Meta-analysis of randomized controlled trials as a method of estimating rare complications of non-steroidal anti-inflammatory drug therapy.

摘要: Abstract The design of randomized controlled trials to assess the efficacy pharmacological measures for prevention gastrointestinal side-effects anti-inflammatory drugs requires an accurate estimate excess risk under conditions. Photocopies 952 trial publications were obtained after scanning titles and abstracts a MEDLINE computer search, 427 excluded obvious reasons, 525 again photocopied obliterating source results. Selection criteria were: presence non-anti-inflammatory drug control group; at least 4 days therapy; 3 without before randomization; no complicating background drugs; mention side-effects; clear differentiation complications. Observer error, with two independent readings, inclusion suitability in study was 19% Methods 9% Results. For 44 aspirin trials, mean therapy duration 357 days; unweighted rate difference between groups ( +/- 1 S.E.M.) ulcer 0.006 0.003, gross haemorrhage 0.002 unspecified gastric symptoms 0.03 0.01. In 123 non-aspirin non-steroidal (NA-NSAID) 67 0.0005 0.0003, 0.007 0.004 0.02 0.005. Risk differences also pooled using DerSimonian Laird method, which weights studies inversely according variance. Using this only (NA-NSAIDs) found be statistically significant. Longer have higher differences. Randomized determine prophylactic against (that is, demonstrate reduction group controls) would require 190 patients each NA-NSAIDs 2-6 months; 950 subjects needed detect 50% reduction. that controls on more than 6 months 700 3346 Such are feasible.