University of Groningen
Laparoscopic versus small-incision cholecystectomy for patients with symptomatic cholecystolithiasis
Keus, F.; de Jong, J. A. F.; Gooszen, H. G.; van Laarhoven, C. J. H. M.
Published in:
Cochrane database of systematic reviews (Online)
DOI:
10.1002/14651858.CD006229
IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below.
Document Version
Publisher's PDF, also known as Version of record
Publication date:
2006
Link to publication in University of Groningen/UMCG research database
Citation for published version (APA):
Keus, F., de Jong, J. A. F., Gooszen, H. G., & van Laarhoven, C. J. H. M. (2006). Laparoscopic versus small-incision cholecystectomy for patients with symptomatic cholecystolithiasis. Cochrane database of systematic reviews (Online), (4), [006229]. https://doi.org/10.1002/14651858.CD006229
Copyright
Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).
The publication may also be distributed here under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license.
More information can be found on the University of Groningen website: https://www.rug.nl/library/open-access/self-archiving-pure/taverne- amendment.
Take-down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately
CochraneDatabase of Systematic Reviews
Laparoscopic versus small-incision cholecystectomy for patients with symptomatic cholecystolithiasis (Review)
Keus F, de Jong J, Gooszen HG, van Laarhoven CHJM
Keus F, de Jong J, Gooszen HG, van Laarhoven CHJM.
Laparoscopic versus small-incision cholecystectomy for patients with symptomatic cholecystolithiasis.
Cochrane Database of Systematic Reviews 2006, Issue 4. Art. No.: CD006229.
DOI: 10.1002/14651858.CD006229.
www.cochranelibrary.com
T A B L E O F C O N T E N T S
1 HEADER . . . .
1 ABSTRACT . . . .
2 PLAIN LANGUAGE SUMMARY . . . .
2 BACKGROUND . . . .
3 OBJECTIVES . . . .
3 METHODS . . . .
6 RESULTS . . . .
Figure 1. . . 7
Figure 2. . . 8
Figure 3. . . 11
Figure 4. . . 12
Figure 5. . . 13
Figure 6. . . 14
14 DISCUSSION . . . . 17 AUTHORS’ CONCLUSIONS . . . . 17 ACKNOWLEDGEMENTS . . . . 17 REFERENCES . . . . 26 CHARACTERISTICS OF STUDIES . . . . 47 DATA AND ANALYSES . . . . Analysis 1.1. Comparison 1 LC versus SIC - high-quality and low-quality trials regarding generation of the allocation sequence, Outcome 1 Mortality. . . 52
Analysis 1.2. Comparison 1 LC versus SIC - high-quality and low-quality trials regarding generation of the allocation sequence, Outcome 2 Intra-operative complications. . . 54
Analysis 1.3. Comparison 1 LC versus SIC - high-quality and low-quality trials regarding generation of the allocation sequence, Outcome 3 Minor complications. . . 55
Analysis 1.4. Comparison 1 LC versus SIC - high-quality and low-quality trials regarding generation of the allocation sequence, Outcome 4 Severe complications (without bile duct injuries). . . 56
Analysis 1.5. Comparison 1 LC versus SIC - high-quality and low-quality trials regarding generation of the allocation sequence, Outcome 5 Bile duct injuries. . . 57
Analysis 1.6. Comparison 1 LC versus SIC - high-quality and low-quality trials regarding generation of the allocation sequence, Outcome 6 Total complications. . . 58
Analysis 1.7. Comparison 1 LC versus SIC - high-quality and low-quality trials regarding generation of the allocation sequence, Outcome 7 Conversions. . . 59
Analysis 1.8. Comparison 1 LC versus SIC - high-quality and low-quality trials regarding generation of the allocation sequence, Outcome 8 Operative time (minutes). . . 60
Analysis 1.9. Comparison 1 LC versus SIC - high-quality and low-quality trials regarding generation of the allocation sequence, Outcome 9 Hospital stay (days). . . 61
Analysis 1.10. Comparison 1 LC versus SIC - high-quality and low-quality trials regarding generation of the allocation sequence, Outcome 10 Convalescence: work leave (days). . . 62
Analysis 1.11. Comparison 1 LC versus SIC - high-quality and low-quality trials regarding generation of the allocation sequence, Outcome 11 Convalescence: normal activity (at home) (days). . . 63
Analysis 2.1. Comparison 2 LC versus SIC - high-quality and low-quality trials regarding allocation concealment, Outcome 1 Mortality. . . 64
Analysis 2.2. Comparison 2 LC versus SIC - high-quality and low-quality trials regarding allocation concealment, Outcome 2 Intra-operative complications. . . 65
Analysis 2.3. Comparison 2 LC versus SIC - high-quality and low-quality trials regarding allocation concealment, Outcome 3 Minor complications. . . 66
Analysis 2.4. Comparison 2 LC versus SIC - high-quality and low-quality trials regarding allocation concealment, Outcome 4 Severe complications (without bile duct injuries). . . 67 Analysis 2.5. Comparison 2 LC versus SIC - high-quality and low-quality trials regarding allocation concealment, Outcome
Analysis 2.6. Comparison 2 LC versus SIC - high-quality and low-quality trials regarding allocation concealment, Outcome 6 Total complications. . . 69 Analysis 2.7. Comparison 2 LC versus SIC - high-quality and low-quality trials regarding allocation concealment, Outcome
7 Conversions. . . 70 Analysis 2.8. Comparison 2 LC versus SIC - high-quality and low-quality trials regarding allocation concealment, Outcome
8 Operative time (minutes). . . 71 Analysis 2.9. Comparison 2 LC versus SIC - high-quality and low-quality trials regarding allocation concealment, Outcome
9 Hospital stay (days). . . 72 Analysis 2.10. Comparison 2 LC versus SIC - high-quality and low-quality trials regarding allocation concealment,
Outcome 10 Convalescence: work leave (days). . . 73 Analysis 2.11. Comparison 2 LC versus SIC - high-quality and low-quality trials regarding allocation concealment,
Outcome 11 Convalescence: normal activity (at home) (days). . . 74 Analysis 3.1. Comparison 3 LC versus SIC - high-quality and low-quality trials regarding blinding, Outcome 1 Mortality. 75 Analysis 3.2. Comparison 3 LC versus SIC - high-quality and low-quality trials regarding blinding, Outcome 2 Intra-
operative complications. . . 76 Analysis 3.3. Comparison 3 LC versus SIC - high-quality and low-quality trials regarding blinding, Outcome 3 Minor
complications. . . 77 Analysis 3.4. Comparison 3 LC versus SIC - high-quality and low-quality trials regarding blinding, Outcome 4 Severe
complications (without bile duct injuries). . . 78 Analysis 3.5. Comparison 3 LC versus SIC - high-quality and low-quality trials regarding blinding, Outcome 5 Bile duct
injuries. . . 79 Analysis 3.6. Comparison 3 LC versus SIC - high-quality and low-quality trials regarding blinding, Outcome 6 Total
