Effect of Isoniazid Intake on Ethionamide Pharmacokinetics and Target Attainment in Multidrug-Resistant Tuberculosis Patients

ABSTRACT

Ethionamide is recommended as part of regimens to treat multidrug-resistant and rifampicin-resistant tuberculosis. This study was conducted to (i) describe the distribution of ethionamide MICs, (ii) describe the pharmacokinetics of ethionamide, and (iii) determine the probability of attaining target area under the concentration-time curve from 0 to 24 h (AUC_0–24)/MIC values associated with suppression of resistant subpopulation and microbial kill. Participants received 15 to 20 mg of drug/kg of body weight of ethionamide daily (in 500- or 750-mg doses) as part of a multidrug regimen. Pretreatment MICs of ethionamide for Mycobacterium tuberculosis sputum isolates were determined using Sensititre MYCOTB MIC plates. Plasma concentrations of ethionamide (measured predose and at 2, 4, 6, 8, and 10 h postdose) were available for 84 patients. A one-compartment disposition model, including a liver compartment capturing hepatic extraction, best described ethionamide pharmacokinetics. Clearance and volume were allometrically scaled using fat-free mass. Isoniazid coadministration reduced ethionamide clearance by 31%, resulting in a 44% increase in AUC_0–24. The median (range) MIC (n = 111) was 2.5 mg/liter (<0.3 to >40 mg/liter). Simulations showed increased daily doses of ethionamide (1,250 mg, 1,500 mg, and 1,750 mg for patients weighing ≤45 kg, 46 to 70 kg, and >70 kg, respectively) resulted in the probability of attaining an area under the concentration-time curve from 0 to 24 h for the free, unbound fraction of a drug (fAUC_0–24)/MIC ratio of ≥42 in more than 90% of patients only at the lowest MIC of 0.3 mg/liter. The WHO-recommended doses of ethionamide do not achieve target concentrations even for the lowest MIC measured in the cohort.

TEXT

Ethionamide has been recommended as a group C antituberculosis drug since 2016 for multidrug-resistant or rifampicin monoresistant tuberculosis (MDR/RR-TB) when one of the preferred group A or B drugs is contraindicated (1). The role of ethionamide in the treatment of MDR-TB is unclear, and its use is often limited by toxicity, most notably gastrointestinal intolerance. Other adverse effects include hepatotoxicity, hypothyroidism, peripheral neuropathy, and psychosis, which may be exacerbated by companion drugs in the treatment regimen (2–4).

Ethionamide and isoniazid (INH) are coadministered during the intensive phase of the shorter WHO regimen (1). Ethionamide is a structural analogue of isoniazid, and the drugs share a primary target, inhA. Mutations in inhA regulatory or coding regions confer low-level resistance to isoniazid and resistance to ethionamide (5, 6). High doses of isoniazid are used to overcome “low-level” isoniazid resistance conferred by the inhA mutations. A drug-drug interaction may further increase isoniazid concentrations (regardless of the acetylator status) when it is taken with ethionamide (7). Among patients harboring Mycobacterium tuberculosis with inhA mutations, whether ethionamide concentrations are sufficient to overcome the higher MICs is not clear.

A dose of 15 to 20 mg of drug/kg of body weight/day of ethionamide in single or divided doses is recommended for the treatment of MDR-TB (8). Ethionamide is metabolized mainly in the liver (9), and food has minimal impact on its pharmacokinetics (10). The efficacy of ethionamide has been linked to the ratio of the unbound area under the 24-h concentration-time curve to MIC (fAUC_0–24/MIC). Hollow fiber system models associated an fAUC_0–24/MIC ratio of 56.2 with 80% of maximal microbial kill and ratios of 42 and 10 with suppression of resistant subpopulations and 1.0 log₁₀ CFU/ml kill, respectively. Monte Carlo simulations demonstrated that daily doses of 20 mg/kg achieve the fAUC_0–24/MIC target of 56.2 for MICs of <2.5 mg/liter only (11). We sought to (i) characterize the pharmacokinetics of ethionamide in hospitalized MDR/RR-TB patients, (ii) describe the distribution of ethionamide MICs in relation to inhA and katG mutations, and (iii) identify ethionamide doses and MICs associated with attaining fAUC_0–24/MIC targets for suppression of resistant subpopulations, 1.0 log₁₀ CFU/ml kill, and 80% microbial kill.

