Disseminated mycobacterial infections (DMI), caused by Mycobacterium tuberculosis (MTB) and M. avium (MAC), are common complications of late-stage human immunodeficiency virus (HIV) infection [1]. Although the use of highly active antiretroviral therapy (HAART) has substantially reduced DMI incidence [2], disease diagnosis still remains a critical problem for physicians. Both pathologies have non-specific clinical features, unhelpful in the differential diagnosis [3, 4], and microbiological isolation and identification are characterised by long turnaround time and uncertain sensitivity that often oblige clinicians to commence anti-mycobacterial therapy based solely upon clinical suspect. In light of these premises, we evaluated whether the introduction of a molecular method for mycobacterial DNA detection may represent a suitable approach for diagnosis. After optimizing a polymerase chain reaction (PCR) hybridization technique, based on the amplification of a 16rRNA gene sequence using pan-Mycobacterium primers and on the hybridization to two probes, designed to differentiate between M. tuberculosis and M. avium [5], we estimated its reliability and usefulness on blood specimens and/or bone marrow (BM) aspirates. Between March 1999 and April 2004, all of the episodes of suspected DMI investigated with PCR assay performed on blood specimen [6] and a corresponding blood culture (radiometric Bectec AFB system) [7], and/or with a PCR performed on BM aspirate [6] and a corresponding BM analysis (i.e. histopathologic, microscopic and culture exam) [7] were included in the study. Clinical charts were reviewed to assess the diagnosis of disseminated MTB and MAC disease. We drew up two clinical diagnostic criteria: (1) suggestive signs and symptoms of DMI (fever, drenching night sweats, anaemia and weight loss) [3], in addition to blood or BM culture confirmation; (2) suggestive signs and symptoms of DMI in addition to at least two other minor signs or symptoms (diarrhoea, abdominal pain, lymphadenopathy, hepatomegaly, splenomegaly) and/or histopathologic patterns suggestive for mycobacterial infection in patients responding positively to anti-TB or anti-MAC therapy, whose clinical findings excluded other diseases. Patients without exhaustive data to confirm or rule out a DMI were excluded. Data were classified in two-by-two contingency tables to evaluate sensitivity, specificity, negative predictive value and positive predictive value of culture and PCR assay performed on different samples, compared to clinical diagnosis. PCR assay was successfully performed on 117 samples from 95 patients, 71 blood specimens and 46 BM aspirates. Among these, there were 11 paired BM–blood specimens collected from 11 patients available for complete analysis. Eur J Clin Microbiol Infect Dis (2008) 27:163–166 DOI 10.1007/s10096-007-0407-0