complications. . . 80 Analysis 3.8. Comparison 3 LC versus SIC - high-quality and low-quality trials regarding blinding, Outcome 8 Operative
time (minutes). . . 81 Analysis 3.9. Comparison 3 LC versus SIC - high-quality and low-quality trials regarding blinding, Outcome 9 Hospital
stay (days). . . 82 Analysis 3.10. Comparison 3 LC versus SIC - high-quality and low-quality trials regarding blinding, Outcome 10
Convalescence: work leave (days). . . 83 Analysis 3.11. Comparison 3 LC versus SIC - high-quality and low-quality trials regarding blinding, Outcome 11
Convalescence: normal activity (at home) (days). . . 84 Analysis 4.1. Comparison 4 LC versus SIC - high-quality and low-quality trials regarding follow-up, Outcome 1
Mortality. . . 85 Analysis 4.2. Comparison 4 LC versus SIC - high-quality and low-quality trials regarding follow-up, Outcome 2 Intra-
operative complications. . . 86 Analysis 4.3. Comparison 4 LC versus SIC - high-quality and low-quality trials regarding follow-up, Outcome 3 Minor
complications. . . 87 Analysis 4.4. Comparison 4 LC versus SIC - high-quality and low-quality trials regarding follow-up, Outcome 4 Severe
complications (without bile duct injuries). . . 88 Analysis 4.5. Comparison 4 LC versus SIC - high-quality and low-quality trials regarding follow-up, Outcome 5 Bile duct
injuries. . . 89 Analysis 4.6. Comparison 4 LC versus SIC - high-quality and low-quality trials regarding follow-up, Outcome 6 Total
complications. . . 90 Analysis 4.7. Comparison 4 LC versus SIC - high-quality and low-quality trials regarding follow-up, Outcome 7
Conversions. . . 91 Analysis 4.8. Comparison 4 LC versus SIC - high-quality and low-quality trials regarding follow-up, Outcome 8 Operative
time (minutes). . . 92 Analysis 4.9. Comparison 4 LC versus SIC - high-quality and low-quality trials regarding follow-up, Outcome 9 Hospital
stay (days). . . 93 Analysis 4.10. Comparison 4 LC versus SIC - high-quality and low-quality trials regarding follow-up, Outcome 10
Convalescence: work leave (days). . . 94
Analysis 5.1. Comparison 5 LC versus SIC - sensitivity and subgroup analyses, Outcome 1 Sensitivity analysis 1: Assuming zero mortality in nonreporting trials. . . 96 Analysis 5.2. Comparison 5 LC versus SIC - sensitivity and subgroup analyses, Outcome 2 Sensitivity analysis 2: Assuming
zero conversions in nonreporting trials. . . 97 Analysis 5.3. Comparison 5 LC versus SIC - sensitivity and subgroup analyses, Outcome 3 Sensitivity analysis 3: Imputing
medians and standard deviations for missing data in operative time (minutes). . . 98 Analysis 5.4. Comparison 5 LC versus SIC - sensitivity and subgroup analyses, Outcome 4 Sensitivity analysis 4: Imputing
medians and standard deviations for missing data in hospital stay (days). . . 99 Analysis 5.5. Comparison 5 LC versus SIC - sensitivity and subgroup analyses, Outcome 5 Sensitivity analysis 5: Imputing
medians and standard deviations for missing data in convalescence: work leave. . . 100 Analysis 5.6. Comparison 5 LC versus SIC - sensitivity and subgroup analyses, Outcome 6 Sensitivity analysis 6: Imputing
medians and standard deviations for missing data in normal activity. . . 101 Analysis 5.7. Comparison 5 LC versus SIC - sensitivity and subgroup analyses, Outcome 7 Sensitivity analysis 7: Omitting
outlier Srivastava in minor complications. . . 102 Analysis 5.8. Comparison 5 LC versus SIC - sensitivity and subgroup analyses, Outcome 8 Sensitivity analysis 8: Omitting
outlier Srivastava in total complications. . . 103 Analysis 5.9. Comparison 5 LC versus SIC - sensitivity and subgroup analyses, Outcome 9 Sensitivity analysis 9: Omitting
outlier Grande in hospital stay (days). . . 104 Analysis 5.10. Comparison 5 LC versus SIC - sensitivity and subgroup analyses, Outcome 10 Sensitivity analysis 10: Total
complications including Redmond. . . 105 Analysis 5.11. Comparison 5 LC versus SIC - sensitivity and subgroup analyses, Outcome 11 Sensitivity analysis 11:
Operative time (minutes) including Redmond. . . 106 Analysis 5.12. Comparison 5 LC versus SIC - sensitivity and subgroup analyses, Outcome 12 Subgroup analysis 1: Influence
antibiotic prophylaxis on total complications. . . 107 Analysis 5.13. Comparison 5 LC versus SIC - sensitivity and subgroup analyses, Outcome 13 Subgroup analysis 2: Influence
surgical experience on total complications. . . 108 Analysis 5.14. Comparison 5 LC versus SIC - sensitivity and subgroup analyses, Outcome 14 Subgroup analysis 3: Influence
cholangiography on operative time (minutes). . . 109 Analysis 5.15. Comparison 5 LC versus SIC - sensitivity and subgroup analyses, Outcome 15 Subgroup analysis 4: Influence
surgical experience on operative time (minutes). . . 110 Analysis 5.16. Comparison 5 LC versus SIC - sensitivity and subgroup analyses, Outcome 16 Subgroup analysis 5: Influence
antibiotic prophylaxis on hospital stay (days). . . 111 Analysis 5.17. Comparison 5 LC versus SIC - sensitivity and subgroup analyses, Outcome 17 Subgroup analysis 6: Influence
surgical experience on hospital stay (days). . . 112 Analysis 5.18. Comparison 5 LC versus SIC - sensitivity and subgroup analyses, Outcome 18 Subgroup analysis 7: Influence
cholangiography on bile duct injuries. . . 113 113 ADDITIONAL TABLES . . . .
124 APPENDICES . . . .
125 WHAT’S NEW . . . .
126 CONTRIBUTIONS OF AUTHORS . . . .
126 DECLARATIONS OF INTEREST . . . .
126 SOURCES OF SUPPORT . . . .
126 NOTES . . . .
126 INDEX TERMS . . . .
[Intervention Review]
Laparoscopic versus small-incision cholecystectomy for patients with symptomatic cholecystolithiasis
Frederik Keus1, Jeroen de Jong2, Hein G Gooszen3, CHJM van Laarhoven4
1Surgery, Diakonessenhuis, Utrecht, Netherlands. 2University Medical Center, Utrecht, Netherlands. 3Department of Surgery G.04.228, University Medical Center, GA Utrecht, Netherlands.4Department of Surgery, University Medical Centre St Radboud Nijmegen, Nijmegen, Netherlands
Contact address: Frederik Keus, Surgery, Diakonessenhuis, Bosboomstraat 1, Utrecht, Utrecht, 3582 KE, Netherlands.
erickeus@hotmail.com.
Editorial group: Cochrane Hepato-Biliary Group.
Publication status and date: Edited (no change to conclusions), published in Issue 1, 2010.
Review content assessed as up-to-date: 10 August 2006.
Citation: Keus F, de Jong J, Gooszen HG, van Laarhoven CHJM. Laparoscopic versus small-incision cholecystectomy for pa- tients with symptomatic cholecystolithiasis. Cochrane Database of Systematic Reviews 2006, Issue 4. Art. No.: CD006229. DOI:
10.1002/14651858.CD006229.
Copyright © 2010 The Cochrane Collaboration. Published by John Wiley & Sons, Ltd.
