Meta-Analysis of Randomised Trials on Laparoscopic Versus Open Surgery for Acute Appendicitis: Has Firm Evidence been Reached?

Excerpt

In their interesting meta-analysis, Ohtani and colleagues have assessed the benefits and harms of laparoscopic surgery versus open surgery for acute appendicitis.1 One of the most patient-centred outcomes of the meta-analysis is wound infection, and wound infection was also the primary outcome in a Cochrane review comparing laparoscopic versus open appendectomy.2 Ohtani and colleagues showed that laparoscopic appendectomy reduced the proportion of patients with wound infection with 54 % when compared with open surgery (relative risk (RR), 0.46; 95 % CI, 0.34–0.64; 32 trials; 4,936 patients).1 However, randomised clinical trials and meta-analyses are at risk of systematic errors due to design problems and risk of bias.3 To minimise the risk of systematic errors, we assessed the importance of proper randomisation, and a very similar risk reduction in wound infection was found when only the trials with adequate randomisation (both adequate generation of allocation sequence and allocation concealment) were included in the meta-analysis (RR, 0.46; 95 % CI, 0.32–0.66; 23 trials; 3,727 patients). In addition to adequate randomisation, other components like adequate blinding, adequate dealing with missing data, and avoidance of selective outcome reporting are important to assess risk of bias in trials. Sufficient blinding in trials comparing laparoscopic appendectomy with open appendectomy is very challenging, if not impossible. When assessing wounds for infection the different types of wounds (number, length and localisation) immediately reveal the type of surgery performed. Only one trial attempted to achieve adequate blinding as at the end of the procedure three wound dressings and an abdominal binder were applied to every patient to blind the patient, the nursing and medical staff and the independent data collector.4 While this blinding method probably approaches the best we can achieve, and may be sufficient for certain outcomes like for example ‘pain assessment’ or ‘length of hospital stay’, it might still not be adequate to assess ‘wound infection’, and these trials might therefore all be at risk of bias regarding wound infection. …