OBJECTIVE To determine whether the total lymphocyte count (TLC) accurately predicts a low absolute CD4+ T-cell count and CD4+ percentage in persons infected with human immunodeficiency virus (HIV). METHODS Retrospective analysis of data collected in the US Air Force HIV Natural History Study. METHODS Military medical center that performs annual medical evaluation of all HIV-infected US Air Force personnel. METHODS A total of 828 consecutive patients with no prior history of zidovudine use, evaluated from January 1985 through July 1991. For patients with multiple observations over time, a single data point within each 6-month interval was included in the analysis (N = 2866). RESULTS The sensitivity, specificity, and likelihood ratio (LR) of the TLC, in the range of 1.00 x 10(9)/L to 2.00 x 10(9)/L, in predicting an absolute CD4+ T-cell count less than 0.20 x 10(9)/L or a CD4+ percentage less than 20% were calculated. In addition, the LR and pretest probability of significant immunosuppression were used to calculate posttest probabilities of a low CD4+ count for a given TLC value. The LR of the TLC in predicting an absolute CD4+ count < 0.20 x 10(9)/L increased from 2.4 (95% confidence interval, 2.2 to 2.5) for all TLCs less than 2.00 x 10(9)/L, to 33.2 (95% confidence interval, 24.1 to 45.7) for all TLCs less than 1.00 x 10(9)/L. The specificity for this prediction increased from 57% to 97% over this range. The LR also increased from 1.4 (95% confidence interval, 1.3 to 1.6) for all TLCs less than 2.00 x 10(9)/L to 9.7 (95% confidence interval, 7.1 to 13.1) for all TLCs less than 1.00 x 10(9)/L in predicting a CD4+ percentage less than 20%. CONCLUSIONS The TLC, between 1.00 x 10(9)/L and 2.00 x 10(9)/L, appears to be a useful predictor of significant immunosuppression as measured by a CD4+ T-cell count less than 0.20 x 10(9)/L in HIV-infected persons. The LR for a given TLC value and the pretest probability of immunosuppression can be used to determine the posttest probability of significant immunosuppression in individual patients. For example, in a patient with a TLC less than 1.50 x 10(9)/L and a pretest probability of 16%, the posttest probability of a low CD4+ T-cell count increases to 53%. In contrast, a TLC greater than 2.00 x 10(9)/L in an individual with a pretest probability of 30% will decrease the posttest probability of a low CD4+ T-cell count to less than 4%. Physicians should find these data useful to help predict the risk for opportunistic infection among HIV-infected persons who present with syndromes that are potentially compatible with opportunistic infection but who have not had recent or prior CD4+ T-cell analysis.