Skip to main content
American Journal of Respiratory and Critical Care Medicine logoLink to American Journal of Respiratory and Critical Care Medicine
. 2021 Mar 30;204(11):1327–1335. doi: 10.1164/rccm.202103-0534OC

A Semimechanistic Model of the Bactericidal Activity of High-Dose Isoniazid against Multidrug-Resistant Tuberculosis: Results from a Randomized Clinical Trial

Kamunkhwala Gausi 1, Elisa H Ignatius 2, Xin Sun 3, Soyeon Kim 4, Laura Moran 5, Lubbe Wiesner 1, Florian von Groote-Bidlingmaier 6, Richard Hafner 7, Kathleen Donahue 8, Naadira Vanker 6, Susan L Rosenkranz 3, Susan Swindells 9, Andreas H Diacon 6, Eric L Nuermberger 2, Kelly E Dooley 2, Paolo Denti, on behalf of the A5312 Study Team1,
PMCID: PMC8786077  PMID: 34403326

Abstract

Rationale

There is accumulating evidence that higher-than-standard doses of isoniazid are effective against low-to-intermediate–level isoniazid-resistant strains of Mycobacterium tuberculosis, but the optimal dose remains unknown.

Objectives

To characterize the association between isoniazid pharmacokinetics (standard or high dose) and early bactericidal activity against M. tuberculosis (drug sensitive and inhA mutated) and N-acetyltransferase 2 status.

Methods

ACTG (AIDS Clinical Trial Group) A5312/INHindsight is a 7-day early bactericidal activity study with isoniazid at a normal dose (5 mg/kg) for patients with drug-sensitive bacteria and 5, 10, and 15 mg/kg doses for patients with inhA mutants. Participants with pulmonary tuberculosis received daily isoniazid monotherapy and collected sputum daily. Colony-forming units (cfu) on solid culture and time to positivity in liquid culture were jointly analyzed using nonlinear mixed-effects modeling.

Measurements and Main Results

Fifty-nine adults were included in this analysis. A decline in sputum cfu was described by a one-compartment model, whereas an exponential bacterial growth model was used to interpret time-to-positivity data. The model found that bacterial kill is modulated by isoniazid concentration using an effect compartment and a sigmoidal Emax relationship (a model linking the drug concentration to the observed effect). The model predicted lower potency but similar maximum kill of isoniazid against inhA-mutated compared with drug-sensitive isolates. Based on simulations from the pharmacokinetics-pharmacodynamics model, to achieve a drop in bacterial load comparable to 5 mg/kg against drug-sensitive tuberculosis, 10- and 15-mg/kg doses are necessary against inhA-mutated isolates in slow and intermediate N-acetyltransferase 2 acetylators, respectively. Fast acetylators underperformed even at 15 mg/kg.

Conclusions

Dosing of isoniazid based on N-acetyltransferase 2 acetylator status may help patients attain effective exposures against inhA-mutated isolates.

Clinical trial registered with www.clinicaltrials.gov (NCT01936831).

Keywords: tuberculosis, isoniazid resistance, early bactericidal activity, inhA mutation, phase 2 clinical trial


At a Glance Commentary

Scientific Knowledge on the Subject

Isoniazid is a first-line drug for tuberculosis treatment. Evidence is accumulating that higher-than-standard doses of isoniazid are effective in patients infected with low- to intermediate-level isoniazid-resistant strains of Mycobacterium tuberculosis. The optimal “high dose” in patients with slow, intermediate, or fast N-acetyltransferase 2 acetylator status is unknown.

What This Study Adds to the Field

We present a model to characterize the association between acetylator status, isoniazid pharmacokinetics, and early bactericidal activity against wild-type and mutated strains. We find that dosing of isoniazid based on bacterial inhA mutations and acetylator status is essential to attain desired isoniazid exposures.

Antimicrobial resistance is a grave threat to global public health, responsible for approximately 700,000 deaths yearly (1). Individuals with drug-resistant tuberculosis (TB) receive less active, poorly tolerated, or hard to access drugs. Detection of resistance is sporadic given that diagnostic platforms and assay cutoffs are not well established or available for many second-line drugs, but we have rapid tests to detect resistance to rifampicin and isoniazid. The primary mechanisms of isoniazid resistance are 1) prevention of activation of the prodrug isoniazid owing to katG mutation and 2) mutation of the promoter region of inhA, leading to overexpression of the isoniazid target InhA (2). InhA mutation tends to result in low-to-intermediate–level phenotypic resistance (3), whereas katG mutation results in high-level resistance that may or may not be overcome with safe isoniazid doses.

There is an urgent need to optimize existing agents until new ones are widely available to overcome the threat of drug resistance (4). Improved treatment, including dose customization for individual patients or patient subgroups, can be driven by exposure–response relationships, a pharmacologic framework that provides an evidence-based method of dose selection. There is accumulating evidence that higher-than-standard doses of isoniazid are effective against low- to intermediate-level isoniazid-resistant strains, but the optimal “high dose” is unknown, making isoniazid a prime candidate for exposure–response modeling to assess its use against drug-resistant strains. For many drugs, there is high interindividual variability in drug exposures, and these variability sources are unknown and unpredictable. For isoniazid, however, the main determinant of drug exposures is N-acetyltransferase 2 (NAT2) status (exposure varies by two- to sevenfold between slow and rapid acetylators [5, 6]), and these genetic polymorphisms can be determined with available technology, albeit not at the point of care.

