This article discusses the design of an ongoing open-label, randomized trial comparing three strategies of initial and subsequent HIV therapy in terms of long-term immunological and virological effect. The three treatment arms are (1) didanosine (ddI) plus stavudine (d4T) plus efavirenz (EFV) followed by zidovudine (ZDV) plus lamivudine (3TC) plus abacavir (ABC) plus nelfinavir (NFV); (2) ddI plus d4T plus NFV followed by ZDV plus 3TC plus ABC plus EFV; (3) ddI plus d4T plus EFV plus NFV followed by ZDV plus 3TC plus ABC plus saquinavir plus ritonavir. The primary objective is to determine whether it is best to start with a protease inhibitor (PI)-containing regimen, a non-nucleoside analogue reverse transcriptase inhibitor (NNRTI)-containing regimen, or with a regimen containing both a PI and an NNRTI. The aim is to recruit over 1000 patients followed for at least 3 years. The entry criteria are broad with no restriction on stage of disease, CD4 count, or HIV viral load. The criteria for therapeutic failure determining change of treatment are not defined and are left to the clinicians, but randomization is stratified by country and by the current criteria used for changing treatment. We describe the rationale behind various aspects of the design and discuss the complexities involved in undertaking such a large trial in HIV-infected patients from 180 clinical sites in 17 countries. Control Clin Trials 2001;22:160–175