Influence of N-acetyltransferase 2 (NAT2) genotype/single nucleotide polymorphisms on clearance of isoniazid in tuberculosis patients: a systematic review of population pharmacokinetic models

Purpose Significant pharmacokinetic variabilities have been reported for isoniazid across various populations. We aimed to summarize population pharmacokinetic studies of isoniazid in tuberculosis (TB) patients with a specific focus on the influence of N-acetyltransferase 2 (NAT2) genotype/single-nucleotide polymorphism (SNP) on clearance of isoniazid. Methods A systematic search was conducted in PubMed and Embase for articles published in the English language from inception till February 2022 to identify population pharmacokinetic (PopPK) studies of isoniazid. Studies were included if patient population had TB and received isoniazid therapy, non-linear mixed effects modelling, and parametric approach was used for building isoniazid PopPK model and NAT2 genotype/SNP was tested as a covariate for model development. Results A total of 12 articles were identified from PubMed, Embase, and hand searching of articles. Isoniazid disposition was described using a two-compartment model with first-order absorption and linear elimination in most of the studies. Significant covariates influencing the pharmacokinetics of isoniazid were NAT2 genotype, body weight, lean body weight, body mass index, fat-free mass, efavirenz, formulation, CD4 cell count, and gender. Majority of studies conducted in adult TB population have reported a twofold or threefold increase in isoniazid clearance for NAT2 rapid acetylators compared to slow acetylators. Conclusion The variability in disposition of isoniazid can be majorly attributed to NAT2 genotype. This results in a trimodal clearance pattern with a multi-fold increase in clearance of NAT2 rapid acetylators compared to slow acetylators. Further studies exploring the generalizability/adaptability of developed PopPK models in different clinical settings are required.


Introduction
Isoniazid, first synthesized by two Prague chemists, Hans Meyer and Josef Mally, in 1912, was subsequently demonstrated to have antitubercular activity in three different laboratories (Squibb and Hoffmann La Roche in the USA and Bayer in West Germany) in 1951 [1,2]. Isoniazid has remained the first-line therapy for tuberculosis (TB) even after 70 years of clinical therapy and the centenary of its synthesis [3]. A considerable proportion of TB population has been reported to have significant isoniazid pharmacokinetic variabilities. The pooled proportion of TB patients with low isoniazid concentration after 2 h of intake (C 2h ) from 26 studies was reported to be 0.43 (95% CI 0.32-0.55) [4]. Patients with lower isoniazid concentrations were associated with poor treatment outcomes, such as delayed sputum culture conversion [5]. Several covariates such as age, gender, food, drug-drug interactions (DDIs), nutritional status, comorbidities, and N-acetyltransferase 2 (NAT2) genotype may account for the pharmacokinetic variability of isoniazid [6].
NAT2 genotype is one of the most important covariates influencing the plasma concentration of isoniazid [7]. Among the three NAT2 acetylator phenotypes, rapid acetylators achieve the lowest and slow acetylators achieve the highest plasma concentration of isoniazid [7]. NAT2 slow Levin Thomas and Arun Prasath Raju contributed equally. acetylator TB patients have been reported to have a comparatively higher early bactericidal activity of isoniazid than rapid acetylators [8]. A metanalysis study of 13 randomized clinical trials reported that NAT2 rapid acetylators were more likely to have microbiological failure (pooled risk ratio [RR], 2.0; 95% confidence interval [CI], 1.5-2.7), adverse drug reaction (ADR) (RR, 2.0; CI, 1.1-3.4), and relapse (RR, 1.3; CI, 0.9-2.0) than slow acetylators [9]. On the other hand, a pooled analysis of 37 studies reported that slow NAT2 acetylators had increased the risk for the development of anti-TB drug-induced liver injury (AT-DILI) compared to non-slow NAT2 acetylators (intermediate and fast NAT2 acetylators) (overall odds ratio (OR) = 3.15 (95% CI 2.58-3.84, heterogeneity measure (I 2 ) = 51.3%, p = 0.000)) in TB patients [10]. Dose stratification of isoniazid among TB patients based on NAT2 genotype reduced the incidence of treatment failure among rapid acetylators and DILI among slow acetylators [11]. Hence, it is imperative to design precise dosage regimen for isoniazid to achieve better clinical outcomes and reduce ADRs in TB patients. The population pharmacokinetics (PopPK) approach has emerged as a potential tool for the optimization of antitubercular therapy (ATT) [12].
PopPK estimates drug pharmacokinetic parameters and has emerged as a powerful approach for identifying the sources and correlates of pharmacokinetic variability in a particular patient population [13]. PopPK approach allows several advantages over traditional pharmacokinetics such as permitting sparse sampling approach, cost effectiveness, concentrations without regard to steady-state conditions; allowing irregularly measured concentrations, estimation of variabilities along with identification of its sources; and combining heterogenous types of data from varying sources [14,15]. PopPK models may undertake a parametric or nonparametric approach [15]. PopPK models using parametric approach have a good explanatory potential and exhibit good flexibility for unusual situations requiring complex models, easier interpretation of covariate effects, and easier for conceptualization and fitting to observations [15,16].
A systematic review (Vietnamese language) reported ten PopPK studies from inception till July 2017. The selection criteria included articles with non-TB populations, without NAT2 genotyping, and with all types of modelling strategies [17]. A meta-analysis by Hong et al. evidenced that the dosenormalized summary estimates of isoniazid maximum [peak] plasma drug concentration (C max ) and area under the plasma concentration-time curve (AUC) of the TB population were significantly lower than that of healthy volunteers (C max and AUC were 36% and 26% lower in adult TB patients compared to healthy volunteers). This implies that currently recommended isoniazid dosages produced less drug exposure in TB patients compared to healthy volunteers. NAT2 acetylator status also significantly influenced the isoniazid pharmacokinetic properties [7]. We presume that several isoniazid PopPK studies exploring NAT2 genotype/single-nucleotide polymorphism (SNP) as a covariate would have been published to address the pharmacokinetic variabilities in isoniazid among different TB populations. Therefore, we performed a systematic review of relevant articles to assess the significance of NAT2 genotype/ SNP on the clearance of isoniazid.

