Pretomanid Pharmacokinetics in the Presence of Rifamycins: Interim Results from a Randomized Trial among Patients with Tuberculosis

Shorter, more potent regimens are needed for tuberculosis. The nitroimidazole pretomanid was recently approved for extensively drug-resistant tuberculosis in combination with bedaquiline and linezolid.

ABSTRACT

Shorter, more potent regimens are needed for tuberculosis. The nitroimidazole pretomanid was recently approved for extensively drug-resistant tuberculosis in combination with bedaquiline and linezolid. Pretomanid may also have benefit as a treatment-shortening agent for drug-sensitive tuberculosis. It is unclear how and whether it can be used together with rifamycins, which are key sterilizing first-line drugs. In this analysis, data were pooled from two studies: the Assessing Pretomanid for Tuberculosis (APT) trial, in which patients with drug-sensitive pulmonary TB received pretomanid, isoniazid, and pyrazinamide plus either rifampin or rifabutin versus standard of care under fed conditions, and the AIDS Clinical Trials Group 5306 (A5306) trial, a phase I study in healthy volunteers receiving pretomanid alone or in combination with rifampin under fasting conditions. In our population pharmacokinetic (PK) model, participants taking rifampin had 44.4 and 59.3% reductions in pretomanid AUC (area under the concentration-time curve) compared to those taking rifabutin or pretomanid alone (due to 80 or 146% faster clearance) in the APT and A5306 trials, respectively. Median maximum concentrations (C_max) in the rifampin and rifabutin arms were 2.14 and 3.35 mg/liter, while median AUC_0–24 values were 30.1 and 59.5 mg·h/liter, respectively. Though pretomanid exposure in APT was significantly reduced with rifampin, AUC_0–24 values were similar to those associated with effective treatment in registrational trials, likely because APT participants were fed with dosing, enhancing pretomanid relative bioavailability and exposures. Pretomanid concentrations with rifabutin were high but in range with prior observations. While pretomanid exposures with rifampin are unlikely to impair efficacy, our data suggest that pretomanid should be taken with food if prescribed with rifampin. (This study has been registered at ClinicalTrials.gov under identifier NCT02256696.)

TEXT

In 2018, over 10 million people had incident tuberculosis (TB), and 1.4 million died (1). A continued challenge in TB control is the duration and complexity of treatment regimens (2). Treatment success rates are still suboptimal and have fallen slightly in recent years, from 86% in 2013 to 82% in 2017 (3). First-line treatment for drug-susceptible TB usually requires 6 months of multidrug therapy. Shorter-duration regimens are needed to decrease logistical burden for patients and programs, improve adherence, and limit the emergence of resistance. Poor adherence, defined as taking 90% or fewer prescribed doses, is known to portend poorer outcomes (4), and treatment default increases over the course of treatment (2); hence the continued interest in shortened TB regimens. One model-based analysis suggested that a 4-month regimen for drug-susceptible TB, priced at 1 U.S. dollar per day, would increase treatment completion and decrease mortality (5). Additionally, a regimen with higher efficacy than the standard first-line combination of isoniazid, rifampin, pyrazinamide, and ethambutol (HRZE) would reduce mortality as well as TB incidence (6).

Earlier efforts at treatment shortening have yielded mixed results, likely due to heterogenous pools of participants as well as the inherent challenge of killing both replicating and dormant Mycobacterium tuberculosis bacilli. Three trials comparing relapse rates of 4-month fluoroquinolone-containing treatment regimens versus standard of care (OFLOTUB, REMoxTB, and RIFAQUIN) showed inferiority of the shortened regimens (4, 7–9). If our goal is to offer all patients the opportunity for shortened treatment duration, a regimen comprising so-called sterilizing agents, with potent activity against dormant (persister) populations of bacilli, may offer the most promise (10).

