While quantitative mycobacterial blood cultures have been accepted as the standard for evaluating response to various Mycobacterium avium complex (MAC) treatment regimens, variability in this methodology has not been evaluated in a rigorous fashion. We thus studied the reproducibility of quantitative MAC cultures by a lysis-centrifugation culture system within and among five institutions. To measure the intralaboratory variation in mycobacterial colony counts, colony counts from duplicate blood specimens collected from 52 AIDS patients with MAC bacteremia were determined. Colony counts ranged from 0 to 50,000 CFU/ml. Nonparametric analyses revealed there was no significant difference in colony counts between the 52 duplicate specimens. The agreement between the intralaboratory paired specimens, as measured by the intraclass correlation coefficient, was 0.997. To measure the interlaboratory variation, multiple 10-ml aliquots from 12 patients were distributed to five institutions and processed within 24 to 32 h by lysis-centrifugation. For the 12 specimens distributed to the five laboratories, two-way analysis of variance for repeated measures revealed no significant difference in an individual patient's colony counts between laboratories (P > 0.2). We conclude that quantitation of mycobacterial colony counts by the lysis-centrifugation system is reproducible within and between institutions. Clinical trials evaluating response to therapeutic interventions for MAC can use multiple laboratories for quantitation of mycobacteremia. Furthermore, a 24- to 32-h delay in processing appeared to have no impact on reproducibility.