Literature search strategy
The review was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement 2020 [18]. A systematic search of PubMed and Embase for articles published in the English language from inception till Feb 3, 2022 was performed to identify PopPK models of isoniazid in TB subjects. Hand searching of articles were also carried out to identify additional relevant articles. Search terms included for the literature search were: "isoniazid" OR "INH" OR "antitubercular" AND "population pharmacokinetic*" OR "PopPK" OR "NONMEM" OR "nonlinear mixed effect* model*" AND "Tuberculosis" OR "TB."

Study selection criteria
Two independent reviewers (L.T. and A.P.R.) performed literature search, identification, and selection. The reviewers reviewed the titles, abstracts, and full text of the articles in accordance with the defined inclusion and exclusion criteria. Any disagreements were resolved by a third reviewer (S.M.). Studies were included based on the following criteria: (i) non-linear mixed-effects modelling approach was used for PopPK analysis; (ii) parametric approach was used; (iii) separate model developed for isoniazid in TB patients; and (iv) NAT2 SNP/genotype was used as a covariate for PopPK model. Conversely, studies that met the following exclusion criteria were removed: (i) model developed for isoniazid other than TB; (ii) study population was non-TB population/latent TB/healthy volunteers, and (iii) isoniazid was given as preventive therapy.

Data extraction
The full text of selected articles was reviewed, and data were extracted using a standardized extraction form, independently by two authors (L.T. and A.P.R.) and were cross checked. Data discrepancies were resolved by S.M.. For all selected articles, the first author, publication year, total study sample size, gender, samples used for isoniazid modelling, age, body weight, percentage of human immunodeficiency virus (HIV) and diabetes mellitus (DM) population, country of the study population, NAT2 SNPs investigated, NAT2 genotype, analytical method, PopPK software, structural model, external validation, PopPK estimates, residual variability, clearance based on NAT2 SNPs/ genotype, and significant covariates affecting PopPK model were extracted. The extracted data was verified by M.S.R. and M.R. Illustration of PRISMA flow chart was created using Microsoft PowerPoint, and illustrations based on selected studies were performed in RStudio [19] using ggplot2 package [20].

