To Err Is Human, to Forgive Is Pharmacodynamic

It was not long after completion of the first controlled clinical trial of a new treatment for tuberculosis (TB) that drug-resistant strains of TB developed (1), dampening the enthusiasm generated by the mortality improvement that had been observed after 6 months of streptomycin (2). This discovery initiated a period of sustained research to identify combination regimens for the treatment of TB that were effective at curing disease and preventing the acquisition of drug resistance (3, 4). Despite a number of options for the treatment of both drug-sensitive and drug-resistant TB, there are large gaps between the proportion of individuals cured when allocated to standard of care in clinical trials (5–8) and comparable figures published annually in the World Health Organization global surveillance reports (9). How robust a new treatment regimen is likely to perform in a programmatic setting in the presence of nonadherence is a critical aspect of drug development; in this issue of the Journal, the work by Stagg and colleagues (p. 193–205) to compare the forgiveness of 6- and 4-month regimens using data from clinical trials is welcome (10).

As the authors note, the relationship between dose-taking and treatment outcomes is complex, but they are to be commended for using causal mediation analysis to tease out the direct and indirect effects of nonadherence. Their finding that shorter regimens are not more vulnerable, per se, to nonadherence is reassuring, particularly in light of recent new collaborations with public, private, and philanthropic funding aimed at further aggressive treatment shortening (11–13). The authors’ work also has relevance for national TB programs, as they highlight the importance of adherence in the continuation phase and not just the intensive phase of treatment.

The results of this analysis are limited by the data available; in particular, overall percentage adherence could be evaluated, whereas patterns of adherence could not. It has been shown that longer periods of consecutive nonadherence and shorter gaps between periods of nonadherence are independently associated with poor outcomes among patients with multidrug-resistant TB (14), but the granular data were not available for Stagg and colleagues to evaluate these findings in these data; this is a lesson for trialists, that they should attempt to capture adherence data with sufficient detail to permit important secondary analyses. Adherence, as expected, was generally high in these clinical trials, so there were little data to inform how robust shorter regimens might be when adherence is lower, as is often observed in programmatic settings.

The finding that the shorter regimens were reasonably robust to missed doses is encouraging but should be viewed through the lens of what we already know about forgiveness in TB and other infectious diseases. Although the slow doubling time of Mycobacterium tuberculosis would theoretically make this infection more forgiving of missed doses, the data suggest that intermittent regimens (twice or thrice weekly) are associated with higher relapse rates (15). There is clearly some threshold at which forgiveness is no longer granted, and it is difficult to find that threshold in cohorts of highly motivated clinical trial participants. Furthermore, forgiveness may well be primarily a pharmacodynamic optimization problem. Current TB treatment includes drugs with relatively short half-lives (30–90 min for isoniazid and 2–3 h for rifampin) that are less robust when doses are missed. It would be interesting to understand how regimens with drugs that have longer half-lives—such as the moxifloxacin (half-life, ∼12 h) and rifapentine (half-life, ∼14 h) regimen used in the recent Study 31/A5349 trial (8)—would fare under real-world conditions of suboptimal adherence. The evolution of HIV therapies over time is informative here; treatment began with pharmacodynamically suboptimal drugs requiring multiple daily doses of several different pills to achieve therapeutic levels, resulting in the frequent emergence of resistance. Now, 25 years later, the current standard of care typically includes drugs with long half-lives and high barriers to resistance, often coformulated into a single pill, that are robust to suboptimal adherence. Pharmacodynamic optimization, even in the setting of a virus that replicates and mutates much more quickly than M. tuberculosis, has even made intermittent therapy for HIV possible (16).

The manuscript by Stagg and colleagues provides encouragement that shorter TB treatment regimens for drug-sensitive disease are not necessarily more fragile than the standard 6-month regimen in the face of imperfect adherence. However, the key lesson is that regimens need to be developed that incorporate our understanding of antimicrobial pharmacodynamics while taking human behavior into account. Forgiveness will not be achieved by forgetting the lessons of the past.