SETTING: City of Manaus, Amazonas, Brazil. OBJECTIVE: To compare estimates of Mycobacterium tuberculosis infection prevalence obtained using traditional tuberculin skin test (TST) criteria and dual skin test (DST) data. METHODS: A total of 1070 schoolchildren received DST with tuberculin and four environmental mycobacteria sensitins. Responses were classified as sensitin-dominant, tuberculin-dominant or non-dominant. Positive predictive values (PPV) were defined using 'narrow' and 'wide' standards based on DST responses. These predictive values were derived for each category of tuberculin indurations, and were used to calculate the prevalence estimates. RESULTS: Using DST data, the estimates of M. tuberculosis prevalence for scar-negative children were 7.4% (M. avium) and 7.8% (M. scrofulaceum) using the 'narrow' standard, and 16.9% (M. avium) and 15.2% (M. scrofulaceum) using the 'wide' standard. The percentage with TST > or =10 mm was 11.5%. Scar-positive children had higher estimates using both the 10 mm cut-off and DST data. CONCLUSION: In settings with a relatively low prevalence of M. tuberculosis infection and high cross-reactivity with environmental mycobacteria, DST can help to assess the validity of traditional thresholds for estimating the prevalence of M. tuberculosis infection. DST data with environmental antigens and tuberculin do not distinguish BCG-induced cross-reactivity.