Diagram showing flow of participants through each stage of the randomized trial. No patients in either group were lost to follow-up, and no patients were excluded from the final analyses. LC indicates laparoscopic cholescystectomy; SIC, small-incision cholecystectomy.
Keus F, Werner JEM, Gooszen HG, Oostvogel HJM, van Laarhoven CJHM. Randomized Clinical Trial of Small-Incision and Laparoscopic Cholecystectomy in Patients With Symptomatic CholecystolithiasisPrimary and Clinical Outcomes. Arch Surg. 2008;143(4):333-334. doi:10.1001/archsurg.143.4.371
To evaluate the primary and clinical outcomes in laparoscopic and small-incision cholecystectomy.
Blinded randomized single-center trial emphasizing methodologic quality and generalizability.
General teaching hospital in the Netherlands.
A total of 257 patients undergoing cholecystectomy for symptomatic cholecystolithiasis.
Laparoscopic cholecystectomy and small-incision cholecystectomy, performed mainly by surgical residents.
Main Outcome Measures
Complications and symptom relief were primary outcome measures; conversion rate, operative time, and hospital stay were secondary outcome measures. Feasibility of performing both procedures by residents was evaluated as well.
In the 257 patients, surgical residents performed 105 laparoscopic and 118 small-incision cholecystectomies. There were no significant differences in complications, conversion rates, and hospital stay. Operative time was significantly shorter with the small-incision technique.
No differences in primary clinical outcome measures were found between laparoscopic and small-incision cholecystectomy in this randomized trial with emphasis on methodologic quality and generalizability. The gold standard status of laparoscopic cholecystectomy is questionable.
isrctn.org Identifier: locator="http://isrctn.org/ISRCTN67485658">ISRCTN67485658
Langenbuch's classic cholecystectomy was the gold standard for treating cholecystolithiasis for more than a century.1 Shortly after surgeons began making incisions smaller to speed recovery,2- 5 laparoscopic cholecystectomy (LC) was introduced and rapidly conquered the world. Its popularity was based on the appeal of the technological innovation as well as enthusiastic marketing by industry, rather than resulting from an evidence-based approach.6,7
This technique was accepted as the gold standard by consensus in 1993, without a high level of evidence for its superiority.8 Now, in the modern era of new laparoscopic developments, LC has become the “ideal model” of implementation and justification for a laparoscopic technique. In newer areas for laparoscopic techniques, advantages over previous practice are still under debate. However, true evidence of the superiority of LC to other procedures, such as small-incision cholecystectomy (SIC) and open cholecystectomy (OC), has yet to be assessed.
Whereas the comparison of LC with OC seems to favor LC,9 conflicting results are found in randomized controlled trials comparing LC with SIC.10- 21 Nearly all trials did not discriminate between primary and secondary outcome measures, and only a few trials included surgical residents performing the operations. Differences in methodologic quality of the studies raise doubts about validity, and expert settings bring into question the generalizability.
Therefore, the principal aim of this study, with emphasis on methodologic quality, was to test the generalizability of the comparable outcome of LC and SIC in a general teaching hospital with predominantly surgical residents performing the operations.
Our aim was to perform a randomized trial focusing on 3 issues: discrimination between primary and secondary outcome measures, methodologic quality, and generalizability.
Mortality and complication rate are primary outcome measures and the ultimate determining factors in deciding between LC and SIC. On the basis of the available literature, we hypothesized that no major differences in primary outcome measures would be found between LC and SIC. All other outcomes are secondary outcomes (eg, cost-minimization analysis) and eventually may be deciding factors for choosing one procedure over the other. However, before it is justified to focus on secondary outcomes, we have to show results of primary outcomes in our study population comparable with those in the literature.
The second issue relates to methodologic quality. Many methods of scoring the intrinsic quality of a randomized trial exist, but none is sufficient. However, only 4 methodologic quality items have proved to be important factors for minimization of bias.22 We have tried to optimize trial quality in these 4 key domains: generation of the allocation sequence, allocation concealment, blinding, and follow-up.
The third issue relates to generalizability. If a small number of highly experienced surgeons perform all operations in a trial, excellent results might be expected. However, our real world includes surgeons without a high level of experience who also perform cholecystectomies and the training of residents, possibly leading to completely different results. To maximize applicability of the trial results, we needed to include residents performing operations.
This article presents the clinical outcomes of the trial, highlights efforts to minimize bias, and discusses the generalizability of our real-world findings.
