

35. McNeely MJ, Boyko EJ. Type 2 diabetes prevalence in AsianAmericans: results of a national health survey. Diabetes Care 2004;27(1):66–9.

36. Khalili H, Chan AT. Is diabetes a risk factor for colorectal cancer? DigDis Sci 2012;57(6):1427–9.

37. Giovannucci E, Ascherio A, Rimm EB, et al. Physical activity, obesity,and risk for colon cancer and adenoma in men. Ann Intern Med1995;122(5):327–34.

38. Chan DS, Lau R, Aune D, et al. Red and processed meat and colorectalcancer incidence: meta-analysis of prospective studies. PLoS One2011;6:e20456.

39. Fung T, Hu FB, Fuchs C, et al. Major dietary patterns and the risk ofcolorectal cancer in women. Arch Intern Med 2003;163(3):309–14.