Pretomanid, the newest antituberculosis medication approved by the U.S. Food and Drug Administration (FDA), is a novel nitroimidazole with activity against both replicating and dormant bacilli. Pretomanid requires activation by deazaflavin (F420)-dependent nitroreductase (Ddn), present in M. tuberculosis, and has various mechanisms of action, including disruption of cell wall mycolic acid synthesis (11). Pretomanid is highly effective in combination with linezolid and bedaquiline, constituting the first 6-month regimen for extensively drug-resistant (XDR) TB (M. tuberculosis resistant to isoniazid, rifampin, fluoroquinolones, and injectable agents) (12). Given its spectrum of action, there is interest in pretomanid as a treatment-shortening agent for drug-sensitive TB (13). A murine model of aerosol TB infection demonstrated a potent treatment-shortening effect of pretomanid combined with either pyrazinamide or rifampin (14). Murine models have described pretomanid activity as time dependent, where the bactericidal activity was best correlated with the cumulative percentage of the dosing interval that the free drug concentration exceeds the MIC (% fT_>MIC). Free drug T_>MIC values of 48 and 77% were associated with 1-log₁₀ killing (bactericidal activity) and 1.59-log₁₀ killing (accounting for 80% of the maximum observed killing effect), respectively (15). The unresolved question is how best to use pretomanid in shortened combination TB treatment regimens in humans.

Pretomanid is extensively metabolized, with seven primary metabolites identified so far (16). Though cytochrome P450 enzyme 3A (CYP3A) is thought to be a minor metabolic pathway, prior data in healthy volunteers showed that, on average, rifampin (a potent inducer of CYP3A) reduced pretomanid exposures by 66% (17). Rifamycins (rifampin, rifabutin, and rifapentine) provide a vital backbone to TB treatment; therefore, the feasibility of combining one with pretomanid is of great interest in potential treatment-shortening regimens. To assess the safety and efficacy of a regimen that incorporates pretomanid with a rifamycin, the Assessing Pretomanid for Tuberculosis (APT) trial was designed (18). Participants with drug-susceptible pulmonary TB were randomized into three arms: pretomanid (Pa), isoniazid (H), and pyrazinamide (Z) plus either rifampin (R) (PaHZR) or rifabutin (Rb) (PaHZRb) or standard of care (HRZE). The hypothesis prior to study commencement was that pretomanid exposures would be lower when pretomanid was given together with rifampin versus rifabutin (rifabutin is a less potent inducer of CYP3A), potentially falling into a range that could reduce microbiologic response. Here, we report results from the planned interim pharmacokinetic (PK) analysis from the APT trial, incorporating data from the earlier healthy-volunteer drug-drug interaction trial, A5306 (17).

RESULTS

Data from 58 participants from the APT trial pretomanid treatment arms and 32 participants from the A5306 trial were included, 16 of whom (in A5306) provided paired pharmacokinetic profiles in the rifampin treatment arm. Baseline characteristics by study and treatment arms are summarized in Table 1. Participants contributed a total of 1,157 PK samples with a total of four pharmacokinetic visits and eight sampling occasions. Five samples (0.8%) had concentrations below the lower limit of quantitation (LLOQ), which was considered to reflect nonadherence in the preceding days; dosing and sampling records from those days were excluded.

TABLE 1.

Baseline characteristics and disease severity of study participants with smear-positive pulmonary TB receiving pretomanid-containing regimens, by treatment arm^a

Characteristic	Value in:
	A5306 trial (n = 32)b	PaHZR regimen (n = 30)	PaHZRb regimen (n = 28)	Total (n = 90)
Male [no. (%)]	18 (56.3)	22 (73.3)	19 (67.8)	59 (65.6)
Age (yr)c	28.5 (25.0–39.2)	31 (23.0–38.3)	30.0 (22.5–33.8)	30.0 (24.0–38.0)
Wt (kg)c	77.0 (69.0–91.3)	53.0 (48.5–58.3)	53.4 (49.0–60.8)	58.0 (51.0–72.9)
FFM (kg)c	53.6 (44.4–60.7)	44.8 (41.5–46.7)	44.0 (39.6–47.9)	46.3 (41.9–51.9)
Albumin (g/liter)c		35.0 (31.0–39.0)	37.0 (33.3–41.8)	36.0 (32.8–40.0)
Female		36.5 (34.0–40.8)	38.0 (36.0–40.5)	38.0 (34.0–40.0)
Male		34.5 (31.0–38.3)	37.0 (33.5–42.0)d	36.0 (33.0–39.0)
HIV positive [no. (%)]		2 (6.7)	1 (3.57)	3 (3.33)
Diabetes [no. (%)]		0 (0)	1 (3.4)	1 (1.11)
Smoking [no. (%)]
Nonsmoker		11 (36.7)	6 (21.4)	17 (18.9)
Current		13 (43.3)	17 (60.7)	30 (33.3)
Quit <1 year ago		5 (16.7)	4 (14.3)	9 (10.0)
Quit >1 year ago		1 (3.33)	1 (3.57)	2 (2.22)
Disease severity [no. (%) with baseline smear grade]
Scanty		1 (3.3)	0 (0)	1 (1.11)
1+		7 (23.3)	8 (27.6)	15 (16.7)
2+		10 (33.3)	13 (44.8)	18 (20.0)
3+		12 (40.0)	13 (44.8)	25 (27.8)
Cavitary disease [no. (%)]		25 (83.3)	24 (85.7)	49 (54.4)

