BACKGROUND: Planning clinical-endpoint trials in patients with HIV remain difficult as long-term follow-up of many patients is required. Cohort studies of patients with HIV can provide key estimates of the likely disease progression, required sample size and follow-up. OBJECTIVES: To verify the assumptions used in designing ESPRIT, a large randomized clinical trial assessing the clinical benefit of interleukin-2 treatment in patients with HIV infection, to use EuroSIDA to mimic the inclusion criterion of ESPRIT in order to compare the observed event rate in ESPRIT with the projected rate in EuroSIDA, and to project the required length of ESPRIT. METHODS: Patients in EuroSIDA who satisfied the ESPRIT recruitment criteria were selected. Patients were followed from baseline to new AIDS or death. RESULTS: The incidence of clinical progression in the selected EuroSIDA patients (N = 4482) was 1.5 per 100 PYFU (95% CI 1.3-1.7), and did not increase with increasing time from baseline, contrary to what was assumed in the design of the ESPRIT trial. In ESPRIT (N = 4150), for which the comparative data remain blinded, the incidence was 1.1 per 100 PYFU (95% CI 0.9-1.3), with no increase over time. The average follow-up required to complete ESPRIT and accrue the 320 events required by protocol would be seven years, 10 months using the projected rates from the EuroSIDA study, and seven years, 11 months if the observed event rate in ESPRIT continued unchanged. LIMITATIONS: Differences between patients recruited to observational studies or clinical trials cannot always be adjusted for. CONCLUSIONS: Event rates in EuroSIDA were similar in the first two years to those used in the design of ESPRIT, but did not increase over time, leading to an increase in the expected duration of ESPRIT. Clinical endpoint trials in HIV infection remain feasible, and large cohort studies are critical to the planning and ongoing assessment of design assumptions in such trials. The underlying assumptions of the clinical trial should be re-examined to ensure the original trial assumptions remain valid.