High Isoniazid Exposures When Administered with Rifapentine Once Weekly for Latent Tuberculosis in Individuals with Human Immunodeficiency Virus

ABSTRACT

Isoniazid pharmacokinetics are not yet well-described during once weekly, high-dose administrations with rifapentine (3HP) for latent tuberculosis infection (LTBI). Fewer data describe 3HP with dolutegravir-based antiretroviral therapy for the treatment of human immunodeficiency virus (HIV). The only prior report of 3HP with dolutegravir reported elevated isoniazid exposures. We measured the plasma isoniazid levels in 30 adults receiving 3HP and dolutegravir for the treatment of LTBI and HIV. The patients were genotyped to determine NAT2 acetylator status, and a population PK model was estimated by nonlinear mixed-effects modeling. The results were compared to previously reported data describing 3HP with dolutegravir, 3HP alone, and isoniazid with neither dolutegravir nor rifapentine. The isoniazid concentrations were adequately described by a one compartment model with a transit compartment absorption process. The isoniazid clearance for slow (8.33 L/h) and intermediate (12 L/h) acetylators were similar to previously reported values. Rapid acetylators (N = 4) had clearance similar to those of intermediate acetylators and much slower than typically reported, but the small sample size was limiting. The absorption rate was lower than usual, likely due to the coadministration with food, and it was faster among individuals with a low body weight. Low-body weight participants were also observed to have greater oral bioavailability. The isoniazid exposures were consistent with, or greater than, the previously reported “elevated” concentrations among individuals receiving 3HP and dolutegravir. The concentrations were substantially greater than those presented in previous reports among individuals receiving 3HP or isoniazid without rifapentine or dolutegravir. We discuss the implications of these findings and the possibility of a drug-drug interaction that is mediated by cellular transport. (This study has been registered at ClinicalTrials.gov under identifier NCT03435146 and has South African National Clinical Trial Registration no. DOH-27-1217-5770.).

INTRODUCTION

Latent tuberculosis infection (LTBI) is estimated to impact approximately one quarter of the world’s population (1). Approximately 5 to 10% of those with LTBI will go on to develop active tuberculosis (TB) disease (2). The risk of progression is more pronounced in immunocompromised individuals, such as those living with human immunodeficiency virus (HIV), for whom the risk of dying from an active TB infection is estimated to be three times that of someone without HIV (3). The administration of TB preventive treatment (TPT) for LTBI among individuals with HIV is complicated, due to the potential for drug-drug interactions (DDI) between LTBI medications and antiretroviral therapy (ART).

A short course TPT regimen of weekly high-dose isoniazid-rifapentine for 3 months (3HP) has been endorsed by the United States Centers for Disease Control and Prevention (US CDC) since 2011 and the World Health Organization since 2018 as an option for treating LTBI, whereas dolutegravir-based ART (dART) has been a preferred first-line therapy for HIV management since 2018 (2, 4). Isoniazid (INH) is an antibiotic that has formed a central component of TPT since it was first used to treat LTBI in 1957 (5). There were initial concerns that rifapentine, a potent inducer of the enzymes responsible for dolutegravir metabolism, could reduce the efficacy of dART. Only recently was 3HP shown to be safe without dose adjustments when coadministered with dART to individuals who previously received efavirenz-based ART (6).

The INH PK are not yet well-described in individuals receiving 3HP and dART. The only prior study of 3HP and dART reported adverse drug reactions that were coincident with elevated INH concentrations and elevated cytokines (7). The INH PK are highly influenced by the N-acetyltransferase 2 (NAT2) genotype, but they are not known to display drug-drug interactions with rifapentine or dolutegravir. Similarly, the INH PK are not known to differ between people with LTBI versus active TB. However, one prior analysis has found that healthy volunteers experience elevated INH concentrations, relative to people with TB or TB and HIV (8). High INH exposures have been linked to an increased risk of hepatotoxicity, though this has not been explored in the context of intermittently-dosed INH. The goal of this study was to characterize the high-dose intermittent INH PK among adults with LTBI and HIV who were receiving 3HP and dART.

RESULTS

Population pharmacokinetic model.

A one-compartment model with transit-compartment absorption was found to adequately describe the INH PK. The final model is presented in Table 1. Though most prior models of INH use a two-compartment structure (9–12), the parameters for a second compartment were inestimable from the available data. The residual errors are best described as additive to the logged response. The allometric scaling of the clearance and volume parameters by total body weight resulted in a better fit than did scaling by the fat-free mass. The interindividual variability was estimated for the central volume, oral bioavailability, mean transit time, and number of transit compartments.

