Disclosure Information: Nothing to disclose.

Disclosures outside the scope of this work: Dr Findlayreceived an honorarium for a lecture from Heibling CResearch AG.

Support: Dr Findlay is supported by the NIHR OxfResearch Centre.

Received July 27, 2016; Revised September 13, 2016; Acce13, 2016.From the Oxford OesophagoGastric Centre, Churchill HosKafsi, Gillies, Maynard) and the Department of EmergencRadcliffe Hospital (Hammer, Gilmour), Oxford UniversityFoundation Trust; and NIHR Oxford Biomedical Researdlay), Churchill Hospital, Oxford, UK.Correspondence address: John M Findlay, BMedSci,MRCS; Oxford OesophagoGastric Centre, Churchill HOX3 7LJ, UK. email: john.findlay@oncology.ox.ac.uk

ª 2016 by the American College of Surgeons. Published by

All rights reserved.

COLLECTIVE REVIEW

Nonoperative Management of Appendicitis inAdults: A Systematic Review and Meta-Analysis

of Randomized Controlled Trials

John M Findlay, BMedSci, BMBS (hons), MRCS, Jihene el Kafsi, MBBS, FRCS, Clare Hammer, MBChB, MRCS,Jeffrey Gilmour, MPhil, FRCS, Richard S Gillies, MD, FRCS, Nicholas D Maynard, MS, FRCS

Appendicitis remains the most common intra-abdominalsurgical emergency, with an annual incidence of 250,000patients in the US, 50,000 patients in the UK, and a life-time risk of 8%.1,2 The vast majority of these cases aremanaged by appendectomy, underpinned by the dogmathat uncomplicated appendicitis inevitably progresses toabscess formation, gangrene, and perforation. However,more routine use of imaging has improved diagnosticaccuracy and identification of the majority of patientswith uncomplicated appendicitis at admission.3

In parallel, there is increasing recognition that the path-ogenesis and natural history of appendicitis is variable.Logically, some patients will have an episode that willnot progress or perhaps may even be self-limiting,4 andantibiotics alone will sometimes suffice. Indeed, thisapproach is necessary in remote environments5 and histor-ically antibiotics have been considered the treatment ofchoice for delayed presentations with an appendix mass.6

Consequently, a number of randomized controlled trials(RCTs) and attendantmeta-analyses have explored the roleof routine nonoperative management with antibiotics.Some have concluded that it represents an equally validalternative to surgery for uncomplicated appendicitis,with an equivalent safety profile and possibly, fewer com-plications,7,8 although this view was not supported bythe most recent Cochrane review of 5 RCTs.9 At present,appendectomy remains the treatment of choice

has previouslyonsulting and

ord Biomedical

pted September

pital (Findlay, ely Surgery, JohnHospitals NHSch Centre (Fin-

BMBS (hons),ospital, Oxford,

814Elsevier Inc.

recommended in the US and Europe.10-12 However, theseguidelines, reviews, and meta-analyses were publishedbefore a recent relevant RCT in 2015.13 Although thisdid not demonstrate noninferiority of antibiotics, the sig-nificant reduction in complications seen with antibioticshas frequently been interpreted as additional evidencethat antibiotics represent a valid alternative to surgery.14,15

The aim of this study was to perform an up-to-date system-atic review and meta-analysis of primary antibiotic therapyvs surgery for the management of acute appendicitis.

METHODS

Literature search

A literature searchwasperformedof the PubMed,EMBASE,and Cochrane Central Register of Controlled Trials in May2016 by 2 authors (JEK andCH), independently using stan-dardized data collection software, in accordance with thePreferred Reporting Items for Systematic Reviews andMeta-Analyses (PRISMA) and Meta-Analysis Of Observa-tional Studies in Epidemiology (MOOSE) guidelines.16,17

The following search terms were used: ([antibiotics ORappendectomy OR appendicectomy OR surgery] AND[trial OR randomized] AND [appendicitis OR appendix]).Bibliographies of retrieved articles were searched. Anydisagreement was resolved by a third author (JMF). Datawere extracted and studies similarly assessed by 2 authors(JMF and JEK). We contacted the corresponding authorsof 3 studies with questions regarding methodology andrequests for data,18-20 although none could be provided.

Inclusion criteria

We included randomized and quasi-randomized prospec-tive controlled trials, randomizing patients aged 16 andolder to either primary antibiotic therapy or appendec-tomy for acute appendicitis.

Endpoints

Successful treatment was defined pragmatically as resolu-tion of the presenting acute episode of appendicular inflam-mation without recurrence: ie removal of the appendixwithout subsequent inflammation of the stump or ileum/

http://dx.doi.org/10.1016/j.jamcollsurg.2016.09.005

ISSN 1072-7515/16

Abbreviations and Acronyms

GRADE ¼ Grading of Recommendations Assessment,Development, and Evaluation

ITT ¼ intention to treatPP ¼ per protocolRCT ¼ randomized controlled trialRR ¼ risk ratio

Vol. 223, No. 6, December 2016 Findlay et al Nonoperative Management of Appendicitis 815

cecum, or significant improvement in clinical/biochemicalevidence of inflammation, not requiring further antibioticsor appendectomy. This was either during the initial episodeor admission, or at 1 year. Recurrent appendicitis wasdefined as a clinical or radiologic diagnosis. Complicatedappendicitis was defined as intra-abdominal pus, entericcontamination, or necrotic/gangrenous appendicitis. For2 studies not stipulating the presence of pus (all categorizedas phlegmonous, gangrenous, or perforated18,19), gangreneor perforation were used.Complications were as defined in studies. Variable pre-

sentation precluded meaningful synthesis of complicationsreported by all studies, so we performed pragmatic analysesfor all, minor, and major complications. Minor complica-tions included wound infections, presumed adhesional/wound pain, urinary catheterization at discharge, minoranesthetic complications (eg tooth injury), and fungalinfections. Major complications constituted death,intra-abdominal abscess/post-intervention peritonitis,enterocutaneous fistula, major medical complications(eg pneumonia, acute coronary syndrome, pulmonaryembolism, Clostridium difficile colitis), incisional hernia,wound dehiscence, or surgery for bowel obstruction.