A B S T R A C T Background
Cholecystectomy is one of the most frequently performed operations. Open cholecystectomy has been the gold standard for over 100 years. Small-incision cholecystectomy is a less frequently used alternative. Laparoscopic cholecystectomy was introduced in the 1980s.
Objectives
To compare the beneficial and harmful effects of laparoscopic versus small-incision cholecystectomy for patients with symptomatic cholecystolithiasis.
Search methods
We searchedThe Cochrane Hepato-Biliary Group Controlled Trials Register (6 April 2004), The Cochrane Library (Issue 1, 2004), MEDLINE (1966 to January 2004), EMBASE (1980 to January 2004), Web of Science (1988 to January 2004), and CINAHL (1982 to January 2004) for randomised trials.
Selection criteria
All published and unpublished randomised trials in patients with symptomatic cholecystolithiasis comparing any kind of laparoscopic cholecystectomy versus small-incision or other kind of minimal incision open cholecystectomy. No language limitations were applied.
Data collection and analysis
Two authors independently performed selection of trials and data extraction. The methodological quality of the generation of the allocation sequence, allocation concealment, blinding, and follow-up was evaluated to assess bias risk. Analyses were based on the
Main results
Thirteen trials randomised 2337 patients. Methodological quality was relatively high considering the four quality criteria. Total complications of laparoscopic and small-incision cholecystectomy are high: 26.6% versus 22.9%. Total complications (risk difference, random-effects -0.01, 95% confidence interval (CI) -0.07 to 0.05), hospital stay (weighted mean difference (WMD), random-effects - 0.72 days, 95% CI -1.48 to 0.04), and convalescence were not significantly different. High-quality trials show a quicker operative time for small-incision cholecystectomy (WMD, high-quality trials ’blinding’, random-effects 16.4 minutes, 95% CI 8.9 to 23.8) while low-quality trials show no significant difference.
Authors’ conclusions
Laparoscopic and small-incision cholecystectomy seem to be equivalent. No differences could be observed in mortality, complications, and postoperative recovery. Small-incision cholecystectomy has a significantly shorter operative time. Complications in elective chole- cystectomy are prevalent.
P L A I N L A N G U A G E S U M M A R Y
Laparoscopic and small-incision cholecystectomy seem equivalent in complications and recovery, but small-incision cholecys- tectomy is quicker to perform
The laparoscopic and the small-incision cholecystectomy are two alternative minimally invasive techniques for removal of the gallbladder.
There are no significant differences in mortality and complications between the two minimal invasive procedures. The laparoscopic and the small-incision operation should be considered equal apart from a quicker operative time using the small-incision technique.
The complications in both techniques are common.
B A C K G R O U N D
Gallstones are one of the major causes of morbidity in western society. It is estimated that the incidence of symptomatic chole- cystolithiasis is up to 2.17 per thousand inhabitants (Legorreta 1993;Steiner 1994) with an annual performance rate of chole- cystectomies of more than 500,000 in USA (Olsen 1991;NIH Consensus 1993;Roslyn 1993). Until the end of the 1980s, open cholecystectomy was the gold standard for treatment of stones in the gallbladder. As incisions for cholecystectomy were short- ened resulting in ’small-incision’ cholecystectomy, morbidity and complications seemed to decline and patients recovered faster. In the early 1970s small-incision cholecystectomy was introduced as a minimal invasive procedure (Dubois 1982;Goco 1983). This technique was introduced in order to decrease surgical trauma and consequently accelerate convalescence. Conflicting data on clin- ical outcome and effectiveness arose from studies evaluating this technique.
Laparoscopic cholecystectomy was introduced in 1985 (Mühe 1986) and rapidly became the method of choice for surgical re- moval of the gallbladder (NIH Consensus 1993) although the
absent. This rising popularity was based on many arguments, in- cluding assumed lower morbidity and complication proportions, and a quicker postoperative recovery compared to open chole- cystectomy and despite an increase in bile duct lesions. Laparo- scopic cholecystectomy seemed superior to small-incision chole- cystectomy (Dubois 1982;Goco 1983;Moss 1986;Merrill 1988;
Ledet 1990;O’Dwyer 1990;O’Dwyer 1992;Assalia 1993a;Olsen 1993; Tyagi 1994; Downs 1996;Assalia 1997;Schmitz 1997;
Seale 1999). However, the aforementioned studies were non-ran- domised observations, which may not provide an adequate assess- ment of intervention effects.
Differences in primary outcomes like mortality and complication proportions (particularly bile duct injuries) are important reasons to choose one of the operative techniques. When these primary outcomes show no significant difference, then secondary outcomes like non-severe complications, pulmonary outcomes, differences in health status related quality-of-life, hospital stay, and differences in cost-effectiveness analysis should help decide which technique is superior.
Up to now, despite the availability of numerous randomised trials
has been conducted in order to compare laparoscopic and small- incision cholecystectomy. This lack of evidence was the main rea- son for writing this systematic review. The objective was to evalu- ate the assumed superiority of the laparoscopic cholecystectomy.
O B J E C T I V E S
To evaluate the beneficial and harmful effects of two different types of cholecystectomy for patients with symptomatic cholecystolithi- asis. To assess whether laparoscopic and small-incision cholecys- tectomy are different in terms of primary (mortality, complica- tions, and relief of symptoms) and secondary outcomes (conver- sions to open cholecystectomy, operative time, hospital stay, and convalescence). If data were present, differences in other secondary outcomes like analgesic use, postoperative pain, pulmonary func- tion, and costs were compared as well.
M E T H O D S
Criteria for considering studies for this review
Types of studies
All randomised clinical trials comparing laparoscopic cholecystec- tomy to small-incision or other kind of minimal-incision open cholecystectomy. Trials were included irrespectively of blinding, number of patients randomised, and language of the article. Quasi- randomised studies were excluded.
Types of participants
Patients with one or more stones in the gallbladder confirmed by ultrasonography or other imaging technique and symptoms at- tributable to them, scheduled for cholecystectomy. Acute chole- cystitis is a disease with different operative results including the number of complications and conversions. Cholecystectomy in patients suffering from acute cholecystitis should be distinguished from cholecystectomy in patients suffering from symptomatic cholecystolithiasis. Therefore, randomised trials only including patients with acute cholecystitis were excluded from this review.
Randomised trials including both symptomatic cholecystolithiasis and acute cholecystitis were included in the review only if the large majority (more than half ) of the included patients were operated
Types of interventions
Any kind of laparoscopic cholecystectomy was assessed versus any kind of small-incision cholecystectomy.
The following classifications of the surgical procedures (based on intention-to-treat) were used:
Laparoscopic cholecystectomy includes those procedures started as a laparoscopic procedure. Any kind of laparoscopic cholecystec- tomy with creation of a pneumoperitoneum (by Veress needle or open introduction) or mechanical abdominal wall lift, irrespective of the number of trocars used.
Only if the words ’small-incision’, ’minimal access’, ’minilaparo- tomy’ or similar as intended terms were mentioned in the primary classification of the procedure, the surgical intervention was clas- sified as a ’small-incision’ cholecystectomy (eg, length of incision of less than 8 cm). The length of incision up to 8 cm was chosen arbitrary as in literature most authors used this length as a cut-off point between small-incision and (conversion to) open cholecys- tectomy.
In all other cases the surgical intervention was classified as ’open cholecystectomy’ and was excluded.
