Title: Influence of dialysis membranes on outcomes in acute renal failure
Abstract: To the Editor: In their recent meta-analysis published in Kidney International, Subramanian, Venkataraman, and Kellum1.Subramanian S. Venkataraman R. Kellum J.A. Influence of dialysis membranes on outcomes in acute renal failure: A meta-analysis.Kidney Int. 2002; 62: 1819-1823Abstract Full Text Full Text PDF PubMed Scopus (105) Google Scholar used a fixed effects model to combine the results of the studies addressing the impact of synthetic dialysis membranes on survival in acute renal failure. In addition, their primary analysis pooled results from an observational study with those from randomized/quasirandomized trials. In our opinion both of these approaches are inappropriate. Fixed effects models assume that the effect of treatment on outcome is homogeneous across the pooled trials. This assumption can be tested statistically by the Q statistic. Using data from the authors' paper, we replicated their analyses. The calculated Q statistic was 13.0 (P = 0.07), indicating significant heterogeneity2.Clarke M. Oxman A.D. Cochrane Reviewers Handbook 4.1.5; Section 8,.in: Clarke M. Oxman A.D. The Cochrane Library. Oxford, 2002Google Scholar. This suggests that a random effects model should be used, resulting in a pooled odds ratio for mortality of 1.28, which is not statistically significant (95% CI, 0.86 to 1.92; P = 0.15). Even if a fixed effects model was used, meta-analyses should include observational studies only when randomized trials are impractical3.Clarke M. Oxman A.D. Cochrane Reviewers Handbook 4.1.5; Section 4,.in: Clarke M. Oxman A.D. The Cochrane Library. Oxford, 2002Google Scholar. This suggests that the trial by Neveu et al [author query: please cite Neveu et al in reference] should be excluded, which results in a pooled odds ratio for mortality of 1.11 (95% CI, 0.82 to1.50). Since meta-analyses cannot replace single large, methodologically rigorous randomized trials, their results are only convincing when they are unchanged across a spectrum of sensitivity analyses. As Subramania, Venkataraman, and Kellum1.Subramanian S. Venkataraman R. Kellum J.A. Influence of dialysis membranes on outcomes in acute renal failure: A meta-analysis.Kidney Int. 2002; 62: 1819-1823Abstract Full Text Full Text PDF PubMed Scopus (105) Google Scholar point out, the pooled odds ratio for mortality was materially affected by the subgroup of cellulose membrane studied. In addition, the results are dependent on the type of analysis performed and on the inclusion of a single observational study, and thus the authors' conclusion (that synthetic membranes reduce mortality) should be viewed with skepticism2.Clarke M. Oxman A.D. Cochrane Reviewers Handbook 4.1.5; Section 8,.in: Clarke M. Oxman A.D. The Cochrane Library. Oxford, 2002Google Scholar.