Extract

We thank Lam and Owen for their interest in our article. In the network analysis we performed, we indeed ignored that a limited set of pairs of treatments were clustered in a multi-group trial. There are theoretical advantages of correcting for multi-arm trials, and a Bayesian approach might be a suitable tool to do so in a flexible way. The advantages become especially clear when there are a large number of multi-arm trials. Our network meta-analysis, however, included only four multi-arm trials.

We disagree with the claim that the analysis performed by Lam and Owen is ‘concordant’ with the neutral results from the Prevention Regimen for Effectively Avoiding Second Stroke (PROFESS) regarding the comparison of aspirin and dipyridamole (A + D) vs. clopidogrel (C). The PROFESS trial showed a hazard ratio (HR) of 0.99 (95% CI 0.92–1.07) with A + D vs. C for the composite endpoint stroke, MI, and vascular death. The neutral results of PROFESS are based on the estimated HR, which is nearly equal to 1, and the small width of the confidence interval. We cannot infer similar clinical efficacy from the network meta-analysis according to the method proposed by Lam and Owen, nor our meta-analysis. Neither our results, OR = 0.84 (0.73–0.97), nor the results from Lam and Owen, OR = 0.86 (0.75–1.01), suggest a point estimate of 1 and a narrow confidence interval around 1.

You do not currently have access to this article.