Refocusing the timeline of tuberculosis through the lens of BCG

Abstract

An estimated 1.7 billion people—23% of the world’s population—have been infected by Mycobacterium tuberculosis. It has been known for over five decades that, once someone has been infected with M. tuberculosis for more than 2–5 years, the risk of developing tuberculosis (TB) disease— barring a second infection or immunosuppressive condition—is small. Nevertheless, numerous publications and textbooks continue to cite a ‘rule of thumb’ that 5% of people with latent TB infection will develop reactivation TB disease after this initial highrisk period. Multiple recent analyses have reaffirmed that this 5% figure is a substantial overestimate. 4 In this issue of Thorax, Trauer et al reanalyse data from 21 clinical trials of Bacille CalmetteGuérin (BCG) vaccination. Their comprehensive analysis provides yet another piece of evidence that the traditional conceptualisation of latent TB infection—as a static state in which people experience a constant, reasonably high risk of reactivation for many years, if not decades—should be abandoned. Prior analyses have estimated that BCGderived protection against TB disease may wane after about 10 years, though longterm followup data are scant. Trauer et al updated a prior metaanalysis of BCG vaccine trials to explore whether this waning of protection more likely reflects the timing of exposure to M. tuberculosis or the timing of reactivation disease. They accomplished this objective by separating trials into two categories of underlying TB burden (declining vs sustained high burden) and of duration of followup (early vs late). Classified in this fashion, results suggested that BCG may have provided high levels of protection during early followup in all trials— protection that may have persisted in settings of declining TB burden (ie, where the risk of repeated exposure was low). But in settings of sustained high TB transmission—where the risk of ongoing infection outweighed the risk of late reactivation—BCG may have offered little (or even zero) longterm protection. This finding is consistent with the hypothesis that the observed waning of BCG protection reflects the timing of exposure, rather than the timing of reactivation. Though it represents only one of many reasonable explanations for the observed data, this hypothesis is beautiful in its simplicity; it need not invoke more complex relationships such as with latitude or nontuberculous mycobacteria. As the authors themselves note, this finding is not surprising, given the low rate of ‘late’ reactivation (ie, progression to TB disease beyond the initial 2–5 years highrisk period following initial infection). In settings of sustained high TB burden, incident TB disease over longer periods of followup largely reflects repeated infection events rather than late reactivation. For example, the annual risk of TB infection in India at the time of the largest BCG trial (1.4%–1.7%) is similar to the remaining lifetime risk of TB disease (1.2%– 2.2%) for an average TBinfected adult living in a lowexposure setting like the presentday USA. But the findings of Trauer et al are difficult to reconcile with the traditional thinking that, once someone is infected with M. tuberculosis, that person will experience a persistently high risk of reactivation disease. If this were the case (assuming a large fraction of people in the BCG trials were exposed to TB in the years following vaccination), the longterm protection afforded by BCG would not depend strongly on the burden of TB in the surrounding community. It is important to interpret these findings in light of two key caveats. First, among trials primarily enrolling adults, Trauer et al’s summary estimate of BCG effect over longer followup did not vary greatly between settings of sustained burden (incidence rate ratio 1.05, 95% CI 0.87 to 1.24) and settings of declining burden (0.86, 95% CI 0.57 to 1.15). Second, among trials primarily enrolling children, only one was categorised by the authors as ‘sustained burden, longer followup’. This trial was performed in Puerto Rico around 1950— at a time when TB mortality fell by more than 90% in the following two decades. If this study was reclassified as being in a region of ‘declining’ burden, the weight of evidence supporting the authors’ primary conclusion would be correspondingly lighter. Despite these caveats—and irrespective of the final conclusions—this reanalysis provides tremendous additional scientific value, in the form of (1) detailed descriptions of each trial contributing to the analysis; (2) rigorous reevaluation (and in some cases, revision) of previously published results; (3) inclusion of novel disaggregated data from the largest contributing trial and (4) open publication of all data and code, including for generation of key figures. These additional descriptions, analyses and data will serve as an invaluable resource to a scientific community that will undoubtedly continue to debate the mechanism of BCG action for decades to come. In the past 5 years, our focus on the timeline of TB infection has sharpened. The natural history of TB is now recognised as a dynamic entity that is best characterised not as a persistent high risk of ‘reactivation’, but rather as a timevarying balance between the activity of the pathogen (which declines rapidly in the first few years after initial infection) and the response of the host (which can be weakened at any time)—a balance in which both the intensity and frequency of infection also play an important role. Trauer et al add to this understanding of TB natural history by viewing the evidence through a different lens: that of BCGinduced protection over time. Their findings hold important implications for the development of novel TB vaccines, which may also need to establish longerterm protection against infection (and progression to disease) if they are to have sustained populationlevel impact. They also highlight the centrality of reducing the burden of TB transmission if we are to end this epidemic that claims the lives of more than a million people every year.