We evaluated the usual baseline characteristics (age, sex, body mass index, and American Society of Anesthesiologists [ASA] score), preoperative biochemistry results, classic diagnostic symptoms, and recovery after complicated gallstone disease. The following outcomes were examined: mortality, complications, symptom relief, operative time, hospital stay, conversion rate, and intraoperative technical aspects. The number of operations performed by residents was recorded as well. Symptom relief was measured by asking all patients whether painful episodes had recurred, as well as whether cholecystectomy had resolved their primary complaint.
Approval from the Medical Ethics Committee for this single-center trial was obtained in September 2000. Patients were recruited between January 1, 2001, and January 31, 2004. If patients referred to the surgical outpatient clinic met the inclusion criteria and no exclusion criteria were present, written informed consent was obtained. Patients were placed on the waiting list for elective cholecystectomy. If patients wanted to reconsider their participation in the trial, they could be excluded on the day of admission, before randomization.
Inclusion criteria were symptomatic cholecystolithiasis (confirmed by ultrasonography) in male or female patients, 18 years or older at recruitment, reasonable to good health (ASA score of 1 or 2), no known relevant allergies, and signed informed consent letter.
Exclusion criteria were age younger than 18 years, choledocholithiasis (icterus, acholic feces, and/or bilirubin level of twice the upper limit of normal), cholangitis, known pregnancy, moderate to severe systemic disease (ASA score of 3 or higher), known cirrhosis of the liver, history of abdominal malignant neoplasm, previous upper abdominal surgery (precluding laparoscopic approach), psychiatric disease, or other factor (eg, lack of knowledge of the Dutch language) that might make follow-up or completion of questionnaires unreliable.
Obesity was recorded but was not an exclusion criterion.14 Recovery after successful endoscopic treatment of choledocholithiasis was not a contraindication. Acute cholecystitis is a different disease with different complication rates, morbidity, and conversion rates and therefore was cause for exclusion.
A random-number table was used for generation of the allocation sequence,23 and the allocation concealment was guaranteed by using sealed envelopes (no blocking, no stratification). To eliminate bias caused by preoperative expectations, patients were randomly assigned in the operation room after induction of anesthesia. A telephone call was placed to the secretarial office and an employee opened an envelope to determine the surgical method. Details of surgery were recorded in a case record form. Otherwise, the procedure was recorded as “trial cholecystectomy.” We did not use sealed envelopes for record keeping.17
All consultant surgeons participating in the trial had experience in LC and were trained in SIC in a pilot phase before the trial. After this learning phase, each consultant surgeon was considered equally familiar with the 2 techniques. Operations were supervised by 1 of the consulting surgeons. Our hospital is a training hospital; thus, residents (from second year on) performed most of the operations, which enabled us to test external validity in a teaching hospital.
Although discussions in the literature on intraoperative cholangiography continue, the policy in our hospital was not to perform cholangiography in any patient undergoing elective cholecystectomy. Performing standardized laboratory tests and, on indication, preoperative endoscopic retrograde cholangiopancreatography is the national policy in the Netherlands, based on the broad availability of high-quality endoscopy facilities.
All patients had a standard anesthesia regimen. Premedication, medications for induction and continuation of anesthesia, and respiration during surgery were standardized. Analgesics and medication for nausea were supplied according to a standard scheme.
All patients had nasogastric intubation during the operation that was removed immediately afterward. Operating time was measured from the first skin incision to the last suture placement for both techniques. In case of technical difficulties or for any other reasons, either technique could be converted to OC. No suction drains were left in the subhepatic space at the end of the procedure. Abdominal wall and skin closure were standardized. The wounds were covered with standard wound dressings, with iodine stains applied to the wound dressings to allow blind assessment during clinical stay.14 We did not use local anesthetic techniques or intercostal nerve blocks because their application in LC and SIC differs, thereby possibly introducing bias. Identical systemic administration of analgesics in both procedures was used, giving less potential for bias.
Open introduction was performed in all patients, regardless of previous abdominal surgery. Pneumoperitoneum was created with intra-abdominal pressures up to 12 mm Hg, and 3 trocars were inserted. The dissection of the cystic duct and artery, identifying the Calot triangle, was performed by means of a 3-point “flag” technique.24 The cystic duct and artery were clipped and transected. After complete dissection, the gallbladder was removed. If conversion to OC was necessary, the reasons for conversion were recorded.