The INHindsight trial was designed to assess the effect of isoniazid dose escalation on its early bactericidal activity (EBA) against inhA-mutated isolates. We previously reported (7) measurable EBA against inhA-mutated isolates at high-dose isoniazid (mean EBAcfu0–7 of 0.17 and 0.22 log10cfu/ml/d, at 10 and 15 mg/kg, respectively, in the inhA group compared with 0.16 log10cfu/ml/d in the drug-sensitive group). TB biomarkers (colony-forming units [cfu]cfu and time to positivity [TTP]) and isoniazid exposure are characterized by large inter- and intraindividual variability, and the measurement of the biomarkers is very noisy (8). To adjust for factors that modulate exposure–response relationships and tease out signal from noise, model-based approaches that can jointly characterize the relationship between isoniazid dose, its exposure, and TB biomarkers are needed. Such an approach becomes even more useful when identifying subgroups that may need dose adjustments (9). Modeling can incorporate complex pharmacologic properties, such as delayed onset of action caused by isoniazid activation, and is capable of simulating different clinical scenarios. This study employs such methods to characterize the association between isoniazid pharmacokinetics (PK) (at high or standard doses) and its EBA against Mycobacterium tuberculosis (M.tb) (drug sensitive and inhA mutated).

Methods

Study Design and Participants

This analysis used isoniazid PK and EBA data collected in the INHindsight trial, a 7-day EBA study conducted in South Africa. It was approved by the local ethics committee and the South African Health Products Regulatory Authority. All participants gave written informed consent. Adults with sputum smear–positive pulmonary TB, treatment-naïve with an infecting strain shown to have inhA mutation or no mutations (control group), were eligible. Chest radiographs were performed to determine cavitary disease status. Further details of the study design were reported previously (7).

Study Procedures

In brief, participants infected with inhA-mutated M.tb were randomized to receive isoniazid doses of 5, 10, or 15 mg/kg, respectively, whereas those with drug-sensitive isolates were administered 5 mg/kg. All participants received isoniazid monotherapy for 7 days; sputum samples were collected daily from 2 days before initiation of isoniazid until Day 7 of treatment, and PK sampling occurred on Day 6. Sputum samples were cultured on solid media to capture cfucfu and in liquid media to capture TTP. Isoniazid minimum inhibitory concentration (MIC) was estimated as previously described (7). NAT2 genotypic information and PK samples were captured and analyzed as described in the Methods of the online supplement. NAT2 was assigned the phenotype fast, intermediate, or slow using the suggestion by Sabbagh and colleagues (10).

PK-Pharmacodynamic Modeling

Isoniazid PK-pharmacodynamics (PD) was modeled using NONMEM version 7.4.3 and ancillary software (11). Full details on model development are available in the Supplemental Methods.

In brief, several structural PK models were tested to describe isoniazid plasma concentrations. The effect of NAT2 genotype on clearance was tested, and allometric scaling was applied on all clearance and volume of distribution parameters (12) to account for body size. M.tb bacterial load in patients was characterized using a compartmental model and a first-order kill rate. Both cfucfu and TTP were used jointly to quantify bacterial load: 1) cfucfu was directly obtained as an observation from the bacterial load compartment, whereas 2) TTP was obtained by modeling the growth of M.tb in Mycobacterial Growth Indicator Tubes (MGIT) using the bacterial load in the patient (cfucfu on solid culture media) as the initial amount of bacteria transferred into the tube, which then grows and is detected after crossing a threshold. To link the PK and PD, the effect of isoniazid exposure was tested on the kill rate using overall area under the concentration–time curve (constant kill), instantaneous isoniazid concentration in plasma, or isoniazid effective concentration in a hypothetical effect compartment (13).

Model development was guided using graphical assessment of the goodness-of-fit plots, likelihood ratio tests (using supervised stepwise approach with forward inclusion [P < 0.05]), and backward elimination (P < 0.01), as described by Mould and Upton (14). Simulation-based diagnostics including visual predictive checks were used for model validation. The PK model was developed first, and then individual PK parameter estimates from the PK model were fixed in the PK-PD model. Monte Carlo simulations were performed with the final model to predict change in cfu from baseline for the drug-sensitive versus inhA-mutated isolates under the three dose levels.

Results

Enrolment and Baseline Characteristics

Among 59 adults included, 16 were control subjects with drug-sensitive M.tb; among the 43 participants with inhA-mutated M.tb, 13, 14, and 16 were randomized to the 5-, 10-, and 15-mg/kg arms. The participants had a median weight of 51 kg, 20% were HIV-positive, and 88% had cavitary disease (Table 1). PK samples were captured in 58 participants, as 1 participant withdrew consent for family reasons on Day 5, providing a total of 517 PK observations. Sputum bacillary load as measured by cfu and TTP was captured in all 59 participants, providing 449 and 524 observations for cfu and TTP, respectively. cfu missingness (15.4%) was due to contamination, whereas TTP missingness (1.3%) was due to leakage. Among the 55 participants who consented to genetic sample collection, intermediate NAT2 acetylators were predominant (54%) and rapid the least prevalent (14%). MIC data were available for 27 participants and ranged from 0.2 to 1 mg/L in the drug-sensitive arm and 0.2 to 4 mg/L in the inhA-mutated arms; the reason for missing MIC is provided in the Supplemental Methods.

Table 1.

