Background To determine the seroprevalence of hepatitis B (HBV), hepatitis C (HCV) and HIV infections in problem drug users (PDU) in Luxembourg. To measure the validity of self-reported test results provided by study participants as well as obtained through the national drug-monitoring system (RELIS). Methods In a cross-sectional multisite study, data were collected by voluntary, anonymous and assisted questionnaires and serological detection of antibodies and antigens. Out of 1169 contacts, 397 participants were recruited within in and out-of-treatment settings (84.2% injecting drug users; IDU). Results The prevalence of antibodies to HIV was 8/272 (2.9%; 95% CI 0.9% to 4.9%), to HCV 245/343 (71.4%; 66.6% to 76.2%), and 67/310 (21.6%; 17.1% to 26.2%) to total HBV antibodies and surface antigen (for IDU 5/202, 218/268 and 59/239, respectively). Specificity of study self-reports was very high for HBV and perfect for HCV and HIV. Sensitivity was 0.224, 0.798 and 0.800, respectively. Kappa scores provided degrees of agreement between serological tests and study self-reports of 0.89 for HIV, 0.65 for HCV and 0.25 for HBV. In contrast to simultaneous cross-sectional self-reports, secondary self-reported data (RELIS) showed high agreement for HIV and HBV infections and provided a good proxy for estimation of HCV seroprevalence. Conclusion HIV testing routines in PDU should be completed at least by HBV and HCV detection given the poor validity of cross-sectional self-reports on hepatitis infections. HIV and hepatitis prevalence estimations in PDU gain by relying on multisite/setting data collection. Research should further investigate the validity of HIV and hepatitis self-reports from routine drug-monitoring systems versus cross-sectional surveys.