Types of outcome measures
The primary outcome measures are mortality, complication pro- portions (intra-operative, severe, bile duct injuries, and total com- plications; except minor complications), and relief of symptoms (pain relief ). Although relief of symptoms is the aim of cholecys- tectomy, some patients continue to suffer from their complaints and have persistent pain. Most important in obtaining a high pro- portion of patients with relief of symptoms is adequate decision making in setting the indication to operate or not. However, it cannot be ruled out that some of this persistent pain should be attributed to the way of incision for cholecystectomy. Therefore it is of interest to include pain relief as a primary outcome.
Secondary outcome measures are all other outcomes assessed in comparing the two operative techniques. We assessed the follow- ing secondary outcomes: conversion proportions to open chole- cystectomy, operative time, hospital stay, convalescence, analgesic use, postoperative pain (visual analogue scale), health status related quality-of-life, pulmonary outcome (pulmonary function tests by flow-volume curves), and cost-effectiveness if data were available.
Search methods for identification of studies We searched the following databases:The Cochrane Hepato-Bil- iary Group Controlled Trials Register (6 April 2004), the Cochrane Database of Systematic Reviews, Database of Abstracts of Reviews of Effects (DARE), the Cochrane Central Register of Controlled Trials (CENTRAL), Health Technology Assessment (HTA) Database, NHS Economic Evaluation Database, all in The Cochrane Library (Issue
to January 2004),The Intelligent Gateway to Biomedical & Phar- macological Information (EMBASE) (1980 to January 2004), ISI Web of Knowledge (Web of Science) (1988 to January 2004), and CINAHL (1982 to January 2004). The search strategies used are provided inAppendix 1.
Our aim was to perform a maximal sensitive search in order to conduct a more complete review. As describing an operation of the gallbladder in medical terms without the word cholecystectomy is impossible, a maximal sensitive search with the term cholecys- tectomy was used. For our MEDLINE search, a more sophisti- cated strategy, advised by the Dutch Cochrane Centre and listed inAppendix 1was used (with help from Geert van der Heijden, Julius Center, Utrecht).
Additional relevant trials were looked for by cross reference check- ing of identified randomised trials. Finally all authors of included trials were requested by letter for additional information on any published, unpublished, or ongoing trials.
Furthermore, during data extraction it turned out that in a large number of trials essential data and information on methods were missing. To improve the quality of the analysis, individual trialists were contacted and asked for missing data.
Data collection and analysis
The review was conducted according to the present protocol (Keus 2004) and the recommendations by theCochrane Handbook for Systematic Reviews of Interventions (Higgins 2005). All identified trials were listed in the ’Characteristics of included studies’ table and an evaluation whether the trials fulfilled the inclusion criteria was made. Excluded trials and the reasons for exclusion were listed as well (table with ’Characteristics of excluded studies’).
Assessment of methodological quality
Inadequate methodological quality in randomised controlled trials carries the risk of overestimating intervention effects (Schulz 1995;
Moher 1998;Kjaergard 2001). Methodological quality, study de- sign, and reporting quality have been recognised as criteria which can restrict bias in the comparisons of interventions (Moher 1998;
Kjaergard 2001). Therefore the methodological quality of ran- domised clinical trials was assessed using the following compo- nents.
Generation of the allocation sequence
• Adequate, if the allocation sequence was generated by a computer or random number table. Drawing of lots, tossing of a coin, shuffling of cards, or throwing dice was considered as adequate if a person who was not otherwise involved in the recruitment of participants performed the procedure.
• Unclear, if the trial was described as randomised, but the method used for the allocation sequence generation was not described.
• Inadequate, if a system involving dates, names, or
These studies are known as quasi-randomised and were excluded from the present review.
Allocation concealment
• Adequate, if the allocation of patients involved a central independent unit, on-site locked computer, or sealed envelopes.
• Unclear, if the trial was described as randomised, but the method used to conceal the allocation was not described.
• Inadequate, if the allocation sequence was known to the investigators who assigned participants or if the study was quasi- randomised.
Blinding
• Adequate, if the trial was described (at least) as blind to participants or assessors and the method of blinding was described. We are well aware that it is very difficult to properly blind trials comparing surgical treatments.
• Unclear, if the trial was described as (double) blind, but the method of blinding was not described.
• Not performed, if the trial was not blinded.
Follow-up
• Adequate, if the numbers and reasons for dropouts and withdrawals in all intervention groups were described or if it was specified that there were no dropouts or withdrawals.
• Unclear, if the report gave the impression that there had been no dropouts or withdrawals, but this was not specifically stated.
• Inadequate, if the number or reasons for dropouts and withdrawals were not described.
Extraction of data
Inclusion and exclusion criteria used in each trial.
The following data on the randomisation procedure have been extracted:
(1) Number of randomised patients
(2) Number of patients not randomised and reasons for non-ran- domisation
(3) Exclusion after randomisation (4) Drop-outs
(5) ’Intention-to-treat’ analysis.
Also information on sample size, single- or multicentre study de- sign, assessment of primary and secondary outcome measures, use of antibiotic prophylaxis, surgical experience, and intra-operative cholangiography was registered (Table 1).
General descriptive data (like sex, age, body mass index (BMI), and American Society of Anaesthesiology (ASA) classification) are supposed to be equally divided due to randomisation (Assmann 2000). These data are presented in Table 2as far as available.
Outcome data on mortality, complications, health-related quality- of-life, pulmonary function, pain, duration of operation, hospital stay, and convalescence were extracted according to availability.
Statistical analysis
With adequate binary data available, a priori presentation in odds
robustness of the odds ratio. From this, results could be presented in relative risk (ratio) (RR(R)) or numbers needed to treat (NNT) by recalculation. However exploring the data showed that for many binary data the outcome was rare or zero in both arms. Odds ratios (OR) and risk ratios (RR) are not estimable in trials with zero events in both arms (Sweeting 2004). Binary outcomes with zero events in both arms can merely be presented in risk differences (RD). Although risk differences are statistically less robust and result in conservative estimates, they are simple measures, easy to understand, and useful for public communication.
For continuous data, authors generally present their results in me- dians with ranges due to suspicion of skewed data. However, for the analysis of data in a meta-analysis, means with their corresponding standard deviations (SD) are needed to calculate mean differences (MD) or weighted mean differences (WMD) with 95% confi- dence intervals (CI). Using means from all trials would ignore a non-Gaussian distribution. Therefore, skewness ratios (mean di- vided by the standard deviation) were calculated first (Higgins 2005, page 96). With a ratio larger than two, skewness is ruled out, whereas skewness is suggested when the ratio is between one and two and a ratio less than one indicates strong evidence of skewness.
In situations where skewness could be ruled out, assumptions on equality of median to mean was made and used in the sensitivity analyses. For trials presenting confidence intervals or standard er- ror of means, we performed a recalculation to a standard deviation (SD) (Higgins 2005, page 90-91). In case no data on standard de- viation was available, we calculated an average standard deviation from those observed in other studies and imputed this value for the standard deviation in the sensitivity analysis (Higgins 2005, page 92).
Results were considered according to the four different criteria of quality. The existence of an overall difference in outcome was clear when all four criteria showed significance. However, when the dif- ferent quality criteria showed contradicting results, then an overall conclusion considering one outcome was not obvious and had to be made individually. In each individual component, results from high-quality trials subgroups were given more weight compared to analyses including all trials or low-quality trials subgroups. Results with confidence intervals that touched, but did not cross, the line of equivalence were considered not significant.