Demographic, Clinical, and Laboratory Characteristics of Participants Stratified by the Four Arms

Characteristics Drug Sensitive inhA Mutated Total
Dosing, mg/kg 5 (n = 16) 5 (n = 13) 10 (n = 14) 15 (n = 16) (N = 59)
Sex, M/F 12/4 10/3 11/3 10/6 43/16
Race*          
 Black 6 (38%) 6 (46%) 4 (29%) 5 (31%) 21 (36%)
 Colored/mixed 9 (56%) 7 (54%) 10 (71%) 11 (69%) 37 (63%)
 White 1 (6%) 0 (0%) 0 (0%) 0 (0%) 1 (2%)
Age, yr 30 (19–58) 31 (20–58) 32 (20–58) 34 (18–55) 32 (18–58)
Body mass index, kg/m2 18.0 (15.2–29.7) 18.1 (16.2–30.9) 17.7 (14.6–23.6) 18.5 (15.4–30.1) 18.1 (16.6–19.4)
HIV status, negative/positive 13/3 11/2 11/3 12/4 47/12
Cavities, yes/no 12/4 11/2 14/0 15/1 52/7
NAT2 acetylation status          
 Rapid 4 (25%) 2 (15%) 0 (0%) 2 (13%) 8 (14%)
 Intermediate 10 (63%) 8 (62%) 6 (43%) 8 (50%) 32 (54%)
 Slow 2 (12%) 1 (8%) 7 (50%) 5 (31%) 15 (25%)
 Missing 0 (0%) 2 (15%) 1 (7%) 1 (6%) 4 (7%)

Definition of abbreviation: NAT2 = N-acetyltransferase 2.

Data are presented as median and range, n, or n (%).

*

Percentages do not always add to 100% because of rounding.

PK-PD Model

The schematic of the final PK-PD model structure is displayed in Figure 1, and a visual predictive check of the PD and PK model are shown in Figure 2 and Figure E2 in the online supplement, respectively, indicating adequate model fit. The PK of isoniazid was best described by a two-compartment model with transit compartment absorption and first-order elimination using a well-stirred liver model. We included the effects of allometric scaling using fat-free mass on disposition parameters and NAT2 genotype on clearance (see Supplemental Results for additional details and parameter values).

Figure 1.


Figure 1.

Schematic representation of the PK-PD model used to describe the PK of isoniazid, the drug-induced decline in bacteria load, and the growth of M.tb in MGIT. cfu = colony-forming units; MGIT = Mycobacterial Growth Indicator Tubes; M.tb = Mycobacterium tuberculosis; MTT = absorption mean transit time; NN = number of absorption transit compartment; PD = pharmacodynamics; PK = pharmacokinetics; TTP = time to positivity.

Figure 2.


Figure 2.

Visual predictive check of the pharmacokinetics-pharmacodynamics model of the cfu and TTP, stratified by arm. The solid and dashed lines are the 10th, 50th, and 90th percentiles of the observations, whereas the shaded areas represent the 95% model-predicted confidence intervals for the same percentiles. The two samples collected before treatment (baseline) have been collapsed together as Day 0. cfu = colony-forming units; TTP = time to positivity.

The decline of bacterial load in patients was best described using a one-compartment model with first-order killing of M.tb modulated by the isoniazid effect. The baseline bacterial load was lower in patients with drug-sensitive TB than those with drug-resistant TB. Among patients with inhA-mutated isolates, a significant difference in baseline bacterial load was also found in participants administered a 5 mg/kg dose; therefore, three baseline cfu parameters were estimated. The model for TTP used exponential growth in the MGIT, as a logistic growth model did not substantially improve the fit and made the model unstable. A delay of about 6 hours in the onset of the growth was observed for sputum samples collected after initiation of isoniazid treatment compared with baseline (P ≪ 0.001), and inhA-mutated isolates were found to grow with a doubling time 4 hours slower than drug-sensitive isolates (P ≪ 0.001).

The bactericidal effect of isoniazid was best described by using an Emax function (an equation linking the drug concentration to the observed effect) driven by “effective” concentration in a hypothetical effect compartment as opposed to the concentration in plasma (P ≪ 0.001). Figure 3A depicts this relationship, and Figure 3B shows the model-predicted profiles of “effective” concentration for a typical participant with intermediate NAT2 acetylator status administered the three isoniazid dose levels. The inhA-mutated isolates had a significantly higher EC50 compared with the drug-sensitive isolates (P ≪ 0.001) as shown in Table 2, reflecting the lower potency of isoniazid against this strain. For the drug-sensitive M.tb arm (5 mg/kg), the bactericidal activity was predicted at or near the maximal kill (Killmax ) throughout the treatment; thus, the model estimated extremely low but unstable values of concentration associated with 80% of maximal kill (EC50 ), which were then fixed to 0.01 mg/L. A sensitivity analysis showed that using this or lower values had no substantial effect on the model results. Testing for different maximal bactericidal effects of isoniazid (Killmax ) between inhA-mutated and drug-sensitive M.tb did not produce a significant difference (P = 0.30), so the same Killmax value was used in both groups. Of note, although we observed a weak positive relationship between higher MIC values and an increase of EC50 , this effect did not reach statistical significance (P = 0.035) and was not included in the final model. It should be noted that MIC was missing in 32 participants (54%).

Figure 3.


Figure 3.

(A) The pharmacokinetics-pharmacodynamics relationship of isoniazid against Mycobacterium tuberculosis (M.tb). The green and the red profiles represent the pharmacokinetics-pharmacodynamics relationship of the drug-sensitive and inhA-mutated M.tb, respectively. The red and green dashed lines represent the concentration associated with 80% of maximal kill (EC80 ) of the inhA-mutated and drug-sensitive M.tb, respectively. (B) Isoniazid concentration profiles in the effect compartment of a typical individual with intermediate N-acetyltransferase 2 acetylator status dosed with 5, 10, and 15 mg/kg doses. The red and green lines represent the EC80 of the inhA-mutated M.tb and drug-sensitive M.tb, respectively.

