Abstract
Background:
Ethambutol, a longstanding treatment for tuberculosis, lacks comprehensive pharmacokinetic data, particularly regarding covariates affecting exposure and their link to treatment outcomes. Using the largest data to date from Study 31/AIDS Clinical Trials Group A5349, we aimed to characterize ethambutol pharmacokinetics, identify risk factors for lower exposure, quantify relationships between exposure and efficacy and safety outcomes, and explore dosing considerations that optimize risks and benefits.
Methods:
We analyzed ethambutol pharmacokinetic data using nonlinear mixed-effect modeling. We evaluated the contribution of ethambutol exposure to efficacy outcomes using parametric time-to-event modeling and safety outcomes using logistic regression.
Results:
In 1,493 participants with 3,077 plasma samples, a one-compartment model with first-order elimination best described ethambutol pharmacokinetics, with diabetes, higher creatinine clearance, and higher weight linked to increased clearance. Higher ethambutol exposure was modestly associated with grade 3 or higher adverse events and a reduced risk of tuberculosis-related unfavorable outcomes. Target exposures for 95% cure were 8.9 µg∙h/mL in the 6-month control regimen and 25.6 µg∙h/mL in the 4-month rifapentine regimen. Simulations showed that reducing doses in middle (1200 to 1000 mg) and highest (1600 to 1200 mg) weight-bands could equalize median exposures, lowering toxicity risk while maintaining efficacy in the control regimen, but might not achieve target exposure in the rifapentine regimen.
Conclusion:
The contribution of ethambutol exposure is regimen-dependent, with a higher efficacy target required for the 4-month rifapentine regimen. Dose reductions for patients above 55 kg may mitigate toxicity in standard therapy but risk compromising efficacy in the shorter regimen.
Keywords: tuberculosis, ethambutol, population pharmacokinetics, dose-response, exposure-response
summary:
In this largest ethambutol pharmacokinetic/pharmacodynamic analysis to date, diabetes and higher weight were liked to lower exposure. Dose reductions for patients above 55 kg may mitigate toxicity in standard therapy but risk compromising efficacy in the shorter regimen.
INTRODUCTION
Tuberculosis (TB) remains the world’s leading infectious killer, with an estimated 10.7 million new cases and 1.23 million deaths in 2024 [1]. For over 50 years, drug-susceptible TB has been treated with a standard regimen consisting of 4 antimicrobial drugs: rifampicin, isoniazid, pyrazinamide, and ethambutol [1].
The Tuberculosis Trials Consortium (TBTC) Study 31/Advancing Clinical Therapeutics Globally (ACTG) A5349 (S31/A5349; NCT02410772), an international, multicenter, randomized, open-label, phase 3, noninferiority trial demonstrated the noninferiority of a 4-month regimen of once-daily rifapentine, isoniazid, pyrazinamide, and moxifloxacin when compared to the 6-month standard of care regimen of once-daily rifampicin, isoniazid, pyrazinamide, and ethambutol (control regimen) [2,3]. A 4-month regimen of rifapentine, isoniazid, pyrazinamide, and ethambutol (rifapentine regimen) failed to achieve noninferiority [3]. S31/A5349 is the first phase 3 TB trial that included pharmacokinetic (PK) sampling for all component drugs, thus providing an exceptional opportunity to build robust population PK models.
Since its FDA approval, ethambutol has been used for over 50 years in the first-line TB treatment regimen [1]. Ethambutol inhibits arabinosyltransferases, enzymes responsible for synthesizing the mycobacterial cell wall component arabinogalactan. This weakens the bacterial cell wall, increasing susceptibility to other TB drugs and reducing bacterial load in the early phase of treatment [4]. At 15–20 mg/kg daily, ethambutol protects against the emergence or expansion of resistance to companion drugs [5]. Optic neuritis is a well-known, dose-dependent adverse event (AE) leading to impaired visual acuity, although ocular toxicity is rare when used at recommended doses for 2 to 3 months [5]. The risk of ocular toxicity is dose-dependent, where incidence in adults exceeds 40% at doses above 50 mg/kg, compared with 0–3% at the 15 mg/kg daily dose [6]. Ethambutol is primarily eliminated by the kidneys, with 50 to 70% of the dose excreted unchanged renally, and has a half-life of approximately 4 hours [7]. Despite the long history of use, literature lacks ethambutol PK information from large patient populations with characterization of covariates to identify individuals at risk for low or high exposure, which may influence optimal dosing for these groups. S31/A5349 contains the largest dataset on ethambutol PK, and therefore, represents an unprecedented opportunity to analyze ethambutol PK data in the context of detailed safety and efficacy data.