Meta-analysis

Two analyses were performed for all comparisons:intention-to-treat (ITT) and per protocol (PP). For theindex admission, it was possible to perform a PP analysisthat included patients crossing over after randomizationin 1 study.18 However, this could not be performed at 1year. For ITT, all patients were included, irrespective ofprotocol violations and follow-up. For PP, only patientscompleting treatment per protocol and follow-up at therelevant time point were included. For 1 study,13 althougha few patients were formally lost to follow-up, the authorswere confident they had identified subsequent appendec-tomies. These patients were therefore included in primarybut not secondary analyses.

Statistical analysis

Heterogeneity was assessed using chi-square (p < 0.05),and quantified using I2. We used random effects models

due to methodologic and comparison heterogeneity,although we performed sensitivity analyses using fixed ef-fects, and individual studies. For binary variables Mantel-Haenszel risk ratio (RR) was used. For the 2 studies18,20

presenting standard error (SEM) rather than standard de-viation (SD), we converted the former to the latter. Onestudy presented length of stay as median and range, ratherthan mean and SD13; the latter were estimated.21 Funnelplots were inspected visually for evidence of asymmetry.All analysis was performed using Review Manager v5.222

and R v3.0.2.23

RESULTS

Literature search

There were 1,146 studies identified, 69 of which wereretrieved for full text evaluation (Fig. 1). Seven RCTswere identified. One was excluded due to subsequentretraction.24

Study characteristics and quality

Six RCTs conducted between 1992 and 2012 wereincluded (Table 1),13,18-20,25,26 2 with noninferiority de-signs.13,25 A total of 1,724 patients were randomized toantibiotics (n ¼ 837) or surgery (n ¼ 887). Studies variedconsiderably in their interventions, inclusion, exclusion,and diagnostic criteria. All reported a maximum 1-yearfollow-up. None were blinded. Randomization was bysealed envelope,13,19 computer generation,25 date ofbirth,18 or an unclear method.20,26 One study18 allowedcrossover after randomization, whereby clinicians or sur-geons could alter treatment. All except 1 study19 describeddropouts and withdrawals.Overall, there were many generic areas of potential bias.

These included major differences in inclusion, exclusion,and diagnostic criteria; differences in antibiotics and ther-apy duration; major variability in surgical approach (openprocedures being used selectively); follow-up methods;and definition and reporting of complications.

Inclusion, diagnostic, and exclusion criteria

Five studies included patients aged 18 and older; 1 studyincluded patients 16 and older.20 Two had upper agelimits of 60 years13 and 50 years.19 One included onlymale patients.19 Two studies mandated CT diagnosis,13,25

and 3 had clinical diagnosis with or without ultrasound/CT18 or ultrasound alone alone.26

Overall, exclusion criteria effectively amounted toradiologic or clinical evidence of perforation or abscessformation. Four studies excluded radiologic13,25 or clinicalsuspicion of perforation or abscess.19,26 One study18 didnot formally exclude such patients, but clinicians could

Records identified through PubMed

(n =965)

Screening

Included

Eligibility

Identification

Records after duplicates removed(n =1,146)

Records screened(n = 1,146)

Records excluded(n = 1,077)

Full-text articles assessed for eligibility

(n = 69)

Full-text articles excluded, with reasons

(n=63)

n=24 not assessing outcome of interest

n = 14 reviewn= 3 meta-analyisis

n= 12 lettern=8 non-randomised n=1 study proposal

n=1 RCT subsequently retracted

Studies included in qualitative synthesis

(n = 6)

Studies included in quantitative synthesis

(meta-analysis)(n = 6)

Records identified through EMBASE

(n =431)

Records identified through Cochrane

(n =508)

Figure 1. Preferred Reporting Items for Systematic Reviews and Meta-Analyses flow diagram. RCT, randomized controlled trial.

816 Findlay et al Nonoperative Management of Appendicitis J Am Coll Surg

alter treatments if they felt it was justified. Two alsoexcluded possible malignancy on CT13,25; 1 excludedany patient with an appendix >15 mm in diameter.26

One study13 excluded patients with a serious systemicillness or the presence of an appendicular fecolith.

Primary antibiotic therapy

Five studies involved a variable period of intravenous an-tibiotics for at least 1,18,25 2,19,26 or 3 days13,20; 1 study25

used oral antibiotics in the absence of nausea or vomiting.Seven to 10 days of oral antibiotics were used thereafter.Variable combinations of penicillins, cephalosporins, car-bopenems, or fluoroquinolones were used, in combina-tions with a nitroimidazole. Criteria for converting toappendectomy varied, but essentially constituted deterio-ration or failure to improve within 24 to 48 hours.