In accordance with the literature, a cutoff point of 8 cm was used to differentiate between SIC and OC.11,12,14,15,17,18,20,21 As part of a separate research question, all patients underwent preoperative ultrasound for location of the gallbladder. We used the craniocaudal position of the mark for incision. The incision was placed over the musculus rectus abdominis. Only standard instruments were used, with no special equipment. Access to the peritoneum was obtained by a muscle splitting (and not transection) technique (comparable to open appendectomy). The gallbladder was dissected by a “fundus-first” technique. If necessary, the gallbladder was punctured to remove its liquid contents. The cystic duct and artery were ligated and the gallbladder was removed. Posterior and anterior fascias were closed separately. After wound closure, the length of the incision was measured; if it exceeded 8 cm, the operation was considered a conversion to OC. The reasons for conversions were recorded.
Early oral intake and mobilization were encouraged. Patients left the hospital as soon as they were able to do so. Although ambulatory cholecystectomy has been proved feasible, it was not practiced in our hospital.25 Hospital stay was defined as the number of postoperative nights in the hospital. Shortly before discharge, wound dressings were removed for wound inspection. Follow-up took place after 2 weeks, 6 weeks, and 3 months. For logistic reasons, we were not able to blind the surgeon at follow-up. Patients were encouraged to resume work and normal daily activity as soon as they felt able to do so.
Assuming no differences in primary outcome measures, sample size calculation was based on differences of costs. For this purpose, the direct costs of the first 50 patients in the trial were calculated to estimate the likely range of differences in costs and their standard deviations. On this basis, we estimated that 120 patients per group would be needed to detect a difference of 10% in direct costs using an α of 0.05 and a β of 0.9. Although such analysis was not the purpose of this trial, a difference of 10% in complications would have been possible to detect with this sample size. Before the reporting of costs would be allowed, the data set of the trial had to be validated in regard to the primary outcome measures (complications and symptom relief). After primary outcomes comparable to those in the literature were demonstrated, it would be justified to report secondary outcomes such as costs. No interim analysis was planned, but the monitoring committee could stop the trial if a substantial difference in complications occurred.
All data were stored in a case record form and subsequently transferred to an Access database. Double data entry was performed to prevent typing errors. Data were subsequently read into SPSS 11.0 (SPSS Inc, Chicago, Illinois) for statistical analysis.
Comparisons were made on an intention-to-treat basis.26 Calculations were made with SPSS 11.0. The χ2 test was used for dichotomous outcomes. For all continuous data, we present our results both as medians with ranges and as means with standard deviations to facilitate their interpretation.
The normality of the data was checked by means of the Kolmogorov-Smirnov test. In case of deviations from normality, data were transformed to normality.27 The Levene test was used for checking the equality of variances. When the condition of normality and equal variances was met, the independent samples t test was used for independent data. When equality of variances was absent or normality could not be reached with transformation, the nonparametric Mann-Whitney test was used.
All trial patients were operated on between January 1, 2001, and March 31, 2004. Initially, 366 patients fulfilled the inclusion criteria and agreed to be included in the trial. A total of 102 patients were not randomized for a variety of reasons (Figure).28 After 264 patients were randomly assigned, another 7 patients were excluded for the following reasons: unwillingness to participate further (2 patients), intraoperative suspicion of malignant neoplasm (2 patients), transfer to another ward not participating in the trial (1 patient), participation in 2 trials (not in line with the Helsinki declaration) (1 patient), and insufficient knowledge of the Dutch language (1 patient). A total of 257 patients were left for analysis (Figure).
The 2 groups did not differ with regard to age, sex, body mass index, and ASA classification (Table 1).29 We evaluated the following classic diagnostic symptoms of cholecystolithiasis in both groups30: severe pain, episodic pain, epigastric pain, pain in the right upper quadrant, pain radiating to the back, pain lasting 1 to 5 hours, awakening at night, and the Murphy sign. There were no significant differences in the presence and the duration of these symptoms. The numbers of patients presenting with complicated gallstone disease and operated on in a later stage also were equally distributed (Table 1). Preoperative blood analysis consisting of erythrocyte sedimentation rate, white blood cell count, and levels of C-reactive protein, alanine aminotransferase, aspartate aminotransferase, γ-glutamyltransferase, alkaline phosphatase, and bilirubin showed no significant differences between groups.
Results are presented including data from converted operations (intention-to-treat) unless otherwise stated.