Table 2.

Final Model Population Parameter Estimates

Parameter Typical Value (95% CI*) Variability (95% CI*)
Baseline cfu for drug sensitive, log10cfu 5.75 (5.19 to 6.09) 0.892 (0.543 to 1.20)
Baseline cfu for inhA, 5 mg/kg, log10cfu 5.87 (5.18 to 6.36)
Baseline cfu for inhA, 10–15 mg/kg, log10cfu 6.63 (6.27 to 7.01)
Additive error in cfu, log10cfu 0.493 (0.42 to 0.565)
Growth rate in MGIT drug sensitive, kgrowth, 1/d 1.65 (1.45 to 1.97)
Growth rate in MGIT for inhA, kgrowth, 1/d 1.19 (1.04 to 1.42)
Delay of growth in MGIT after INH treatment, h 5.85 (0.971 to 11.2) 14.2 (10.4 to 17.3)
MGIT detection threshold, log10cfu 9.31 (8.93 to 9.78)
Proportional error in TTP, % 13.3 (11.1 to 14.9)
EC50 for drug sensitive, mg/L 0.01 fixed
EC50 for inhA, mg/L 0.544 (0.125 to 3.22)
Gamma, ϒ (·) 2.88 (1.04 to 7.93)
Killmax, 1/d 0.42 (0.355 to 0.907) 62.0§ (52.0 to 66.0)
T1/2E0, half-life of effect compartment delay, h 17.0 (3.94 to 31.3)
Correlation of the cfu–TTP errors, % −36.7 (−52.5 to −22.4)

Definition of abbreviations: cfu = colony-forming units; CI = confidence interval; EC50 = 50% concentration at which 50% of the maximum induced kill by isoniazid is achieved; INH = isoniazid; MGIT = Mycobacterial Growth Indicator Tubes; TTP = time to positivity.

The ε-shrinkage for cfu = 8% and TTP = 17%.

Coefficient of variation was calculated by (eω21), where ω2 represent the variance.

*

Estimated from nonparametric bootstrap of the final model (n = 200, stratified on arm).

The between-subject variability of the parameter estimate expressed as SD (in the log10 scale).

The between-occasion variability of the parameter expressed as approximate coefficient of variation (CV%).

§

The between-subject variability of the parameter expressed as approximate coefficient of variation (CV%).

The effect compartment introduced a “delay” with a half-life of around 17 hours before isoniazid concentration in plasma is active against M.tb. This delay also causes a certain degree of “accumulation” of the effective drug concentration over the first days of treatment, as illustrated in Figure 3B. Although this has little to no effect on the kill of drug-sensitive M.tb (for which the effective isoniazid concentrations were above the EC80 throughout treatment), this delay matters for inhA-mutated M.tb: for participants in the 15- and 10-mg/kg arms, the effective concentration profile only achieved stable values above the EC80 (and thus near maximal kill) after 24 and 48 hours, respectively. Of note, the typical effective concentration profile for the 5-mg/kg arm was below the EC80 threshold throughout the observed period. This reflects the poor bactericidal effect of the 5-mg/kg dose, as elucidated in the simulation below.

Dosing Simulation

The simulated effect of dose escalation on the drop of log10cfu from baseline for a typical individual (51-kg body weight and 44-kg fat-free mass) is illustrated in Figure 4. For drug-sensitive M.tb, no apparent difference in cfu decline was detected among the three NAT2 genotypes, and on Day 7, a drop of 1.3 log10cfu (0.186 log10cfu drop daily) was predicted. For inhA-resistant M.tb, on the other hand, NAT2 acetylator status made a difference in terms of bacterial kill: slow NAT2 acetylators achieve cfu drops comparable with the drug-sensitive arm at 10 mg/kg, intermediate NAT2 acetylators need 15 mg/kg, and fast NAT2 acetylators do not achieve the cfu drop even with 15 mg/kg, the highest dose tested.

Figure 4.


Figure 4.

Boxplot of the simulated drop in log10cfu, stratified by arm, across the 7 days the participants were on isoniazid monotherapy for a typical individual weighing 51 kg with a fat-free mass of 44 kg. Orange, green, and blue boxplots represent fast, intermediate, and slow NAT2 acetylation status. The dashed red line represents the median drop in log10cfu for the drug sensitive (the reference group) after 7 days of isoniazid monotherapy. cfu = colony-forming units; FAT = fat mass; FFM = fat-free mass; NAT2 = N-acetyltransferase 2; WT = weight.

Discussion

Our study adds to the accumulating evidence on the usefulness of high-dose isoniazid in multidrug-resistant (MDR)-TB treatment (3, 1517), providing additional insight into its activity against inhA-mutated isolates with low-level isoniazid resistance (2). The model predicts that against inhA-mutated isolates, a dose of 10 mg/kg for slow, 15 mg/kg for intermediate, and >15 mg/kg for fast NAT2 acetylators produce EBA comparable to that observed in the drug-sensitive group administered the standard dose of 5 mg/kg. The World Health Organization (WHO) recommends 10 mg/kg against MDR-TB (18), which our results suggest might be suboptimal against inhA-mutated isolates for the rapid and intermediate acetylators.

