Study Design, Participant Enrollment, Randomization, and Follow-up

Full details of the parent study design, eligibility criteria, enrollment and randomization, safety monitoring, and study outcomes have been previously published (15). In brief, we enrolled participants >12 years of age with newly diagnosed pulmonary TB during January 2016–October 2018. We randomly assigned enrolled participants in a 1:1:1 ratio to 1 of the 3 regimens (i.e., control, rifapentine, or rifapentine/moxifloxacin). We administered study drugs once daily, by directly observed therapy, on >5 of 7 days/week.

The study protocol required diabetes screening before enrollment. Hemoglobin A1c (HgbA1c) was the preferred test. If such testing was not available, we collected readings of fasting blood glucose (defined as no caloric intake for >8 hours) or nonfasting blood glucose. A prior diagnosis of diabetes at the time of TB diagnosis was self-reported by the study participants and verified with medical documentation when available. Concomitant medications taken during the study were routinely recorded by the study sites on the concomitant medications case report form and thereafter coded and characterized by using the WHO Drug Dictionary’s anatomic therapeutic classification system (18). We used the WHO Drug Dictionary’s standardized drug groupings to identify class 2 category drugs used in diabetes (18).

Because we used different approaches in different sites for capturing data on diabetes, we developed a consensus definition of diabetes. We classified participants as having diabetes if any of the following case selection criteria were met at baseline: a prior diagnosis of diabetes, receipt of insulin or any other diabetes medications, HgbA1c >6.5%, fasting blood glucose >126 mg/dL, or nonfasting blood glucose ≥200 mg/dL.

All participants had study visits at baseline, at weeks 2, 4, 8, 12, 17, 22, and 26, and at months 9, 12, 15, and 18 after randomization (15). During study visits, we evaluated participants for adverse events, collected blood samples for complete blood count and biochemical analyses through week 22, and collected sputum samples for mycobacterial culture through the follow up. We collected adverse event reports through the 18 months of the study follow-up period. We graded adverse events severity on the basis of Common Terminology Criteria for Adverse Events criteria version 4.03 (19).

The study was approved by the CDC institutional review board. Each participating institution provided for the review and approval of protocol and its informed consent documents by a local institutional or ethics committee or relied formally on the CDC institutional review board’s approval. All participants provided written informed consent. The study data were monitored by a data safety monitoring board coordinated by the study sponsor.

Definitions of Outcomes

The primary efficacy outcome was TB disease–free survival 12 months after randomization. For each participant, we assigned a primary outcome status of favorable, unfavorable, or not assessable, as described previously; we further classified unfavorable outcomes as TB-related or not TB-related (15). We defined time to stable culture conversion as the time to the first of 2 consecutive negative sputum cultures without an intervening positive culture.

The primary safety outcome was the proportion of participants with grade >3 adverse events during treatment (with onset up to 14 days after the last dose of study medication). Tolerability was a secondary safety outcome and was defined as premature discontinuation of the assigned regimen for any reason other than microbiologic ineligibility.

Analysis Populations

The microbiologically eligible analysis population included participants with culture-confirmed TB without resistance to isoniazid, rifampin, and fluoroquinolones. The assessable analysis population excluded those without an assessable outcome. We considered microbiologically eligible and assessable as primary analysis populations. Secondary analysis populations included participants who completed >75% and >95% of treatment doses (2 per protocol analysis populations), and all participants randomized (intention to treat). We included all randomized participants that started study treatment in safety analyses.

Statistical Analysis

We used descriptive statistics to summarize the demographic and clinical characteristics among participants with diabetes. For primary efficacy and safety secondary subgroup analysis, we calculated the risk difference between the regimens and their respective 95% CIs.

Pharmacokinetics

We sampled all participants who underwent randomization for pharmacokinetic analysis. All participants had 1–3 sparse pharmacokinetic samples (timepoints were at 0.5, 5, and 24 hours postdose), and at some sites participants had 7 intensive pharmacokinetic sampling (timepoints were at 0.5, 3, 5, 9, 12, and 24 hours postdose), conducted during weeks 2–8 of TB treatment. We determined plasma concentrations of rifapentine, 25-desacetyl-rifapentine, rifampin, isoniazid, pyrazinamide, ethambutol, and moxifloxacin by using validated high-performance liquid chromatography assays. We developed population pharmacokinetic models for each of the 6 drugs, and we calculated the individual area under the concentration time curve from 0–24 hours (AUC_0–24h) and maximal plasma concentration (C_max) for each drug (20). We compared AUC_0–24h and C_max for each drug by using t-tests by diabetes.

