VALIDATION AND REFINEMENT OF THE ACTG STAGING SYSTEM FOR AIDS-ASSOCIATED KAPOSI'S SARCOMA (KS)

Abstract

Background: A KS staging classification that correlates features of the disease with prognosis would facilitate the identification of optimal therapeutic strategies. Recognition of the intimate link between KS and HIV-1 infection led, in 1988, to proposal by the ACTG Oncology Committee of a KS staging system that classified patients (pts) according to tumor extent (T), immune system status (I), and evidence for HIV-1-associated systemic illness (S). Methods: HIV seropositive pts with biopsy-proven KS were staged as either good- or poor-risk using TIS criteria prior to entering ACTG KS treatment trials and were followed for survival. Associations between survival and the TIS variables were evaluated by both univariate (logrank test) and multivariate (Cox proportional hazard regression) analyses. Results: 294 pts enrolled in 8 trials from 4/89 to 1/95. Median f/u was 11 months, and 47.3% of pts were dead at the time of analysis. The numbers of pts with good and poor risk features for each of the TIS variables were 104 and 185 for T, 84 and 210 for I (using 200 CD4 cells/ µL as a cutoff), and 53 and 240 for S. Logrank tests showed a significantly shorter estimated median survival for pts in the poor risk category for each of the TIS variables. Cox proportional hazards regression yielded a final model that included I, T and an interaction variable (IxT). Examination of crude death rates in different CD4 strata suggested that a CD4 count of 150/ µL was a better cutpoint for predicting survival than 200/ µL, and the model selected using the 150 cell cutoff included only I and T, without a significant interaction variable. Among pts with poor risk T stage, significant survival differences were not detected for pts with and without pulmonary KS. Conclusions: The ACTG TIS staging system for AIDS-associated KS is a valid predictor of survival, but a modified system, using 150 CD4 cells/µL to distinguish good- and poor-risk I stage, yields a simplified and more predictive model. Tumor stage added to CD4 counts in predicting survival in pts whose immune function was least impaired. Additional staging refinements may be possible by including other variables such as viral load, markers of immune activation, or more specific poor-risk tumor features.