Here, we characterized ethambutol PK to (1) identify subpopulations at risk for low or high drug exposure, (2) characterize PK-safety relationships, (3) describe PK-efficacy relationships with respect to TB-related unfavorable outcomes, and (4) perform simulations to suggest potential dosing considerations across subpopulations.
METHODS
Study Design and Pharmacokinetic Sampling
The trial was approved by the U.S. CDC Institutional Review Board (IRB). Each participating institution provided for the review and approval of this protocol and its informed consent documents by a local IRB or ethics committee or relied formally on the U.S. CDC IRB approval.
Participants received 800 mg (<55kg), 1200 mg (55–75kg), and 1600 mg (>75kg) oral doses of ethambutol (administered as individual 400 mg tablets) for 8 weeks. The rifapentine regimen was administered within 1 hour after food intake, and the control regimen was administered under fasted conditions. Semi-intensive sampling was performed on a small number of participants in the rifapentine regimen with samples taken at 0.5, 6-, and 12-hours post-dose. The remaining participants were sampled sparsely with time points at 0.5- and 5-hours post-dose. Plasma concentrations of ethambutol were determined using validated liquid chromatography coupled with tandem mass spectrometry. Detailed methods have been published previously [2,3].
Modeling Software and Methods
PK data were randomly split into a training cohort (67%) for model development and validation cohort (33%) for model validation, where clinical site was balanced between the two cohorts. Data were analyzed using nonlinear mixed-effects modeling with NONMEM (version 7.41, Icon Development Solutions, Dublin, Ireland and Perl-speaks-NONMEM (PsN; version 5.18.2). Data manipulation and visualization were carried out using R (version 3.6.1). Parameters were estimated using first-order conditional estimation method with interaction [8,9]. One- or two-compartment models with first-order elimination were explored as a structure model. First- or zero-order absorption models with either lag time or transit compartment absorption were tested. Interindividual variability was assumed to follow a log-normal distribution.
Variables, such as weight, age, sex, self-reported race, treatment arm, lung cavitation, extent of disease, smear grade, HIV status, diabetes status, and renal function were evaluated as potential covariates on PK parameters including apparent clearance (CL/F), bioavailability (F), and apparent volume of distribution (V/F). Covariates were selected through stepwise procedure with forward selection (P<0.05) and backward elimination (P<0.01) and included based on statistical significance, scientific plausibility, and clinical relevance.
Pharmacokinetic-Efficacy and Pharmacokinetic-Safety Analyses
Steady state area under the concentration-time curve from 0–24 hours (AUC0–24h,ss) and maximal plasma concentration (Cmax) were used as the markers of ethambutol exposure. The primary efficacy outcome was time to TB-related unfavorable outcomes over 12 months of follow-up, while the primary safety outcome was any grade 3 or higher (grade 3+) AEs throughout the on-treatment period. Additional safety outcomes included treatment-related grade 3+ AEs, serious AEs, premature treatment discontinuation for reasons unrelated to microbiological ineligibility, and death. The association between exposure and the primary efficacy outcome was examined using parametric time-to-event models adjusted for treatment arm to determine the exposure required for a 95% sustained cure rate. The relationship between exposure and each safety outcome was evaluated using logistic regression, with both unadjusted and adjusted models accounting for risk factors associated with the primary efficacy outcome including treatment arm, age, sex, clinical site, and HIV status.
Dosing Simulations
Dosing simulations were performed with the final population PK model to evaluate potential dose adjustments for participants in the middle and highest weight bands when compared to the standard WHO guidelines [5]. Dosing regimens were optimized to maximize the percentage of participants achieving the target exposure required for efficacy while minimizing the risk of grade 3+ AEs.
RESULTS
Data Characteristics
The ethambutol PK dataset consisted of 1493 participants of whom 27 were semi-intensively sampled in the rifapentine regimen (Figure 1). Participant demographics and clinical factors were similar across the two ethambutol-containing arms and between the model development and validation datasets (Table 1, Table S1). Furthermore, only 7 samples were below the limit of quantification (lower limit of quantification=0.01 µg/mL) and were excluded from the PK analysis.
Figure 1. Observed ethambutol plasma concentrations over time after dose, stratified by treatment arm.
Circles represent individual observations and solid lines represent medians by dose level. Data is colored by dose, dark red for 800 mg, blue for 1200 mg, and yellow for 1600 mg. Panels represent the 4-month rifapentine regimen (left) and 6-month control regimen (right).
Table 1.