Appendectomy

All studies except one18 included either open or laparo-scopic procedures, performed at the discretion of the sur-geon. One did not present this information18; excludingthis, open procedures accounted for 71.4% (584 of818). No further operative details were provided. For pa-tients randomized to surgery, antibiotics were limited to asingle prophylactic dose in 3 studies13,18,25 and were notstipulated by 3 studies. Three18,25,26 stipulated an unde-fined postoperative course in the presence of complicatedappendicitis; this was unclear in 2,19,20 while 1 did not usepostoperative antibiotics.13

Endpoints

All studies presented the number of patients randomizedto antibiotics subsequently treated for failure or recurrent

Table 1. Studies Included

First author,y Population, n, ITT (PP) Diagnostic Exclusions Antibiotic therapy Operation Potential risks of bias

Salminen,13

2015Age 18 to 60 y; Finland

(2009-2012); AB: 257(256); S: 273 (272)

CT Fecolith; possibleperforation; abscess;possible tumor;significant systemicillness

Hospital: ertapenem IV(3 d); discharge:levofloxacin þmetronidazole (7 d)

Open: 257; lap: 15; AB: x1preop dose � postop

Partly incomplete follow-up;laparoscopy at discretion ofsurgeon; no postoperativeantibiotics, even in presenceof contamination;exclusion: age > 60 y andsignificant systemic illness;incomplete description ofsome complications

Vons,25 2011 Age >18 y; France (2004-2007); AB: 123 (106);S: 120 (98)

CT Perforation, abscess,possible tumor/appendix >15 mm

Hospital: co-amoxiclav po/iv (�1 d); discharge: co-amoxiclav po (8 d total)

Open: 41; lap: 78; AB:single preop dose �postop

Laparoscopy at discretion ofsurgeon; oral antibioticsunless nauseated orvomiting

Hansson,18

2009Age >18 y; Sweden (2006-

2007); AB: 202 (102)(including crossover115); S: 167 (130)(including crossover:219)

Clinical �US/CT Nil formal; surgeons/patients alteredtreatment as thoughtappropriate

Hospital: cefuroxime þmetronidazole iv (�1d); discharge:ciprofloxacin þmetronidazole po (10d total)

Open: not specified; lap:not specified; AB: singlepreop dose � postop

Quasi-randomized; cross-overafter randomization atpatient/surgeon request;laparoscopy at discretion ofsurgeon; incomplete andincompletely describedfollow-up and data;nonstandardizedinvestigations

Turhan,20

2009Age >16 y; Turkey (2005-

2006); AB: 107 (107);S: 183 (183)

Clinical � US and CT(antibiotic group only)

Nil formal Hospital: ampicillin þgentamicin þmetronidazole iv (3 d);discharge: ampicillin þmetronidazole po (10 d)

Open 150; lap: 133; AB:not specified

Randomization methodunclear; significant disparityin patients; allocationconcealment: unclear;laparoscopy at discretion ofsurgeon; drop-outs notspecified; use of US/CTonly in AB group, withoutreporting negativeappendectomy rate;postoperative antibioticsunclear

Styrud,19 2006 Males age 18 to 50 y;Sweden (1996-1999);AB: 128 (128); S: 124(124)

Clinical Suspected perforation; Creactive protein <10mg/L

Hospital: cefotaxime þtinidazole iv (2 d);discharge: ofloxacin þtinidazole po (10 d)

Open: 116; lap: 8; AB: notspecified

Females excluded; incompletedescription ofrandomization; laparoscopyat discretion of surgeon;complications notdescribed; postoperativeantibiotic therapy unclear;drop-outs not specified

Eriksson,26

1995Age >18 y; Sweden (1992-

1994); AB: 20 (20); S:20 (20)

Clinical þ US Generalized peritonitis Hospital: cefotaxime þtinidazole iv (2 d);discharge: ofloxacin þtinidazole po (8 d)

Open: 20; AB only ifperforation/spillage

No prophylactic antibiotics

AB, antibiotics; ITT, intention-to-treat group; lap, laparoscopic; po, per os; postop, postoperative; PP, per protocol group; preop, preoperative; S, surgery; US, ultrasound scan.

Vol.223,No.6,Dece

mber2016

Findlayet

alNonoperative

Manage

mentofAppendicitis

817

818 Findlay et al Nonoperative Management of Appendicitis J Am Coll Surg

appendicitis. One18 excluded patients undergoing surgeryfrom PP analysis without histologic evidence of appendi-citis. However, we included these in the absence of asimilar mechanism for identifying false positives amongpatients treated with antibiotics, and its reflection of clin-ical practice.Complications were variably reported and defined,

with a high risk of reporting bias. One study reportedevery complication encountered18; 1 stipulated onlywound infections.19 The other 4 studies13,18,20,25 reportedvariable complications including intra-abdominalabscesses and enterocutaneous fistulae; 3 studies13,18,25

also reported wound hernias, postoperative pain/obstruc-tive symptoms (subjective and unqualified), with 1 study18

reporting more extensive minor complications. However,it is unclear whether additional complications were seen.One study defined wound infection.13 None stratifiedcomplication severity.

Directness

In addition to introducing bias, many factors above alsolimited directness, ie applicability to the population of in-terest. These included the over-representation of openprocedures (laparoscopy representing the approach ofchoice in the US and Europe,10-12 with less pain, fewercomplications, and shorter length of stay27,28), and varia-tions in inclusion, exclusion, and diagnostic criteria(such as major comorbidities and routine CT, whichmay not reflect standard practice10).

