Statistical problems in epidemiologic studies of the natural history of disease

The development of effective disease prevention and treatment programs depends on an understanding of the natural history of disease. A conceptual framework is presented for disease natural history and consists of an asymptomatic period of disease followed by a period of symptomatic disease. The focus is on epidemiologic studies for identifying risk factors of the onset of asymptomatic disease, for identifying cofactors of progression to symptomatic disease, and for estimating the duration of the asymptomatic period. The strengths and limitations of various epidemiologic study designs and sources of epidemiologic data are considered for characterizing disease natural history. Issues in the interpretation and analysis of natural history parameters of disease estimated from cross-sectional, prevalent cohort, cohort, and matched case-control studies are considered. The issues and analytic methods are illustrated with studies of the acquired immunodeficiency syndrome (AIDS) and cervical cancer. Based on these analytic methods, an estimate of the incubation period distribution of AIDS is given.