Summary of patient characteristics and pharmacokinetic data included in the analysis.
| 2HRZE/4HR (n=723) | 2HPZE/2HP (n=770) | Total (n=1493) | Missing | |
|---|---|---|---|---|
|
| ||||
| PHARMACOKINETIC SAMPLING | ||||
| Semi-Intensive sampling Participants | N=0 (0 samples) | N=27 (187 samples) | N=27 (187 samples) | -- |
| Sparse Sampling Participants | N=723 (1422 samples) | N=743 (1468 samples) | N=1466 (2890 samples) | -- |
| Below Limit of Quantification | 2 samples | 5 samples | 7 samples | -- |
|
| ||||
| DEMOGRAPHIC FACTORS | ||||
| Age [years] | 31 (13 – 77) | 31 (14 – 81) | 31 (13 – 81) | 0 (0) |
| Male Sex | 511 (70.7) | 545 (70.8) | 1056 (70.7) | 0 (0) |
| Height [cm] | 167 (140 – 200) | 167 (145 – 191) | 167 (140 – 200) | 0 (0) |
| Weight [kg] | 53 (40 – 122) | 53 (40 – 97) | 53 (40 – 122) | 0 (0) |
| BMI [kg/m 2 ] | 19 (13 – 45) | 19 (13 – 41) | 19 (13 – 45) | 0 (0) |
| Race | ||||
| Black | 518 (71.6) | 549 (71.3) | 1067 (71.5) | |
| Mixed/Multi-racial | 86 (11.9) | 111 (14.4) | 197 (13.2) | |
| Asian | 91 (12.6) | 88 (11.4) | 179 (12.0) | |
| White | 11 (1.5) | 7 (0.9) | 18 (1.2) | |
|
| ||||
| CLINICAL FACTORS | ||||
| African Clinical Site | 537 (74.3) | 577 (74.9) | 1114 (74.6) | |
| Dose | ||||
| 800 mg | 425 (58.8) | 454 (59.0) | 879 (58.9) | |
| 1200 mg | 278 (38.5) | 293 (38.1) | 571 (38.2) | |
| 1600 mg | 20 (2.8) | 23 (3.0) | 43 (2.9) | |
| Baseline Chest X-Ray Cavity present | 537 (74.3) | 562 (73.0) | 1099 (73.6) | |
| Baseline Chest X-Ray Class | 9 (0.6) | |||
| No cavities | 182 (25.2) | 202 (26.2) | 384 (25.7) | |
| Cavities < 4cm | 243 (33.6) | 239 (31.0) | 482 (32.3) | |
| Cavities ≥ 4cm | 294 (40.7) | 324 (42.1) | 618 (41.4) | |
| Baseline Chest X-Ray Extent | 9 (0.6) | |||
| Lesions < 1/4 thoracic area | 113 (15.6) | 136 (17.7) | 249 (16.7) | |
| Lesions 1/4 to < 1/2 thoracic area | 319 (44.1) | 328 (42.6) | 647 (43.3) | |
| Lesions ≥ 1/2 thoracic area | 287 (39.7) | 301 (39.1) | 588 (39.4) | |
| Baseline Smear Grade | 2 (0.1) | |||
| Negative | 22 (3.0) | 36 (4.7) | 58 (3.9) | |
| Scanty | 118 (16.3) | 128 (16.6) | 246 (16.5) | |
| Grade 1 | 179 (24.8) | 176 (22.9) | 355 (23.8) | |
| Grade 2 | 220 (30.4) | 222 (28.8) | 442 (29.6) | |
| Grade 3 | 183 (25.3) | 207 (26.9) | 390 (26.1) | |
| Karnofsky Score | 90 (60 – 100) | 90 (60 – 100) | 90 (60 – 100) | 0 (0) |
| HIV | 57 (7.9) | 66 (8.6) | 123 (8.2) | 1 (0.1) |
| Diabetes | 52 (7.2) | 48 (6.2) | 100 (6.7) | 0 (0) |
Abbreviations: BMI, body mass index; HIV, human immunodeficiency virus.
Data are expressed as n (%); continuous variables are expressed as median (range).
Ethambutol Pharmacokinetic Model
A one-compartment distribution model with first-order elimination and first-order absorption with lag time best described the ethambutol PK data (Figure S1). The final PK model demonstrated an appropriate fit to the observed data (Figure S2). The final PK parameters for the full cohort were well estimated with no significant differences from the estimates from the training and validation cohorts (Table 2, Table S2).
Table 2.