Efficacy

Seventy-five of 834 patients (8.99%; 95% CI 7.49 to11.1) randomized to antibiotics underwent appendec-tomy during their initial treatment (74 of 720 [10.2%]PP). Antibiotics were less effective than surgery: ITTRR 0.92 (95% CI 0.87 to 0.97; p ¼ 0.002), PP (0.91[95% CI 0.87 to 0.96]; p ¼ 0.0002), and crossover(0.90 [95% CI 0.86 to 0.94]; p < 0.00001; Table 2;Fig. 2). This constituted failure in 8.00% (95% CI

Table 2. Successful Treatment of (Presumed) Appendicitis

Analysis Effect size, RR (95% CI) n

Successful treatment at initial episode

ITT 0.92 (0.87e0.97) 6

PP 0.91 (0.87e0.96) 6

PP þ crossover 0.90 (0.86e0.94) 6

Successful treatment at 1 y

ITT 0.75 (0.65e0.88) 6

PP 0.71 (0.67e0.76) 6

GRADE, Grading of Recommendations Assessment, Development, and Evalua

3.00% to 13.0%) and 9.00% (95% CI 4.00% to13.0%), respectively.For patients treated successfully with antibiotics, recur-

rence within 1 year necessitating surgery (n ¼ 130) or an-tibiotics (n ¼ 5) occurred in 135 of 759 (ITT 17.8%[95% CI 15.2% to 20.7%]) and 119 of 613 (PP 19.0%[95% CI 16.5 to 22.7]). On meta-analysis, risks were15.0% (95% CI 8.00% to 22.0%) and 17.0% (95% CI11.0% to 24.0%), respectively.Cumulatively, at 1 year, meta-analysed risks of antibi-

otics failure were 25.0% (95% CI 12.0% to 35.0%),and 29.0% (95% CI 23.0% to 34.0%; p < 0.00001;Table 2; Fig. 2). Neither study attempting to do sodemonstrated noninferiority.13,25 GRADE (Grading ofRecommendations Assessment, Development, and Evalu-ation) levels of evidence were low (PP), and very low (ITTand PP with crossover [Table 3]).

Risk of complicated appendicitis at surgery

One study18 presenting only data for patients crossingover was excluded. Overall, there were no significant dif-ferences in either index episode or 1-year risk of compli-cated appendicitis at surgery (Table 4; eFigs. 1 and 2). Forthe index admission, 24 of 632 (3.80%) patients treatedwith antibiotics had complicated appendicitis (ITT),compared with 74 of 719 (10.3%) treated with surgery.Per-protocol rates were 23 of 618 (3.72%) and 65 of693 (9.38%), respectively. At 1 year, these rates were 40of 632 (6.33%) and 74 of 719 (10.3%), and 36 of 619(5.98%) and 65 of 593 (9.38%), respectively.However, for patients in whom antibiotics failed,

complicated appendicitis was considerably more likely:PP RR 6.21 (95% CI 1.68 to 23.0; p ¼ 0.006;Table 3; eFig. 3). As a proportion of patients undergoingsurgery, 25 of 54 (46.3%; ITT) and 23 of 65 (35.4%; PP)had complicated appendicitis. At 1 year, significance didnot persist, with recurrent episodes no more likely to becomplicated: 16 of 117 (13.7%; ITT), and 14 of 108(12.7%; PP). Again GRADE levels were low and verylow.

I2, % Chi-square p Value Model GRADE quality

80 24.42 0.002 Random Very low

74 19.41 0.0002 Random Low

64 14.08 <0.00001 Random Very low

91 57.28 <0.00001 Random Very low

30 7.19 <0.00001 Random Low

tion; ITT, intention-to-treat group; PP, per protocol group; RR, risk ratio.

Figure 2. One-year efficacy (per protocol analysis).

Vol. 223, No. 6, December 2016 Findlay et al Nonoperative Management of Appendicitis 819

Risk of complications

One study did not present sufficient data to assess minorand major complications individually.19 Another did notprovide sufficient data to perform PP meta-analysis.18

Overall, there were no differences in complications at 1year (Table 4; eFig. 4). On sensitivity analysis using fixedeffects, antibiotics were associated with fewer complica-tions, although random effects were mandated by severestudy heterogeneity (I2 >76%), and significance wasreliant on considerable rates of undefined adhesional/obstructive symptoms and wound infections after surgeryin 1 study.13 Similarly, there were no differences in minor(eFig. 5) or major complications (eFig. 6; Table 4). TheGRADE recommendations were of very low quality,due to risks of bias and directness discussed earlier, plusa lack of definitions of complications and possible selec-tive reporting.

Length of stay

Antibiotics were associated with a longer length of stay onPP analysis: mean difference 0.48 days (95% CI 0.10 to0.85 days; n ¼ 4 studies; p ¼ 0.01; Table 5; eFig. 7),but significance was dependent on 1 low quality study20

(GRADE recommendation very low).

Length of sick leave

Antibiotics were associated with a shorter sick leave on ITTand PP plus crossover analyses (mean difference -2.13 days(95% CI -3.85 to -0.41); n ¼ 4 studies; p ¼ 0.020), butnot on PP analysis (n ¼ 2 studies; p ¼ 0.070; GRADErecommendation very low; Table 5; eFig. 8).