^aPa, pretomanid; H, isoniazid; Z, pyrazinamide, R, rifampin; Rb, rifabutin; FFM, fat-free mass.

^bSixteen participants received rifampin.

^cData are reported as median (IQR).

^dThe median was imputed for 2 missing values.

Pretomanid PK was best described by a one-compartment disposition model with first-order elimination and dynamic transit compartment absorption. The typical value of clearance (CL/F) for the median patient receiving pretomanid alone or with rifabutin was 3.90 liters/h. Treatment arm by study was a significant covariate when tested on clearance (a drop of 244 points in objective function value [OFV]; 2 degrees of freedom; P < 0.001). Patients receiving rifampin in the APT trial had 80% higher clearance, while patients in the A5306 trial had 146% higher clearance, resulting in 44.4 and 59.4% reduction in exposure in terms of overall area under the concentration-time curve (AUC), respectively. Fat-free mass was used to allometrically scale clearance, while volume of distribution was scaled by total body weight via power relationship; the exponent was fixed to 1.0 and 0.75 for volume and clearance, respectively. The model determined that pretomanid in the A5306 study had 42.2% lower bioavailability, 53% faster absorption, and 34.5% shorter absorption mean transit time (ΔOFV, −159; 2 degrees of freedom; P < 0.001). The model did not detect a significant effect of rifabutin on pretomanid clearance. The residual unexplained variability was best described by combined error model incorporating both proportional and additive components. A visual predictive check showing adequate model fit to the data is depicted in Fig. 1. The final model parameters are summarized in Table 2, and model-estimated secondary pharmacokinetic parameters are shown in Fig. 2.

FIG 1.

Visual predictive check for pretomanid concentration (in milligrams per liter) versus time (time since observed dose), stratified by study and treatment arms. Circles represent original data, dashed and solid lines are the 5th, 50th, and 95th percentiles of the original data, and the shaded areas are the corresponding 95% confidence intervals for the same percentiles, as predicted by the model. Vertical yellow lines on the x axis represent bins for sampling time points. An appropriate model is expected to have most observed percentiles within the simulated confidence intervals.

TABLE 2.

Final population pharmacokinetic model parameters^a

Parameter	Typical value (95% CIb)
CL (liters/h)c	3.91 (3.65–4.17)
Vc (liters)c	90.9 (86.7–95.7)
MTT (h)	1.16 (1.02–1.30)
NN	6 (fixed)
Ka (1/h)	0.592 (0.516–0.681)
F	1 (fixed)
Proportional error (%)	7.85 (7.18–8.65)
Additive error (mg/liter)	0.0359 (0.0251–0.0480)
APT rifampin effect on CL (%)	+80 (63.2–98.7)
A5306 study effect (%)
Rifampin effect on CL	+146 (126–171)
Kad	+53.6 (35.9–75.4)
MTTd	−34.9 (−26.4 to −43.0)
F	−41.9 (−46.3 to −36.9)
Between-subject variability (%)e
CL	19 (15.7–23.0)
F	12.5 (7.20–17.5)
Between-visit variability (%)e,f
CL	11.3 (8.83–14.2)
Between-occasion variability (%)e,f
MTT	56.8 (49.6–66.2)
Ka	39.9 (25.5–53.2)
F	19.8 (17.0–23.2)

^aCL, clearance; V_c, central volume of distribution; MTT, absorption mean transit time; NN, number of transit compartment; K_a, absorption rate constant; F, bioavailability.