Study Population

Of 2,516 randomized participants in the full study, 181 (7.2%) we classified as having diabetes (Table 1; Appendix Figure). Among 181 participants who were classified as having diabetes, 83 (45.8%) reported a prior diabetes diagnosis at enrollment. Participants with diabetes were from study sites in 12 countries (Brazil, Haiti, India, Kenya, Malawi, Peru, South Africa, Thailand, Uganda, United States, Vietnam, and Zimbabwe). The percentage of participants with diabetes among the enrolled sites was 19.3% (17/88) in sites located in South America, 15.5% (45/290) in Asia, 5.7% (104/1832) in Africa, and 4.9% (15/306) in North America.

We examined baseline demographics and clinical characteristics of participants with diabetes by regimen (Appendix Table 1). Overall, 67.4% were male and 32.6% female, the median age was 46 years, 16 (8.8%) participants were HIV-positive, 132 (72.9%) had baseline cavitation on chest radiograph, and the median body mass index was 21 kg/m². A total of 146 (80.1%) of 181 participants had available HgbA1c results at baseline (median value 7%). Sixty-two participants (34.3%) reported receiving medications for diabetes. Among the 83 participants reporting a prior diagnosis of diabetes at baseline, 8 (9.6%) reported having diabetes mellitus type 1 and 73 (88.0%) reported having diabetes mellitus type 2; for 2 (2.4%), the type of diabetes was unknown. Twenty participants were classified as having diabetes on the basis of blood glucose test results only.

Compared with participants without diabetes, participants with diabetes were older (median age 46 vs. 30 years); more often reported Asian race (25.9% vs. 10.3%), White race (8.4% vs. 1.0%), >1 race (18.1% vs. 13.0%), or Hispanic ethnicity (13.3% vs. 2.4%); were more often enrolled at study sites in Asia (25.9% vs. 10.1%) or South America (9.6% vs. 3.1%); had higher bodyweight (56 vs. 53 kg); had smaller (<4 cm) cavity size (44.6% vs. 32.2%); and had lower (negative to 1+) smear positivity grade (55.4% vs. 42.2%) (all p<0.005) (Appendix Table 2). The presence of baseline cavitation on chest radiograph was similar (72.9% of participants with diabetes had cavitary disease vs. 72.3% participants without diabetes). We observed a shorter time-to-detection in liquid media in participants with diabetes compared with those without diabetes (8.27 days vs. 8.82 days; p = 0.03).

Efficacy

We included 166 participants with diabetes (91.7%) in the microbiologically eligible analysis population and 155 (85.6%) participants in the assessable analysis population (Figure 1; Appendix Table 3). Among participants in the microbiologically eligible population, unfavorable outcomes occurred in 26.3% of participants in the control regimen and 13.8% of participants in the rifapentine/moxifloxacin regimen, indicating a risk difference from control of –12.5% (95% CI –27.0% to 1.9%). Unfavorable outcomes occurred in 29.4% of participants in the rifapentine regimen, indicating a risk difference from control of 3.1% (95% CI –13.8% to 20.0%). For the assessable analysis population, unfavorable outcomes occurred in 17.6% of participants in the control regimen and 12.3% of participants in the rifapentine/moxifloxacin regimen, indicating an absolute difference from control of –5.4% (95% CI –18.9% to 8.1%). Unfavorable outcomes occurred in 23.4% of participants in the rifapentine regimen, indicating an absolute risk difference from control of 5.8% (95% CI –10.2% to 21.8%). The percentage of participants with TB-related unfavorable outcome was 5.3% in the control arm, 3.4% in the rifapentine/moxifloxacin regimen, and 19.6% in the rifapentine regimen in the microbiologically eligible population and 5.9% in the control arm, 3.5% in the rifapentine/moxifloxacin regimen, and 21.3% in the rifapentine regimen in the assessable population (Appendix Table 3). We observed no cases of acquired TB drug resistance in participants with diabetes.