RESULTS

Patient characteristics.

Eighty-four participants were treated with ethionamide and had pharmacokinetic data available for analysis. These included 63 of 111 participants who had pretreatment ethionamide MICs determined. The participants characteristics are summarized in Table 1. Approximately two-thirds of the participants with pharmacokinetic data were HIV positive, and a similar proportion received a regimen including both isoniazid and ethionamide on the pharmacokinetic sampling day. More than half of the participants had a body mass index (BMI) of <18.5 kg · m⁻². The characteristics for participants who had pretreatment MICs was similar to those who had pharmacokinetic data only (Table 1). The proportion of participants receiving concomitant isoniazid was lower among the those with ethionamide pharmacokinetic data, as isoniazid replaced ethionamide in some patients based on the line probe assay results for katG and inhA mutations and the attending clinician’s discretion. Ethionamide was also discontinued due to adverse events in some patients. In our cohort, 91/144 (63%) experienced gastrointestinal upset (nausea and/or vomiting) with 28/144 (19%) experiencing grade 2 or above symptoms.

TABLE 1.

Participant characteristics at recruitment

Characteristica	Participants with pharmacokinetic data (median [range] or n [%])	Participants with pretreatment MICs (median [range] or n [%])
No. of participants	84	111
No. (%) of female participants	44 (52.4)	65 (58.6)
Wt (kg)	45.0 (32.0–83.0)	47 (32.0–84.0)
Ht (m)	1.6 (1.4–1.8)	1.7 (1.4–1.8)
FFM (kg)	40.3 (29.5–56.2)	40.1 (23.5–59.7)
BMI (kg · m−2)	16.9 (11.2–32.4)	17.1 (11.2–32.4)
Age (yr)	35.2 (19.5–68.9)	37.2 (18.8–68.9)
ALT (U/liter)	11.0 (5.0–75.0)	17.0 (4.0–74.0)
HIV status
Negative	29 (34.5)	41(36.9)
Positive	53 (63.1)	68 (61.3)
Unknown	2 (2.4)	2 (1.8)
No. (%) of participants who received crushed tablets	8 (9.5)
No. (%) of participants who received isoniazid	56 (66.7)b	95 (85.6)

^aFFM, fat-free mass; BMI, body mass index; ALT, alanine aminotransferase.

^bReceived isoniazid on the pharmacokinetic sampling day.

Pharmacokinetic model.

A one-compartment disposition model with first-order elimination and absorption described ethionamide pharmacokinetics well. The model fit was improved by the inclusion of a liver compartment to capture hepatic extraction and resulted in a 5.4-point drop in objective function value (OFV). The delay in absorption was best described by a transit compared to a lag time (a 28-point lower OFV), and the latter was superior to a simple first-order absorption model (a 90-point drop in OFV; 2 degrees of freedom (df); P < 0.001). The typical values for the volume of the liver compartment and hepatic plasma flow were fixed to 1 liter and 90 liters/h, respectively, and the unbound fraction of ethionamide was fixed to 70% (7). Allometric scaling was included on clearance and volume of distribution parameters using fat-free mass (FFM), which was slightly better than total body weight and retained in the model to maintain consistency with other pharmacokinetic models of drugs used in the study. Table 2 shows the parameter estimates of the final model and the 95% confidence intervals. The values for clearance and volume of distribution are reported for an individual with an FFM of 40.7 kg, the median FFM in the cohort. Ethionamide intrinsic clearance was reduced by 31% in participants who also received isoniazid as part of treatment (a 17-point drop in OFV; 1 df; P < 0.001). The effect of isoniazid coadministration was confirmed in a one-compartment model without a well-stirred liver component where isoniazid increased relative bioavailability and reduced oral clearance. The effect of HIV infection was not significant in our model, and we did not detect an effect of efavirenz-based (n = 18) or lopinavir/ritonavir-based (n = 9) antiretroviral therapy on ethionamide pharmacokinetics. We did not detect a change in the delay or rate of absorption of ethionamide when crushing the antituberculosis drugs before administering them together in water (12, 13). However, as only eight patients received crushed ethionamide tablets, the finding is not conclusive. Between-subject variability (BSV) was included on clearance, and between-occasion variability (BOV) was supported on absorption rate constant, mean transit time, and relative bioavailability. Between-visit variability (BVV) was not supported by the model. The additive component of the residual unexplained variability (RUV) was bounded to be at least 20% of the lower limit of quantification (LLOQ), and the model estimated a proportional component of 14%. Eighty-seven (16%) concentration measurements were reported as below the LLOQ.