Apart from comparisons in the four individual quality criteria, an- other comparison was performed with trials divided into low-bias risk trials (high methodological quality) and high-bias risk trials (low methodological quality). Trials that were assessed as adequate regarding all four methodological criteria were considered low-bias risk trials. All trials that were not assessed as adequate with regard to all the four methodological criteria were considered high-bias risk trials.
Bias detection
We have used funnel plots to provide a visual assessment of whether treatment estimates were associated with study size. The presence
Macaskill 2001) varies with the magnitude of the treatment effect, the distribution of study size, and whether a one- or two-tailed test is used (Macaskill 2001).
Both the random-effects model (DerSimonian 1986) and the fixed-effect model (DeMets 1987) for pooling effect estimates were explored.
• In case of no discrepancy (and no heterogeneity) the fixed- effect models were presented.
• In case of discrepancy between the two models (ie, one giving a significant intervention effect and the other no significant intervention effect) both results were reported.
Discrepancy will only occur when substantial heterogeneity is present.
• Most weight was put on the results of the fixed-effect model if the meta-analysis included one or more large trials, provided that they had adequate methodology. (By large trials we refer to those that outnumber the rest of the included trials in terms of numbers of outcomes and participants (ie, more than half of all included events and participants)).
• Otherwise, most weight was put on the results of the random-effects model as it incorporated heterogeneity. The reason for this was that the random-effects model increases the weight of small trials. Small trials however are more often than large trials conducted with unclear or inadequate methods (Kjaergard 2001).
• In situations of excessive heterogeneity we refrained from reporting a pooled estimate when inappropriate.
The main focus of looking at heterogeneity in meta-analysis is to discriminate true effect modifiers from other sources of het- erogeneity. Heterogeneity was calculated by the Cochrane Q test and quantified by measuring I2(Higgins 2002). If excessive het- erogeneity occurred, data were re-checked first and then adjusted.
Extreme outliers were excluded (and tested in sensitivity analyses) when adequate reasons were available. If excessive heterogeneity still remained, depending on the specific research question, alter- native methods were considered: subgroup analysis and meta-re- gression if appropriate.
Subgroup analysis
Subgroup analyses were performed to compare the effects of the interventions according to the methodological quality of the trials (adequate compared to unclear/inadequate). Furthermore, causes of heterogeneity (defined as the presence of statistical heterogeneity by chi-squared test with significance set at P-value < 0.10 and measured by the quantities of heterogeneity by I2(Higgins 2002)) were explored by comparing different groups of trials stratified to level of experience of the surgeon and other factors that may explain heterogeneity.
Sensitivity analyses were performed assuming zero mortality and zero conversions for missing data. Sensitivity analyses were per- formed imputing medians and using average standard deviations
ity analyses were performed as well. Subgroup analyses were per- formed testing the influence of antibiotic prophylaxis, surgical ex- perience and intra-operative cholangiography on operative time, complications and hospital stay. These subgroup and sensitivity analyses were performed as far as data were available.
The statistical package (RevMan Analyses) provided by The Cochrane Collaboration was used (RevMan 2003). The statistical analyses were performed by FK and CL.
R E S U L T S
Description of studies Searches and trial identification
For the search strategies used and the number of hits we refer to Appendix 1.
The search was conducted in The Cochrane Hepato-Biliary Group Controlled Trials Register (840 hits, 65 selected) and The Cochrane Library, Issue 1, 2004 with the following results: the Cochrane Database of Systematic Reviews (33 hits, none were se- lected), the Database of Abstracts of Reviews on Effects (DARE) (17 hits, 5 selected), the Cochrane Central Register of Controlled Trials (CENTRAL) (1343 hits, 146 selected), the Health Tech- nology Assessment (HTA) Database (11 hits, 4 selected), and the NHS Economic Evaluation Database (43 hits, 6 selected).
The search further comprised the following databases: The Na- tional Library of Medicine (MEDLINE) (8354 hits, 347 selected), The Intelligent Gateway to Biomedical & Pharmacological Infor- mation (EMBASE) (685 hits, 131 selected), ISI Web of Knowl- edge (Web of Science) (1163 hits, 148 selected), and CINAHL (740 hits, 9 selected).
Altogether, the search resulted in 13229 hits. The first selection process was performed based on the title of the publications. In each step of selection, we included the publication in case of any doubt. The total number of selections by title from this group of 13229 publications was 911 hits. After correction for duplicates, 586 remained.
The abstracts of these 586 publications were reviewed indepen- dently by two reviewers (FK and JJ) in order to evaluate whether the study should be included in the review. Differences between FK and JJ were discussed with CL. A total of 428 publications could be rejected based on their abstract. Initially, trials which did not clearly mention whether they were randomised clinical trials or not, were given the benefit of the doubt. If appropriate, they were excluded later on. Eventually, 158 publications were selected for further evaluation and these are all listed in this review with reasons for in- or exclusion.
A total of 123 publications were excluded (see table with ’Charac- teristics of excluded studies’). A total of 35 publications describing
’Characteristics of included studies’ andTable 1). Critical appraisal and data extraction of these 13 trials were done by FK, JJ, and CL separately. Any disagreements were solved in several consensus meetings.
The study by Redmond (Redmond 1994) is excluded because of no correct randomisation. However this study must be considered a borderline case, therefore the results of this study were included in sensitivity analyses on total complications and operative time. For data management reasons this study is listed in the characteristics of included studies table.
Of the 13 included trials, one trial was only described in short (comment:Tate 1993). Therefore only limited information was obtained from this trial. As no language restrictions were used, one publication (Secco 2002) was translated. Double publications of the trial results by the same research group are listed in the references of included studies, and are considered one trial (eg, McMahon 1994;Bruce 1999). After contacting individual trialists additional data and information were obtained from 3 out of 13 included trials (see acknowledgement).
Patient characteristics
All included trials used similar inclusion criteria, ie, patients with symptomatic cholecystolithiasis who were scheduled for elective cholecystectomy. The extensiveness in which exclusion criteria were described varied among the trials, but nearly all trials excluded acute cholecystitis. Trials with exclusively acute cholecystitis as in- clusion criterion for cholecystectomy were excluded. Trials that included minorities of patients with acute cholecystitis in addition to patients with symptomatic cholecystolithiasis were included.
Trial designs
Only one trial used a three-arm design (Coelho 1993). All other trials used a two arm, parallel-group design.
Surgical interventions
Usually laparoscopic cholecystectomy was not further specified.
Some trials stated that a four trocar technique was used, creating a pneumoperitoneum by using carbon dioxide insufflation with a maximum intraperitoneal pressure of 12 to 15 millimetres mer- cury. As noted before, an incision length of 8 centimetres was taken as the cut-off point between small-incision and open chole- cystectomy. Some trials using the small-incision technique did not mention the size of the incision. We classified these trials as a small-incision cholecystectomy and not an open cholecystectomy, based on how the author labelled the operation procedure. Two trials performed small-incision cholecystectomy by a 5 centime- tres midline incision, the others by a transverse subcostal incision, some with muscle splitting, others by transsection of the rectus or oblique muscles.