TABLE 1.

Parameter values for the final pharmacokinetic model

Parameter	Population mean (% RSEa)	95% CIb	% IIVc (% RSEa)
Mean parameters
Acetylation: fast/intermediate (L/h)	12 (3.79)	(11.1, 12.9)
Acetylation: slow (L/h)	8.33 (5.44)	(7.45, 9.22)
Central vol (L)	40 (5.33)	(35.8, 44.1)	17.6 (25.4)
Mean transit time (MTT, h)	1.46 (4.81)	(1.33, 1.6)	7.86 (72.2)
No. transit compartments	4.19 (22.2)	(2.37, 6.01)	104 (18.1)
Bioavailability (F)	1 Fixed		25.1 (11.5)
Allometric scaling exponent
Wt on V	1 Fixed
Wt on CL	0.75 Fixed
Wt on MTT	0.611 (44.5)	(0.0781, 1.14)
Wt on F	−0.491 (31.8)	(−0.798, −0.185)
Residual variability (log scale)
Additive error SDd (mg/L)	0.144 (12.1)	(0.11, 0.178)

^aRSE, relative standard error.

^bCI, confidence interval.

^cIIV, interindividual variability.

^dSD, standard deviation.

The residual diagnostics following the structural model development indicated a negative correlation between the participant weight and the normalized prediction distribution errors on observations with simulation (NPDEYS) value. Consequently, we additionally examined the allometric scaling of the oral bioavailability and mean transit time parameters based on participant body weight, with the exponent being estimated from the data. We observed significant decreases in the objective function value (OFV) that were associated with separately scaling the oral bioavailability (ΔOFV = −5.787, P = 0.016), mean transit time (ΔOFV = −5.217, P = 0.022), and both simultaneously (ΔOFV = −10.318, P = 0.00575, df = 2). The testing of sex, CD4 count, and prior TB infection revealed no significant covariate effects. Consequently, only the NAT2 acetylator status and participant body weight were included in the final model as covariates. Intermediate acetylators (IA) and rapid acetylators (RA) were estimated to have similar coefficients of 12.21 L/h and 11.05 L/h, respectively, with RA having a marginally slower clearance. Due to the low number of RA (N = 4) and the previously reported overlap in clearance rates between RA and IA, we attributed this observation to chance. The final model combined the IA and RA categories, resulting in a nonsignificant increase in the OFV (ΔOFV = 2.201). The final model visual predictive checks and residuals are shown in Fig. 1. The visual predictive checks displayed adequate agreement between the model predictions and the observations. The NPDEYS residuals followed a standard normal distribution, as expected, and they were assessed via Wilcoxon signed-rank, Fisher, and Shapiro-Wilk tests. No associations were noted between the residual values and the model predictions, time, total body weight, or NAT2 acetylator status.

FIG 1.

Visual predictive checks (VPC) and distribution of normalized prediction distribution errors on observations with simulation (NPDEYS) residuals for the final pharmacokinetic model, based on 1,000 simulated data sets from each model. The VPC solid lines indicate the median observed values, whereas the dashed lines indicate the 5th and 95th percentiles. The areas shaded in gray and purple show the ranges of values associated with the medians and the 5th and 95th percentiles in the simulated data sets. The NPDEYS residuals show associations between simulation-based residuals and predicted values, time, total body weight, and NAT2 acetylator type.

Our model estimates the clearance rates for IA/RA to be approximately 44% greater than those for slow acetylators (SA), at 12 L/h and 8.33 L/h, respectively. These are similar to the clearance rates reported in (11), in which a 32% increase in INH clearance between slow (9.2 L/h) and intermediate (12.1 L/h) acetylators was estimated in a population of individuals who were coinfected with HIV and TB disease. Also, (11) reported a greater (129%) increase in clearance associated with rapid acetylators (21.1 L/h). An increase of this magnitude was not observed among the RA in this study; however, this may have been due to the small number of RA (N = 4).