Among intraoperative factors, we found more adhesions (χ2 = 9.15, P = .002) and intraoperative bile leakage (by gallbladder perforation) (χ2 = 10.26, P = .001) in the LC group. Correspondingly, the skin was left open for secondary wound healing more often in the LC group (χ2 = 31.69, P < .001). Intraoperative stone loss, presence of inflammation, and identification of the cystic duct and the common bile duct were not statistically different between the 2 groups. The cystic artery (P = .005) and Calot triangle (P < .001) were identified more frequently in the LC group, and in the SIC group a combined ligation of the cystic duct and artery was performed more frequently (P < .001).
Surgical residents performed 105 LCs (87.5%) and 118 SICs (86.1%). Operative time was shorter for SIC (60 vs 72 minutes; U = 6013.0; P < .001) (Table 2). Conversion rates were similar, with similar distribution of reasons for conversion (Table 3). Total incision length of scars, measured in 229 patients, appeared to be not statistically different between the LC and SIC groups.
There were no deaths in the trial. There were 5 intraoperative complications in the LC group and 3 in the SIC group (Table 4). Most importantly, 1 common bile duct injury occurred in each group (treated by T-drainage and hepaticojejunostomy). Considering postoperative complications, we did not find a difference in the number or severity between the 2 groups.
The numbers of complications in the “resident-resident” operative team subgroups were compared in an exploratory subgroup analysis. These numbers (3 and 4 in the LC and SIC groups, respectively) were not different from complication rates in operative teams that included a surgeon.
The follow-up rate between the groups was not statistically different. Follow-up in the LC and SIC groups, respectively, was 88.3% and 94.9% at 6 weeks, 80.0% and 81.0% at 3 months, and all patients showed up at either their 6-week or their 3-month follow-up appointment. All patients who did not show up at their 2-week or 6-week follow-up appointment appeared not to have any problems at their next scheduled follow-up appointment.
Whether symptom relief was achieved and whether symptoms had recurred was ambiguously reported by some patients. With unclear cases included, it appeared that 9.2% and 10.2% of patients in the LC and SIC groups, respectively, experienced failure of symptom relief (Table 2).
Postoperative complaints at follow-up were evaluated and appeared comparable between the LC and SIC groups, including dietary complaints (26 vs 30 patients; P = .96), diarrhea (17 vs 15; P = .44), fatigue (13 vs 6; P = .049), and complaints suggesting the presence of common bile duct stones (6 vs 9; P = .59). There was also no statistical difference in the number of patients undergoing endoscopic retrograde cholangiopancreatography or magnetic resonance cholangiopancreaticography postoperatively (2 vs 5; P = .24).
There were no differences in baseline characteristics between the 2 groups of patients, suggesting an effective randomization process. We found no significant differences in primary and secondary outcome measures between LC and SIC: complication rate (both intraoperatively and postoperatively), symptom relief, conversion rate, and hospital stay. Operative time was shorter in the SIC group. The results of this study compare well with those in the literature but add to it in terms of methodologic quality and real-world general teaching hospital setting.
Results of previous randomized controlled trials have been contradictory. Trials in the literature together include more than 2000 patients. Most trials found no difference in complication rates between LC and SIC,10,12,14,16 whereas in 2 trials complication rates were lower with SIC.15,17 However, complication rates vary substantially between the different trials, without unambiguously showing lower complication rates in trials with expert settings. Differences in intrinsic validity probably also play a role in varying complication rates.
Most trials found a shorter operative time for SIC,14,15,17 whereas 2 other trials did not.10,12 Results for hospital stay10,12- 17,19,20 and convalescence10,14,16,17 are conflicting as well. Overall, most authors conclude that there is no difference between the 2 procedures.12- 15,17 Considering these conflicting but numerous data, pooled estimates by meta-analysis are needed to derive strong conclusions.31
Minimizing bias gives considerable strength to results. Although it is not possible to be sure that all patients really were blinded to their treatment, expectations of ward personnel influencing postoperative recovery probably is a much more important factor in convalescence.
Generalizability includes extrapolation of study results to other patient categories or other surgical units. We believe that extrapolation of these results (surgery performed largely by residents) to other surgical units is possible without increased risk. However, extrapolation to other patient categories (ie, those with ASA scores of 3 or 4, or patients with cholecystitis) is dangerous because it uses assumptions that may not hold true. Additional research exploring the appropriateness of extrapolation of these results to other patient categories is necessary. Unlike the evidence in expert studies,14 we have shown the feasibility of SIC and LC in a general teaching hospital with mainly surgical trainees (86% and 88%) performing the operations. We believe that this reflects a real-world situation and adds to the generalizability.