Antibiotic prophylaxis administered at induction of anaesthesia was explicitly mentioned in some trials. In others the explicit omis- sion of antibiotic prophylaxis was mentioned, but most trials did not report on its use. Information on surgical experience (one or a few highly experienced surgeons performing all operations or also involving registrars) and intra-operative cholangiography (at-
tempted in all or only in selected patients) was recorded as well.
Outcome measures
A problem considering relief of symptoms and pain is how this outcome is defined and measured. Apart from differences in mea- surement, very few trials reported on this outcome. Therefore, we were unable to report results considering relief of symptoms and pain. Nearly all trials reported on complications, operative time, and hospital stay. We analysed conversion proportions between both techniques. Duration of sick leave was another commonly examined outcome. Not all trials clearly mentioned mortality. In trials in which mortality remained unclear, only in a sensitivity analysis the assumption that no patient had died was made when mortality was not mentioned (mean and most probable outcome of a trial). Some trials included mortality in their complications;
we considered them separately. Because of the wide range of the types of complications described, we classified (subcategorised) all complications into four subcategories (intra-operative, minor, se- vere, or bile duct injury) in addition to a total complication pro- portion (Table 3). Each complication was classified twice: once in one of the four subcategories (intra-operative, minor, severe, or bile duct injury) and once again in the total complication pro- portion. Consequently, all bile duct complications were registered separately from all other complications (and not counted in the severe and minor subcategories). Likewise, all intra-operative com- plications (except from the bile duct injuries) were categorised sep- arately from other minor and severe complications.
The following outcomes were reported in one trial only or in dif- ferent ways: bowel function, immunological parameters and cy- tokines, acute phase proteins, and acute phase hormones. Pain scores and analgesic use as well as health status related quality-of- life were frequently examined outcomes. Due to the great vari-
ation in the way these were measured and reported, it appeared impossible to pool results.
Considering pulmonary function there is some limited data avail- able from randomised trials. However, considering the inconsis- tency in the type of effect measure reported, as well as the difference in moments in time the outcome was measured, and the statistical problems that arise in pooling these results, we decided to refrain from reporting these results. Our intention was to cover costs (an important secondary outcome) as well, but although costs were described in several trials, it was reported in a lot of different ways.
Moreover, as different points of view were taken in those analy- ses, and regarding the cultural differences (Vitale 1991) as well as differences in local costs, we decided that reporting this outcome would not be meaningful.
Risk of bias in included studies
We evaluated the internal validity of the trials by considering the four quality components, resulting in the following number of high-quality (ie, adequate) trials. Information that was not men- tioned in a trial, was scored ’unclear’. When necessary informa- tion about randomisation, blinding procedure, or follow-up was unclear or missing, the authors were contacted to obtain specific additional information on these issues. Trials of which no response was received, remained classified as ’unclear’ trials.
We assessed the quality of the 13 included trials as follows: gener- ation of allocation sequence was adequate in three trials (23.1%), allocation concealment in eight trials (61.5%), blinding in four trials (30.8%), and follow-up in eight trials (61.5%) (Table 4) (Figure 1;Figure 2).
Figure 1. Methodological quality graph: review authors’ judgements about each methodological quality item presented as percentages across all included studies.
Figure 2. Methodological quality summary: review authors’ judgements about each methodological quality item for each included study.
A comparison dividing trials into low-bias risk trials (adequate methodological quality in all four criteria) versus high-bias risk trials could not be performed as only one trial was considered low- bias risk.
Effects of interventions
We conducted five analyses: four comparisons based on the four methodological quality components including the subgroups high- and low-quality trials, and a fifth comparison containing sensitivity and subgroup analyses. Background data of all trials on age, sex, body mass index (BMI), and American Society of Anaes- thesiology (ASA) classification are shown inTable 2.
We identified a total of 13 randomised trials comparing laparo- scopic versus small-incision cholecystectomy, including one trial described in a comment (Tate 1993) and one unpublished trial (Keus 2006). A total of 1175 and 1162 patients were included in the laparoscopic and small-incision groups respectively. Data are presented inTable 1together with data on antibiotic prophylaxis, performance of cholangiography, and experience of the surgeon.
We defined a complication as something the author called a com- plication and refrained from our own interpretation, except for a non-aesthetic scar (Secco 2002): we agreed on this not being regarded as a complication.
In the analyses, there were no significant differences in mortal- ity, minor complications, severe complications, and bile duct in- juries considering all trials, neither in the subgroups high-quality and low-quality trials, nor between the fixed-effect model and the random-effects model. As ’concealment of allocation’ is consid- ered the most important component of methodological quality, all subgroup results considering this aspect (except for the fore- mentioned results that were not significantly different) were pre- sented in additionalTable 5.
Sensitivity analyses were performed assuming zero mortality and zero conversions, imputing medians and standard deviations for missing data, omitting one outlier in complications and hospi- tal stay, and adding results from the borderline trial of Redmond (Redmond 1994). Subgroup analyses were performed testing the influence of antibiotic prophylaxis, surgical experience and intra- operative cholangiography on operative time, complications and hospital stay. Other sensitivity and subgroup analyses were not performed as necessary data were missing or analyses were consid- ered inappropriate.
Mortality
Mortality was only explicitly reported in seven trials. Mortality was not reported in six trials (Coelho 1992a;Tate 1993;Bruce 1999;
Srivastava 2001;Grande 2002;Secco 2002). No significant differ- ences were identified in analyses stratifying trials for all four qual-
effect model has been applied. Sensitivity analysis (5-1) assuming zero mortality in non-reporting trials, led to the same result (risk difference 0.00, 95% CI -0.01 to 0.01). In both treatment groups one patient died (resulting in mortality proportions of 0.09% in both laparoscopic and small-incision cholecystectomy).
Intra-operative complications
Complications were explicitly reported in all but one trial. Intra- operative complications were 13.1% and 7.6% in the laparoscopic and small-incision group, respectively. There was severe hetero- geneity (up to 98%), therefore the random-effects model has been applied resulting in no significant effects in all subgroups in all four comparisons (risk difference 0.01, 95% CI -0.03 to 0.06).
The majority of the events were accounted for by the trial of Ros (Ros 2001) (most of the intra-operative complications in this trial were gallbladder perforations).
Minor complications
The minor complication proportions were 8.3% and 9.2% in the laparoscopic and small-incision group, respectively. The random- effects model has been applied and there were no significant dif- ferences present in the four quality components in the subgroups (risk difference all trials, random-effects -0.01, 95% CI -0.04 to 0.02). Performing a sensitivity analysis (5-7) by omitting the out- lier (Srivastava 2001) led to disappearance of heterogeneity, but still showed no significant difference (risk difference, fixed-effect 0.00, 95% CI -0.02 to 0.03).
Severe complications
The severe complication proportions were 4.0% and 4.2% in the laparoscopic and small-incision group, respectively. As heterogene- ity was present (up to 78%), the random-effects model has been applied. There were no significant differences in the four quality components present, neither in the high-quality and in the low- quality subgroups (risk difference all trials, random-effects 0.01, 95% CI -0.01 to 0.03).
Bile duct injury
The bile duct injury proportions were 1.2% and 1.9% in the la- paroscopic and small-incision group, respectively. The difference was mainly caused by eight cases of bile leakage with unknown origin and conservative treatment in the small-incision group (five cases from one trial). As no heterogeneity was present, the fixed- effect model has been applied, and there were no significant dif- ferences present in all four quality components (risk difference all trials -0.01, 95% CI -0.02 to 0.00).