The pharmacokinetic summary statistics derived from the final model are presented in Fig. 2. The median (interquartile range [IQR]) 24 h area under the curve (AUC_0–24) was estimated to be 123 (97.5 to 139) h·mg/L among the SA and 75.1 (67.5 to 100) h·mg/L among the IA/RA. The target maximum concentration (C_max) values of 9 to 15 mg/L have previously been reported during the therapeutic drug monitoring of bi-weekly 900 mg INH (13). The C_max values in this study (DOLPHIN) were substantially higher, with median C_max values of 21.8 (14.4 to 25.0) mg/L among the SA and 15.2 (11.3 to 20.2) mg/L among the IA/RA. The time to maximum concentration (T_max) was 2.4 (2.1 to 3.0) h and 2.7 (2.2 to 3.4) h for the SA and the IA/RA, respectively. These values are later than those that are commonly reported for INH, but they are consistent with previous reports when INH is coadministered with a high-carbohydrate meal, which has been shown to significantly delay INH absorption (14).

FIG 2.

Pharmacokinetic summary statistics for N = 30 subjects enrolled in DOLPHIN, stratified by NAT2 acetylator status. The estimates are based on individual-level predictions from the final population pharmacokinetic model. AUC_0–24, area under the individually predicted concentration time curve from 0 to 24 h; C_max, maximum concentration; T_max, time of maximum concentration (hours).

Comparisons to previously published data.

The final model results are compared to previously published data in Fig. 3 and 4. Panel A of Fig. 3 compares the concentrations from the three subjects reported by (7) to 90% prediction intervals from the final DOLPHIN population PK model. Solid lines represent the observed concentrations, whereas solid and dotted lines represent the 50th and 5th/95th percentiles of the predicted concentrations, based on 1,000 simulations for each subject. The model predictions appear consistent with the observed data for Subjects 1 and 4, both of whom are slow acetylators, though the model overpredicts the concentrations for Subject 2, who is an intermediate acetylator.

FIG 3.

(A) Comparison of observed isoniazid (INH) concentrations to DOLPHIN model predictions for three individuals reported by Brooks et al. (2018) (7) as receiving 3HP plus 50 mg dolutegravir daily. Solid points display the reported data values. Solid and dotted lines represent the 50th and 5th/95th percentiles, respectively, of 1,000 model-simulated concentrations. The AUC_0–24 and C_max estimates were reported by Brooks et al. (7). (B) Comparison of the 6 h INH concentrations reported by Lee et al. (2019) (15) among individuals receiving 3HP without dolutegravir. The mean (SD) body weight-adjusted (mg/kg) dose among N = 84 individuals reported by Lee et al. (15) was 14.0 (2.2) mg/kg, compared to 12.84 ± 2.3 mg/kg among the N = 30 participants in DOLPHIN. The NAT2 acetylator status was not reported by Lee et al. (15). SA, slow acetylator; IA, intermediate acetylator; RA, rapid acetylator.

FIG 4.

Comparison of the dose-adjusted and weight-adjusted estimates of AUC_0–24 and C_extmax between DOLPHIN and a meta-analysis by Hong et al. (2020) (8). The estimates are stratified by NAT2 acetylator type for both studies and disease status for Hong et al. (8). SA, slow acetylator; IA, intermediate acetylator; RA, rapid acetylator; HV, healthy volunteers.

Panel B of Fig. 3 compares the 6 h INH concentrations observed in DOLPHIN to those published by (15). We observe substantially greater concentrations in DOLPHIN. The median (IQR) INH concentrations at 6 h in DOLPHIN were 8.0 (6.6 to 9.0) mg/L, compared to 2.1 (1.1 to 3.5) mg/L in (15). The subjects included in (15) received weight-adjusted doses to a maximum of 900 mg, whereas all of the DOLPHIN participants received fixed doses of 900 mg. However, this is unlikely to account for the differences in concentrations, as the subjects in (15) received higher weight-adjusted doses, with an average of 14.0 (2.2) mg/kg, compared to 12.8 (2.3) mg/kg in DOLPHIN. The NAT2 status was not reported by (15), and it is possible that a greater proportion of RA may account for these differences.

Fig. 4 compares the AUC and C_max estimates from DOLPHIN to the meta-analysis estimates reported by (8). The mean AUC and C_max estimates are greater in DOLPHIN across all NAT2 categories and TB disease states. Within disease states, the AUC and C_max values from DOLPHIN were more similar to those of healthy volunteers, rather than those of individuals with TB disease or TB/HIV. Notably, the analysis by (8) excluded INH PK administered with meals, which have previously been shown to decrease C_max values by 42% (14). This suggests that greater differences between the C_max values may have been observed, had the DOLPHIN participants received INH in a fasted state.

