Background: We performed a meta-analysis on granisetron in the prevention of postoperative nausea and vomiting (PONV) and further investigated whether total results and the dose-response characteristics may be significantly affected by a single centre.
Methods: Systematically searched randomised controlled trials (RCT) using granisetron for the prevention of PONV after general anaesthesia were included in the analysis. The pooled relative risks (RR) and numbers needed to treat (NNT) with their corresponding 95%-confidence intervals (CI) were calculated. For all centres, one dominating centre and other centres pooled, comparisons were performed according to all doses, low dose (<or=20 microg/kg) and high dose (>20 microg/kg) granisetron.
Results: A total of 27 RCT with 2938 patients were included in the analysis. RR (CI) to suffer from PONV with granisetron when all comparisons were considered was 0.46 (0.39-0.54), 0.7 (0.6-0.81) and 0.34 (0.28-0.41) for all doses, low and high dose, respectively. RR of the dominating centre (1867 patients) were significantly better compared to the remaining centres (1071 patients), with 0.41 (0.34-0.49) and 0.60 (0.49-0.73), respectively. In the dominating centre low dose granisetron was ineffective with a RR of 0.84 (0.68-1.04), while high dose granisetron led to a strong decrease with a RR of 0.30 (0.26-0.36). In contrast, the RR of other centres pooled for low and high dose granisetron were comparable with 0.62 (0.49-0.79) and 0.56 (0.42-0.75), respectively.
Conclusions: Overall results and dose-response characteristics of meta-analyses may be significantly altered by one dominating centre. Further, if data of a dominating centre do not appear to be valid for other centres, it may seem advisable to either exclude them from the analysis or to perform sub-group analyses so that results without the data from the dominating centre are available.