摘要：When I first heard of meta-analysis in the mid-1980s,I was suspicious.After slaving for 3 years on a case-control study of renal adenocarcinoma that yielded just 3 papers, the idea of someone else getting a quick publication by quantitatively pooling my findings with other studies seemed parasitic, if not plagiarist. On the other hand, I was aware that reviews of scientific literature often simply bolstered the authors’ opinions by citing supportive studies and glossing over, dismissing or overlooking unsupportive studies. My suspicion was reduced by Meir Stampfer’s important paper pooling results of streptokinase trials. I decided meta-analysis might be appropriate for drug trials. However, I was troubled by the conflict between the apparent goal of metaanalysis –- to achieve statistical significance or tight confidence intervals — and Rothman’s and others’ criticism of p-values. Meta-analysis seemed to me to be a fancy name for combining results to get a better p-value.