In sensitivity analysis limited to the 83 participants with prior diabetes diagnosis, proportions of unfavorable outcome were slightly higher than in analysis of all participants classified as having diabetes, but differences between regimens were similar (Appendix Table 4). Participants with diabetes had higher overall proportion of unfavorable outcomes compared with participants without diabetes (microbiologically eligible population, 22.9% vs. 15.4%; assessable population, 17.4% vs. 11.4%).

Time to Culture Conversion

We found no statistically significant difference in time to stable sputum culture conversion to negative in participants with diabetes treated with each of the experimental regimens compared with the control regimen: rifapentine/moxifloxacin regimen hazard ratio 1.3 (95% CI 0.9–1.9) in liquid media and 1.4 (95% CI 1.0–2.1) on solid media; rifapentine regimen hazard ratio 1.0 (95% CI 0.7–1.5) in liquid media and 1.1 (95% CI 0.8–1.7) on solid media (Figure 2). CIs were wide, and the point estimates of the hazard ratios were similar to those previously reported for the whole study population (15). In the microbiologically eligible population in participants with diabetes, culture conversion was achieved by the 8-week follow-up visit in liquid media in 62.3% in the control arm, 84.2% in the rifapentine/moxifloxacin arm, and 75.1% in the rifapentine arm and, on solid media, in 67.3% in the control arm, 91.2% in the rifapentine/moxifloxacin arm, and 85.6% in the rifapentine arm.

Safety and Tolerability

Of 178 participants with diabetes included in the safety analysis population, 24.7% experienced grade >3 adverse events during treatment (31.6% in the control arm, 23.1% the in the rifapentine/moxifloxacin arm, and 19.6% in the rifapentine arm) (Table 2). The difference in proportion of participants with grade 3–5 adverse events between the control and rifapentine/moxifloxacin arm was –8.7% (95% CI –24.5 to 7.1), and the difference between the control and rifapentine arm was –11.0% (95% CI –26.7 to 4.8).

Serious adverse events during treatment were experienced by 14% participants with diabetes (17.5% in the control arm, 10.8% in the rifapentine/moxifloxacin arm, and 14.3% in the rifapentine arm) (Table 2). Two deaths (3.5%) occurred in participants in the control arm and none in the rifapentine or rifapentine/moxifloxacin arms. Six participants permanently discontinued study treatment (6.2% the in the rifapentine/moxifloxacin arm and 3.6% in the rifapentine arm) (Table 2). The percentage of participants that had any transaminase value during treatment of >5-fold the upper limit of normal was highest in the rifapentine/moxifloxacin arm (6.2%, 4/65) compared with the rifapentine arm (3.6%, 2/56) and the control arm (3.5%, 2/57) (Table 2). No participants in the control regimen had any transaminase value of >10-fold the upper limit of the reference range, compared with 3.6% in the rifapentine arm and 3.1% in rifapentine/moxifloxacin arm. The most frequent adverse events among participants with diabetes were hepatitis (n = 14), hypertension (n = 9), and diabetes mellitus under inadequate control (n = 8) (Appendix Table 5). One case of peripheral neuropathy was reported in a participant in the rifapentine arm (Appendix Table 5).

Discontinuation of assigned treatment for any reason (tolerability) in microbiologically eligible analysis population was 19.3% in the control arm, 13.8% in the rifapentine/moxifloxacin arm, and 13.7% in the rifapentine arm (Table 2). The difference in proportion of discontinuation of assigned treatment for any reason between the control and rifapentine/moxifloxacin arm was –4.9% (95% CI –18.0% to 8.2%), and the difference between the control and rifapentine arm was –4.7% (95% CI –18.4% to 9.0%).

In a sensitivity safety analysis limited to participants with prior diabetes diagnosis, point estimates of grade >3 adverse events were higher than in analysis of all participants classified as having diabetes but showed similar findings across the regimens (Appendix Table 6). The proportion of participants with grade >3 adverse events was higher in participants with diabetes compared with those without diabetes (24.7% vs. 16.9%; p = 0.01).

Pharmacokinetics

We compared model-estimated mean AUC_0–24h and C_max in participants classified with diabetes with those of participants without diabetes for each of the study drugs (Table 3). Rifamycin (rifampin and rifapentine) AUC_0–24h and C_max were similar among participants with diabetes and participants without diabetes. Participants with diabetes compared with participants without diabetes had lower AUC_0–24h values for moxifloxacin and ethambutol and higher C_max values for pyrazinamide, but the magnitude of these differences was modest.