Models of Survival in HIV Infection and Their Use in the Quantification of Treatment Benefits

Because acquired immunodeficiency syndrome (AIDS) is a shifting endpoint and sufficient follow-up data now allow modeling of survival time (i.e., time from human immunodeficiency virus (HIV) seroconversion to death), the authors evaluated non-parametric and parametric models of mortality with the use of data from 554 seropositive participants in the Vancouver Lymphadenopathy-AIDS Study. The authors then applied these models to quantify treatment benefits at the national level in Canada, using back-calculation and forward-projection based on death registries. The study revealed that the lognormal model better describes survival time than the Weibull model. Relative to observations prior to 1987, later observations (in the era of treatment) revealed a statistically significant change in disease progression: the median survival time increased from 10.1 to 12.0 years, but no further survival improvements were observed in the early 1990s. Concurrent with the increase in availability of treatment, the authors have observed pronounced treatment benefits at the national level: prior to 1995, approximately 1,500 deaths were prevented and 4,200 person-years of life were saved. Also, mortality rates were observed to level off in the mid-1990s due to the shape of the historical HIV infection curve and the accumulating availability of treatment in Canada. Am J Epidemiol 1998; 148: 487–96.