In a short training phase, surgeons were taught the SIC technique. After this introductory phase, all participating surgeons were considered equally familiar with both techniques. In reality, they probably were more experienced with LC. Explorative analysis of this theoretical difference in experience by comparing early with later results did not show any difference in outcome.
Opportunities to learn the OC technique are few because conversion rates are too low to provide enough numbers. Small-incision cholecystectomy with extended incision if necessary appears to be a valuable alternative. An advantage of this strategy is that it familiarizes residents with the open approach and precludes the performance of an unknown procedure. It seems useful to add the small-incision technique to the repertoire of the general surgeon.
At the time the trial started, general policy in the Netherlands (and in our hospital) was not to perform intraoperative cholangiography. Discussions on selective use vs standard or no cholangiography continue. Recent research shows the benefit of intraoperative cholangiography.32 Possibly this policy should be reconsidered on the basis of these findings, both in our hospital and on a national level. This issue is currently being studied by a national committee on guideline development of diagnosis and treatment of gallstones.
Hospital stay in this study might be considered long compared with everyday practice. Although ambulatory cholecystectomy has proved feasible with both the laparoscopic and small-incision techniques,33,34 it was not the policy in this hospital at the start of the trial. Both cultural factors and differences in expectations of patients, as well as arrangements with insurance companies at that time, were responsible. In addition, it should be remembered that hospital stay is a surrogate marker for recovery and that it is influenced by many factors, some unknown.
Evidence of superiority of LC to SIC is lacking, and conflicting data arise from the existing randomized trials.10- 21It is remarkable, however, that the acceptance of LC was extremely rapid: within 3 years, from 1989 to 1992, the rate of laparoscopy for cholecystectomy in the United States changed from 0% to 80%.35 Moreover, it became the treatment of choice by consensus of the National Institutes of Health in 1993. This is especially remarkable because there had been no possibility of acquiring an adequate level of evidence within this very short period.8 Analysis afterward showed that access to positive information, more favorable adoption costs–related conditions, and the role of the “early adopters” had been most important in this process.6
Whereas in the 1980s, the “age of optimism,” technical innovation dominated decision making in medical science, currently health technology assessments and cost-effectiveness analyses of treatments play a dominant role.7 It was, therefore, interesting to compare the presumed clinical superiority of LC, the ideal example of laparoscopic innovation, with more basic minimally invasive techniques such as SIC. With budget restrictions in mind, it is worthwhile looking at other factors that may play a role in future decisions on laparoscopic surgery.
This randomized controlled trial with emphasis on methodologic quality and generalizability shows no benefit in clinical outcome measures of LC compared with SIC. The question arises of which other measures do differ to persuade us of superiority of one minimally invasive technique over the other. In extending these conclusions to the broad discussion on laparoscopic gastrointestinal surgery, should we be looking for other outcome measures, such as cost-effectiveness? Or should we look for other factors such as fast-track recovery programs that seem to outweigh some of the effects of laparoscopic surgery, rendering the surgical approach of secondary interest?
Surgical decision making should be evidence based. A systematic review of all randomized controlled trials on this topic should derive the highest possible evidence and surgeons should dare to act accordingly.
Correspondence: Frederik Keus, MD, Department of Surgery, Diakonessenhuis, Bosboomstraat 1, 3582 KE Utrecht, the Netherlands.
Accepted for Publication: December 21, 2006.
Author Contributions: Drs Keus and van Laarhoven had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Keus, Oostvogel, and van Laarhoven. Acquisition of data: Keus, Werner, and van Laarhoven. Analysis and interpretation of data: Keus, Gooszen, and van Laarhoven. Drafting of the manuscript: Keus and Werner. Critical revision of the manuscript for important intellectual content: Gooszen, Oostvogel, and van Laarhoven. Statistical analysis: Keus and van Laarhoven. Obtained funding: Keus and van Laarhoven. Administrative, technical, and material support: Keus, Werner, and van Laarhoven. Study supervision: Keus, Gooszen, Oostvogel, and van Laarhoven.
Financial Disclosure: None reported.
Funding/Support: This study was supported by Foundation “Gasthuisraad” (a foundation for scientific research in the Midden Brabant region of the Netherlands).
Additional Contributions: Eric Buskens, PhD (Julius Center for Health Sciences and Primary Care, University Medical Center, Utrecht), provided statistical help. Mike Clarke, PhD (UK Cochrane Centre, Oxford, England), and Jolanda de Vries, PhD (Department of Psychology and Health, Tilburg University, Tilburg), provided advice in preparing the manuscript.