As shown in a previous study, the outcome of multidrug-resistant tuberculosis (MDR-TB) in France is not satisfactory (41% success rate). To improve the management and outcome of patients with MDR-TB, a prospective intervention study was carried out in 1998-1999 by National Reference Centre (NRC) staff. NRC staff contacted (i) the microbiologists reporting MDR-TB patients in order to offer second-line drug testing and (ii) physicians to help tailor treatment regimens based on World Health Organization guidelines. A total of 45 MDR-TB patients were followed, including 33 for whom NRC staff successfully interacted with the physicians in charge and 12 patients managed without NRC collaboration. The crude proportion of favourable outcomes was 58%, and this proportion reached 70% for patients managed in collaboration with the NRC. Among the latter 33 patients, 32 were treated with at least three active drugs for a median duration of 12 months. In multivariate analysis, HIV-positive status was associated with treatment failure (hazard ratio (HR) 4.3), whereas NRC intervention was associated with a favourable outcome (HR 0.2). However, even in the NRC intervention group the median duration of treatment was shorter than planned (median 12 months) and 12% of patients were lost to follow-up. This intervention study provides encouraging results and should be continued until reference teams and a directly observed treatment strategy (DOTS) are implemented to improve patient management further.