We found that isoniazid’s bactericidal action against inhA-mutated isolates was slightly delayed compared with plasma levels, which was accounted for using an effect compartment; this delay agrees with a previous descriptive analysis of these data (7). Some mechanistic reasons can be provided for this delay; namely, the concentration of isoniazid must first accumulate to a particular threshold in the effect compartment before it is effective against the inhA-mutated isolates, leading to a delay in the killing of the M.tb. The effect compartment model accounts for any delay that might lead to a lag in observing the dose–response dynamic. Isoniazid has numerous potential sources of delay (13). First, as M.tb bacilli are mostly in the lung tissue and lesions, the drug must first distribute from plasma to this target site before a response is observed; this would lead to a delay in drug response across all isolates. Second, overexpression of isoniazid target by inhA promoter mutations may lengthen the time required for isoniazid–nicotinamide adenine dinucleotide (NAD) adducts to saturate available inhA targets to a sufficient concentration for the effect to be observed. Even though the residence time of isoniazid in the lung tissue and lesions is short (19, 20), the residence time of the activated isoniazid–NAD adduct (isoniazid prodrug) at the target site on inhA is relatively long (21, 22). Third, inhA mutation causes loss in microbial fitness (23), which might lead to mutated isolates replicating more slowly; therefore, the killing effect of isoniazid, which is presumably growth rate dependent (24), may be slower in onset.

The model predicted an EC80 of 0.75 mg/L, which falls within the range of inhA-mutated MIC of 0.2–1 mg/L (3). A trend was observed with increasing EC50 as MIC increased, but the effect did not reach statistical significance and was not included in the final model. The exclusion of MIC from the final model was also driven by the large variability associated with the determination of MIC (25), the fact that observed MIC values depend on the dilution series and are actually interval censored data, and the lack of MIC data for approximately 54% of the participants. Finally, MIC results can take days or longer to obtain, and very few laboratories have the capacity to perform this test, which ultimately limits the real-world utility of including MIC in a model offering clinical dosing strategies.

A delay in the onset of the bacterial growth in MGIT was observed after the bacteria were exposed to isoniazid. This is possibly because of the postantibiotic effect associated with isoniazid (26, 27) or the fact that, to better survive drug exposure, bacteria may become metabolically dormant and take a longer time to start growing in MGIT (28). This is also in agreement with results by Bowness and colleagues (29), which showed that two samples with the same cfu had shorter TTP when captured at an earlier versus a later occasion during treatment. cfu counts are primarily dependent on the number of viable bacteria in a sample, whereas TTP measurements are dependent on both the number of viable bacteria and how fast they grow. M.tb exposed to increasing sizes and/or numbers of isoniazid pulses in vitro show successively longer postantibiotic effects (and presumably greater disconnect between cfu and TTP). This might also relate to the slower apparent onset of isoniazid activity against inhA mutants (especially in MGIT) since they “see” smaller pulses than drug-sensitive-M.tb and require cumulative effects of multiple small pulses (e.g., more saturation of drug target), as suggested by Awaness and colleagues (26), to exert the same killing and postantibiotic effect.

The estimated growth rate in our model is in line with what is expected for M.tb in vitro and what have been previously published (30). A difference between drug-sensitive and inhA-mutated M.tb growth rate in MGIT was observed. This might have been brought about by the loss in microbial fitness (23), resulting in the inhA-mutated isolates growing at a slower rate compared with the drug-sensitive isolates. This finding has been extensively explored by le Roux and colleagues (31).

During the model development process, several descriptors of effective concentration were tested, including area under the concentration–time curve; however, the model favored instantaneous concentration in the effect compartment as the main driver. Apart from delay in the onset of drug action, this might be due to the fact that the majority of the active drug (isoniazid–NAD adduct) remains in the mycobacterial cell and is not able to cross the cell envelope (32). Therefore, the amount of bound target does not decline in parallel with the plasma concentration of isoniazid, the prodrug. Thus, the time required for the prodrug to reach the threshold required for bacterial kill can be reduced with a higher peak concentration (22). Given this reliance on peak concentration, fluctuating isoniazid concentration in the effect compartment drives bacterial kill and may explain the model preference.

Individuals respond very differently to treatment depending on weight, age, genetic factors, administration conditions (fasting vs. fed), and comorbidities. Therefore, the one-size-fits-all dosing strategy long employed for programmatic ease can be harmful, especially in individual clinical scenarios in which current doses of TB drugs are on the steep part of the dose–response curve. Examples include rifampicin (33), which is systematically underdosed, particularly in underweight people and people living with HIV (34), as well as bedaquiline, a potent new drug for which black populations experience 50% lower exposures (35). These observations and our finding of different optimal doses against inhA-mutated isolates, dependent on NAT2 acetylator, highlight the importance of tailoring treatment to ensure similar exposures and treatment outcomes in a diverse global population. Ideally, NAT2 genotyping would be available as a package similar to point-of-care PCR testing platforms for bacterial resistance genotyping. We can only hope that our results will be an incentive for industry to provide such technology in future.

Limitations

Our study has several limitations, many of which we believe have been alleviated by using a model-based approach for data analysis. cfu and TTP data are known to be very noisy (8); this was mitigated by jointly modeling the two biomarkers and characterizing their correlation within each observation. Missing MIC information in 54% of the participants may have limited our power to detect an effect, but when testing the available MIC values, these were not found to provide useful information after stratifying by drug-sensitive M.tb versus inhA, which was obtained with genetic testing. Furthermore, our model was unable to characterize an exposure–response relationship in the drug-sensitive arm. This may be because all participants in this arm received the standard 5-mg/kg dose, which has been reported to achieve close to the maximum EBA (EBA90) in most patients irrespective of acetylator status (36).

