Only incomplete data are available to guide decision on anti-HIV treatment. A British HIV Association consensus is that guidance must draw on other evidence besides the randomised trial. Marker studies, work on disease pathogenesis and viral dynamics, and expanding knowledge of resistance patterns mean that the approach to therapy is constantly evolving. There is a need for well-informed dialogue between HIV-infected patient and physician to achieve rational, individualized treatment. However, the following broad principles have a wide consensus amongst HIV-treating physicians in the UK: (1) treatment should be offered before substantial immunodeficiency ensues; (2) initial treatment should include combinations of at least two drugs; (3) switches in therapy should involve substitution or addition of at least two new agents; (4) viral load and CD4 measurements are essential; (5) reduction in viral load to below the detection level of a sensitive assay represents the optimal treatment response and failure to achieve or sustain this control should prompt consideration of therapy modification. This response seems to be achieved most reliably with combinations of two nucleoside analogues plus a third agent (a protease inhibitor, a non-nucleoside reverse-transcriptase inhibitor, or a third nucleoside analogue) or of two protease inhibitors.