This study evaluated the validity of bone marrow (BM) and blood specimens for the diagnosis of disseminated mycobacterial infections (DMIs). From 1990 to February 1997, all specimens were processed with the lysis-centrifugation procedure; thereafter (until December 2001), they were processed with the BACTEC Myco/F Lytic system. Twenty-three paired BM-blood specimens with mycobacteria in at least one specimen were studied from 23 patients. The strains isolated were 14 Mycobacterium avium complex (MAC) and nine M. tuberculosis complex (MTBC). Blood specimens had a statistically significant greater sensitivity for the isolation of MAC than BM (100% vs. 57.1%, respectively), whereas sensitivity for the isolation of MTBC was equal for the two specimen types (66.7%). Although not statistically significant, the times required to detect mycobacteria from blood specimens were lower than those from BM in the MycoF/Lytic system. Overall, blood cultures represented a more sensitive and less invasive alternative to BM cultures for the diagnosis of disseminated mycobacteriosis caused by MAC, especially when the MycoF/Lytic system was used, but provided no advantage for the diagnosis of DMI caused by MTBC.