Data quality assessment
The quality assessment of isoniazid based PopPK studies was carried out using an adopted checklist that was developed from (i) previously published clinical pharmacokinetics [21], (ii) population pharmacokineticpharmacodynamic guidelines [22], and (iii) two studies that developed a combination of the (i) and (ii) checklist [23,24]. The combined modified checklist consisted of 46 criteria categorized into five domains: the title, abstract, background/ introduction, methods/results, and discussion/conclusion, as shown in Table 1. A score of 1 was given for each criterion, if the relevant information was identified from the study, else zero point was given. All the 12 isoniazid PopPK studies were assessed based on these criteria. The compliance rate of each study was calculated using the following equation and reported in percentage.

Literature search
A total of 155, 122, and 2 articles were identified from Pub-Med, Embase, and hand searching of articles, respectively. A total of 95 duplicate articles were removed. After title and abstract screening, 37 articles were available for full texts. Among these 37 articles, 25 articles were excluded for the following reasons, including (i) articles did not have NAT2 genotype or no SNP information of NAT2 genotype was available (n = 18), (ii) NAT2 genotype was carried out; however, it was not included as a covariate for isoniazid PopPK modelling building (n = 1), (iii) the patient population of the study was not having TB (n = 2), (iv) full text of the articles was not available (n = 1), (v) was not a PopPK study (n = 1), and (vi) nonparametric approach was used for isoniazid PopPK modelling (n = 2). A total of 12 articles remained for the systematic review. The PRISMA flow diagram detailing the selection of PopPK studies of isoniazid for the systematic review is shown in Fig. 1.
Compliance rate(%) = Total number of criteria met Total number ofcriteria that are applicable for the study × 100

Quality assessment of selected literatures
All the studies had a compliance rate of above 80% for the quality of the PopPK study (range: 81.8-92.8%). Criterions with minimum compliance were a plot of concentration versus time/effects (8.3%), schematic of the final model (16.6%), equations for all model structures and covariate relationships (20%), external validation (25%), information regarding the specific body weight used in drug dosing and pharmacokinetic calculations (n = 41.6%), and co-administration of drugs (42.8%) as shown in Table 1.
One study was conducted in the Mexican population [30]. The number of TB patients and total number of samples with isoniazid ranged from 29 to 454 and 141 to 1814 respectively as shown in Table 2. Fifty percent of studies had less than or equal to hundred TB patients.

Sampling procedure
Some of the studies have taken minimal number of samples from TB patients, whereas few others have taken five or more samples from a patient on a single day and/or different days. Gao et al. 2021 [26] Cho et al. 2021 [27] Jing et al. 2020 [28] Sundell et al. 2020 [29] Huerta-García et al. 2020 [30] Sekaggya-Wiltshire et al.

Title
The title identifies the drug(s) and patient population(s)

Bioanalytical methods
Liquid chromatography with tandem mass spectrometry (LC-MS/MS) was the most preferred analytical method to determine the isoniazid concentration in the samples (n = 10). On the other hand, Huerta-García et al. and Sekaggya-Wiltshire et al. used high-performance liquid chromatography (HPLC) and high-performance liquid chromatography-ultraviolet (HPLC-UV), respectively, as shown in Table 3.
Other covariates influencing isoniazid clearance was coadministration of efavirenz in TB patients with HIV [31] and post menstrual age among pediatric TB population [34]. The equation describing the influence of all these covariates on isoniazid clearance is shown in Table 3. Gender and CD4 cell count were reported to affect isoniazid bioavailability among patients co-infected with TB and HIV [29]. Body mass index was a significant covariate affecting apparent volume of the central or plasma compartment in a two-compartment model (V 1 ) [30]. Different fixed dose combination (FDC) tablet formulations of ATT affected the bioavailability and absorption of isoniazid [31].

Study bias
A potential source of bias involved in this systematic review is the differences in the SNPs panel used for inferring NAT2 genotype. Additionally, diverse genotyping methodologies for screening NAT2 genotyping were used in different studies.  [42]. Despite intake of weight band-based dose of isoniazid, we identified significant variations in the isoniazid disposition characteristics within a study population and across various studies, suggesting the need for more population-specific isoniazid PopPK models for providing precision therapy of isoniazid. Several isoniazid PopPK models suggest higher isoniazid doses than the conventional weight-based isoniazid doses, particularly among rapid acetylators [39,43] [28].

