Spurious precision? Meta-analysis of observational studies.

摘要: In previous articles we have focused on the potentials, principles, and pitfalls of meta-analysis randomised controlled trials.1 2 3 4 5 Meta-analysis observational data is, however, also becoming common. a Medline search identified 566 (excluding those published as letters) in 1995 indexed with medical subject heading (MeSH) term “meta-analysis.” We randomly selected 100 these examined them further. Sixty reported actual meta-analyses, 40 were methodological papers, editorials, traditional reviews (1). Among about half based studies, mainly cohort case-control studies interventions or aetiological associations. View this table: Characteristics from keyword “meta-analysis” The trial is principal research design evaluation interventions. However, hypotheses—for example, relating common exposures to occurrence disease—cannot generally be tested experiments. Does breathing other people's tobacco smoke cause lung cancer, drinking coffee coronary heart disease, eating diet rich saturated fat breast cancer? Studies such “menaces daily life”6 use designs examine presumed biological mechanisms laboratory. situations risks involved are small, but once large proportion population exposed, potential public health implications associations—if they causal—can striking. Analyses role effectiveness research.7 The evidence available clinical trials will rarely answer all important questions. Most conducted establish efficacy safety single agent specific situation. Owing limited size trials, less adverse effects drugs may only detected …