Study selection
As reported in Figure 1, there were 294 studies found from the five databases, with another 14 articles from other sources. After reading the titles and abstracts and eliminating duplicates, 35 articles were selected for full-text reading, of which 15 were included in the analysis 15,16,27-39.
Study characteristics
In total, three studies compared immediate cholecystectomy performed within 24 hours of admission to delayed cholecystectomy 27-29 and six studies reported a comparison between early and delayed cholecystectomy, with early timing within 48 and 72 hours of admission 15,30-34. However, the studies that compared immediate to delayed cholecystectomy included in the immediate group, patients with symptoms persisting up to 1 week 27 or up to 96 hours 28, while in one study, it was not possible to assess the exact timing related to the onset of symptoms 29. Moreover, six additional studies were found that compared immediate cholecystectomy, performed within 72 hours of the onset of symptoms, to delayed cholecystectomy 16,35-39. The inclusion among these studies of one study in which patients were randomized to be operated on before and after 72 hours of the onset of symptoms required discussion 36. As all patients received medical treatment, the timing of 72 hours after the onset of symptoms was considered to be delayed. The characteristics and the results of the studies reporting data on immediate versus delayed cholecystectomy are summarized in Table 2.
Table 2: Characteristics of studies that reported post-operative complications in comparing immediate cholecystectomy performed within 24 hours of admission or within 72 hours of the onset of symptoms compared to delayed cholecystectomy.
In total, 20 studies were excluded for the following reasons: the absence of or uncertainty about randomization 40–44, lack of reporting or uncertainty about timing 40, 45–48,52, lack of reporting or uncertainty about diagnostic criteria for acute cholecystitis 40, 44, 45, 49-51,53, limitation of the population study to the first episode and inclusion of biliary colic cases 51, the inclusion of only patients with symptoms persisting for more than 72 hours 52, exclusion of elderly patients 53, lack of reporting post-operative complications or lack of specifying whether the reported complications were primarily post-operative 40, 47, 48, 53–57, or the study reported interim results of a randomized study 58.
Risk of bias within studies
Only five studies reported a computer-generated randomization sequence allowing a low risk of allocation concealment bias 27,28,32,35,38, seven studies reported an odd-even simple randomization method 16,29,30,33,34,36,39 and three studies did not report data on randomization method 15,31,37, with, as a result, an uncertain or high risk of selection bias. None of the studies reported any blinding. While the absence of information on blinding of operators and patients could be considered to be at low risk of bias (being hardly feasible in surgical trials), the absence of blinding of the outcome assessment may be of concern and the risk should be considered as uncertain 59. The risk of bias related to the missing outcome data, the measurement of the outcome and selective reporting was considered low as for all the studies.
Synthesis of results
The risk of post-operative complications was not significantly different after immediate cholecystectomy performed within 24 hours of admission compared to delayed cholecystectomy with RR 1.89 [95% CI 0.76; 4.71], as well as for early compared to delayed cholecystectomy with RR 1.32 [95% CI 0.82; 2.11]. Heterogeneity was moderate in both comparisons. The test for differences between sub-groups did not provide a significant result with p = 0.49 (Figure 2).
Within the second analysis, pooling data showed a statistically significant reduction in the rate of post-operative complications with RR 0.57 [95% CI 0.37; 0.89] after immediate cholecystectomy performed within 72 hours of the onset of symptoms compared to delayed cholecystectomy. No heterogeneity was found. For the sub-group comparison, eight studies comparing early with delayed cholecystectomy, the early timing of which was reported to be up to 1 week of the onset of symptoms, were selected among the included studies 15,27,28,30-34. Pooled data from these studies did not show a significant difference between early and delayed cholecystectomy with RR 1.28 [95% CI 0.84; 1.97], heterogeneity was found to be moderate. Moreover, the comparison between the two sub-groups, immediate versus delayed and early versus delayed, showed a statistically significant difference with p = 0.01, giving strength to the results on immediate cholecystectomy (Figure 3).