DISCUSSION

We have reported a population PK model for INH developed from the participants of the DOLPHIN trial, during which INH was administered for LTBI as a part of the 3HP protocol alongside dolutegravir-based ART for the treatment of HIV. Two prior population PK models have been developed for INH during 3HP: Mathad et al. (16) studied 3HP administered to pregnant and postpartum women, and Lee et al. (2022) (17), who studied 3HP administered to an Asian population with LTBI. No previous population PK models of INH have been reported for 3HP plus dolutegravir. The only prior study to evaluate the combination of 3HP with dolutegravir reported elevated INH in two out of three patients, both of whom experienced adverse drug reactions (7).

Our study observed similarly elevated INH concentrations, with INH exposures that were consistent with or greater than those of the three patients receiving 3HP plus dART in (7). In contrast, we did not observe adverse drug reactions in DOLPHIN that were attributed to HP administration. 1 participant (an IA) had a grade 2 flu-like reaction following a HP dose that resolved within 24 h, though the INH concentrations were not collected for this participant (6). Compared to the data from the individuals receiving 3HP alone, we observed much greater differences in INH exposures. The median 6 h concentration in DOLPHIN was approximately four times the median concentration of previously reported data. Further, all of the 6 h concentrations from DOLPHIN fell in the upper quartile of the values reported by (15). After adjusting for dosage and total body weight, the AUC and C_max estimates were also greater than those observed when INH was administered either alone or in combination with rifampicin, pyrazinamide, or ethambutol.

Collectively, these results suggest the possibility for a drug-drug interaction between isoniazid and dolutegravir, though the basis for such an interaction is not clear. Dolutegravir is a potent inhibitor of the organic cation transporter (OCT)2, for which INH was found to be a substrate in a recent study (18). OCT2 is a membrane transporter that is primarily expressed in the kidneys and small intestine, and it is central to the transport of molecules from the blood to the urine (19, 20). A portion of INH is secreted unchanged in the urine, raising the possibility that the inhibition of renal transport via OCT2 would result in higher plasma concentrations (21). This has been previously observed with the anti-diabetic drug metformin, which experiences dose-dependent increases in plasma concentrations when coadministered with dolutegravir in a process that is likely mediated, in part, by OCT2 (22). However, unlike metformin, INH does not depend primarily on renal elimination; instead, it is primarily transformed in the liver and intestines into N-acetylisoniazid by NAT2 (21).

Interestingly, reductions in renal INH clearance have been linked to reduced hepatic clearance via diminished NAT2 activity. The specific mechanism by which this occurs is unclear; however, there is strong evidence that renal impairment can influence the hepatic biotransformation of nonrenally cleared drugs (23). Specific to INH, a study (24) observed significantly greater AUC values and the decreased clearance of INH in individuals with chronic renal failure, relative to healthy volunteers. Further, several of the subjects with chronic renal failure underwent kidney transplantation, after which they showed significant increases in nonrenal clearance. A later study (25) in rats observed that chronic renal failure resulted in a 50% reduction in NAT2 activity. Finally, in a study of patients receiving 3HP for LTBI (15), it was observed that reduced renal function was associated with higher INH concentrations, acetyl-isoniazid concentrations, and systemic drug reactions. Lee et al. (2022) (15) further observed an association between creatinine clearance and the clearance of acetyl-isoniazid. If the renal clearance of DOLPHIN participants was reduced by the inhibition of OCT2 by DTG, the participants may have also experienced diminished NAT2 activity, which could have caused the slow clearance rates and high plasma concentrations that were observed in our study.

In addition to high INH exposures, we observed associations between total body weight and oral bioavailability and mean transit time parameters, with lower-weight individuals tending to have a greater bioavailability and more rapid absorption. These associations have not been previously reported but are biologically plausible. Gastrointestinal physiology is known to differ substantially between individuals and can have a substantial impact on the absorption and bioavailability of orally administered drugs (26). Further, obesity has been observed to have a profound impact on gastrointestinal physiology (27), implying a possible link between weight and absorption. INH doses were also administered with meals, which may have influenced absorption processes. The coadministration of INH with food has been reported to increase the time to maximum concentration, as was observed in our study, though it has also been reported to decrease the maximum concentrations (14). The clearance rates were consistent with published estimates (11) for slow and intermediate acetylators. Few RA were present in our sample; however, those that were present had clearance rates that were comparable to those of the IA.

