Thompson W D, Kelsey J L, Walter S D
Am J Epidemiol. 1982 Nov;116(5):840-51. doi: 10.1093/oxfordjournals.aje.a113475.
This study compares matched and unmatched case-control designs in terms of the precision with which one can estimate an exposure-disease association while controlling for the effects of a confounding variable. Provision is made for the cost of the matching process by calculating the reduction in the number of controls that can be studied for fixed study cost. The purpose is to provide epidemiologists with guidelines and quantitative procedures for making rational decisions as to which is the more appropriate study design for specific research problems. The results indicate that when the cost of the matching process is negligible, a matched design is usually more efficient than an unmatched one. The difference in efficiency is generally slight, however, and is found to depend primarily on the strength of the confounder-disease association, the prevalence of exposure, and the strength of the exposure-disease association. When the cost of the matching process is not negligible, a matched design is likely to be less efficient than an unmatched one.
本研究比较了匹配和非匹配病例对照设计在控制混杂变量影响时估计暴露与疾病关联的精度。通过计算在固定研究成本下可研究的对照数量的减少来考虑匹配过程的成本。目的是为流行病学家提供指导方针和定量程序，以便就针对特定研究问题哪种研究设计更合适做出合理决策。结果表明，当匹配过程的成本可忽略不计时，匹配设计通常比非匹配设计更有效。然而，效率差异通常较小，并且发现主要取决于混杂因素与疾病关联的强度、暴露的患病率以及暴露与疾病关联的强度。当匹配过程的成本不可忽略不计时，匹配设计可能比非匹配设计效率更低。
