Reply to Moonan and Weis, Seaworth et al, and Nunn and Phillips

To the Editor—We thank Drs Moonan and Weis for their comments [1] on our recent publication [2]. First, regardless of its excellent pedigree, public health policy that is applied worldwide should be examined using the best scientific methods available. Thus, we restated the effect of directly observed therapy (DOT) versus self-administered therapy (SAT) on microbiological outcome as falsifiable hypotheses, and tested the hypotheses in the laboratory. We found that multidrug-resistant tuberculosis was not encountered with any degree of nonadherence, except transient isoniazid monoresistance, whereas failure occurred only at >50% nonadherence; studies in murine tuberculosis by others have demonstrated the same [3, 4]. Next, we investigated the same question using the highest level of clinical evidence for therapeutic maneuvers: randomized controlled clinical trials (RCTs). We rated the quality of each study based on best rules of evidence, such as GRADE criteria, that address the randomization and allocation concerns Moonan and Weis raise. Four of the RCTs had a score of 4, and one had 3, with 5 as the highest possible score. According to GRADE criteria, further evidence to identify RCTs as high or moderate quality are risk of bias, imprecision, inconsistency, indirectness, and publication bias [5]. All 5 of these were addressed. The most direct outcome measure is microbiologic outcome. Our results demonstrated high consistency in influence analyses; that is, no one study demonstrated statistically superior microbiologic failure with DOT versus SAT, and showed no publication bias. Finally, the perennial issue raised in defense of DOT is quality of DOTS programs. We included measures of DOTS program performance and demonstrate that adherence was improved, as intended. Clearly, the DOTS programs in the studies selected performed well when measured by improved adherence. Moreover, our subgroup and sensitivity analysis directly addressed quality of DOTS programs; results were shown in Supplementary Figure 2 [2]. Thus, it is very unlikely that the quality of the RCTs or even the observational studies could be blamed for the nonsuperiority of DOT. In fact, if we applied the same quality of evidence rules to the “look-back” and quasi-scientific studies that have been used to support DOTS, there would be little to no quality evidence in support. Any global health policy that has an impact on millions of people, and is sold by public health authorities as a panacea, should be, and will be, tested in the laboratory and in the clinic based on the best rules of evidence regardless of consensus by “experts.” Caveat venditor.

Nunn and Phillips [6] raise the issue of a possible double publication. A close review convinced us that the studies were sufficiently different and took place in different locales; although some overlap could not be ruled out. However, in the sensitivity analysis, neither the omission of each one of the studies from the analysis (Supplementary Figure 1) [2], nor the omission of Ormerod et al's study, changed the effect size and conclusion. Moreover, the type of DOT supervision was well described in Table 1 [1].

Seaworth et al [7] raise the interesting issue of moving away resources from patient care. We agree. Indeed, the thrust of our paper was that resources should be concentrated elsewhere, not away from tuberculosis care, but away from DOT. It is a mistake to confuse DOT for intensive care: most infectious disease care is individualized and intensive without DOT. More intensive patient care, including close follow-up of patients for adverse events, among others, and measurement of drug concentrations, is where resources should go. That more focused and individualized care is, in our opinion, superior to programs that try to fit every tuberculosis patient on earth onto Damastes’ bed. Indeed, the poorer outcomes among rural patients under DOTS suggest more harm than benefit. Primum non nocere!