Terry M B, Neugut A I, Schwartz S, Susser E
Division of Epidemiology, Joseph L. Mailman School of Public Health, Columbia University, New York, NY, USA.
Am J Epidemiol. 2000 Feb 15;151(4):339-45. doi: 10.1093/oxfordjournals.aje.a010212.
When a risk factor influences the intermediary but not progression to the endpoint, it has been shown that the relative risk estimate for the causal intermediate is identical to that for the endpoint under a single pathway framework. When there are multiple pathways, the relative risk estimate for the endpoint is reduced. The authors examine how the reduction of the effect size from a risk factor's association with the causal intermediate to that with the endpoint relates to the proportion of endpoint cases arising through other pathways, and the measure of effect used. For multiple pathways, all measures of effect are reduced and the reduction increases as the proportion of endpoint cases arising through other pathways increases. For single pathways, the relative rate ratio and odds ratios are reduced. In particular, the reduction in the odds ratio may be dramatic because of the commonness of causal intermediates relative to the endpoint. Comparisons of causal intermediate studies with those for the endpoint should consider the influences of multiple pathways, the prevalence of the causal intermediate, the measure of effect used, and the multiple effects a risk factor may have along the pathway when interpreting the differences observed across the causal chain.
当一个风险因素影响中间变量但不影响向终点的进展时，研究表明，在单一途径框架下，因果中间变量的相对风险估计与终点的相对风险估计相同。当存在多条途径时，终点的相对风险估计会降低。作者研究了从风险因素与因果中间变量的关联到与终点的关联，效应大小的降低如何与通过其他途径产生的终点病例比例以及所使用的效应度量相关。对于多条途径，所有效应度量都会降低，并且随着通过其他途径产生的终点病例比例增加，降低幅度也会增大。对于单一途径，相对率比和优势比会降低。特别是，由于因果中间变量相对于终点的普遍性，优势比的降低可能会很显著。在解释因果链中观察到的差异时，因果中间变量研究与终点研究的比较应考虑多条途径的影响、因果中间变量的患病率、所使用的效应度量以及风险因素在途径中可能产生的多种效应。