TABLE 2.

Population parameter estimates of ethionamide pharmacokinetics

Parameter	Estimate (95% CIa)
Intrinsic clearance (liter/h)b	61.9 (52.7, 70.4)
Vol (liter)b	62.0 (56.4, 66.5)
Absorption rate constant (h−1)	0.998 (0.772, 1.31)
Prehepatic bioavailability (%)	100, Fixed
Mean transit time (h)	0.967 (0.683, 1.24)
No. of transit compartments	6, Fixed
Change in clearance due to isoniazid (%)	−30.6 (−41.0, −16.2)
Between-subject variability (%) for clearance	35.2 (24.1, 41.1)
Between-occasion variability (%) for:
Bioavailability	19.8 (13.9, 24.4)
Absorption rate constant	52.2 (38.9, 69.7)
Mean transit time	67.3 (47.4, 92.2)
Residual unexplained variability
Proportional (%)	14.8 (12.7, 16.4)
Additive (mg/liter)c	0.00626, Fixed

^aObtained from a nonparametric bootstrap (n = 300). CI, confidence interval.

^bEstimates reported for an individual with a fat free mass (FFM) of 40.7 kg (the median FFM in the cohort).

^cAdditive was error fixed to 20% of LLOQ (0.00626) for all samples, and an extra 50% of LLOQ was added on imputed values.

Figure 1 shows a visual predictive check stratified by isoniazid intake, showing a suitable fit to the data. Table 3 shows the median (range) of model-derived individual AUC_0–24 for all profiles, stratified by the administered dose and whether a participant received isoniazid. Model-derived steady-state ethionamide AUC_0–24 was approximately 50% higher in participants who received 750 mg than in those who received a 500 mg dose. Simulations applying the WHO-recommended once daily doses predicted comparable exposure across the five weight bands (30 to 35 kg, 36 to 45 kg, 46 to 55 kg, 56 to 70 kg, and >70 kg) with corresponding median (interquartile range) AUC_0–24 of 15.0 mg · h/liter (11.4 to 19.8), 13.3 mg · h/liter (10.1 to 17.5), 17.6 mg · h/liter (16.4 to 23.1), 15.9 mg · h/liter (12.1 to 20.9), and 18.9 mg · h/liter (14.4 to 25.0).

FIG 1.

Visual predictive check stratified by isoniazid (INH) intake. Open circles represent the observed concentrations. The middle continuous line is the 50th percentile of the observed data, and upper and lower dashed lines are the 97.5th and 2.5th percentiles of the observed data, respectively. The shaded regions represent the 95% confidence interval of the 2.5th, 50th, and 97.5th percentiles.

TABLE 3.

Summary of individual model-derived steady-state AUC_0–24

Parameter	No.	Median AUC (minimum; maximum)
All participants	84	17.2 (6.1; 66.4)
Isoniazid coadministered
No	28	12.2 (6.1; 37.1)
Yes	56	19.1 (6.6; 66.4)
Ethionamide dose
500 mg	62	16.1 (6.1; 66.4)
750 mg	22	24.4 (13.1; 35.4)
Isoniazid intake and ethionamide dose
Isoniazid, ethionamide 500 mg	41	17.7 (6.6; 66.4)
No isoniazid, ethionamide 500 mg	21	11.1 (6.1; 37.1)
Isoniazid, ethionamide 750 mg	15	30.7 (13.3; 35.4)
No isoniazid, ethionamide 750 mg	7	16.9 (13.1; 26.7)

Distribution of MICs.

