Brookmeyer R, Gail M H
Department of Biostatistics, Johns Hopkins University School of Hygiene and Public Health, Baltimore, Maryland 21205.
Biometrics. 1987 Dec;43(4):739-49.
Several natural history studies of the Acquired Immunodeficiency Syndrome (AIDS) have been based on the follow-up of individuals infected with the AIDS virus prior to enrollment. The natural time scale for studying the preclinical course of AIDS is the time since first infection; however, in these studies the time at infection was a random unknown quantity. The biases inherent in using follow-up time instead of time from infection are investigated for estimation of both the cumulative distribution function and the hazard ratio for proportional hazards models with both fixed and time-dependent covariates. Although the magnitudes of the biases depend on the shape of the epidemic curve, a number of bounds on the biases are established. These results are useful for interpreting prevalent cohort studies and then comparing them with studies on newly infected individuals in order to assess consistency of results across studies.
几项关于获得性免疫缺陷综合征（艾滋病）的自然史研究是基于对入组前感染艾滋病病毒个体的随访。研究艾滋病临床前期病程的自然时间尺度是自首次感染以来的时间；然而，在这些研究中，感染时间是一个随机的未知量。对于具有固定协变量和时间依存协变量的比例风险模型，在估计累积分布函数和风险比时，研究了使用随访时间而非感染时间所固有的偏差。尽管偏差的大小取决于流行曲线的形状，但还是建立了一些偏差的界限。这些结果有助于解释现患队列研究，并将其与新感染个体的研究进行比较，以评估不同研究结果的一致性。