^b95% confidence intervals (CI) were obtained with a sampling importance resampling technique using PsN software.

^cAllometric scaling was used for CL and V_c; typical values are reported for the typical median patient with a weight of 58 kg and an FFM of 45 kg and not receiving rifampin under APT study conditions.

^dStudy effect was modeled on K_a and MTT⁻¹ with the same study effect parameter, theta (θ).

^eBetween-subject variability, between-visit variability, and between-occasion variability were assumed to be log-normally distributed and are reported as approximate CV (in percent).

^fA visit is any PK sampling visit as defined in the protocol, whereas an occasion is any dosing event followed by at least one sampling point.

FIG 2.

Box-and-whisker plots showing individual model-predicted pretomanid AUC_0–24 and C_max on day 14 in the APT study, stratified by treatment arm. The dots represent individual values, the box is the interquartile range (IQR), and the whiskers show 2.5th and 97.5th percentiles. Data were available for 57 patients at day 14.

Probability of target attainment.

The probability of target attainment for pretomanid is depicted in Fig. 3, using the target for 1-log₁₀ bactericidal activity (48% fT_>MIC) (15). Pretomanid coadministered with rifampin or rifabutin under fed conditions showed a favorable probability of target attainment (PTA) at the recommended dose of 200 mg daily. The PTA was above 90% in both the rifampin and rifabutin arms at a MIC of 0.03125 or 0.0625 mg/liter. When the target corresponding to 1.59-log₁₀ bactericidal activity (77% fT_>MIC) was used, the PTA was above 90% at a MIC of 0.03125 or 0.0625 mg/liter.

FIG 3.

Probability of target attainment (PTA) versus MIC, assuming 15% protein binding in each treatment arm at steady state. A 1-log₁₀ killing (48% fT_>MIC) represents the PK/pharmacodynamic (PK/PD) target of bactericidal activity, while a 1.59-log₁₀ (77% fT_>MIC) PK/PD target represents 80% of maximum observed killing. Calculations were based on 10,000 simulations of the median TB patient PK observed in the APT study.

DISCUSSION

In this analysis, we evaluated the pharmacokinetics of pretomanid, when administered with isoniazid, pyrazinamide, and either rifampin or rifabutin, for the treatment of smear-positive pulmonary tuberculosis or when administered with or without rifampin in healthy volunteers and found that pretomanid exposures were reduced by nearly half when pretomanid was given with rifampin versus being given with rifabutin or alone. Pretomanid is currently approved only for the treatment of XDR and treatment-intolerant or -refractory MDR TB but holds potential as a treatment-shortening agent for drug-susceptible TB; hence the need to study pretomanid in combination with first-line drugs.

The APT trial was originally designed to determine whether pretomanid, when combined with a rifamycin-containing first-line regimen, demonstrated enhanced microbiologic activity and thus showed treatment-shortening potential for drug-susceptible TB. Results from the A5306 study, in which healthy volunteers received 200 mg pretomanid either alone or in combination with 600 mg rifampin, had shown that pretomanid exposures were reduced markedly, with AUC_0–24 falling from 34.5 to 13.4 mg·h/liter (66%) (17). The pretomanid exposures observed in A5306, when pretomanid was given with rifampin, were comparable to those seen when pretomanid was given at a much lower dose, 50 mg, without rifampin. From dose-finding early bactericidal activity (EBA) studies, doses of 50 mg given without food showed decreased activity compared to doses of 100 mg or higher (19). Therefore, there was concern that coadministration of pretomanid with rifampin could lower pretomanid exposures into a potentially subtherapeutic range. For this reason, the package insert currently states that pretomanid should not be combined with any rifamycins (although prior to the present study, pretomanid had not been studied in combination with any rifamycin other than rifampin).

