Abstract

BACKGROUND:

AIDS-associated Kaposi's sarcoma remains common in individuals with HIV-1 infection in the era of highly active antiretroviral therapy (HAART). We developed a simple model for predicting mortality on the basis of clinical characteristics present at the time of diagnosis of Kaposi's sarcoma.

METHODS:

Of 5873 individuals with HIV-1 infection, 326 (6%) developed Kaposi's sarcoma; for 262 (80%) this was their first AIDS-defining illness. We did univariate and multivariate Cox regression analyses to identify covariates predictive of overall survival and validated our model with an independent data set of 446 patients with Kaposi's sarcoma.

RESULTS:

In the primary model, we developed a prognostic score from 0 to 15 starting at 10. Having Kaposi's sarcoma as the AIDS-defining illness (-3 points) and increasing CD4 count (-1 point for every complete 100 cells per mm3) improved prognosis; age of 50 years or older (2 points) and having another AIDS-associated illness at the same time (3 points) conveyed a poorer prognosis. In individuals with prognostic scores of 0, 5, 10, and 15, probability of survival at 1-year was 0.993, 0.967, 0.834, and 0.378, and at 5 years was 0.984, 0.918, 0.631, and 0.084, respectively. Increasing prognostic score by 1 increased 1-year death hazard ratio by 40% (95% CI 28-53%; bootstrapped hazard ratio 1.39, 1.25-1.51). The index had concordance of 76.8% (71.7-82.3).

INTERPRETATION:

We identified four prognostic factors that can be used to obtain an accurate prognostic index at diagnosis of AIDS-associated Kaposi's sarcoma. This index is widely applicable and can be used to guide therapeutic options.