Comparing in a subgroup analysis (5-18) the bile duct injuries from trials that explicitly mentioned they did attempt intra-oper- ative cholangiography in all patients versus the bile duct injuries
tine intra-operative cholangiography in all patients (but only in selected or no patients), did not show an influence of cholangiog- raphy on the bile duct injury proportion (regression coefficient 0.452, 95% CI -1.003 to 1.907, P = 0.54).
Total complications
Complications were not reported in one trial (Tate 1993). All complications (and frequencies) were listed (Table 3).
The total complication proportions were 26.6% and 22.9% in the laparoscopic and small-incision group respectively (Table 3).
There were 19 re-operations in the laparoscopic group and 18 in the small-incision group. Reoperation proportions were 1.6% in both groups. Because of heterogeneity the random-effects model has been applied and no significant difference was present (risk difference all trials -0.01, 95% CI -0.07 to 0.05). Also in the low-quality and high-quality subgroups there were no significant differences present applying the random-effects model.
Complications in trials with three or more quality components assessed as high-quality (Majeed 1996;Ros 2001;Keus 2006) were summarised and presented in additionalTable 6. Complication proportions in both treatment groups were comparable, but overall complication proportions were higher compared to complications in additionalTable 3including all trials.
Performing a sensitivity analysis (5-8) by omitting the outlier ( Srivastava 2001) led to a reduction of heterogeneity (but it was
still present: 55%) and also showed no significant difference in the random-effects model (risk difference 0.02, 95% CI -0.02 to 0.07).
As the study of Redmond was considered a borderline case, the complications of this study were included in a sensitivity analysis (5-10). There was no significant difference present between both groups including this study.
Comparing the total complication proportions from trials that ex- plicitly mentioned the use of antibiotic prophylaxis versus the to- tal complication proportions from trials that explicitly mentioned that they did not use antibiotic prophylaxis or trials that did not mention the use of antibiotic prophylaxis, did not show a signifi- cant influence of antibiotic prophylaxis in a subgroup analysis (5- 12) (regression coefficient -0.220, 95% CI -0.557 to 0.118, P = 0.20).
Comparing in a subgroup analysis (5-13) the total complication proportions from trials that explicitly mentioned that one or a few experienced surgeons performed all the operations versus the to- tal complication proportions from trials that explicitly mentioned that also registrars performed the operations, did not show an in- fluence of surgical experience (regression coefficient -0.327, 95%
CI -0.784 to 0.129, P = 0.16).
Funnel plot of total complications is shown inFigure 3, showing no bias.
Figure 3. Funnel plot on laparoscopic versus small-incision cholecystectomy regarding generation of the allocation sequence considering total complications, including 95% confidence interval lines. There is some suspicion of bias considering the absence (in the lower right part of the figure) of small trials favoring the
small-incision technique.
Conversion
Conversions were not reported in five trials (Coelho 1992a;Kunz 1992;Tate 1993;Bruce 1999;Grande 2002). The conversion pro- portions were 13.4% and 16.1% in the laparoscopic and small- incision group, respectively. As heterogeneity was present, the ran- dom-effects model has been applied. All comparisons did not show significant differences in conversion proportions, neither in the low-quality nor in the high-quality subgroups (risk difference all trials, random-effects 0.00, 95% CI -0.05 to 0.04). Analysis of conversions in the ’blinding’ comparison were not performed, as
the decision to convert (or not to convert) could not have been made under blinded conditions.
In a sensitivity analysis the assumption was made that trials that did not mention converted operations had zero conversions. Re- sults showed no significant difference (5-2). Including these trials resulted in overall conversion proportions of 12.2% and 14.8%
respectively. This difference was not significant (risk difference - 0.02, 95% CI -0.05 to 0.01).
Funnel plot of conversions is shown inFigure 4, showing no bias.
Figure 4. Funnel plot on laparoscopic versus small-incision cholecystectomy regarding generation of the allocation sequence considering conversions, including 95% confidence interval lines. No arguments for bias.
Operative time
Because of heterogeneity, the random-effects model was applied.
Meta-analyses in all comparisons showed that small-incision chole- cystectomy was significantly faster to perform than laparoscopic cholecystectomy ((WMD all trials, random-effects 9.20 minutes, 95% CI 2.06 to 16.35); high-quality trials ranging from 10.03 minutes to 16.36 minutes). Low-quality subgroups showed no significant difference.
All available data were presented inTable 7. In a sensitivity anal- ysis (5-3) including the assumptions on standard deviations and medians considering skewness, there was a significant shorter op- erative time in the small-incision group (WMD, random-effects 9.71 minutes, 95% CI 2.34 to 17.08). As the study of Redmond was considered a borderline case, the results of operative times of this study were included in a sensitivity analysis (5-11). The previous findings of a shorter operative time in the small-incision group did not change.
Comparing in a subgroup analysis (5-14) the operative time from trials that explicitly mentioned they did attempt intra-operative cholangiography in all patients versus the operative time from tri- als that explicitly mentioned they did not attempt routine intra- operative cholangiography in all patients (but only in selected or no patients) (and also including the assumptions on standard de- viations and medians) did not show an influence of cholangiog- raphy on operative time (regression coefficient 11.381, 95% CI - 9.645 to 32.407, P = 0.29). Comparing in a subgroup analysis (5- 15) the operative time from trials that explicitly mentioned that one or a few experienced surgeons performed all the operations versus the operative time from trials that explicitly mentioned that also registrars performed the operations (and also including the assumptions on standard deviations and medians) did not show an influence of surgical experience on operative time (regression coefficient -14.151, 95% CI -32.364 to 4.063, P = 0.13).
Funnel plot of conversions is shown inFigure 5, showing no bias.
Figure 5. Funnel plot on laparoscopic versus small-incision cholecystectomy regarding generation of the allocation sequence considering operative time, including 95% confidence interval lines. No arguments for bias.
Hospital stay
Heterogeneity was present (up to 87%), therefore the random- effects model has been applied resulting in significant differences analysing all trials and in the low-quality trials subgroups (WMD, random-effects -0.85, 95% CI -1.35 to -0.34). However, there were no significances in the high-quality trials subgroups (except for the high-quality trials regarding adequate follow-up). As high- quality trials results are more reliable, a real difference is probably not present.
All available data were presented inTable 8. In a sensitivity analysis (5-4) we made assumptions on standard deviations and medians considering skewness. However, the second assumption did not hold: skewness was suggested in two trials (Kunz 1992;Majeed 1996), and there was evidence of skewness in three trials (Ros 2001;
Srivastava 2001;Keus 2006). Including these assumptions on all trials (ignoring skewness), again a significant shorter hospital stay was suggested in the laparoscopic group (WMD, random-effects -0.96 days, 95% CI -1.41 to -0.51). Excluding the trials with
evidenced skewness (as these are likely to introduce severe bias) also suggested a shorter hospital stay in the laparoscopic group (WMD, random-effects -1.10 days, 95% CI -1.68 to -0.51).
Performing a sensitivity analysis (5-9) by omitting the outlier (Grande 2002) and including the previous mentioned assump- tions showed a significant difference in the random-effects model (WMD, random-effects -0.59 days, 95% CI -0.98 to -0.21).