Isoniazid bactericidal activity in drug-sensitive M.tb is known to be greatest during the first 2 days of treatment and then to decrease (37). To capture this, our model would have required additional components, such as a decline in kill rate with time or a tolerance model, but the available data were inadequate to support more complexity. Despite this, our model was able to describe the overall maximum kill achieved by the standard dose against drug-sensitive M.tb. The overall drop in cfu predicted by our model is comparable with previously reported isoniazid EBA (38) of drug-sensitive M.tb after 5 days of 300-mg isoniazid monotherapy, and the biphasic kill curve has not been as prominent in more recent studies (39). One shortfall of EBA studies is their limited ability to generalize results to later treatment periods, as bactericidal activity might differ then. Similarly, our model may have limited ability to predict the role of isoniazid later in treatment, when fewer actively metabolizing bacilli remain.

Conclusions

For isoniazid, unlike most other TB drugs, we currently have the opportunity to use simple molecular tests to detect and characterize levels of drug resistance, which is the major factor to determine the necessary drug exposure. Since resistance can be detected earlier, dose adjustments can occur at the start of treatment based on the presence of inhA mutation.

The study results show that a dose of 15 mg/kg against inhA-mutated M.tb has a high probability of achieving EBA similar to that of a standard dose against drug-sensitive M.tb. NAT2 genotype information could help customize isoniazid dose selection against inhA-mutated M.tb, as slow acetylators could be dosed at 10 mg/kg (consistent with WHO recommendations) and intermediate acetylators at 15 mg/kg, whereas fast acetylators are likely to require more than 15 mg/kg. Such tailored dosing will ensure similar exposures across NAT2 genotypes, minimizing toxicity risks associated with higher exposures and potential for reduced efficacy. Based on these data, the current dose for treatment for MDR-TB of 10 mg/kg as recommended by WHO may be suboptimal for fast and intermediate acetylators. However, the safety and tolerability of 15 mg/kg or higher doses must be investigated in long-term studies.

Acknowledgments

Acknowledgment

The authors wish to gratefully acknowledge Professor Cedric Werely for genotyping participant blood specimens to determine NAT2 acetylator status and Sachiko Miyahara for her input in this project. K.G. acknowledges her Ph.D. funders, the Virtual consortium, whose aim is to investigate the challenges of tuberculosis treatment for individuals on second-line antiretroviral therapy while promoting African leadership and capacity building, and her colleagues at the University of Cape Town Pharmacometrics division. Computations were performed using facilities provided by the University of Cape Town’s Information and Communication Technology Services High Performance Computing team (hpc.uct.ac.za).

Footnotes

Supported by TASK Applied Sciences grant #UM1AI069521; the National Center for Medical Rehabilitation Research grant AI068632; and the Division of AIDS, National Institute of Allergy and Infectious Diseases, NIH grant U01 AI068632. The University of Cape Town Clinical PK Laboratory is supported in part via the AIDS Clinical Trial Group; by the National Institute of Allergy and Infectious Diseases of the NIH under award numbers UM1 AI068634, UM1 AI068636, and UM1 AI106701; and the Infant Maternal Pediatric Adolescent AIDS Clinical Trials Group. E.H.I. is supported by T32 GM066691-17 and K.E.D. is supported by K24AI150349 both from the National Institute of Allergy and Infectious Diseases of the NIH. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

Author Contributions: S.K., L.M., R.H., S.L.R., S.S., A.H.D., E.L.N., K.D., and K.E.D. designed the study. L.W., F.v.G.-B., N.V., and A.H.D. conducted the study. K.G., X.S., and P.D. performed the analyses. K.G., E.H.I., E.L.N., K.E.D., and P.D. drafted the manuscript. All authors provided input and reviewed and approved the final manuscript.

This article has an online supplement, which is accessible from this issue’s table of contents at www.atsjournals.org.

Originally Published in Press as DOI: 10.1164/rccm.202103-0534OC on August 17, 2021

Author disclosures are available with the text of this article at www.atsjournals.org.