A total of 10 studies reported data on biliary injury 15,27,28,30-34,36,38, but only four were from the immediate cholecystectomy sub-groups 27,28, 36,38. When considering immediate cholecystectomy performed within 24 hours of admission compared to delayed cholecystectomy 27,28, the RR was not estimable because one study reported no biliary injury 27 and one reported only one case in the immediate cholecystectomy group 28. When considering immediate cholecystectomy performed within 72 hours of the onset of symptoms 36,38, no significant difference was found with RR 0.23 [95% CI 0.04; 1.34].
All included studies reported data on conversion. No significant difference was found when comparing immediate cholecystectomy performed within 24 hours to delayed cholecystectomy with RR 1.38 [95% CI 0.75; 2.54], while immediate cholecystectomy performed within 72 hours of the onset of symptoms appeared to significantly reduce the rate of conversion compared to delayed cholecystectomy with RR 0.53 [95% CI 0.32; 0.89].
Reported mortality was very low and data were not sufficient to perform a meta-analysis on this variable.
Sensitivity analysis
The sensitivity analysis performed by applying the random-effects model revealed similar results compared to the fixed-effect model, for both the definitions of immediate cholecystectomy respectively within 24 hours of admission and within 72 hours of onset of symptom. For the former, RR of immediate versus delayed cholecystectomy was 1.61 [95% CI 0.44; 5.85], RR of early versus delayed cholecystectomy was 1.36 [95% CI 0.73; 2.55] and p = 0.02 for the sub-group comparison. For the latter, RR of immediate versus delayed cholecystectomy was 0.58 [95% CI 0.37; 0.90], RR of early versus delayed cholecystectomy was 1.27 [95% CI 0.78; 2.06] and p = 0.82 for the sub-group comparison.
According to the literature finding, no other sensitivity analysis was feasible for comparisons based on immediate cholecystectomy performed within 24 hours of admission. The two further planned sensitivity analysis were therefore limited to the comparisons based on immediate cholecystectomy performed within 72 hours of the onset symptoms. By excluding the study, which was a matter of discussion 36, sensitivity analysis confirmed the results on the risk of post-operative complications with RR of immediate versus delayed cholecystectomy of 0.59 [95% CI 0.38; 0.92], RR of early versus delayed cholecystectomy of 1.28 [95% CI 0.84; 1.97] and p = 0.01 for the sub-group comparison.
When investigating the effect of the inclusion of studies that had been omitted because of incomplete information regarding the exact timing of cholecystectomy, the absence of reported criteria for the diagnosis of acute cholecystitis, or other methodological aspects of the studies, the sensitivity analysis confirmed the results of the main comparison with RR of immediate versus delayed cholecystectomy of 0.57 [95% CI 0.37; 0.89], RR of early versus delayed cholecystectomy of 1.09 [95% CI 0.82; 1.44] and p = 0.02 for the sub-group comparison.
Quality of evidence
As the present meta-analysis only included randomized studies, the level of evidence should first be high according to the GRADE rule. The only domain that should be considered in rating down the quality of evidence has been the potential risk of bias. None of the dedicated domains could allow the level of evidence to be rated up. The risk of bias involved not only the method of randomization, allocation concealment and blinding but also the lack of pathological confirmation of a diagnosis of acute cholecystitis. While non-acute cholecystitis is expected at the time of surgery in the delayed group, an unknown rate of non-acute cases may be included in the immediate group because of a diagnostic error that may lead to an overestimation of the benefit of an immediate cholecystectomy.
A large magnitude of the effect was not found, the dose-response gradient was not applicable and no potential residual confounders would decrease the magnitude of the effect. The funnel plot shown in Figure 4 illustrates the low risk for potential publication bias in this study. Overall, the quality of evidence of this meta-analysis should be considered moderate.