Competing risk adjustment reduces overestimation of opportunistic infection rates in AIDS

Abstract 

To illustrate the importance of adjusting the estimates of cumulative incidence of acquired immunodeficiency syndrome (AIDS) related illnesses for competing risk of other causes of death, we compared unadjusted and adjusted (for competing events) incidence estimates for four AIDS illnesses: pneumocystis cavinii pneumonia (PCP), mycobacterium avium complex (MAC), cytomegalovirus (CMV), and esophageal candidiases. The study population was patients followed by the Johns Hopkins Hospital AIDS Service between 1989 to 1995. Ratios of 4 year unadjusted incidence estimates to 4 year adjusted incidence estimates for the four diseases ranged from 1.38 to 1.86, corresponding to cumulative death rates of 61% to 69%. For CMV, the ratios of 4 year unadjusted to adjusted incidence estimates for five groups of patients ranged from 1.5 to 2.33, corresponding to cumulative death rates of 48% to 78%. We conclude that ignoring the competing risk of death can result in substantial overestimation of disease occurrence, which may give misleading results in estimating and comparing the occurrence of a disease of interest.

Keywords:  Natural history of AIDS, Kaplan-Meier estimates, Cox proportional hazards model, Competing risks, Multiple decrement model