In this analysis, pretomanid concentrations in the APT study were measured at steady state (days 14, 28, 56, and 84), and, as predicted, coadministration with rifampin significantly reduced pretomanid exposures. On average, patients receiving rifampin had pretomanid AUC reduced by about 44.4% compared to patients receiving rifabutin, driven by an 80% increase in clearance. Though the observed reduction in pretomanid exposure in these results is significant, it is slightly smaller in magnitude than that observed in the A5306 cohort (44.4% versus 59.4%). More importantly, the absolute individual model-predicted values of AUC_0–24 and maximum concentration (C_max) in the APT rifampin arm are similar to those in the pretomanid-only arm of the A5306 study and to those seen in trials of the now-registered dose, wherein pretomanid was given without a rifamycin (19, 20). In these earlier trials, the 200-mg dose, which was subsequently approved based on robust clinical efficacy, produced a pretomanid AUC of 36 mg·h/ml (19, 20). Even with the relative reduction in exposures, the rifampin arm in the APT trial produced exposures that were similar, in absolute terms, to those seen in registration trials (Table 3). If the exposures seen in the APT trial can be replicated in clinical practice, it would seem to suggest that pretomanid could be coadministered with rifampin without a compromising effect.

TABLE 3.

Summary of PK parameters from the APT trial compared to prior studies, among both healthy volunteers and TB patients

Study (reference)	Population	Fed	Regimen	AUC0–24 (mg·h/liter)	Cmax (mg/liter)	CL/F (liters/h)
CL-009 (21)a	Healthy	No	Pa	28.1 (8.0)	1.1 (0.2)	7.6 (2.5)
		Yes	Pa	51.6 (10.1)	2.0 (0.3)	3.9 (0.8)
APT (this study)b	Patient	Yes	Pa-HZR	30.1 (23.5–35.3)	2.14 (181–2.42)	6.78 (6.35–8.39)
			Pa-HZRb	59.5 (48.0–65.2)	3.35 (2.93–3.74)	3.79 (3.62–4.54)
A5306 (this study)c	Healthy	No	Pa-R	13.4 (9.94–18.2)	1.07 (0.731–1.43)	14.7 (10.5–20.3)
			Pa	34.5 (26.3–45.8)	1.98 (1.41–2.59)	4.48 (3.46–5.50)
CL-010 (29)d	Patient	No	Pa50	12 (1.4)	0.75 (1.4)
			Pa100	17 (1.5)	1.1 (1.4)
			Pa200	36 (1.4)	2.1 (1.3)

^aPK data were collected after a single 200-mg dose of pretomanid. Values are means (standard deviations).

^bPK data were collected on day 14 (data from 57 patients were available). Drug doses were as follows: pretomanid (Pa), 200 mg; isoniazid (H), 300 mg; pyrazinamide (Z), 1,000 to 2,000 mg; rifampin (R), 600 mg; rifabutin (Rb), 300 mg. Values are medians (IQR) determined by model-estimated individual exposure metrics.

^cPK data were collected on day 7. The pretomanid dose was 200 mg and the rifampicin dose was 600 mg. Values are medians (IQR) determined by model-estimated individual exposure metrics.

^dPK data were collected on day 14. The pretomanid doses were 50, 100, and 200 mg, as indicated. Model-generated clearance across doses (ranging from 50 to 1,200) was estimated at 4.8 liters/h. Values are geometric means (geometric standard deviations).

One reason for the observed pretomanid exposures in the rifampin-containing arm of the APT cohort being comparable with pretomanid given without rifamycins in A5306 and other efficacy trials is likely the effect of food on pretomanid absorption. Unlike in the licensing trials, in which participants were fasting, participants in the APT trial were given a standard meal prior to dosing, usually consisting of coffee and toast with butter. Interestingly, this simple breakfast appeared to enhance absorption and exposures to a degree similar to that seen with the high-fat meal provided in the earlier food effect trials. Earlier phase I trials observed a significant food effect on pretomanid exposures, with C_max and AUC increasing to 176% and 188% in the fed versus the fasting state (21). By incorporating data from A5306, the healthy volunteer trial of pretomanid plus rifampin in which participants were dosed in a fasting state, we were able to more fully explore the hypothesis that food effect drove the acceptable pretomanid exposure levels, even in the presence of rifampin. As expected, A5306 participants had faster absorption and reduced bioavailability, fitting with the fasting-state conditions. Interestingly, the model combining APT and A5306 data showed different rifampin induction effects on pretomanid PK between the two studies. Rifampin coadministration was associated with a 59.4% decrease in pretomanid AUC in A5306, versus a 44.4% decrease in APT. This differential induction effect could be attributed to several factors, including administration with food and the use of a fixed-dose combination (FDC) in APT, both of which are known to reduce rifampin exposure (22, 23).

