BACKGROUND: Data on the short-term risk of disease progression in HIV-1-infected children are needed to address the question of when to begin combination antiretroviral therapy. We estimated 12-month risks of progression to AIDS and death, by age and most recent measurement of CD4 T-cell percentage (CD4%) or viral load, in children receiving no antiretroviral therapy or zidovudine monotherapy only. METHODS: We undertook a meta-analysis of individual longitudinal data for 3941 children from eight cohort studies and nine randomised trials in Europe and the USA. Estimates of risk were derived from parametric survival models. FINDINGS: 997 AIDS-defining events were recorded over 7297 person-years of follow-up in the analysis of CD4%, and 284 events over 2282 person-years in the viral load analysis, corresponding to 568 deaths (9087 person-years) and 129 deaths (2816 person-years), respectively. In children older than 2 years, risk of death increased sharply when CD4% was less than about 10%, or 15% for risk of AIDS, with a low and fairly stable risk at greater CD4%. Children younger than 2 years had worse outlook than older children with the same CD4%. Risk of progression increased when viral load exceeded about 10(5) copies per mL, although this association was more gradual compared with CD4%. Both markers had independent predictive value for disease progression; CD4% was the stronger predictor. INTERPRETATION: This information is important for paediatricians making decisions, and for researchers designing trials, about when to initiate or restart antiretroviral therapy.