References

  • 1.Review on Antimicrobial Resistance. Antimicrobial resistance: tackling a crisis for the health and wealth of nations. London: Review on Antimicrobial Resistance. 2014.
  • 2. Karakousis PC, Williams EP, Bishai WR. Altered expression of isoniazid-regulated genes in drug-treated dormant Mycobacterium tuberculosis. J Antimicrob Chemother . 2008;61:323–331. doi: 10.1093/jac/dkm485. [DOI] [PubMed] [Google Scholar]
  • 3. Lempens P, Meehan CJ, Vandelannoote K, Fissette K, de Rijk P, Van Deun A, et al. Isoniazid resistance levels of Mycobacterium tuberculosis can largely be predicted by high-confidence resistance-conferring mutations. Sci Rep . 2018;8:3246. doi: 10.1038/s41598-018-21378-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4. Mouton JW, Ambrose PG, Canton R, Drusano GL, Harbarth S, MacGowan A, et al. Conserving antibiotics for the future: new ways to use old and new drugs from a pharmacokinetic and pharmacodynamic perspective. Drug Resist Updat . 2011;14:107–117. doi: 10.1016/j.drup.2011.02.005. [DOI] [PubMed] [Google Scholar]
  • 5. Zabost A, Brzezińska S, Kozińska M, Błachnio M, Jagodziński J, Zwolska Z, et al. Correlation of N-acetyltransferase 2 genotype with isoniazid acetylation in Polish tuberculosis patients. BioMed Res Int . 2013;2013:853602. doi: 10.1155/2013/853602. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6. Donald PR, Sirgel FA, Botha FJ, Seifart HI, Parkin DP, Vandenplas ML, et al. The early bactericidal activity of isoniazid related to its dose size in pulmonary tuberculosis. Am J Respir Crit Care Med . 1997;156:895–900. doi: 10.1164/ajrccm.156.3.9609132. [DOI] [PubMed] [Google Scholar]
  • 7.Dooley KE, Miyahara S, von Groote-Bidlingmaier F, Sun X, Hafner R, Rosenkranz SL, et al. A5312 Study Team Early bactericidal activity of different isoniazid doses for drug resistant TB (INHindsight): a randomized open-label clinical trial Am J Respir Crit Care Med 20202011416–1424.. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8. Rockwood N, du Bruyn E, Morris T, Wilkinson RJ. Assessment of treatment response in tuberculosis. Expert Rev Respir Med . 2016;10:643–654. doi: 10.1586/17476348.2016.1166960. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9. Ogasawara K, Breder CD, Lin DH, Alexander GC. Exposure- and dose-response analyses in dose selection and labeling of FDA-approved biologics. Clin Ther . 2018;40:95–102.e2. doi: 10.1016/j.clinthera.2017.11.012. [DOI] [PubMed] [Google Scholar]
  • 10. Sabbagh A, Darlu P, Vidaud M. Evaluating NAT2PRED for inferring the individual acetylation status from unphased genotype data. BMC Med Genet . 2009;10:148. doi: 10.1186/1471-2350-10-148. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11. Keizer RJ, Karlsson MO, Hooker A. Modeling and simulation workbench for NONMEM: tutorial on Pirana, PsN, and Xpose. CPT Pharmacometrics Syst Pharmacol . 2013;2:e50. doi: 10.1038/psp.2013.24. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12. Anderson BJ, Holford NHG. Mechanism-based concepts of size and maturity in pharmacokinetics. Annu Rev Pharmacol Toxicol . 2008;48:303–332. doi: 10.1146/annurev.pharmtox.48.113006.094708. [DOI] [PubMed] [Google Scholar]
  • 13. Upton RN, Mould DR. Basic concepts in population modeling, simulation, and model-based drug development: part 3-introduction to pharmacodynamic modeling methods. CPT Pharmacometrics Syst Pharmacol . 2014;3:e88. doi: 10.1038/psp.2013.71. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14. Mould DRDR, Upton RN. Basic concepts in population modeling, simulation, and model-based drug development-part 2: introduction to pharmacokinetic modeling methods. CPT Pharmacometrics Syst Pharmacol . 2013;2:e38. doi: 10.1038/psp.2013.14. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15. Katiyar SK, Bihari S, Prakash S, Mamtani M, Kulkarni H. A randomised controlled trial of high-dose isoniazid adjuvant therapy for multidrug-resistant tuberculosis. Int J Tuberc Lung Dis . 2008;12:139–145. [PubMed] [Google Scholar]
  • 16. Rieder HL, Van Deun A. Rationale for high-dose isoniazid in the treatment of multidrug-resistant tuberculosis. Int J Tuberc Lung Dis . 2017;21:123–124. doi: 10.5588/ijtld.16.0619. [DOI] [PubMed] [Google Scholar]
  • 17. Piubello A, Harouna SH, Souleymane MB, Boukary I, Morou S, Daouda M, et al. High cure rate with standardised short-course multidrug-resistant tuberculosis treatment in Niger: no relapses. Int J Tuberc Lung Dis . 2014;18:1188–1194. doi: 10.5588/ijtld.13.0075. [DOI] [PubMed] [Google Scholar]
  • 18.World Health Organization. Frequently asked questions about the implementation of the new WHO recommendation on the use of the shorter MDR-TB regimen under programmatic conditions. Geneva: World Health Organization 2016. [accessed 2020 Feb 20]. Available from: https://www.who.int/tb/areas-of-work/drug-resistant-tb/treatment/FAQshorter_MDR_regimen.pdf.