The clinical implications of high INH exposures during weekly administration are unclear. The primary adverse drug reactions (ADR) of INH therapy include peripheral neurotoxicity and hepatotoxicity. Peripheral neurotoxicity is believed to develop due to the competitive inhibition of pyridoxine (vitamin B₆) by INH, and it is associated with high INH exposures and slow NAT2 acetylation (28). Peripheral neuropathy is typically preventable with pyridoxine supplementation. However, in practice, pyridoxine is rarely given to patients receiving INH-containing TB prophylactic therapy. The connection between INH exposures and hepatotoxicity is more tenuous and remains debated (29). INH therapy can cause hepatotoxicity due to an accumulation of hepatotoxic metabolites, namely, hydrazine and acetylhydrazine. Evidence suggests that this accumulation and the corresponding risk of hepatic injury are more pronounced for individuals with reduced NAT2 acetylation (21, 30). Slow NAT2 acetylators typically experience greater INH exposures, and it has been proposed that an AUC_0–24 value of greater than 55 mg*h/L may be used as a predictor of hepatotoxicity during daily INH therapy (31).

Few data describe how the risks of INH-induced ADR vary with respect to cumulative exposures, and guidance regarding safe and effective target exposures during weekly INH is sparse. A published study (13) has recommended targeting peak concentrations of 9 to 15 mg/L during the biweekly administration of 900 mg INH, although it does not describe the basis for this recommendation. Several clinical studies have compared 3HP to 9 months of daily isoniazid (9H) and have found comparable rates of ADR in adults (32), children (33), and persons with HIV (34). However, they did not compare INH PK. Further comparisons have noted fewer instances of hepatoxicity but more instances of a flu-like syndrome in 3HP than in 9H (35, 36). ADR during 3HP have been linked to greater 24-hour concentrations and NAT2 mutations (15). Recent work (17) has revealed an association between the INH C_max, but not the AUC, and the ADR and particularly flu-like symptoms among individuals receiving 3HP for LTBI. Similar associations with ADR were not observed for the AUC or C_max of acetyl-isoniazid, rifapentine, or 25-desacetyl-rifapentine. Further work is needed to determine the relationship between INH exposures and ADR and to establish safe and effective target exposures during weekly INH therapy.

Limitations.

The primary limitation of this analysis is that the data do not allow us to identify a cause for the evidently elevated INH concentrations. All of the INH doses were coadministered with dolutegravir and rifapentine, as well as with meals. Therefore, we were unable to assess the impact of each of these factors. Additional limitations included a relatively small sample size and the absence of INH metabolite concentrations. Together, these would have allowed for a more detailed model and would have indicated whether specific routes of elimination were impacted, relative to prior data. Finally, INH was measured on only one occasion, and we could not observe whether the observed increases in INH exposure were time-dependent.

Conclusion.

We observed unexpectedly high exposures of INH when it was administered as a part of weekly high-dose INH and rifapentine. This may have implications for the treatment of LTBI. Further study is needed to discern the mechanism of these high exposures, to examine the possibility of a drug-drug interaction with INH that is mediated by cellular transport, and establish guidelines for INH exposures during weekly therapy.

MATERIALS AND METHODS

Study design.

The DOLPHIN trial (ClinicalTrials.gov ID: NCT03435146; South African National Clinical Trial Registration: DOH-27-1217-5770) was a phase 1/2 single-arm trial that was designed to assess the PK and safety of the coadministration of dART with 3HP for the treatment of LTBI. The primary analysis of DOLPHIN characterized the dolutegravir PK and determined that individuals could safely receive dART and 3HP and maintain HIV virologic suppression without dose adjustments. These results have been previously reported, as have details describing the drug administration and PK sampling of dolutegravir and rifapentine (6). DOLPHIN enrolled HIV positive adults with undetectable viral loads. At the time of enrollment, participants were transitioned from 600 mg efavirenz daily to a once-daily, fixed-dose tablet of 50 mg dolutegravir, 300 mg tenofovir disoproxil fumarate, and 200 mg emtricitabine. Weekly HP was initiated during the ninth week of dART treatment at doses of 900 mg INH and 900 mg rifapentine to allow for the complete washout of efavirenz. The INH and rifapentine doses were formulated in 300 mg and 150 mg single tablets, respectively, and they were administered with a standardized meal of cereal with milk, bacon, a fried egg, and a sandwich or toast option with butter. Semi-intensive INH PK were collected from the N = 30 participants on day 72 at 1, 2, 6, and 10 h postdose for a total of 120 INH measurements.