Final parameter estimates for the ethambutol population pharmacokinetic model.
| Parameters | Estimate | RSE (%) |
|---|---|---|
|
| ||
| CL/F (L/h)a | 62.0 | 1.5% |
| V/F (L) | 386 | 3.4% |
| Ka (1/h)b | 0.596 | 7.7% |
| ALAG (h) | 0.229 | 3.6% |
| Weight exponent on CL/F | 0.519 | 14.4% |
| CrCL exponent on CL/F | 0.267 | 17.4% |
| % Increase in CL/F for diabetes | 13.1 | 38.9% |
| % Decrease in F for mixed racec | 7.95 | 24.8% |
| % Increase in F for 4-month rifapentine regimenc | 15.9 | 13.5% |
| % Decrease in Ka for 4-month rifapentine regimen | 30.1 | 13.6% |
| IIV CL/F (CV%) | 16.8 | 14.7% |
| IIV V/F (CV%) | 30.5 | 14.2% |
| IIV Ka (CV%) | 90.3 | 3.3% |
| Proportional residual error (%) | 30.1 | 2.1% |
Abbreviations: CL/F, apparent clearance; V/F, apparent volume of distribution; F, bioavailability; Ka, absorption rate constant; ALAG, absorption lag time; CrCL, creatinine clearance estimated using the Cockcroft–Gault equation; IIV, inter-individual variability; CV, coefficient of variation; RSE, relative standard error.
Parameter estimates are based on the administered dose of ethambutol dihydrochloride (salt form).
CL/F (L/h) = 62.0 x (Weight / 53 kg)(0.519) x (CrCL / 112 mL/min)(0.267) x (1 + 0.131 x DM), where DM = 1 for participants with diabetes and DM = 0 for participants without diabetes.
Ka (1/h) = 0.596 x (1 – 0.301 x ARM), where ARM = 1 for 4-month rifapentine regimen and ARM = 0 for 6-month control regimen.
F = (1 – 0.0795 x RACE) x (1 + 0.159 x ARM), where RACE = 1 for mixed race and RACE = 0 for Black or Asian race, and ARM = 1 for 4-month rifapentine regimen and ARM = 0 for 6-month control regimen.
Effect of Covariates on Ethambutol Pharmacokinetics
Diabetes, body weight, and creatinine clearance were identified as significant covariates influencing CL/F. Participants with diabetes had a 13.1% higher CL/F, while body weight and creatinine clearance were positively associated with CL/F (Table 2). In addition, participants in the 4-month rifapentine regimen exhibited a 15.9% higher F and a 30.1% lower Ka than those in the 6-month control regimen, while participants who self-reported mixed race exhibited an 8.0% lower F than those who self-reported Black or Asian race. The impact of significant covariates on dose-normalized ethambutol exposure is illustrated in Figure 2.
Figure 2. Model-derived dose-normalized ethambutol exposure stratified by significant covariates (body weight, creatinine clearance, diabetes status, and treatment arm).
All AUC values are dose normalized to the 800 mg dose. The dashed red line represents median dose-normalized ethambutol AUC (13.4 µg∙h/mL) in the overall study population. Dots represent median exposure and error bars represent the 2.5th to 97.5th percentiles of the exposure distribution.
Pharmacokinetic-Efficacy and Pharmacokinetic-Safety Analyses
Ethambutol exposures are associated with any grade 3+ AEs, where events increased with increased ethambutol exposure (AUC) tertiles in both rifapentine and control regimens. Using the observed data, a 5-mg·h/L increase in ethambutol AUC0–24h,ss was significantly associated with grade 3+ AEs in univariable analysis (unadjusted OR, 1.30; 95% CI: 1.06–1.59, P = 0.011). In our multivariable analysis adjusted for arm, arm and pyrazinamide AUCss, arm and isoniazid AUCss, a 5-mg·h/L increase in ethambutol AUC0–24h,ss was significantly associated with grade 3+ AEs (adjusted OR, 1.48; 95% CI: 1.18–1.85, P = 0.00055; adjusted OR, 1.45; 95% CI: 1.15–1.81, P = 0.0013; adjusted OR, 1.48; 95% CI: 1.18–1.84, P = 0.00057, respectively). When considering treatment-related grade 3+ AEs association with 5-mg·h/L increase in ethambutol AUC0–24h,ss, the observed pattern was similar in both unadjusted and adjusted analyses, although there was no evidence for a relationship between this outcome and ethambutol exposure. (Figure 3, Table S3, Table S4).
Figure 3. Relationship between ethambutol exposure and the occurrence of any grade 3 or higher adverse events.
Participants were stratified by tertiles of ethambutol area under the concentration-time curve (AUC). The proportion of participants experiencing grade ≥3 adverse events increased across AUC tertiles in both the 6-month control (2HRZE/4HR) regimen (left) and the 4-month rifapentine (2HPZE/2HP) regimen (right). For each tertile, the percentage of participants experiencing any grade 3 or higher adverse events is shown with the corresponding number of events reported in parentheses.
The most serious AE caused by ethambutol is ocular toxicity. We found that 1.8% (N=29) of unique participants encountered AEs related to ocular toxicity, where one participant encountered bilateral optic neuropathy, and one participant encountered left eye grade 3 decline in visual acuity (Figure S3). These two participants had exposures close to median ethambutol AUC (15.8 µg∙h/mL). Both events were assessed as possibly related to ethambutol by site clinicians and both were resolved.