Post-intervention pain

Vons and colleagues25 found no differences in the dura-tion of pain (visual analogue scale � 4), unlike Hanssonand associates,18 who reported longer duration with sur-gery. Meta-analysis was not possible. Two studies assessedthe visual analogue scale at variable time points, using

different scales. Salminen and coworkers13 reportedmore pain at discharge and at 1 week with surgery. Eriks-son and colleagues29 also found significantly lower visualanalogue scale scores at up to 10 days, associated withless morphine consumption.

Inflammatory response

Eriksson and colleagues29 reported a significant reductionin white cell count for up to 10 days with antibiotics,although C reactive protein levels were not different. Alower temperature was also found on days 1 and 2.

Incidental tumors

In 5 of 843 (0.59%) patients undergoing surgery, forwhom this outcome was reported, incidental tumorswere identified: 3 malignancies requiring right hemi-colectomy, 3 neuroendocrine tumors, and 1 adenomawith dysplasia.

Cost of treatment

Two studies reported costs of treatment, although for theprimary episode only. Hansson and associates18 reportedsignificantly lower costs with antibiotics vs surgery (17.7%reduction on ITT; 51.2% on a PP plus cross-over basis).No difference was reported by Turhan and coauthors.20

DISCUSSIONThere continues to be great interest in nonoperative man-agement of appendicitis, particularly because 1 recentmeta-analysis7 and subsequent RCT13 concluded antibi-otics to be as safe asdif not safer thandsurgery. In thissystematic review and meta-analysis of 6 RCTs, we foundantibiotics to successfully treat the initial episode of un-complicated acute appendicitis in 91% of patients. Thesepatients had a marginally longer length of hospital stay(although heavily influenced by 1 low quality study20),but a quicker return to work and less pain. However,

Table 3. Grading of Recommendations Assessment, Development, and Evaluation Analysis

Analysis Population Risk of bias Directness Consistency Precision Publication bias Overall quality

Successful treatment

ITT 6 studies; 1,724 patients High; very seriouslimitations*,y,z,x,k

Serious limitations{,# No limitations No limitations None detected Very low

PP 6 studies; 1,546 patients(1,440 at 1 y)

Moderate; seriouslimitationsy,z,k

Serious limitations{,# No limitations No limitations None detected Low

PP þ crossover 6 studies; 1,648 patients High; very seriouslimitations*,y,z,x,k

Serious limitationsx,k,{ No limitations No limitations None detected Very low

Complicated appendicitis

ITT 6 studies; 1,641 patients High; seriouslimitationsy,z,k **

Serious limitations{,# No limitations No limitations None detected Very low

PP 6 studies; 1,546 patients(1,440 at 1 y)

Moderate; seriouslimitationsy,z,k,**

Serious limitations{,# No limitations No limitations None detected Low

PP þ crossover 6 studies; 1,648 patients High; very seriouslimitations*,y,z,x,k,**

Serious limitationsx,k No limitations No limitations None detected Very low

Complications

Overall

ITT 6 studies; 1,724 patients High; very seriouslimitations*,y,z,x,k,yy

Serious limitationsx,{,# No limitations No limitations None detected Very low

PP 4 studies; 1,229 patients High; serious limitations# Serious limitations{,# No limitations No limitations None detected Very low

Minor and majorcomplications

ITT 5 studies; 1,468 patients High; very seriouslimitations*,y,z,x,k,yy,zz

Serious limitations{,# No limitations No limitations None detected Very low

PP 4 studies; 977 patients High; seriouslimitations*,y,z,x,k,yy

Serious limitation{,# No limitations No limitations None detected Very low

Length of stay and sickleave

ITT 6 studies; 1,724 patients High; very seriouslimitations*,y,z,x,k

Serious limitationsx,{,# No limitations No limitations None detected Very low

PP 4 studies; 768 patients High; very seriouslimitations*,y,z,x,k

Serious limitationsx,{,# No limitations No limitations None detected Very low

*Randomization by date of birth.18yUnclear randomization.20zVariable/incomplete follow-up.13,18xHigh protocol violations/cross-over/withdrawal rate.18e¼different diagnostic criteria between arms.20kVariable diagnostic inclusion criteria.{Variable antibiotic therapy.#Variable definitions of complicated appendicitis.**Undefined complications.yyPossible selective reporting of complications.zzLikely underlying differences in discharge criteria.

820

Findlayet

alNonoperative

Manage

mentofAppendicitis

JAm

CollSurg

Table 4. Risk of Complicated Appendicitis at Operation

Analysis Risk ratio (95% CI) n I2 Chi-square p Value Model GRADE quality

Cohort risk

Complicatedappendicitis atsurgery: initialepisode

ITT 0.44 (0.13e1.42); 0.38 (0.24e0.60) 5* 75 15.99 0.170; <0.0001 Random; fixed Very low

PP 0.48 (0.12e1.90); 0.41 (0.26e0.67) 5* 74 15.38 0.290; 0.003 Random; fixed Low

PP þ crossover 0.36 (0.11e1.16); 0.27 (0.18e0.41) 6 78 22.77 0.090; <0.0001 Random; fixed Very low

Complicatedappendicitis atsurgery at 1 y

ITT 0.66 (0.19e2.26); 0.62 (0.42e0.92) 5* 83 23.01 0.510; 0.020 Random; fixed Very low

