In Italy, antiretroviral combination therapy was adopted in mid-1995 and protease inhibitors in mid-1996. To conduct a prospective, population-based, observational study to evaluate the effect of these therapies on the survival of persons with AIDS (PWA). PWA living in the Tuscany region diagnosed between 1985 and 31 March 1997 (National AIDS Registry) were studied. Information on antiretroviral drugs, prophylactic treatment, CD4 cell count, and AIDS-defining illnesses was collected for PWA still alive at 1 January 1996 and those diagnosed thereafter (analysis cut-off date, 30 November 1997). Kaplan-Meier curves were calculated by year of diagnosis. A Cox model was then used to estimate the adjusted (by sex, age, HIV exposure category, CD4 cell count, type and number of AIDS-defining illnesses) relative hazard (RH) of death by year of diagnosis and calendar date (considered as a time-dependent variable). Similar analyses were repeated for PWA diagnosed after 1989, having been stratified by disease-specific AIDS condition. A final analysis was performed for PWA still alive at 1 January 1996 or diagnosed thereafter for estimating the effect of single, double and triple combination therapy (time-dependent variables), having adjusted for the above variables and for prophylactic treatment. A total of 1683 (79.5%) out of 2118 PWA died before 1 December 1997. Use of more potent combination therapies, including protease inhibitors, greatly increased during 1997. Median survival was 2.9, 12.3, 13.4, 11.4 and 17.6 months for diagnoses before 1987, in 1987-1990, 1991-1993, 1994 and 1995, respectively; an estimated 62% of those diagnosed in 1996-1997 had survived 15 months after diagnosis. The Cox model showed a trend of decrease of RH for calendar time starting in the first half of 1996, compared with 1994. When stratifying by specific AIDS-defining disease there was no statistically significant evidence that the improved overall survival was due to increased survival only for certain diseases. The final multivariate analysis for the 771 PWA still alive at 1 January 1996 or diagnosed thereafter estimated significant RH < 1.0 for double and triple therapy (RH, 0.61 and 0.36, respectively) compared with no therapy. A significant reduction in risk of death after AIDS was observed from the second half of 1996, apparently due to the widespread use of antiretroviral combination therapies.