Our analysis identified ethambutol exposure thresholds associated with favorable outcomes. In the control regimen, an ethambutol AUC of 8.88 µg∙h/mL (90% CI, <1.00–13.3 µg∙h/mL) was estimated to achieve 95% durable cure. In the rifapentine regimen, this target was higher at 25.6 µg∙h/mL (90% CI, 21.4–52.7 µg∙h/mL). Lower ethambutol exposures were significantly associated with earlier and increased incidence of TB-related unfavorable outcomes (Figure 4, Table S5). The relationship between ethambutol AUC and the hazard of unfavorable outcomes followed an exponential trend, where decreasing ethambutol exposure led to a sharp increase in risk.
Figure 4. Ethambutol exposure associated with primary efficacy and safety outcomes.
Parametric time-to-event models estimated ethambutol area under the concentration-time curve (AUC) targets associated with 95% durable cure of 8.9 µg∙h/mL (90% confidence interval (CI), <1.00–13.3 µg∙h/mL) in the 6-month control (2HRZE/4HR) regimen and 25.6 µg·h/mL (90% CI, 21.4–52.7 µg·h/mL) in the 4-month rifapentine (2HPZE/2HP) regimen. Logistic regression analysis demonstrated that increasing ethambutol AUC was associated with an increased probability of any grade 3 or higher adverse events (odds ratio (OR) 1.30 for every 5 µg∙h/mL increase in ethambutol AUC; 95% CI 1.06–1.59; P = 0.011).
Safety outcomes were also evaluated in relation to ethambutol exposure. A PK-safety model predicted the probability of experiencing any grade 3+ AEs based on ethambutol AUC. Logistic regression analysis showed that higher ethambutol exposures were significantly associated with an increased probability of severe AEs. Specifically, for every 5 µg∙h/mL increase in ethambutol AUC, the odds of experiencing grade 3+ AEs increased by 30% (unadjusted odds ratio (OR) 1.30, 95% CI 1.06–1.59, P = 0.011). The integration of efficacy and safety highlighted the trade-off between achieving optimal ethambutol exposure for TB treatment and minimizing toxicity risks (Figure 4).
Dosing Simulations
Median simulated exposures for participants receiving 800 mg, 1200 mg, and 1600 mg daily doses were 14.3, 18.5, and 21.0 µg∙h/mL, respectively. Two dosing strategies were evaluated: (1) the WHO standard weight-based dosing strategy and (2) an alternative dosing in which participants in the middle weight band received 1000 mg rather than 1200 mg and those in the highest weight band received 1200 mg rather than 1600 mg. Dosing simulations demonstrated that under the alternative dosing strategy, median exposures in the middle and highest weight bands would be 15.4 and 15.7 µg∙h/mL, respectively (Figure S4). Compared with WHO weight-band dosing, the alternative dosing produced consistent exposures across weight bands while maintaining exposures above the estimated target associated with 95% durable cure in the control regimen and lowering the risk of grade 3+ AEs. However, under the alternative dosing, exposures remained below the estimated target associated with 95% durable cure in the rifapentine regimen (25.6 µg∙h/mL).
DISCUSSION
In this study, we characterized ethambutol PK using the largest dataset for ethambutol PK to date, where ethambutol had 6-fold variability in AUC. We showcase several findings: (1) participants with high body weight, high creatinine clearance, and/or with diabetes are at risk for low ethambutol exposure through increased CL/F, (2) ethambutol exposure is associated with efficacy and safety outcomes, and (3) individuals in the highest and middle weight-bands for the control regimen may benefit from dose reductions, reducing risk of grade 3+ AEs without increasing risk of TB-related unfavorable outcomes, however, dose reductions would not be beneficial for the rifapentine regimen. Hence, weight-adjusted ethambutol dosing remains appropriate, but we suggest an ethambutol dose range of 800–1200 mg rather than 800–1600 mg.
Our analyses suggest that ethambutol exposure is more critical in the novel short 4-month rifapentine regimen, which exhibits a steeper exposure-efficacy curve than in the 6-month standard regimen. In the rifapentine regimen, higher ethambutol exposure was associated with fewer TB-related unfavorable outcomes, but also with an increased moderate risk of grade 3+ AEs. In contrast, in the control regimen, the relationship between ethambutol exposure and TB-related unfavorable outcomes was less pronounced. The exposure target required to optimize treatment success was higher in the rifapentine regimen (25.6 µg∙h/mL) than in the control regimen (8.9 µg∙h/mL), highlighting greater reliance on ethambutol in the rifapentine regimen. Importantly, exposures typically achieved with an 800 mg dose aligned with the target in the control regimen while maintaining a modest safety profile. However, the increased risk of AEs with higher ethambutol exposure remains a concern. Specifically, for every 5 µg∙h/mL increase in ethambutol AUC, the odds of grade 3+ AEs rose by 30%. This translates to a 51% relative increase in risk for individuals in the 90th percentile of ethambutol exposure compared to those in the 10th percentile (Figure 3). While these findings emphasize the need for careful dose optimization in the rifapentine regimen, the clinical significance of this risk increase warrants further evaluation.