Figure 2 shows the distribution of MICs of ethionamide stratified by inhA and katG mutations in M. tuberculosis isolates from 111 participants in the study cohort. The median MIC was 2.5 mg/liter. The distribution of MICs in M. tuberculosis isolates from patients who had ethionamide pharmacokinetic data was similar to the rest of the cohort (median = 2.5 mg/liter). The inhA mutation was detected in 59% of the isolates, and katG was detected in 14% of the isolates. There was a trend in which M. tuberculosis isolates with inhA mutation only had higher ethionamide MICs than those with the katG mutation only. The median MICs for the isolates with inhA mutation and those with katG mutation, respectively, were 5 mg/liter and 1.2 mg/liter (P = 0.054).

FIG 2.

Distribution of MIC stratified by genetic mutations for isoniazid.

Probability of target attainment.

Figure 3 shows the probability of attaining the target fAUC_0–24/MIC > 10 and 42 up to an MIC value of 10 mg/liter. Higher MIC values were not plotted since the probability was zero for all of the doses evaluated. The recommended dose of ethionamide (15 to 20 mg/kg daily) is associated with a probability of target attainment (PTA) of >90% for the target fAUC_0–24/MIC > 10 only at MICs of ≤0.6 mg/liter. For the same target, none of the evaluated doses achieves a PTA of >90% at MICs ≥ 2.5 mg/liter. Only doses above 1,250 mg attained a PTA of >90% for the target suppressing resistant subpopulations (fAUC_0–24/MIC > 42) at the lowest measured MIC of 0.3 mg/liter. None of the evaluated doses achieved a PTA of >90% using the fAUC_0-24/MIC > 56.2 target (not shown).

FIG 3.

Probability of attaining fAUC_0–24/MIC ratio of 10 (solid lines) and 42 (dashed lines).

DISCUSSION

Our findings provide evidence against the use of ethionamide in patients with MDR/RR-TB based on failure to meet a variety of pharmacokinetic-pharmacodynamic (PKPD) targets with the currently recommended doses evaluated against the MICs observed in our cohort. As ethionamide is poorly tolerated at the recommended doses, the increased doses suggested by our simulations to attain the PKPD targets are unlikely to be feasible (2, 4, 14). These findings cast doubt on the role of ethionamide against drug-resistant tuberculosis. We also identified a novel drug-drug interaction between ethionamide and isoniazid, resulting in increased ethionamide exposure, which could increase the risks of ethionamide-associated toxicity.

We developed a population pharmacokinetic model of ethionamide in participants hospitalized for second-line tuberculosis treatment. Body size affected the disposition parameters, which were well described by allometric scaling with fixed exponents of 3/4 and 1 for clearance and volume parameters, respectively. Concomitant intake of ethionamide and isoniazid resulted in higher bioavailability and reduced oral clearance of ethionamide, hence our approach to include the effect on hepatic clearance and thus first-pass metabolism to account for both phenomena. Isoniazid increased the AUC_0–24 of ethionamide by 44%. The mechanism underlying this interaction is unclear. Although the two drugs have a similar structure, they have distinct metabolic pathways. In a previous analysis of ethionamide pharmacokinetics, a trend showing increased ethionamide exposures was observed in children treated with isoniazid. However, the effect could not be robustly characterized due to the limited number of children who received both ethionamide and isoniazid (15). Conversely, the effect of HIV infection on ethionamide pharmacokinetics previously reported in children (15, 16) was not significant in our model. While the elimination half-life derived from the final parameter estimates of our model was around 1 h, slightly lower than the previously reported values of 1.4 to 3 h (9, 10, 17, 18), model-derived estimates of AUC_0–24 (presented in Table 3) are within the range of values reported in patients with tuberculosis and in healthy volunteers (10, 18).

We reported the distribution of ethionamide pretreatment MICs. Our median MIC of 2.5 mg/liter was higher than those reported for isolates from Tanzania and Thailand (11, 19). Our finding of higher ethionamide MICs in isolates with inhA mutations is not unexpected and potentially impacts PKPD target attainment. Although isolates with inhA mutations displayed a range of MICs above and below the critical concentration of 5 mg/liter, the fAUC_0–24/MICs achieved in our cohort did not meet the predefined PKPD targets for efficacy and suppression of resistance derived from hollow-fiber models in the vast majority of patients regardless of the isoniazid resistance mutation status of their isolates.

