OBJECTIVE To compare the clinical response among patients who initiate protease inhibitor therapies with different virological potency. METHODS We analysed patients who started indinavir, ritonavir or saquinavir hard gel capsule (hgc) as part of at least triple therapy during prospective follow-up within the EuroSIDA study. METHODS Changes in plasma viral load (pVL) and CD4 cell count from baseline were compared between treatment groups. Time to new AIDS-defining events and death were compared in Kaplan--Meier models, and Cox models were established to further assess differences in clinical progression (new AIDS/death). Adjustment was made for differences in baseline parameters, in particular pVL, CD4 cell count, and region of Europe. RESULTS A total of 2708 patients (median follow-up: 30 months) were included, of which 556 started ritonavir (21%), 1342 indinavir (50%), and 810 saquinavir hgc (30%). The three groups were fairly evenly balanced at baseline regarding CD4 count, previous diagnosis of AIDS and pVL, After 12 months, the median changes in CD4 cell count were 90, 96 and 74 x 10(6) cells/l, respectively;P < 0.001, the proportions of patients with pVL < 500 copies/ml were 47, 54 and 41%; P < 0.001, and the proportions with clinical progression were 11.9, 9.2 and 11.9%, respectively; P = 0.20 (log-rank test). In multivariate models the relative risk of clinical progression for indinavir compared with saquinavir hgc was: 0.77 (0.60--0.99); P = 0.043, and for ritonavir 0.83 (0.62--1.11); P = 0.20. CONCLUSIONS Saquinavir hgc was associated with an inferior long-term clinical response relative to indinavir, which was consistent with the observed differences in virological and immunological responses.