The developed population PK model using a dataset that comprises 1493 participants underscores the robustness and comprehensiveness of our research in understanding ethambutol PK. Previous investigators described a two-compartment disposition model with first-order elimination, and transit compartment absorption for ethambutol [10,11]. However, in our analysis, a two-compartment model did not adequately describe the data, as indicated by a lack of improvement in the objective function value and high parameter uncertainty (data not shown). Instead, our dataset supported a one-compartment model, suggesting that ethambutol PK may be better characterized with a simpler structural model in this diverse study population. Our application of a one-compartment model agrees with the findings among South African children with TB, North American adults with TB, and co-infected adults with TB/HIV in Rwanda [12–14]. Furthermore, we found that adding ALAG best described ethambutol’s absorption phase, consistent with findings among adult healthy volunteers [7,15,16] . Median ethambutol exposure among our study participants was slightly lower when compared with older studies, despite similar weight-based dosing. Previous studies reported median AUC values ranging from 15 to 25 µg∙h/mL in participants receiving 15–20 mg/kg ethambutol [11,12,15,16]. However, the most recent study by Hung et al. reported median AUC values of 13.02, 16.03, and 20.0 µg·h/mL among participants with normal, mild, and moderate renal impairment, respectively, which are comparable to the median AUC of 15.3 µg∙h/mL observed in our study [17] (Table S6).
Consistent with previous population PK analyses, body weight was identified as a significant covariate of ethambutol CL/F [10,14,17–20]. In the present analysis, several parameterizations of the weight effect were evaluated, including an exponential function with an estimated weight effect, a power function with an estimated weight exponent, and a fixed allometric exponent of 0.75. Although all evaluated approaches produced comparable CL/F estimates across the interquartile range of observed body weights (48–59 kg), the empirically estimated power function provided the best overall description of the weight–clearance relationship while minimizing overprediction of clearance at higher body weights (Figure S5). Therefore, the power function was selected to describe the effect of body weight on CL/F in the final model. Given the large and diverse Study 31 population, the estimated weight effect was considered well supported by the data, supporting the generalizability of the final model.
In addition to body weight, creatinine clearance was identified as a significant predictor of ethambutol CL/F, supporting the importance of renal function in ethambutol elimination. This finding is consistent with the predominantly renal elimination pathway of ethambutol and with previous reports demonstrating reduced ethambutol exposure among individuals with impaired renal function [17]. In contrast, HIV infection was not found to affect ethambutol exposure in our study population, despite previous reports suggesting lower exposures among people living with HIV [10,13]. Similar to our findings, Denti et al. reported that HIV status and sex were not significant predictors of ethambutol PK [11]. Treatment arm was identified as a significant covariate on both F and Ka. Because food intake differed systematically between treatment arms, these findings may partly reflect a food effect on ethambutol absorption. However, previous studies have reported minimal effects of food on overall ethambutol exposure [7], and food intake was confounded with treatment arm in Study 31. Therefore, the observed treatment arm effects should be interpreted cautiously. Diabetes remained a significant covariate on CL/F even after accounting for body weight and renal function. Exploratory analyses further suggested that the observed diabetes effect was not solely attributable to differences in body weight or metformin use (Figure S6). The observed influence of diabetes on ethambutol CL/F likely reflects early-stage alterations in renal physiology in individuals with diabetes, specifically affecting active tubular secretion and transporter activity [21]. Because ethambutol is subject to active renal secretion, these diabetes-associated changes—which often precede significant changes in glomerular filtration [22]—offer a mechanistic explanation for the significant effect of diabetes on ethambutol PK.
Using the largest TB population in literature to date, our analyses reveal interesting results. In both old and new studies, it was shown that the addition of ethambutol to the current TB treatment regimen was achieved primarily to protect against bacterial resistance to the core anti-TB drugs [17,23]. Our analysis demonstrates that higher ethambutol exposure in the rifapentine regimen is associated with a reduced risk of TB-related unfavorable outcomes. Additionally, while increased ethambutol exposure was modestly associated with a higher risk of grade 3+ AEs, this relationship remained consistent in both adjusted and unadjusted analyses. We also estimated the ethambutol exposure levels required to achieve 95% sustained cure at 12 months, which differed between regimens: 25.6 µg∙h/mL for the rifapentine regimen and 8.9 µg∙h/mL for the control regimen.