Simulations by Al-Shaer et al. (20) showed that ethionamide doses of 1,500 mg per day achieve a PTA of ≥90% for the fAUC_0–24/MIC target of 10 at an MIC of ≤1 mg/liter. However, their model did not include known covariates, such as body weight or fat-free mass, that contribute to variability in drug exposure. For our simulations, we applied the weight bands used for the treatment of MDR-TB and report that for patients in the 46- to 70-kg weight bands, ethionamide in an increased daily dose of 1,500 mg results in a PTA of ≥90% for the index associated with 1.0 log₁₀ CFU/ml kill. We also show that to achieve the target fAUC_0–24/MIC associated with suppression of resistance subpopulation, the 1,500 mg daily dose of ethionamide is effective at MICs of ≤0.3 mg/liter. Considering the distribution of MIC reported here, suppression of resistant subpopulations is only achievable in less than 2% of the population. This is of concern, as subtherapeutic ethionamide exposure can promote the emergence of acquired drug resistance, including that of companion drugs (11). Increasing the dose of ethionamide to suppress resistant subpopulations might increase the incidence of nausea and vomiting associated with the use of a drug, which is already poorly tolerated. At standard doses, ethionamide and its prodrug prothionamide have been repeatedly associated with gastrointestinal toxicity (21, 22). Ethionamide also penetrates poorly into lung tissues, and increasing the dose is unlikely to improve drug exposure in the lungs (23).

Our study has some limitations. The PKPD targets that we used are based on data from in vitro monotherapy models and may not reflect the exposure and organism susceptibility at the site of infection in patients on multidrug regimens. The MIC distribution in our study, which included a high proportion of patients with inhA mutations, may not be broadly applicable (24). There could be an effect of faster absorption in patients taking ethionamide in crushed form, but our study may have been underpowered to detect this effect. The Sensititre plate used to determine ethionamide MICs has low sensitivity compared to that of the mycobacterium growth indicator tube, and we may have underestimated the MIC values (25). Lastly, there is a need for clinical studies to validate the targets derived in hollow-fiber models and assess the risks and benefits of higher doses of ethionamide evaluated in our study. These factors do not abrogate concerns about the very wide gap between the PKPD targets and the fAUC_0–24/MIC estimates in our study.

In conclusion, we report the effect of isoniazid on pharmacokinetics of ethionamide in a cohort of South African patients treated for MDR/RR-TB and describe the distribution of ethionamide MICs. Ethionamide exposures associated with the current dosing are insufficient to achieve prescribed PKPD targets.

MATERIALS AND METHODS

Study design and patient recruitment.

The PODRtb study (ClinicalTrials registration no. NCT02727582) was an observational study exploring pharmacokinetic-pharmacodynamic relationships in patients treated for MDR/RR-TB. Details of the study design and patient recruitment have been reported before (12). Briefly, we recruited patients with RR- or MDR-TB aged ≥18 years at the Brooklyn Chest and DP Marais TB hospitals in Cape Town, South Africa. Most patients were hospitalized for a minimum of 3 months, and they all received daily therapy as inpatients under supervision. The standard treatment regimen at the time of study included the following: moxifloxacin, terizidone (d-cycloserine), kanamycin, pyrazinamide, ethambutol (in patients likely to be susceptible), and ethionamide and/or isoniazid. The choice to use ethionamide and/or isoniazid was at the discretion of the treating clinician. For 81 of 84 participants on ethionamide, line probe assay results, identifying katG and inhA mutations in the pretreatment sputum culture were available, indicating high- and low-level resistance to isoniazid, respectively (26). Patients received weight-adjusted daily doses of ethionamide on an empty stomach: those weighing 33 to 50 kg and >50 kg received 500 mg and 750 mg, respectively. Four patients with a weight of 32 kg received a 500-mg dose.

Specimen collection and drug quantification.

