CD8+ T lymphocytes may mediate important host responses to human immunodeficiency virus (HIV) infection by human leukocyte antigen (HLA)-restricted cytotoxicity and production of soluble HIV suppressor factors. CD8+ lymphocytes are also important for the suppression of many latent pathogens responsible for opportunistic disease in HIV-infected patients. There has been no systematic analysis of the responses of CD8+ lymphocyte counts to antiretroviral therapy. We compared CD8+ lymphocyte responses in seven trials of nucleoside or non-nucleoside analog reverse transcriptase inhibitors and in two trials of ritonavir, a HIV protease inhibitor. Nucleoside analog and non-nucleoside analog reverse transcriptase inhibitor monotherapy resulted in no substantial changes in CD8+ counts relative to baseline or placebo. Combination nucleoside analog therapy resulted in variable peak responses (-145 to +240 cells/mm3), which remained significantly above baseline for 0 to 12 weeks. In contrast, ritonavir monotherapy caused a peak increase of 892 CD8+ cells/mm3, which remained significantly above baseline for 32 weeks. There was a significant correlation (Rs 0.61, p = 0.01) between the peak CD4+ cell and CD8+ responses to each therapy, but no significant correlation between the peak viral load responses and peak CD8+ cell responses. These findings suggest that the greater CD8+ response seen with ritonavir may be due to its specific inhibition of HIV protease and also that the CD8+ response is dependent on new CD4+ cell production. The CD8+ lymphocyte proliferation observed with protease inhibitor therapy could result in improved suppression of HIV replication by the immune system and should be confirmed in a prospective trial comparing protease inhibitors with both nucleoside and non-nucleoside analog therapies.