Available PopPK studies on isoniazid
Several studies did not have information regarding the co-administration of drug(s). Physiologically based pharmacokinetic (PBPK) analysis has shown the likelihood of DDIs between CYP2C19 and CYP3A4 substrates with isoniazid, more likely among NAT2 slow acetylators [44]. TB patients have been reported to have a considerable number of clinically important potential DDIs (pDDIs) [45]. As coadministration of another drug with isoniazid may represent a potential covariate influencing the isoniazid drug levels, we recommend further isoniazid PopPK studies to include this covariate for testing in the model building approach. Limited studies have explored the potential effects of comorbidities such as HIV infection and DM on isoniazid PopPK. Lower exposures of isoniazid were reported in TB patients with these comorbidities [43,46,47]. Hence, future PopPK modelling studies may involve testing these comorbidities, along with several others, such as cardiovascular diseases, epilepsy, anemia, vitamin deficiencies, asthma/COPD, and the drugs used in their management as covariates for the development of isoniazid PopPK model. The pharmacokinetic parameters were reported to be comparatively less in non-fasting TB patients when compared to TB patients who were in a fasting state when isoniazid was taken [48,49]. Few studies have reported whether the patient had taken the drug in a fasting/or a non-fasting state. This information may be potentially valuable when the sample taken is pre-dose of isoniazid. However, when subsequent samples are taken further, information regarding the extent/duration of the fasting state after intake of isoniazid, along with the type of food taken after intake of a dose, may be required to arrive at more conclusive evidence of the effect of fasting state and type of food on isoniazid exposure.
Except for one study [25], all the NAT2 genotype/SNPbased isoniazid PopPK studies have described isoniazid disposition characteristics by a two-compartment model, as shown in Table 3. An isoniazid PopPK study conducted among the Indian pediatric TB population also described isoniazid disposition by one-compartment model [50]. Significant variabilities in CL/F, V 1 , V 2 , K a and Q estimates were observed across the studies, highlighting the need for population and individual specific NAT2 genotype-derived isoniazid PopPK model for precise dose calculation. We recommend future studies, particularly with larger sample sizes, divide the sample population into two groups, one for model development and another for model validation. External validation would aid in ensuring the generalizability and reproducibility of the model before its implementation in clinics or policy making.
The most common SNP panel that was used for deriving the NAT2 genotype was the six SNP panels of c.282C > T, c.341 T > C, c.481C > T, c.590G > A, c.803A > G, and c.857G > A. SNPs at c.341 T > C, c.590G > A, and c.857G > A of the NAT2 gene were investigated by all the studies that derived NAT2 genotype, except for a study which had investigated only one SNP. We recommend future isoniazid PopPK studies to assess the NAT2 genotype, including six or seven SNP panels with a specific focus on SNP investigations at c.341 T > C, c.590G > A, and c.857G > A. This provides more insights into the impact of genotype on the pharmacokinetics of isoniazid. Pharmacogenomic-guided therapy (PGT) of isoniazid could represent a cost-effective strategy for managing TB in countries such as Brazil, South Africa, and India [57]. NAT2 genotype was a significant covariate influencing the pharmacokinetics of isoniazid in all the studies. Most of the studies reported a trimodal clearance pattern for isoniazid to the three NAT2 genotypes. Two pediatric and one adult TB population studies combined fast and intermediate acetylator phenotypes and reported isoniazid clearance as a bimodal pattern. Studies have shown to have significant isoniazid clearance differences between rapid and intermediate NAT2 acetylators. Therefore, we recommend the classification of NAT2 genotype into three categories and consequent description of isoniazid clearance in TB patients in a trimodal pattern for more precise calculation of isoniazid dosage regimen.