Some of the strengths of this study include the large sample size and the comprehensive characterization of participants. Compared to previous reports on ethambutol PK, our study benefits from a more diverse and well-defined participant group, with detailed clinical, demographic, and PK data. This includes information on comorbidities, treatment regimens, and relevant covariates, which provide a more complete understanding of how various factors influence ethambutol exposure and its outcomes. Some of the limitations include the small number of PK samples collected at each visit and small intensive trough sampling (6.2%), limiting the characterization of the PK curve, as well as the precision of the individual estimates of exposure. In addition, semi-intensive sampling only occurred among study participants in the rifapentine regimen, and less than 3% of participants received ethambutol at the 1600 mg dose.
In conclusion, we presented the largest ethambutol PK and pharmacodynamic analysis to date. Our model identified participants at risk for low exposure including individuals with higher weight, higher creatinine clearance and/or diabetes. Our simulations suggested that participants in the middle and highest weight bands may benefit from lowering ethambutol doses by reducing exposure-associated AE risk while still meeting the exposure target for the 6-month control regimen, but not for the 4-month rifapentine regimen.
Supplementary Material
ACKNOWLEDGMENTS
The authors are saddened by the passing of William C. Whitworth. We gratefully acknowledge his valuable contributions to the coordination of pharmacokinetic sample shipping, quality assurance, and data management activities, which were essential to the completion of this work. The authors thank all the S31/A5349 trial participants for making this work possible, and the trial networks Tuberculosis Trials Consortium (TBTC) and AIDS Clinical Trials Group (ACTG) for their support.
FUNDING
Funding support for this trial was provided by the U.S. Centers for Disease Control and Prevention, National Center for HIV/AIDS, Viral Hepatitis, STD, and TB Prevention, Division of Tuberculosis Elimination (contracts 200-2009-32582, 200-2009-32593, 200-2009-32594, 200-2009-32589, 200-2009-32597, 200-2009-32598, 75D30119C06702, 75D30119C06701, 75D30119C06703, 75D30119C06222, 75D30119C06225, 75D30119C06010); and by the National Institute of Allergy and Infectious Diseases of the National Institutes of Health under Award Numbers UM1 AI068634, UM1 AI068636 and UM1 AI106701. Sanofi (Paris, France, and Bridgewater, New Jersey, USA) donated rifapentine and all other study drugs, supported shipping of study drugs to all sites, and provided funding support for pharmacokinetic testing and preparation of the final Clinical Study Report in this collaborative study. The Division of Tuberculosis Elimination at the U.S. Centers for Disease Control and Prevention served as the data and coordinating center for the study.
Footnotes
Consortium Identifiers: Tuberculosis Trials Consortium Study 31. AIDS Clinical Trials Group A5349
DISCLAIMER
The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention (CDC), the National Institute of Allergy and Infectious Diseases, or the U.S. Department of Health and Human Services.
Potential conflicts of interest: The authors: No reported conflicts of interest. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest.
Data Availability Statement:
The standardized, individual data for the S31/A5349 trial (NCT02410772) that support the development of this study are currently in the preparation of data deidentification. Once deidentified data are ready, the full dataset, including data dictionaries, will be made available, likely in a data repository format with specific access rules. Please note that no time estimates for the release of the full dataset are available at present.