Comparing in a subgroup analysis (5-16) hospital stay from trials that explicitly mentioned that they did use antibiotic prophylaxis versus the hospital stay from trials that explicitly mentioned that they did not use antibiotic prophylaxis (and also including the assumptions on standard deviations and medians) did not show an influence of antibiotic prophylaxis on hospital stay (regression coefficient -0.573, 95% CI -1.457 to 0.312, P = 0.21). There was also no influence of surgical experience on hospital stay (5-17) (regression coefficient 0.665, 95% CI -0.131 to 1.460, P = 0.11).
Funnel plot of conversions is shown inFigure 6, showing no bias.
Figure 6. Funnel plot on laparoscopic versus small-incision cholecystectomy regarding generation of the allocation sequence considering hospital stay, including 95% confidence interval lines. No arguments for bias.
Convalescence
Unfortunately convalescence was not always reported and making the distinction between return to work and to normal activity (at home) was difficult. As these can be regarded separately, we made different comparisons.
Convalescence from work leave
Due to heterogeneity, the random-effects model has been applied resulting in no significant difference between both techniques con- sidering work leave. However, only three trials described return to work.
Convalescence to normal activity
There is heterogeneity present, therefore the random-effects model has been applied. Analysis of all trials showed no significant dif- ference. Although, the low-quality trials subgroups showed signif- icant differences in two of the four methodological quality com- ponents (’concealment of allocation’ and ’blinding’) favouring the laparoscopic technique, the high-quality trials subgroups on con- trary, did not show significant differences in the four comparisons.
Conclusions must be drawn carefully as only four trials are in- volved.
Assuming in a sensitivity analysis that convalescence (both con- sidering work leave (5-5) and normal activity (5-6)) is symmetri- cally distributed and therefore making assumptions on standard
deviations and medians, there were neither significant differences in convalescence on work leave (WMD, random-effects -5.86, 95% CI -12.31 to 0.59) nor in convalescence on normal activity (WMD, random-effects -7.42, 95% CI -15.94 to 1.10).
D I S C U S S I O N
This systematic review contains at least four major findings. First, the comparison of the clinical outcome of laparoscopic cholecys- tectomy to small-incision cholecystectomy has been well tested in randomised clinical trials with over 2300 patients randomised to the two techniques in trials of relatively high methodological quality. Secondly, laparoscopic cholecystectomy carried a bile duct injury rate not significantly different from small-incision chole- cystectomy. Thirdly, the total numbers of patients with complica- tions were high and not significantly different for the two proce- dures. Fourthly, laparoscopic cholecystectomy took significantly more time to perform than small-incision cholecystectomy.
No comparison could be performed dividing trials into low-bias risk trials (trials with adequate methodological quality in all four
quality criteria) versus high-bias risk trials, suggesting that surgi- cal research considering cholecystectomy can be improved impor- tantly.
High-quality trials are more likely to estimate the ’true’ effects of the interventions (Schulz 1995; Moher 1998; Jüni 2001;
Kjaergard 2001;Egger 2003). In this review high-quality trials were more likely to show a neutral effect or a negative effect of la- paroscopic surgery, whereas trials with unclear or inadequate qual- ity tended to show more often a positive effect or a neutral effect of laparoscopic surgery. These observations are in concordance with other studies showing linkage between unclear/inadequate methodological quality to significant overestimation of beneficial effects and underreporting of adverse effects. The small trials fa- voring laparoscopic surgery can be regarded as an example of ei- ther low methodological quality or publication bias (illustrated in funnel plot:Figure 3laparoscopic versus small-incision cholecys- tectomy considering total complications). The other funnel plots (Figure 4;Figure 5; andFigure 6) do not suggest bias.
We identified a total of 13 trials comparing laparoscopic versus small-incision cholecystectomy with different focuses on outcome measures. Mortality in both procedures was near 0% in this sys- tematic review, which is in concordance with the literature includ- ing non-randomised series (Downs 1996). The total complication proportions are 26.6% and 22.9% in the laparoscopic and small- incision group, respectively (Table 3). Excluding gallbladder per- foration (as some surgeons would not regard it being a complica- tion) results in complication proportions decreasing to 17.0% and 17.5%, respectively. These percentages are higher than figures (up to 5% overall complication proportions) known from other series and reviews including non-randomised series, suggesting underre- porting of these observational studies (Southern Surg 1991;Litwin 1992;Deveney 1993;Deziel 1993;Deziel 1994;Wherry 1994;
Downs 1996). The situation in the real world, however, may be even worse due to the positive effect that randomised clinical trials may have on quality of care.
All four methodological quality subgroup analyses showed no sig- nificant difference considering intra-operative and total compli- cation proportions. In subgroup analysis of complications, surgi- cal experience, the use of antibiotic prophylaxis, or performing an intra-operative cholangiography revealed no significant difference (in regression coefficients; STATA®) for both groups. There is also no significant difference in conversion proportions. This is in line with the results from our sensitivity analysis, including all trials in the model.
Operative time was shorter in small-incision cholecystectomy con- sidering all four methodological subgroup analyses and despite heterogeneity, both in the high-quality trials and all trials sub- groups, while the low-quality trials subgroups in all four com- ponent analyses do not show this difference. This is one of the
more favourable effect of the laparoscopic technique, while high- quality trials show more often a positive effect of the small-inci- sion technique. In our sensitivity analysis we make assumptions on values for missing data (calculating average standard deviations and using medians in absence of skewness). Based on fulfillment of the condition of exclusion of the presence of skewness (Table 7; Table 8), means are supplemented by medians and averaged standard deviations are calculated, in order to pool estimates. In the subsequent sensitivity analysis with these adjustments, again a shorter operative time for the small-incision group was found.
As imputing data is dangerous, these results have to be interpreted with care.
Regarding hospital stay, the high-quality trials showed no signif- icant difference between the two operations with the random-ef- fects model, while low-quality trials as well as all trials together do show a significant difference between the two operations with the random-effects model. As methodological high-quality trials pro- duce more reliable results, we conclude that there is no significant difference in hospital stay. In our sensitivity analysis on hospital stay we make the same assumptions on average standard devia- tions and using medians for means. Our sensitivity analysis sug- gests that the duration of hospital stay in the laparoscopic group was shorter. However, skewed data are included in the analysis.
Remembering that in general including skewed data are expected to falsely increase differences, combined with the fact that it is a post-hoc hypothesis generating analysis, makes that the conclu- sion that hospital stay is shorter after laparoscopic cholecystectomy most uncertain. In subgroup analyses no evidence was found of antibiotic prophylaxis or surgical experience influencing hospital stay.
One might find hospital stay long compared to daily life practice.
Probably study conditions and different practice over time are re- sponsible for a longer hospital stay compared to current daily life practice. Apart from these reasons, there might be other reasons including cultural differences and others for differences in hospital stay. However, we have to remember that hospital stay is a sur- rogate marker for convalescence and because of numerous factors influencing its length it does not necessary reflect objective differ- ences between two operative procedures. Differences in hospital stay in open studies may represent bias, unless the type of surgery is blinded. Differences in results have to be interpreted with care.
Considering convalescence we tried to make distinction between work leave and return to normal activity (at home). Unfortunately only three trials reported data in each category. However, it seems that there is no significant difference between both groups in con- valescence in work leave, while return to normal activity seems quicker in the laparoscopic group. It must be emphasized however, that these conclusions are only based on the results of three trials.
In our sensitivity analysis on convalescence considering normal ac- tivity we make assumptions on standard deviation values in three