An additional distinction between APT and the earlier study of pretomanid with rifampin is that APT is a treatment trial of participants with pulmonary TB, while the prior drug-drug interaction (DDI) study was among healthy volunteers. However, patients in APT were healthy overall, except for having pulmonary TB, and prior work demonstrated negligible differences in the PK of other TB drugs among outpatient and hospitalized patients (24). There also remains the potential that some unidentified drug-drug or metabolic effect affected the observed pretomanid exposures, either with the companion drug or with rifabutin. Though isoniazid is an inhibitor of some CYP2C/E and CYP3A enzymes (25), there is no current evidence to support an interaction between pretomanid and isoniazid. Additionally, a prior healthy-volunteer trial explored potential pharmacogenetic effects on pretomanid PK and found no influence of SLCO1B1 521 or rs4149032, CYP3A5, or CYP2B6 polymorphisms (17). Ultimately, the observed AUC in APT in a fed state seems comparable to exposures seen with a standard 200-mg dose in a fasting state, which produced potent early bactericidal activity in earlier trials.

One limitation of our study is that we did not explore the disposition of pretomanid on rifamycins, as we did not measure concentrations of rifampin and rifabutin. However, we would expect any effect of pretomanid on rifamycin metabolism to be minimal. Pretomanid is a time-dependent inhibitor of CYP3A4/5 in vitro, but a prior healthy-volunteer study found no clinical interaction between pretomanid and midazolam (a traditional probe drug for CYP3A4/5) (26). Additionally, though pretomanid is an inhibitor of organic anion-transporting polypeptide (OAT) 3 (27), rifampin and rifabutin metabolism is mediated by OAT1 (28). Finally, though our model used the published range of MICs (15), there have been some trial patients with MICs above the range described here (personal communication with TB Alliance).

Interestingly, the rifabutin-containing arm showed an apparent increase in pretomanid exposure compared to historical data of pretomanid given alone. The pretomanid AUC in this analysis is 59.5 mg·h/liter, a value between the exposures seen with 200 mg and 600 mg when pretomanid was given as monotherapy on an empty stomach in the dose-finding trial (AUC, 36 and 69 mg·h/liter, respectively) (20, 29). Given that pretomanid clearance is relatively unchanged between earlier monotherapy trials (4.8 liters/h) and the rifabutin-containing arm in APT (3.79 liters/h), this suggests that higher pretomanid AUC and C_max in the rifabutin-containing arm of APT may indeed be driven by food effect. This is further supported by the fact that pretomanid exposures in the presence of rifabutin in APT are quite similar to those observed after a single dose of pretomanid in a fed state. Whether these higher exposures in the rifabutin arm present safety issues is not yet known, but interim data have been reviewed by an independent data safety monitoring board (DSMB) and did not lead to any safety-related recommendations for protocol modification. Also, these values fall squarely within the range of exposures of pretomanid from other trials.

Our simulations showed that favorable PTA can be achieved when pretomanid is coadministered with rifamycins under fed conditions. Pretomanid achieved a PTA of more than 90% with rifabutin for 1.59-log₁₀ bactericidal activity (near-maximal killing effect). Pretomanid given with rifampin achieved a similar PTA for 1-log₁₀ bactericidal activity. The predicted PTAs should be interpreted while keeping in mind that the simulation was performed for the median patient in our cohort and that patients with different weights are expected to have different exposure, which might lead to slightly different PTA results. Also, these are PTA results for a single drug which is usually given as part of combination regimen for TB treatment, and the PTA results might be different due to synergism or antagonism effect of other administered TB drugs.

Conclusion.

In this pharmacokinetic analysis of the APT trial, bolstered by healthy volunteer data from A5306, coadministration with rifampin increases pretomanid clearance compared to rifabutin, resulting in reduced pretomanid AUC by 40 to 50%. Pretomanid AUC in the fed state of this trial, even in the presence of rifampin, is comparable to exposures seen in a fasting state, which were associated with excellent early bactericidal activity. Additionally, pretomanid AUC with rifabutin in a fed state is slightly higher than that observed in pretomanid monotherapy in a fasting state, a finding likely explained by food effect.

