Vaccines for tuberculosis, both protective and therapeutic, have existed for some time. Disagreement on their efficacy has existed for just as long. Preparations derived from an attenuated strain of M. bovis, the bacillus of Calmette and Guérin (known collectively as BCG), have been administered to humans for more than 70 years. The reported efficacy of BCG has varied widely and ranged from almost 0% to 80% [3, 4]. To complicate these analyses, BCG is not a single vaccine preparation and is probably best thought of as a family of vaccine preparations or "BCGs." In addition to variation in the vaccines studied, the conditions of the studies have also varied. The ages and nutritional states of study participants and the prevalence of other (nontuberculous) but potentially cross-reacting mycobacteria in the locations where the studies were conducted all potentially confound these analyses.

Nonetheless, a recent meta-analysis [5] of all randomized, controlled trials assessing the efficacy of BCG concluded that these vaccines were modestly effective, reducing the risk for active tuberculosis an average of 50%. Moreover, vaccination seemed particularly useful for preventing the most serious clinical forms of tuberculosis (for example, meningitis and disseminated disease) and death. This was not to be the final word on the subject, however. A new study [6], conducted in Malawi, considered the efficacy of single and multiple BCG vaccinations and the use of a combined BCG and killed M. leprae preparation for prevention of tuberculosis and leprosy. This elaborate randomized, controlled trial assessed more than 100 000 persons of all ages, considered HIV infection, and had several years of follow-up; it did not specifically control for previous tuberculous infection. Although vaccination seemed to provide substantial protection against leprosy, it afforded no protection against pulmonary tuberculosis at any age and only a slight tendency toward a reduction in lymph node disease among participants who had had multiple vaccinations (as judged by the presence of a BCG scar on study entry). The reasons suggested to explain the general lack of protection from tuberculosis recapitulated many of the same explanations that have been offered for similar results in previous studies. Clearly, this question remains unresolved. Results with therapeutic vaccines remain even more preliminary [7].

After all these years, how should we proceed with vaccination for tuberculosis? It is important to recognize that even if currently available preventive vaccines are effective, they are unlikely to benefit persons with previous tuberculous infections or those with HIV infection-two groups that account for a large proportion of cases worldwide. Indeed, because currently available vaccines use live attenuated bacilli, vaccination of HIV-infected persons may carry some risk. Prophylactic treatment with isoniazid [8] or an alternate agent, such as rifampin in the setting of infection caused by an isoniazid-resistant strain [9], seems to offer the best opportunity for reducing the risk for active disease in persons previously infected with M. tuberculosis. Such an approach also seems effective in HIV-infected persons and may, by preventing active tuberculosis, also delay the progression of the underlying HIV infection [10]. It should also be noted that because BCG vaccination can cause conversion of the tuberculin skin test result, vaccination may make it more difficult to conduct prophylaxis programs.

Unfortunately, preventive therapy has its own limitations, including the risk for adverse drug reactions, the need to sustain drug administration for months to derive benefit, and the requirement that the infecting strain of tubercle bacillus be susceptible to an available drug. When risk for tuberculosis is especially great and one or another of these limitations is a factor, administering the BCG vaccine to tuberculin skin test-negative persons may be a viable option [11]. Such a strategy seems most reasonable when the risk for infection with M. tuberculosis is high or ongoing or if the potential for severe tuberculosis, if the disease occurs, is also high. Thus, candidates for BCG vaccination would be children and health care workers with ongoing exposure to isoniazid- or rifampin-resistant patients, in addition to persons (particularly children) who are continually exposed to an infected, drug-sensitive person and for whom separation or long-term preventive therapy cannot be provided.

With further research, it is reasonable to hope that better-defined antigens, particularly those that stimulate protective T-cell immune responses or expression of cytokines that augment immunogenicity, will permit the development of safer, more effective vaccines for tuberculosis. Until that time, with respect to truly effective vaccines for tuberculosis, the proverbial glass remains half empty.