PP 0.68 (0.17e2.77); 0.66 (0.44e0.99) 5* 80 20.12 0.590; 0.040 Random; fixed Low

Risk if undergoingoperation

Complicatedappendicitis atsurgery at initialepisode

ITT 4.14 (1.23e13.96) 5* 86 29.23 0.020 Random Very low

PP 3.96 (1.13e13.91) 5* 82 21.74 0.030 Random Low

PP þ crossover 3.13 (1.20e8.13) 6 81 26.67 0.020 Random Very low

Complicatedappendicitis atsurgery at 1 y

ITT 2.18 (0.60e7.92); 2.30 (1.60e3.31) 5* 86 28.18 0.230; <0.0001 Random; fixed Very low

PP 2.23 (0.50e9.93) 5* 84 25.56 0.290; <0.0001 Random; fixed Low

Complicatedappendicitis atsurgery, recurrentepisode

ITT 1.46 (0.31e6.75) 5* 83 22.91 0.620 Random Very low

PP 1.56 (0.22e11.25) 5* 81 20.73 0.660 Random Low

PP þ crossover 1.38 (0.35e5.41) 6 78 22.89 0.650 Random Very low

*Excluding Hansson and colleagues,18 2009.GRADE, Grading of Recommendations Assessment, Development, and Evaluation; ITT, intention-to-treat group; PP, per protocol group.

Vol. 223, No. 6, December 2016 Findlay et al Nonoperative Management of Appendicitis 821

patients in whom antibiotics failed were much more likelyto have complicated appendicitis, and we saw no signifi-cant difference in complications at 1 year. Although therewas a nonsignificant trend toward fewer minor complica-tions with antibiotics, this was heavily influenced bysurprisingly high rates of undefined and subjective com-plications in some studies.13,18 After the initial episode,there was a substantial risk of recurrence within a year,meaning that antibiotics failed cumulatively in 29% at1 year. Importantly, the risk thereafter is unknown.Unsuspected tumors were identified in 0.6% of patientstreated by surgery, which would not have been identifiedotherwise. Overall, due to potential bias and considerableheterogeneity, these findings were of low and very low

GRADE quality. Ultimately, however, the control groupof appendectomy was poorly representative of currentpractice, primarily consisting of open rather than laparo-scopic procedures.Our findings are similar in some respects to those in pre-

vious meta-analyses. The 2011 Cochrane review under-taken by Wilms and colleagues9 concluded similarly thatthe evidence base was insufficient to recommend nonoper-ative management. By contrast, other meta-analyses7,30

concluded antibiotics to be effective and associated withfewer complications, albeit with a high recurrence rate.Mason and associates30 also found less pain and a shorterlength of stay. However, we found no such significantreduction in complications, despite including the recent

Table 5. Risk of Complications at 1 Year, Length of Stay, and Sick Leave

Analysis Risk ratio (95% CI) n I2 Chi-square p Value Model GRADE

Cohort risk, allcomplications (majorand minor)

ITT 0.53 (0.21e1.33); 0.51 (0.37e0.70) 6 80 24.45 0.170; <0.0001* Random; fixed Very low

PP 0.41 (0.130e1.30); 0.27 (0.17e0.44) 5y 76 16.40 0.130; 0.0001* Random; fixed Very low

Cohort risk, minorcomplications

ITT 0.59 (0.19e1.84); 0.52 (0.36e0.70) 5 82 27.10 0.360; 0.0004* Random; fixed Very low

PP 0.56 (0.09e3.34); 0.28 (0.16e0.50) 4z 79 14.50 0.530; <0.0001* Random; fixed Very low

Cohort risk, majorcomplications

ITT 0.55 (0.28e1.09) 5 0 2.08 0.090 Random Very low

PP 0.47 (0.08e2.65) 3* 32 2.93 0.390 Random Very low

Length of hospital stay

ITT 0.20 (e0.11e0.51); 0.27 (0.18e0.35) 6 71 17.09 0.200; <0.0001 Random; fixed Very low

PP 0.48 (0.10e0.85); 0.52 (0.30e0.74) 4x 52 6.30 0.010; <0.0001 Randomk; fixed Very low

PP þ crossover 0.11 (e0.61e0.82); 0.21 (0.01e0.41) 4x 91 43.89 0.770; 0.040 Random; fixed Very low

Length of sick leave

ITT �2.13 (e3.85 - e0.41) 4 39 3.26 0.020 Random Very low

PP �1.47 (�3.08 � 0.13) 2 7 1.07 0.070 Random Very low

PP þ crossover �2.43 (�4.73 - �0.07) 3 65 5.77 0.040 Random Very low

GRADE, Grading of Recommendations Assessment, Development, and Evaluation; ITT, intention-to-treat group; PP, per protocol group.*Significance due to inclusion of Salminen and colleagues,13 2015.yExcluding Hansson and colleagues,18 2009.zExcluding Hansson and colleagues,18 2009; and Styrud and colleagues,19 2006xVons and colleagues,25 2011; Hansson and colleagues,18 2009; Styrud and colleagues,19 2006; and Eriksson and colleagues,26 1995.kDependent upon Turhan and colleagues,20 2009.