MATERIALS AND METHODS

Study population.

Adults (age, >18 years) with drug-susceptible pulmonary tuberculosis were recruited from the hospital and clinics of University of Cape Town in Cape Town, South Africa, between October 2014 and June 2019. Participants were eligible if they weighed 40 to 80 kg, had a Karnofsky score of ≥60, had acid-fast bacilli on a stained smear from expectorated sputum, and (if HIV positive) had a CD4 count of >350 cells/mm³ without plans to start antiretroviral therapy during the experimental phase of the trial (17, 30). Participants were excluded if they were pregnant or breast-feeding or had central nervous system TB, a history of neuropathy or epilepsy, a baseline QT interval corrected by Fridericia’s formula (QTcF) of >450 ms, lens opacity, resistance to first-line TB regimen components, or more than 5 days of TB treatment. Participants were also excluded if they had the following laboratory parameters: alanine aminotransferase (ALT) of >3 times the upper limit of normal (ULN), total bilirubin of >2 times the ULN, creatinine above the ULN, hemoglobin of <7.0 g/dl, or a platelet count of <100,000/mm³. Women of child-bearing potential were required to use two methods of contraception through 1 week after last dose of study medication. Male participants were required to use similar methods with their female partners to prevent pregnancy. The study was approved by the Institutional Review Board of Johns Hopkins University School of Medicine and University of Cape Town’s Human Research Ethics Committee and registered with clinicaltrials.gov (NCT02256696).

Experimental protocol. (i) APT study design.

APT is an open-label, three-arm, phase 2B randomized clinical trial to assess the mycobactericidal activity of 200 mg pretomanid when substituted for ethambutol and given together with first-line TB treatment (isoniazid, pyrazinamide, and a rifamycin). The experimental phase of treatment is 12 weeks, after which all participants transition to standard of care (SOC) to complete 24 weeks of TB therapy. Participants in arm 1 receive isoniazid, rifampin, pyrazinamide, and pretomanid for 8 weeks, followed by isoniazid, rifampin, and pretomanid for 4 weeks (8HRZPa/4HRPa). Participants in arm 2 receive the same treatment, except that 300 mg rifabutin daily is substituted for rifampin (8HRbZPa/4HRbPa). Participants in arm 3 receive standard of care (8HRZE/4HR). All study medication administration is directly observed, and pretomanid dosing occurs after a standardized meal. Intensive sampling for pretomanid plasma pharmacokinetics (predose and then 1, 2, 5, 8, and 24 h postdose) is performed on day 14. Sparse sampling for pretomanid PK is performed predose and 2 and 4 h postdose on days 28, 56, and 84. Serial sputum cultures are collected at baseline and on days 7, 14, 21, 28, 42, 56, 70, and 84. A pharmacokinetic dosing occasion is any dosing record with at least one sampling point, and a pharmacokinetic visit is any PK sampling visit as defined by the protocol.

(ii) Follow-up.

Participants are followed for a total of 12 months. Laboratory and clinical assessments are performed weekly for the first month and then biweekly for the subsequent 2 months and include targeted symptom and adverse event assessment. On days 7, 14, 28, 42, 56, and 84, blood is collected to measure complete blood count, creatinine, ALT, and total bilirubin. An electrocardiogram (ECG) is performed on days 7 and 28; follow-up visual acuity and color perception is assessed on day 28.

Drug concentration analysis.

Blood samples were collected in EDTA tubes and centrifuged at 1,500 to 2,000 × g for 10 min, and plasma was then transferred to cryovials and frozen at −70°C within 1 h. Pretomanid plasma concentrations were quantified using liquid chromatography-tandem mass spectrometry (LC-MS/MS) in the Division of Clinical Pharmacology, University of Cape Town. The calibration range was 0.01 (the lower limit of quantification [LLOQ]) to 8 mg/liter. Precision (expressed as coefficient of variation) was lower than 9% at all concentrations with accuracies of 95.2 to 110%.

Pharmacokinetic and statistical analyses.