The INH concentrations were measured via liquid chromatography-tandem mass spectrometry by the Clinical Pharmacology Analytical Laboratory at the Johns Hopkins University School of Medicine. INH was isolated from 0.025 mL plasma via protein precipitation and was analyzed using an API 4000 mass analyzer (SCIEX, Redwood City, CA, USA) interfaced with a Waters ACQUITY UPLC system (Waters Corporation, Milford, MA, USA). The analytical run time was was 2.5 min. The primary linearity of the assay was 50.0 to 10,000 ng/mL. Samples above 10,000 ng/mL were diluted to a final result. The assay was validated in accordance with the US Food and Drug Administration, Guidance for Industry, Bioanalytical Method Validation recommendations. The interassay accuracy and precision across the analytical measuring range of the assay ranged from −4.46% to 4.35% and from 6.54% to 11.5%, respectively. Stability challenges in the whole blood (6 h) and plasma (24 h) were acceptable, and negligible matrix effects were observed during the assay validation. The DOLPHIN trial was approved by the Johns Hopkins Medicine Institutional Review Board (reference IRB00143711), the University of the Witwatersrand Human Research Ethics committee (reference number 170705), and the South African Health Products Regulatory Authority (reference number 20170811). Participants provided written informed consent.

NAT2 classification.

NAT2 acetylator status was determined using a four-single nucleotide polymorphism (SNP) genotyping panel of 191G > A (rs1801279), 341T > C (rs1801280), 590G > A (rs1799930), and 857G > A (rs1799931). The NAT2 genotypes were classified into “rapid,” “intermediate,” and “slow” phenotypes, as previously described (37). Participants were classified as “rapid acetylators” (RA) if they had no variant alleles, “intermediate acetylators” (IA) if they were heterozygous for one variant allele, and “slow acetylators” (SA) if they were either homozygous for one or more variant alleles or heterozygous for two or more alleles. This categorization resulted in 10 (33%) SA, 16 (53%) IA, and 4 (13%) RA. These proportions are similar to the proportions reported by (9) in N = 172 South African adults with TB: 34%, 43%, and 18% for SA, IA, and RA, respectively. When models differentiated between only two levels of acetylation status (slow versus rapid), IA and RA were grouped together as rapid acetylators. The baseline participant characteristics and the summaries of INH concentrations are reported in Table 2. The NAT2 genotypes and associated acetylation phenotypes are reported in Table 3.

TABLE 2.

Baseline covariates and isoniazid concentrations for N = 30 DOLPHIN trial participants

Characteristic	No. (%) or median (Min–Max)
Baseline covariates
Female	17 (56.7%)
Age (yrs)	43 (23, 57)
Wt (kg)	72.8 (53, 101.6)
Ht (cm)	1.66 (1.36, 1.77)
Fat-free mass	49.06 (34.15, 59.03)
CD4 count (cells/mm3)	637 (259, 998)
INH concn (mg/L)
1 hour	3.36 (0.08, 14.96)
2 hour	15.45 (2.12, 28.8)
6 hour	7.98 (3.68, 20.27)
10 hour	2.46 (0.75, 7.34)

TABLE 3.

NAT2 genotypes and the associated acetylation phenotype of the DOLPHIN trial participants

Genotypea	191G > A	341T > C	590G > A	857G > A	No. (%)
Rapid acetylators
NAT2*4/*4	GG	TT	GG	GG	4 (13.3%)
Intermediate acetylators
NAT2*4/*5D	GG	CT	GG	GG	10 (33.3%)
NAT2*4/*6B	GG	TT	AG	GG	4 (13.3%)
NAT2*4/*14A	AG	TT	GG	GG	2 (6.7%)
Slow acetylators
NAT2*4/*5E	GG	CT	AG	GG	3 (10%)
NAT2*4/*5S	GG	CT	GG	AG	1 (3.3%)
NAT2*4/*14D	AG	TT	AG	GG	3 (10%)
NAT2*6B/*6B	GG	TT	AA	GG	2 (6.7%)
NAT2*14A/*14A	AA	TT	GG	GG	1 (3.3%)

^aGenotypes heterozygous for zero or one variant allele were classified as rapid acetylators and intermediate acetylators, respectively. Genotypes either homozygous for one or more variant allele or heterozygous for two or more alleles were classified as slow acetylators.

