OBJECTIVES: Describe age-specific mortality patterns of HIV-infected adults in African communities before introduction of HAART. METHODS: Mortality data (deaths and person-years observed) for HIV-positive subjects aged 15-65 from six African community studies in five different countries were pooled, combining information from 1793 seroconverters and 8534 HIV positive when first tested. Age-specific mortality hazards were modelled using parametric regression based on the Weibull distribution, to investigate effects of sex, and site-specific measures of mean age at incidence, crude mortality rate of uninfected, and measures of epidemic maturity. RESULTS: The combined studies yielded a total of 31 777 person-years of observation for HIV-positive subjects, during which time 2602 deaths were recorded. Mortality rates rose almost linearly with age, from below 50/1000 at ages < 20 years, up to 150/1000 at 50 years +. There was no significant difference between men and women in level or age pattern of mortality. Weibull regression analysis suggested that intersite variation could be explained by HIV prevalence trend, and by the ratio of HIV proportional mortality to current HIV prevalence. A model representation was constructed with a common age pattern of mortality, but allowing the level to be adjusted by specifying HIV prevalence indicators. CONCLUSION: The linear age trend of mortality in HIV-infected populations was satisfactorily represented by a Weibull function providing a parametric model adaptable for representing different levels of HIV-related mortality. This model might be simpler to use in demographic projections of HIV-affected populations than models based on survival post-infection.