  • 19. Strydom N, Gupta SV, Fox WS, Via LE, Bang H, Lee M, et al. Tuberculosis drugs’ distribution and emergence of resistance in patient’s lung lesions: a mechanistic model and tool for regimen and dose optimization. PLoS Med . 2019;16:e1002773. doi: 10.1371/journal.pmed.1002773. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20. Prideaux B, Via LE, Zimmerman MD, Eum S, Sarathy J, O’Brien P, et al. The association between sterilizing activity and drug distribution into tuberculosis lesions. Nat Med . 2015;21:1223–1227. doi: 10.1038/nm.3937. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21. Rawat R, Whitty A, Tonge PJ. The isoniazid-NAD adduct is a slow, tight-binding inhibitor of InhA, the Mycobacterium tuberculosis enoyl reductase: adduct affinity and drug resistance. Proc Natl Acad Sci USA . 2003;100:13881–13886. doi: 10.1073/pnas.2235848100. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22. Abel zur Wiesch P, Clarelli F, Cohen T. Using chemical reaction kinetics to predict optimal antibiotic treatment strategies. PLoS Comput Biol . 2017;13:e1005321. doi: 10.1371/journal.pcbi.1005321. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23. O’Sullivan DM, McHugh TD, Gillespie SH. Mapping the fitness of Mycobacterium tuberculosis strains: a complex picture. J Med Microbiol . 2010;59:1533–1535. doi: 10.1099/jmm.0.019091-0. [DOI] [PubMed] [Google Scholar]
  • 24. Unissa AN, Subbian S, Hanna LE, Selvakumar N. Overview on mechanisms of isoniazid action and resistance in Mycobacterium tuberculosis. Infect Genet Evol . 2016;45:474–492. doi: 10.1016/j.meegid.2016.09.004. [DOI] [PubMed] [Google Scholar]
  • 25. Wen X, Gehring R, Stallbaumer A, Riviere JE, Volkova VV. Limitations of MIC as sole metric of pharmacodynamic response across the range of antimicrobial susceptibilities within a single bacterial species. Sci Rep . 2016;6:37907. doi: 10.1038/srep37907. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26. Dickinson JM, Mitchison DA. In vitro and in vivo studies to assess the suitability of antituberculous drugs for use in intermittent chemotherapy regimens. Tubercle . 1968;49:66–70. doi: 10.1016/s0041-3879(68)80051-0. [DOI] [PubMed] [Google Scholar]
  • 27. Awaness AM, Mitchison DA. Cumulative effects of pulsed exposures of mycobacterium tuberculosis to isoniazid. Tubercle . 1973;54:153–158. doi: 10.1016/0041-3879(73)90035-4. [DOI] [PubMed] [Google Scholar]
  • 28. Caño-Muñiz S, Anthony R, Niemann S, Alffenaar JC. New approaches and therapeutic options for Mycobacterium tuberculosis in a dormant state. Clin Microbiol Rev . 2017;31:e00060-e17. doi: 10.1128/CMR.00060-17. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29. Bowness R, Boeree MJ, Aarnoutse R, Dawson R, Diacon A, Mangu C, et al. The relationship between Mycobacterium tuberculosis MGIT time to positivity and cfu in sputum samples demonstrates changing bacterial phenotypes potentially reflecting the impact of chemotherapy on critical sub-populations. J Antimicrob Chemother . 2015;70:448–455. doi: 10.1093/jac/dku415. [DOI] [PubMed] [Google Scholar]
  • 30. Svensson RJ, Svensson EM, Aarnoutse RE, Diacon AH, Dawson R, Gillespie SH, et al. Greater early bactericidal activity at higher rifampicin doses revealed by modeling and clinical trial simulations. J Infect Dis . 2018;218:991–999. doi: 10.1093/infdis/jiy242. [DOI] [PubMed] [Google Scholar]
  • 31.le Roux SP, Upton C, Vanker N, Dooley KE, Diacon AH.Resistance-conferring Mycobacterial mutations and quantification of early bactericidal activity Am J Respir Crit Care Med 2021203 635–637.. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32. Bardou F, Raynaud C, Ramos C, Lanéelle MA, Lanŕelle G, et al. Mechanism of isoniazid uptake in Mycobacterium tuberculosis. Microbiology (Reading) . 1998;144:2539–2544. doi: 10.1099/00221287-144-9-2539. [DOI] [PubMed] [Google Scholar]
  • 33. Nabisere R, Musaazi J, Denti P, Aber F, Lamorde M, Dooley KE, et al. Pharmacokinetics, SAfety/tolerability, and EFficacy of high-dose RIFampicin in tuberculosis-HIV co-infected patients on efavirenz- or dolutegravir-based antiretroviral therapy: study protocol for an open-label, phase II clinical trial (SAEFRIF) Trials . 2020;21:181. doi: 10.1186/s13063-020-4132-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34. Guiastrennec B, Ramachandran G, Karlsson MO, Kumar AKH, Bhavani PK, Gangadevi NP, et al. Suboptimal antituberculosis drug concentrations and outcomes in small and HIV-coinfected children in India: recommendations for dose modifications. Clin Pharmacol Ther . 2018;104:733–741. doi: 10.1002/cpt.987. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35. McLeay SC, Vis P, van Heeswijk RPG, Green B. Population pharmacokinetics of bedaquiline (TMC207), a novel antituberculosis drug. Antimicrob Agents Chemother . 2014;58:5315–5324. doi: 10.1128/AAC.01418-13. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36. Donald PR, Parkin DP, Seifart HI, Schaaf HS, van Helden PD, Werely CJ, et al. The influence of dose and N-acetyltransferase-2 (NAT2) genotype and phenotype on the pharmacokinetics and pharmacodynamics of isoniazid. Eur J Clin Pharmacol . 2007;63:633–639. doi: 10.1007/s00228-007-0305-5. [DOI] [PubMed] [Google Scholar]
  • 37. Jindani A, Aber VR, Edwards EA, Mitchison DA. The early bactericidal activity of drugs in patients with pulmonary tuberculosis. Am Rev Respir Dis . 1980;121:939–949. doi: 10.1164/arrd.1980.121.6.939. [DOI] [PubMed] [Google Scholar]
  • 38. Donald PR, Diacon AH. The early bactericidal activity of anti-tuberculosis drugs: a literature review. Tuberculosis (Edinb) . 2008;88:S75–S83. doi: 10.1016/S1472-9792(08)70038-6. [DOI] [PubMed] [Google Scholar]
  • 39. De Jager V, van der Merwe L, Venter A, Donald PR, Diacon AH. Time trends in sputum Mycobacterial load and two-day bactericidal activity of isoniazid-containing antituberculosis therapies. Antimicrob Agents Chemother . 2017;61:e02088-e16. doi: 10.1128/AAC.02088-16. [DOI] [PMC free article] [PubMed] [Google Scholar]

Articles from American Journal of Respiratory and Critical Care Medicine are provided here courtesy of American Thoracic Society

RESOURCES