PROJECTING DISEASE WHEN DEATH IS LIKELY

Projecting disease incidence, prevalence, and net morbidity is often n eeded when individuals are likely to die, either disease free or after the disease has developed. Examples of this include remission of canc er or heart disease in elderly people who can die from these or other causes and occurrence of a particular acquired immune deficiency syndr ome illness in human immunodeficiency virus type 1 (HIV-1) disease. De ath is not an ancillary event but, rather, indicates either an end to disease morbidity or an end to risk to ever develop that disease. Thus , time to disease survival analyses that censor disease-free individua ls at death can produce misleading results. This paper describes sever al useful quantifications of disease and death for this setting. A par adigm that utilizes Kaplan-Meier functions to estimate these quantitie s is introduced. The approach anchors on a four-stage disease/death mo del: stage A, living without disease; stage B, dead without ever devel oping disease; stage C, developed the disease and living; and stage D, dead after developing the disease. An application is made to projecti ng cytomegalovirus disease in a cohort of HIV-1-infected users of zido vudine and Pneumocystis prophylaxis from the Multicenter AIDS Cohort S tudy (MACS) during 1989-1993. At 3 years after a CD4(+) count below 10 0/mu l, a man had an 18.7%, 46.3%, 5.3%, or 29.9% chance, respectively , to be in stage A, B, C, or D. This man, on average, had 0.28 years o f cytomegalovirus morbidity during these 3 years.