Main

For BCG strains used in more than one clinical trial, we plotted efficacy and induced tuberculin reactivity against in vitro passage number. Four out of five strains have apparently lost efficacy while maintaining tuberculin reactivity (Fig. 1). Considering that attenuation during 230 in vitro passages was used by Calmette and Guérin to produce their vaccine, it is not surprising that protective efficacy of BCG waned over the next 1,000 passages. In 1921 Calmette declared this vaccine a “virus fixé”2, but there is ample evidence that this live bacterial vaccine is far from fixed, with daughter strains differing in mycolic acid composition, protein transcription and genotype1. Over the course of this evolution of BCG strains, there was a secular trend towards decreasing tuberculosis mortality, so public acceptance of adverse effects diminished.

Figure 1: Efficacy (a) and tuberculin conversion rates (b) of BCG substrains plotted against in vitro passage number.
figure 1

Strains: Pasteur (○), Phipps (+), Denmark (□), Tice (▵) and Frappier (×). Estimates of efficacy and tuberculin reactivity are from a recent meta-analysis of BCG9 and a recent review1, with the exception that results in a North American Indian trial have been separated by villages receiving BCG-Pasteur and BCG-Phipps10. Estimates less than zero are plotted as 0% because confidence limits were large and spanned zero. Passage number of strains are those reported in the studies; if not specified, passage number was interpolated from data before and after the trial.

Full size image

Meanwhile, a selective pressure to maintain tuberculin reactivity resulted from the misperception that it served as a proxy for protection and the fact that it could be readily assessed. Thus, studies were performed with the goal of decreasing adverse reactions while retaining tuberculin conversion3. More recent work has shown that tuberculin conversion and protective immunity can be dissociated4, and that virulence appears to correlate with protective efficacy in animal models5,6. This trade-off had already been noted in 1952 by Rene Dubos who wrote, “Very properly, the greatest emphasis is always placed on administering the vaccine in such a manner that no unpleasant reaction ensues, but because vaccination must be safe for all, it is probably ineffective for many⃛”7. The detailed manner in which these selective pressures affected vaccine development cannot be deduced from the historical record, but it is noteworthy that by 1956, half of international vaccine producers (20 of 40) had changed strain at some point8.

There are three important implications of this hypothesis. First, estimates of BCG efficacy using meta-analyses may significantly overestimate current efficacy by averaging the effects of current and ancestral vaccines. An accurate assessment of currently available strains is essential in the design of rational vaccine policy. Second, comparison within strains of recent and more ancestral forms (such as retention lots from vaccine trials) may provide insight into vaccine characteristics that correlate with efficacy and tuberculin conversion. Resulting insights may be exploited to increase protection and decrease tuberculin conversion rates of new vaccines. Third, in the absence of increased understanding of the fundamental determinants of virulence and protection, it is likely that increased efficacy will be achieved only at the cost of increased adverse reactions.