Changes in CD4 lymphocyte counts are widely used in monitoring human immunodeficiency virus (HIV)-infected patients for disease progression. However, random fluctuations may obscure clinically significant changes. CD4 cell counts from 1020 untreated subjects with asymptomatic HIV infection monitored by standardized methods for up to 2 years were assessed. The within-subject coefficient of variation averaged 25% but was higher in subjects with lower counts; in 6% of subjects the count was half or double the one obtained 8 weeks before. Proportionate rates of decline, which had negligible correlation with the baseline count, averaged 14.3%/year but varied considerably between subjects: An estimated 29% had increasing trends. Declines were greater in HIV p24-positive subjects and those with higher lymphocyte percentages or lower platelet counts or hemoglobin levels. With such high variation, changes between single counts should be interpreted cautiously. Using multiple counts and other markers may provide more precise assessment of immune status.