Pretomanid plasma concentrations were analyzed using nonlinear mixed-effects modeling (NONMEM 7.4 with FOCE-I [Icon Development Solutions, Ellicott City, MD, USA]). PsN (Department of Pharmaceutical Biosciences, Uppsala University, Uppsala, Sweden), Xpose, and Pirana (Certara USA, Inc., Princeton, NJ, USA) (31) were used for model run execution and tracking. Data preparation and statistical comparisons were performed using R statistical software (version 3.6.3; https://www.r-project.org/) and Stata (version 15; Stata Corporation, College Station, TX, USA).

Different structural models describing typical pharmacokinetic characteristics were evaluated. To describe the absorption process, we tested first-order absorption with lag or chain-of-transit compartments (32); for the disposition process, we tested one- and two-compartment models with first-order elimination. We tested proportional, additive, and combined error models to describe residual unexplained variability. The typical value of relative oral bioavailability (F) was fixed to 1; thus, the clearance and volume parameters are reported relative to oral F.

The model included between-subject and between-occasion variability on the pharmacokinetic parameters assuming log-normal distribution. Allometric scaling (33) was applied to disposition parameters to account for effect of body size: we tested total body weight (TBW), fat-free mass (FFM), and body fat proportion as body size descriptors. The individual values of FFM were derived from observed TBW, height, and sex (34). Fat mass was obtained as the difference between TBW and FFM. Different covariates were tested on the PK parameters, including, age, sex, HIV, serum creatinine, albumin levels, sex, smoking status, and treatment arm. The development of the model and the inclusion of covariates were based on physiological plausibility, inspection of diagnostic plots, including visual predictive checks (35), and decreases in the objective function value (OFV), which was assumed to follow a chi-square distribution. The statistically significant cutoff for an additional degree of freedom (inclusion of one additional parameter) was an OFV drop of at least 3.84 points, corresponding to a P value of <0.05.

To evaluate percent fT_>MIC and obtain the probability of target attainment (PTA) for each treatment arm at specific MICs, a simulation of 10,000 replicates was performed using final parameter estimates of the typical median patient with TB observed in the APT study for each treatment arm, assuming protein binding to be 85% (personal communication with TB Alliance) and a MIC range of 0.015 to 0.25 mg/liter. The corresponding PTA was calculated as the proportion of simulated replicates with an fT_>MIC of at least 48 or 77%. (15).

Interim analysis.

An interim analysis was planned for when 50% of patients had completed treatment to evaluate safety, efficacy, and PK. An unadjusted hazard ratio for time to culture conversion was estimated using the Cox proportional hazards model, and a hazard ratio of at least 1.20 was required for an experimental arm to continue enrollment. The full sample size of 183 participants was based on a two-sided type 1 error of 5%, a power of 90%, a target hazard ratio of ≥1.9, and 15% nonevaluable or lost to follow-up. The interim analysis was powered at 95% to ensure an end-of-study power at 90%. The interim PK analysis was intended to contextualize interim efficacy results and, if necessary, to inform dosing. Aggregated safety data as well as PK data were reviewed by the study team and data safety monitoring board (DSMB), and arm-specific efficacy and safety data were reviewed in a closed session by the DSMB. Despite the coronavirus disease 2019 (COVID-19) pandemic, study activities have resumed, and complete enrollment is anticipated.

A5306 trial data.

In order to fully evaluate the effect of rifampin on pretomanid pharmacokinetics, taking into account the contribution of food, permission was obtained to analyze data from the AIDS Clinical Trials Group 5306 (A5306) trial (17). This phase I study sequentially assigned healthy volunteers to receive pretomanid with either lopinavir-ritonavir, efavirenz, or rifampin. In the rifampin arm, participants received pretomanid for 7 days (period 1), rifampin for 7 days (period 2), and then combination therapy with rifampin and pretomanid for 7 days (period 3). All study medications on PK sampling days were taken on an empty stomach. Pretomanid was given 200 mg once daily; adherence was assessed by pill counts and medication diaries. Pharmacokinetic samples for pretomanid were obtained predose and 1, 2, 3, 4, 5, 6, 8, 10, 12, and 24 h postdose at the ends of periods 1 and 3. Plasma concentrations for pretomanid were quantified using LC-MS/MS. The bioanalytical method was validated over a linear range of 0.01 to 10 mg/liter. The interday precision (coefficient of variation [CV], in percent) ranged from 2.65 to 4.71%, and the intraday precision (CV) was 1.58 to 6.28% (17).