In Italy tuberculosis (TB) surveillance is hampered by several problems (inconsistency of data and underreporting). A tuberculosis surveillance system was established in Varese Province (VSS) in 1992. The aim of the present study was, using VSS, to estimate: 1) coverage and validity of data collected by the national Compulsory Surveillance System (CSS); 2) validity of diagnosis and risk factors for tuberculosis; 3) tuberculosis incidence. A specifically designed form, including the same items as the CSS form and 10 more questions on risk factors and diagnostic issue, was completed in all the existing health facilities in the province for each new tuberculosis case diagnosed, collected by the study co-ordinator during monthly supervision after a review of clinical records, and stored in a database at the co-ordinating centre. Routine notifications collected at national level by CSS for the province were nominally linked with VSS data. To evaluate coverage and data validity, a comparison was made between VSS and CSS data after removal of duplicates. Quality of data, risk factors and incidence were evaluated on VSS data (gold standard). CSS coverage was 63% with a significant underreporting of extrapulmonary cases. Within CSS, 20% of data were missing (particularly diagnostic information: 42-72%). According to VSS, smear was performed on 88% of cases and culture on 66%. Half of the cases had no risk factors. The overall incidence of tuberculosis in Varese Province was 15 per 100,000. We estimated the coverage and validity of the data collected by CSS, validity of diagnosis, risk factors for and incidence of tuberculosis and proposed several suggestions to improve CSS nationwide.