822 Findlay et al Nonoperative Management of Appendicitis J Am Coll Surg

study by Salminen and coworkers,13 which reported a sig-nificant reduction with antibiotics. This may be explainedsomewhat by differences in reporting and inclusion (weincluded all complications reported), and also bymethodo-logic and statistical heterogeneity. There were also signifi-cant differences in definitions of complications and whatpreventative measures were taken.This is illustrated by the study by Salminen and col-

leagues,13 which reported extremely high rates (8.5% foreach) of wound infection and postoperative symptomsattributed to adhesions. However, 94.5% of procedureswere performed open (substantially increasing the riskof both27), postoperative antibiotics were not used evenin the presence of contamination, and no details were pro-vided (in any studies) as to methods of wound protection,irrigation, specimen extraction, and closure. Similarly,diagnostic criteria were not provided, and severity wasnot stratified; these issues also applied to presumed adhe-sional symptoms.Similarly, although we saw a longer length of stay with

antibiotic therapy, this was strongly influenced by 1 study20

stipulating a longer (3-day) admission for antibiotics. Thisstudy was notable due to its lack of detail regarding

randomization, which resulted in its exclusion byVaradhanand coauthors7; we included it with caution (as did Wilmsand coworkers9) because it was explicitly described as anRCT, and sensitivity analyses showed it did not signifi-cantly alter our findings, other than for this length of stayanalysis. However, no studies provided data regarding theadditive stay and sick leave of patients with recurrentappendicitis, to allow a cumulative comparison at 1 year.Additionally, although we found that surgery was associ-ated with more pain in the first week, no studies provideddetails regarding the use of perioperative analgesia orregional blockade. These issues are reflected in the levelof evidence and recommendation being “very low” forour secondary endpoints; ie we are very uncertain aboutthe estimate.At 1 year, more than two-thirds of patients had been

successfully treated with antibiotics, but this cannot befully interpreted without data regarding recurrence there-after. It is plausible that patients suffering appendicitis aremore prone to the condition, and therefore run a substan-tially increased risk of recurrence.4 Long-term results ofthese and future RCTs are therefore crucial before anti-biotic therapy can be recommended.

Vol. 223, No. 6, December 2016 Findlay et al Nonoperative Management of Appendicitis 823

Beyond the limitations of individual studies, there arethose specific to this meta-analysis. We used an inclusivesearch strategy in order to identify all relevant articles;however, it is possible that some were missed. Statistically,we used a different endpoint to define “successful” treat-ment than did other studies. We defined it pragmatically,as resolution of the acute episode of presumed inflamma-tion without subsequent recurrence. In contrast to othermeta-analyses, we believed that complications may occurdespite complete resolution of the primary ileocecal in-flammatory process, while the presence of abdominalpain or inflammation is similarly complex and may notrelate to any ongoing ileocecal inflammation. Similarly,we did not exclude patients undergoing negative appen-dectomies from analysis because this does not representthe population of interest on an ITT basis and wouldbias analysis against surgery because no such mechanismexists for identifying false positives in patients treatedwith antibiotics.

CONCLUSIONSWe found that overall, nonoperative management ofpatients with uncomplicated appendicitis was successfulin the immediate term in 91%, but that this reduced to71% at 1 year. The risk thereafter is unknown. In addi-tion, we found that nonoperative management was associ-ated with less pain in the first week after treatment and aquicker return to work, but would have missed occulttumors in a small number of patients. We found noconvincing evidence of a reduction in complications.Selective primary antibiotic therapy can be discussedwith patients on this basis, although with the caveat thatthese data relate primarily to open procedures, and derivefrom a poor evidence base overall, with numerous areas ofbias. Laparoscopic appendectomy, therefore, remains thetreatment of choice overall for appendicitis, and has yetto be compared with antibiotics by an RCT. FutureRCTs should, in addition, focus on standardized outcomesreporting and study methodology, and investigating mech-anisms for more personalized use of antibiotics or surgery.

Author Contributions

Study conception and design: Findlay, HammerAcquisition of data: Findlay, Hammer, El KafsiAnalysis and interpretation of data: Findlay, Hammer, ElKafsi, Gilmour, Gillies, Maynard

Drafting of manuscript: Findlay, Hammer, El Kafsi,Gilmour, Gillies, Maynard

Critical revision: Findlay, Hammer, El Kafsi, Gilmour,Gillies, Maynard

REFERENCES

1. Hospital Episode Statistics Analysis, Health and Social CareInformation Centre. Hospital Episode Statistics, AdmittedPatient Care - England, 2014-15. Health and Social CareInformation Center; 2015. Available at: http://content.digital.nhs.uk/catalogue/PUB19124.

2. Addiss DG, Shaffer N, Fowler BS, et al. The epidemiology ofappendicitis and appendectomy in the United States. Am JEpidemiol 1990;132:910e925.

3. Kim K, Kim YH, Kim SY, et al. Low-dose abdominal CT forevaluating suspected appendicitis. N Engl J Med 2012;366:1596e1605.

4. Bhangu A, Soreide K, Di Saverio S, et al. Acute appendicitis:modern understanding of pathogenesis, diagnosis, and man-agement. Lancet 2015;386:1278e1287.

5. Campbell MR, Johnston SL 3rd, Marshburn T, et al. Nonop-erative treatment of suspected appendicitis in remote medicalcare environments: implications for future spaceflight medicalcare. J Am Coll Surg 2004;198:822e830.

6. Senapathi PS, Bhattacharya D, Ammori BJ. Early laparoscopicappendectomy for appendicular mass. Surg Endosc 2002;16:1783e1785.

7. Varadhan KK, Neal KR, Lobo DN. Safety and efficacy of an-tibiotics compared with appendicectomy for treatment of un-complicated acute appendicitis: meta-analysis of randomisedcontrolled trials. BMJ 2012;344:e2156.

