A single CD4+ cell count (CD4) measurement is often used to stage HIV-1 infection, decide when to initiate prophylactic therapy and inform patients, and may soon even define AIDS onset. Documentation of the reliability and validity of employing CD4 for the above purposes in a population-based setting is needed. We utilized data from 4,954 homosexual/bisexual men followed over 6 years, with CD4 testing at 6 month intervals, to study the timing of CD4-based staging of HIV-1 disease and quantify and evaluate the potential impact of CD4 measurement error. The median time from seroconversion to first CD4 test below 500 x 10(6)/L or clinical AIDS was 1.70 years, and the first CD4 test below 200 x 10(6)/L or clinical AIDs was 5.29 years. The time from first testing less than 500 x 10(6)/L to clinical AIDS in untreated men was 5.55 years. With confirmatory retesting, these times were significantly lengthened. The 95% confidence ranges for the true CD4 state in individuals with measured CD4 of 500 and 200 x 10(6)/L are at least (297 x 10(6), 841 x 10(6)/L) and (118 x 10(6), 337 x 10(6)/L), respectively. Without confirmatory retesting, individuals with true CD4 remaining at 700 x 10(6) and 280 x 10(6)/L have at least a 40% chance for one of five CD4 measurements to fall below guideline limits of 500 x 10(6) and 200 x 10(6)/L, respectively. Confirmatory retesting can reduce these probabilities to as low as 4%.(ABSTRACT TRUNCATED AT 250 WORDS)