Pharmacokinetic data analysis.

Population PK models were fit using the nonlinear mixed-effects modeling software NONMEM, version 7.5. All of the models were fit using the first-order conditional estimation procedure with eta-epsilon interaction. Posterior predictive simulations were conducted using Perl-Speaks-NONMEM, version 5.2.6. R (version 4.0.4, https://www.r-project.org/) was used for all of the data preparation (puzzle package, version 0.0.1), data analysis, numerical and graphical summaries, and manuscript preparation.

Model construction began with the identification of an appropriate structural model. We considered one-compartment and two-compartment disposition models with a transit compartment absorption process. NAT2 was automatically entered into the structural model as a predictor of the central compartment clearance. The clearance and volume parameters were allometrically scaled by the total body weight or fat-free mass and were normalized by the median values of 73 kg and 49 kg, respectively. The clearance parameters were scaled using an exponent of 0.75, whereas the volume parameters were scaled with an exponent of 1, as previously recommended (38). The residual error models that were considered were either additive plus proportional to the response or additive to the response on the natural log scale.

Following the development of the structural model, we sequentially tested the associations of sex, CD4 count, and prior TB infection with the model parameters. Sex and CD4 count were considered due to prior associations with PK parameters in existing models (10, 11), whereas prior TB infection was included to explore the possibility that the INH PK may have been impacted by prior disease status. The covariates were related to the model PK parameters: P, by the function $P={\theta }_{\mathit{TV}}\times (1+{\theta }_{\mathit{eff}}\times \mathit{CO}{V}_i)$, in which θ_TV is the typical value within the population and ${\theta }_{\mathit{eff}}\times 100$ is the percent change in θ_TV that is associated with individual i, having covariate value COV_i. Covariate effects were retained in the model if physiologically plausible and were regarded as statistically significant by a drop in the objective function value (OFV) at P = 0.05. The OFV was assumed to follow a ${\chi }^2$ distribution for the comparisons of nested models, which required an OFV decrease of 3.84 for a one degree of freedom test.

Model performance was assessed using simulation-based diagnostics. Both the structural and final model fits were assessed by visual predictive checks from 1,000 data sets that were simulated according to the DOLPHIN study design. The final model performance was further evaluated through normalized prediction distribution errors on observations with simulation (NPDEYS), the calculation of which is described in (39). Briefly, the means and standard deviations of simulated values are calculated for each unique person-time concentration. Both the observed and simulated values are standardized by subtracting the mean and dividing by the standard deviation. The NPDEYS are calculated by evaluating the inverse normal cumulative density function at the proportion of simulated values below each observed point. Under the null hypothesis that the observed data are consistent with the model predictions, the NPDEYS will follow a standard normal distribution. The hypotheses that the median and variance of the NPDEYS were equal to zero and one, respectively, were assessed via Wilcoxon signed-rank and Fisher’s tests, respectively. The normality of the residuals was assessed via a Shapiro-Wilk test. The calculation of the NPDEYS is described in the Appendix.

Comparisons to previously published data.

Given the limited data reporting high-dose INH PK and the prior report (7) of elevated INH concentrations, we sought reference values from the literature to describe 3HP plus dART and 3HP alone. The data for 3HP plus dART were collected by digitizing the INH concentrations reported by (7) for three healthy volunteers, named Subjects 1, 2, and 4, at 2, 3, 4, 5, 6, 8, 10, and 24 h postdose. Subjects 1 and 4 were observed to have elevated INH concentrations in addition to adverse drug reactions. The data for 3HP alone were provided by (15) in the form of 6-hour and 24- hour INH concentrations from N = 84 individuals for the treatment of LTBI. The concentrations from both studies were digitized using WebPlotDigitizer (https://apps.automeris.io/wpd/) to facilitate their comparisons to the model predictions and data from DOLPHIN. We could not reliably digitize the 24-hour INH concentrations reported by (15). Consequently, these were not included for comparison. We were unable to identify any previous estimates of the INH AUC_0–24 and C_max values during 3HP. Instead, we compared the DOLPHIN AUC_0–24 and C_max values to the weight-adjusted and dose-adjusted estimates published by (8), who conducted a meta-analysis of 90 studies that reported INH PK. In the constituent studies, INH was administered alongside a combination of rifampicin, pyrazinamide, and ethambutol, though it was never administered with rifapentine or dolutegravir.