DESIGN: Mortality within the Swiss HIV Cohort Study for the years 1990-2001 was compared with the mortality of the general Swiss population. METHODS: Standardized mortality ratios (SMR) and life tables were calculated for strata defined by combinations of gender and HIV transmission group. The effect of dropouts was investigated with a sensitivity analysis and by analysing CD4 cell counts before dropout. RESULTS: During the study period 10 977 individuals had at least one cohort visit with a median observation time of 46 months. A total of 3630 patients died and 2290 dropped out. SMR decreased from 79.3 [95% confidence interval (CI), 77.2-81.5] before the introduction of highly active antiretroviral treatment (HAART) in 1996 to 15.3 (95% CI, 14.2-16.4) thereafter. For persons who acquired HIV infection by injecting drug use (IDUs), the SMR decreased from 98.2 (95% CI, 94.9-103.5) to 40.9 (95% CI, 37.0-44.8) after 1996; for all other HIV transmission groups the SMR decreased from 69.2 (95% CI, 66.9-71.6) to 9.4 (95% CI, 8.5-10.4). Thus, IDUs had significantly lower survival in comparison with other patient groups after 1996. Patients who had started HAART during the time period in which this treatment was available, had even lower SMRs. CONCLUSIONS: Although overall survival has improved considerably since the introduction of HAART, cohort life expectancy remains below that of the Swiss population. We noted, however, substantial differences in mortality among subgroups, and the results indicate that the additional risk related to injection drug use before 1996 had been masked by HIV-associated mortality.