Isoniazid pharmacokinetic-pharmacodynamic studies
Isoniazid clearance directly determines the isoniazid AUC 0-24 of each patient, and the AUC 0-24 values influence the specific degree of bactericidal activity [58]. In vitro model studies have shown that AUC 0-24 /minimum inhibitory concentration (MIC) ratio was the pharmacokineticpharmacodynamic index that well described the bactericidal activity of isoniazid [58,59]. In a pharmacokineticpharmacodynamic model derived from a longitudinal cohort study conducted among Malawian drug-sensitive TB patients on standard therapy, higher isoniazid exposure correlated with increased bacillary clearance from sputum during the first 2 months of ATT. Higher isoniazid C max , C max /MIC, and AUC 0-24 /MIC correlated with treatment success at the end of treatment [60]. Simulation results from a combined PBPK/pharmacodynamic model suggest that rational adjustment of isoniazid doses requires consideration of the regional prevalence of NAT2 acetylator status for increasing the treatment efficacy and reducing the probability of adverse events, treatment failure, and the emergence of drug resistance [61]. A pharmacokineticpharmacodynamic model assessing relationship between drug exposure and survival among adult TB meningitis patients reported that isoniazid exposure was associated    [64]. Higher isoniazid concentration was associated with a faster time to culture conversion among adult PTB patients with DM [46]. Hence, it is evident that clinical outcome and bactericidal activity in TB patients is dependent on isoniazid exposure (AUC 0-24 ), which is determined by clearance.

Influence of other covariates on isoniazid disposition
In all the identified studies, anthropometric variables such as body weight, lean body weight, body mass index (BMI), and fat-free mass were significant covariates affecting the pharmacokinetics of isoniazid. The effect of these variables was accounted for either by standard allometric scaling or by estimation of coefficients. Pregnancy is associated with an increase in fat-free mass [65]. Pregnancy has been reported to increase the clearance of isoniazid by 26% [66]. Abdelwahab et al. had reported a comparatively high isoniazid clearance of 97.1 L/h, 75.7 L/h, and 29 L/h for NAT2 rapid, intermediate, and slow acetylators, respectively [36]. Further PopPK studies are warranted to assess whether an increase in isoniazid doses is required in the pregnant TB population, particularly among NAT2 rapid acetylators. We recommend future PopPK studies to assess anthropometric variables and test them as potential covariates to characterize the disposition of isoniazid. In PopPK studies conducted among TB patients with HIV infection, CD4 cell count, gender, efavirenz, and FDC formulation of ATT also influenced isoniazid disposition. Decrease in isoniazid exposure has been previously reported when it was concomitantly administered with efavirenz [67]. Post menstrual age was reported as a significant covariate for the pediatric TB population by Panjasawatwong et al [34]. Previous reports have evidenced the significant role of enzyme maturation for each of the three NAT2 genotype groups on isoniazid clearance [68].

Limitations
Most of the studies included in our systematic review did not report the ethnicity of the TB patients. In the absence of definitive information on ethnicity, we stratified the population of each study by country, as shown in Table 2. Significant interethnic and intraethnic variabilities exist in the frequency of SNPs in the NAT2 gene [69]. Hence, extrapolations of NAT2 genotype data should not be carried out even within a population. The current study was restricted to articles published in English; therefore, inferences from studies published in other languages are missed. While this review details about the different PopPK models for isoniazid, it does not address the generalizability of these models which might help in using of these models for informed dosing decisions in clinical settings across populations.

Conclusion
The PopPK modelling approach incorporating several potential covariates including NAT2 genotype, anthropometric measures, and other clinical variables could provide a thrust for precise optimization of individual dosing regimens of isoniazid in TB patients in this era of precision therapy. All the studies have reported that NAT2 genotype/SNP was a significant covariate affecting the clearance of isoniazid. Most of the PopPK studies conducted in adult TB patients reported a twofold or threefold increase in isoniazid clearance for NAT2 rapid acetylators compared to NAT2 slow acetylators. NAT2 genotype-based isoniazid PopPK studies are required from several parts of the world, particularly from high TB burden Asian countries and western populations, as well as in pediatric and pregnant TB populations for initiating a pharmacogenomic-guided therapy to TB patients for improving clinical outcomes and reducing adverse drug reactions. Further studies exploring the generalizability of the available models by integrating them or systematic external evaluation could help in identifying the adaptability of models to specific populations across the globe and to facilitate implementation of available models in clinical practice.