8. Rocha LL, Rossi FM, Pessoa CM, et al. Antibiotics aloneversus appendectomy to treat uncomplicated acute appendi-citis in adults: what do meta-analyses say? World J EmergSurg 2015;10:51.

9. Wilms IM, de Hoog DE, de Visser DC, et al. Appendectomyversus antibiotic treatment for acute appendicitis. CochraneDatabase Syst Rev 2011;[11]:CD008359.

10. Ireland AoSoGBa. Commissioning Guide: Emergency GeneralSurgery (Acute Abdominal Pain). London: Royal College ofSurgeons of England; 2014.

11. Surgeons SoAGaE. Guidelines for Laparoscopic Appendec-tomy. Los Angeles: Society of American Gastrointestinal andEndoscopic Surgeons; 2009.

12. Sauerland S, Agresta F, Bergamaschi R, et al. Laparoscopy forabdominal emergencies: evidence-based guidelines of the Eu-ropean Association for Endoscopic Surgery. Surg Endosc2006;20:14e29.

13. Salminen P, Paajanen H, Rautio T, et al. Antibiotic therapy vsappendectomy for treatment of uncomplicated acute appendi-citis: The APPAC randomized clinical trial. JAMA 2015;313:2340e2348.

14. Pollara G, Marks M. Treatment of uncomplicated acuteappendicitis. JAMA 2015;314:1402.

15. Livingston E, Vons C. Treating appendicitis without surgery.JAMA 2015;313:2327e2328.

16. Moher D, Liberati A, Tetzlaff J, et al. Preferred reporting itemsfor systematic reviews and meta-analyses: the PRISMA state-ment. BMJ 2009;339:b2535.

17. Stroup DF, Berlin JA, Morton SC, et al. Meta-analysis ofobservational studies in epidemiology: a proposal for report-ing. Meta-analysis Of Observational Studies in Epidemiology(MOOSE) group. JAMA 2000;283:2008e2012.

18. Hansson J, Korner U, Khorram-Manesh A, et al. Randomizedclinical trial of antibiotic therapy versus appendicectomy as

824 Findlay et al Nonoperative Management of Appendicitis J Am Coll Surg

primary treatment of acute appendicitis in unselected patients.Br J Surg 2009;96:473e481.

19. Styrud J, Eriksson S, Nilsson I, et al. Appendectomy versusantibiotic treatment in acute appendicitis. a prospective multi-center randomized controlled trial. World J Surg 2006;30:1033e1037.

20. Turhan AN, Kapan S, Kutukcu E, et al. Comparison of oper-ative and non operative management of acute appendicitis.Ulus Travma Acil Cerrahi Derg [Turkish J Trauma EmergSurg] 2009;15:459e462.

21. Wan X, Wang W, Liu J, Tong T. Estimating the sample meanand standard deviation from the sample size, median, rangeand/or interquartile range. BMC Medical Research Methodol-ogy 2014;14:135.

22. Review Manager (program). 5.3 version. Copenhagen: TheCochrane Collaboration; 2014.

23. R Core Team. R: A language and environment for statisticalcomputing. Vienna, Austria: R Foundation for StatisticalComputing; 2013.

24. Malik AA, Bari SU. Conservative management of acute appen-dicitis. J Gastrointest Surg 2009;13:966e970.

25. Vons C, Barry C, Maitre S, et al. Amoxicillin plus clavulanicacid versus appendicectomy for treatment of acute uncompli-cated appendicitis: an open-label, non-inferiority, randomisedcontrolled trial. Lancet 2011;377:1573e1579.

26. Eriksson S, Granstrom L. Randomized controlled trial ofappendicectomy versus antibiotic therapy for acute appendi-citis. Br J Surg 1995;82:166e169.

27. Sauerland S, Jaschinski T, Neugebauer EA. Laparoscopicversus open surgery for suspected appendicitis. CochraneDatabase Syst Rev 2010;[10]:CD001546.

28. Wei B, Qi CL, Chen TF, et al. Laparoscopic versus openappendectomy for acute appendicitis: a metaanalysis. SurgEndosc 2011;25:1199e1208.

29. Eriksson S, Josephson T, Styrud J. A high degree of accuracy ispossible in the diagnosis of appendicitis. Laboratory tests,ultrasonography and computerized tomography are of greatvalue. Lakartidningen 1999;96:3058e3061.

30. Mason RJ, Moazzez A, Sohn H, et al. Meta-analysis of ran-domized trials comparing antibiotic therapy with appendec-tomy for acute uncomplicated (no abscess or phlegmon)appendicitis. Surg Infect 2012;13:74e84.

Vol. 223, No. 6, December 2016 Findlay et al Nonoperative Management of Appendicitis 824.e1

APPENDIX

eFigure 1. Index admission risk of complicated appendicitis (per protocol analysis).

eFigure 2. Two-year risk of complicated appendicitis (per protocol analysis).

eFigure 3. Index admission risk of complicated appendicitis for patients operated on (per protocol analysis).

eFigure 4. One-year complications (per protocol analysis).

eFigure 5. One-year risk of minor complications (per protocol analysis).

eFigure 6. One-year risk of major complications (per protocol analysis).

eFigure 7. Length of stay (per protocol analysis).

eFigure 8. Length of sick leave (per protocol analysis).

824.e2 Findlay et al Nonoperative Management of Appendicitis J Am Coll Surg