BCG has been administered to over 4 billion persons worldwide, but its efficacy in preventing tuberculosis in adults has been highly variable. One hypothesis for its variability is that different strains of BCG vary in protective efficacy, and moreover, that evolutionarily early strains are more efficacious than the more attenuated evolutionarily late strains, which lack region of deletion 2. To examine this hypothesis, we tested six widely used BCG strains--the evolutionarily early strain BCG Japanese, two evolutionarily late strains in DU2 Group III (BCG Danish and Glaxo), and three evolutionarily late strains in DU2 Group IV (BCG Connaught, Pasteur, and Tice)--in the guinea pig model of pulmonary tuberculosis. With the exception of BCG Glaxo, which had relatively poor efficacy, we found no substantial differences in efficacy between the early strain and the late strains, and only small differences in efficacy among late strains. BCG Tice was the most efficacious BCG vaccine, with significantly fewer Mycobacterium tuberculosis in the lung and spleen than BCG Danish and BCG Japanese, although absolute differences in the organ burden of M. tuberculosis among these three vaccines were small (< or =0.2 log). BCG Tice and Pasteur were not significantly different. rBCG30, a recombinant BCG Tice vaccine overexpressing the M. tuberculosis 30 kDa major secretory protein (Antigen 85B), was more potent than any BCG vaccine (P < 0.0001 for differences in organ burden). Our study shows that late strains are not less potent than an early strain and argues against strain differences as a major factor in the variability of outcomes in BCG vaccine trials.