Blood samples for pharmacokinetic analysis were collected between 2 and 6 weeks after MDR-TB treatment initiation. On the pharmacokinetic sampling day, blood samples were collected predose and at 2, 4, 6, 8, and 10 h postdose. A subset of the participants was selected nonrandomly to provide blood samples on a second pharmacokinetic sampling day 2 weeks after the first pharmacokinetic sampling day. These participants received the drugs in a crushed form on the second pharmacokinetic sampling day. Plasma was prepared from blood samples by centrifugation and stored at −80°C until analysis. The plasma samples were processed with a protein precipitation extraction method using acetonitrile, followed by tandem mass spectrometry detection as previously reported (15). The LLOQ of the assay was 0.0313 mg/liter.

MIC determination.

Sputum samples were collected before the start of treatment for MIC determination. Following decontamination of specimens, M. tuberculosis bacilli were processed for culture on liquid medium and growth was monitored in a Bactec MGIT 960 instrument. The MICs were determined using a commercially available Sensititre MYCOTB plate (TREK Diagnostics, Cleveland, OH, USA), following the manufacturer’s instructions. The MIC test range was from 0.3 to 40 mg/liter.

Data analysis.

A population pharmacokinetic model was developed in NONMEM version 7.4.3 (27) using the algorithm first-order conditional estimation with eta-epsilon interaction (FOCE-I). Automation of methods implemented in NONMEM was performed using Perl-speaks-NONMEM version 4.9.0 software (28). R version 3.6.3 and RStudio version 1.1.463 software were used for data management and generating graphical output (29, 30). One- and two-compartment disposition models with linear absorption and elimination were evaluated, and a delay in absorption was described using a lag time or a transit compartment model. Inclusion of a liver compartment to mechanistically describe the first-pass effect and hepatic clearance was evaluated (31). Bioavailability was fixed to 100% and allowed to vary between individuals and between doses within an individual.

Random effects were included on the pharmacokinetic parameters assuming a lognormal distribution. BSV was explored on disposition parameters, and BOV was investigated on absorption parameters. We also considered the effect of including BVV on clearance and volume parameters in addition to BSV. An occasion was defined as one dosing interval (all concentration measurements after a dose were assigned to the same occasion) while a visit was defined by a set of consecutive sampling events typically providing pharmacokinetic data for two dosing intervals. RUV was described using an error model with an additive and a proportional component. Concentration measurements below the LLOQ were replaced by half of the LLOQ. The lower limit of the additive component of the RUV was set at 20% of the LLOQ for all values, and an extra 50% of the LLOQ was added to the additive component of the error model on all imputed values. The extra error was included to assign a relatively lower weight during estimation to the imputed values. Allometric scaling was included in the initial phase of model development and the exponents for clearance and volume were fixed to 3/4 and 1, respectively (32, 33). The effects of other plausible covariates, including tablet crushing, concomitant tuberculosis drugs, HIV status, and antiretroviral drugs (efavirenz or lopinavir/ritonavir), were evaluated using a stepwise covariate search method, and the most significant covariate was included at each step. Model adequacy was assessed using goodness of fit plots, scatterplot of fitted versus observed concentrations, and visual predictive checks. A nonparametric bootstrap (n = 500) was used to determine the uncertainty of the parameters of the final model.

Monte Carlo simulations.

We performed Monte Carlo simulations to assess the exposure achieved in each dosing weight band using daily doses of 500 mg (30 to 45 kg), 750 mg (46 to 70 kg), and 1,000 mg (>70 kg) as recommended by WHO and evaluated the probability of target attainment. A simulation data set (n = 13,475) was created using demographic data of 1,225 TB patients as previously described (12). For each patient in our simulation data set, we simulated 100 pharmacokinetic profiles in NONMEM version 7.4.3 software using parameter estimates of the final model. We also explored higher doses by increasing the doses of ethionamide up to 2,000 mg, assuming a tablet strength of 250 mg. Using the distribution of ethionamide MIC for the current study, we then described the probability of attaining the fAUC_0–24/MIC targets of >10, 42, and 56.2 associated with 1.0 log₁₀ CFU/ml kill, suppression of resistant subpopulation, and 80% of maximum bacterial kill, respectively (11).

Ethics.

Ethics approval for the study was granted by the Human Research Ethics Committee of the University of Cape Town, Cape Town, South Africa (106/2016). All patients provided written informed consent in a language of their choice (English, Afrikaans, or Xhosa) before participant recruitment.