SETTING: Under-ascertainment and under-reporting of tuberculosis (TB) hampers surveillance and control. Case detection is improved by record linkage of case registers and under-reporting can be estimated by capture-recapture (CR) analysis. OBJECTIVES: To assess the completeness of the TB registration systems and estimation of TB incidence and under-reporting in the Piedmont Region of Italy in 2001. METHODS: Record linkage of the 'physician notification system', the TB laboratory register and the hospital records register, and subsequent three-sample CR analysis. RESULTS: Record linkage identified 657 TB cases; CR analysis estimated 47 (95%CI 31-71) unrecorded cases. Under-reporting of the 'physician notification system' was estimated at 21% (95%CI 20-23). The overall estimated TB incidence rate was 16.7 cases per 100000 population (95%CI 16.3-17.3), varying according to the subset investigated: 12.7 for individuals from low TB prevalence countries and 214.1 for immigrants from high TB prevalence countries; 13.1 and 25.8 for persons aged or = 60 years, respectively; and 32.1 in Turin, the regional capital and 10.8 in the rest of the region. CONCLUSIONS: When multiple recording systems are available, record linkage and CR analysis can be used to assess TB incidence and the completeness of different registers, contributing to a more accurate surveillance of local TB epidemiology.