REFERENCES
- 1.Global tuberculosis report 2025 [Internet]. [cited 2026 Jan 22]. Available from: https://www.who.int/publications/i/item/9789240116924
- 2.Dorman SE, Nahid P, Kurbatova EV, et al. High-dose rifapentine with or without moxifloxacin for shortening treatment of pulmonary tuberculosis: Study protocol for TBTC study 31/ACTG A5349 phase 3 clinical trial. Contemp Clin Trials. 2020; 90:105938. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Dorman SE, Nahid P, Kurbatova EV, et al. Four-Month Rifapentine Regimens with or without Moxifloxacin for Tuberculosis. N Engl J Med. 2021; 384(18):1705–1718. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Chauhan M, Barot R, Yadav R, et al. The Mycobacterium tuberculosis Cell Wall: An Alluring Drug Target for Developing Newer Anti-TB Drugs-A Perspective. Chem Biol Drug Des. 2024; 104(3):e14612. [DOI] [PubMed] [Google Scholar]
- 5.WHO consolidated guidelines on tuberculosis: module 4: treatment and care [Internet]. [cited 2026 Feb 13]. Available from: https://www.who.int/publications/i/item/9789240107243
- 6.Leibold JE. The ocular toxicity of ethambutol and its relation to dose. Ann N Y Acad Sci. 1966; 135(2):904–909. [DOI] [PubMed] [Google Scholar]
- 7.Peloquin CA, Bulpitt AE, Jaresko GS, Jelliffe RW, Childs JM, Nix DE. Pharmacokinetics of ethambutol under fasting conditions, with food, and with antacids. Antimicrob Agents Chemother. 1999; 43(3):568–572. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Mould DR, Upton RN. Basic concepts in population modeling, simulation, and model-based drug development-part 2: introduction to pharmacokinetic modeling methods. CPT Pharmacomet Syst Pharmacol. 2013; 2(4):e38. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Byon W, Smith MK, Chan P, et al. Establishing best practices and guidance in population modeling: an experience with an internal population pharmacokinetic analysis guidance. CPT Pharmacomet Syst Pharmacol. 2013; 2(7):e51. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Jönsson S, Davidse A, Wilkins J, et al. Population pharmacokinetics of ethambutol in South African tuberculosis patients. Antimicrob Agents Chemother. 2011; 55(9):4230–4237. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Denti P, Jeremiah K, Chigutsa E, et al. Pharmacokinetics of Isoniazid, Pyrazinamide, and Ethambutol in Newly Diagnosed Pulmonary TB Patients in Tanzania. PloS One. 2015; 10(10):e0141002. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Hiruy H, Rogers Z, Mbowane C, et al. Subtherapeutic concentrations of first-line anti-TB drugs in South African children treated according to current guidelines: the PHATISA study. J Antimicrob Chemother. 2015; 70(4):1115–1123. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Zhu M, Burman WJ, Starke JR, et al. Pharmacokinetics of ethambutol in children and adults with tuberculosis. Int J Tuberc Lung Dis Off J Int Union Tuberc Lung Dis. 2004; 8(11):1360–1367. [PubMed] [Google Scholar]
- 14.Sundell J, Bienvenu E, Birgersson S, Äbelö A, Ashton M. Population Pharmacokinetics and Pharmacogenetics of Ethambutol in Adult Patients Coinfected with Tuberculosis and HIV. Antimicrob Agents Chemother. 2020; 64(2):e01583–19. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Lee CS, Gambertoglio JG, Brater DC, Benet LZ. Kinetics of oral ethambutol in the normal subject. Clin Pharmacol Ther. 1977; 22(5 Pt 1):615–621. [DOI] [PubMed] [Google Scholar]
- 16.Lee CS, Brater DC, Gambertoglio JG, Benet LZ. Disposition kinetics of ethambutol in man. J Pharmacokinet Biopharm. 1980; 8(4):335–346. [DOI] [PubMed] [Google Scholar]
- 17.Hung TM, Jayanti RP, Lee HY, et al. Recommendation of pharmacokinetics/pharmacodynamics target of ethambutol to suppress tuberculosis resistance: A population pharmacokinetics study on a large prospective cohort. Int J Antimicrob Agents. 2023; 62(2):106840. [DOI] [PubMed] [Google Scholar]
- 18.Hall RG, Swancutt MA, Meek C, Leff RD, Gumbo T. Ethambutol pharmacokinetic variability is linked to body mass in overweight, obese, and extremely obese people. Antimicrob Agents Chemother. 2012; 56(3):1502–1507. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.McIlleron H, Wash P, Burger A, Norman J, Folb PI, Smith P. Determinants of rifampin, isoniazid, pyrazinamide, and ethambutol pharmacokinetics in a cohort of tuberculosis patients. Antimicrob Agents Chemother. 2006; 50(4):1170–1177. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Horita Y, Alsultan A, Kwara A, et al. Evaluation of the Adequacy of WHO Revised Dosages of the First-Line Antituberculosis Drugs in Children with Tuberculosis Using Population Pharmacokinetic Modeling and Simulations. Antimicrob Agents Chemother. 2018; 62(9):e00008–18. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Dostalek M, Akhlaghi F, Puzanovova M. Effect of diabetes mellitus on pharmacokinetic and pharmacodynamic properties of drugs. Clin Pharmacokinet. 2012; 51(8):481–499. [DOI] [PubMed] [Google Scholar]
- 22.Umanath K, Lewis JB. Update on Diabetic Nephropathy: Core Curriculum 2018. Am J Kidney Dis Off J Natl Kidney Found. 2018; 71(6):884–895. [DOI] [PubMed] [Google Scholar]
- 23.Horsburgh CR, Barry CE, Lange C. Treatment of Tuberculosis. N Engl J Med. 2015; 373(22):2149–2160. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The standardized, individual data for the S31/A5349 trial (NCT02410772) that support the development of this study are currently in the preparation of data deidentification. Once deidentified data are ready, the full dataset, including data dictionaries, will be made available, likely in a data repository format with specific access rules. Please note that no time estimates for the release of the full dataset are available at present.