Context: The duration of protection from tuberculosis of BCG vaccines is not known.

Objective: To determine the long-term duration of protection of a BCG vaccine that was previously found to be efficacious.

Design: Retrospective record review using Indian Health Service records, tuberculosis registries, death certificates, and supplemental interviews with trial participants.

Setting and participants: Follow-up for the period 1948-1998 among American Indians and Alaska Natives who participated in a placebo-controlled BCG vaccine trial during 1935-1938 and who were still at risk of developing tuberculosis. Data from 1483 participants in the BCG vaccine group and 1309 in the placebo group were analyzed.

Main outcome measures: Efficacy of BCG vaccine, calculated for each 10-year interval using a Cox regression model with time-dependent variables based on tuberculosis events occurring after December 31, 1947 (end of prospective case finding).

Results: The overall incidence of tuberculosis was 66 and 138 cases per 100 000 person-years in the BCG vaccine and placebo groups, respectively, for an estimate of vaccine efficacy of 52% (95% confidence interval, 27%-69%). Adjustments for age at vaccination, tribe, subsequent BCG vaccination, chronic medical illness, isoniazid use, and bacille Calmette-Guérin strain did not substantially affect vaccine efficacy. There was slight but not statistically significant waning of the efficacy of BCG vaccination over time, greater among men than women.

Conclusion: In this trial, BCG vaccine efficacy persisted for 50 to 60 years, suggesting that a single dose of an effective BCG vaccine can have a long duration of protection.