Multidrug-resistant TB (MDR-TB), defined as tuberculosis caused by Mycobacterium tuberculosis (Mtb) isolates resistant to isoniazid and rifampicin, remains a major public health threat. According to the World Health Organization (WHO), India faces one of the highest global burdens of both TB (26%) (World Health Organization, 2020a) and MDR-TB (27%). (World Health Organization, 2020a). Linezolid, an oxazolidinone with demonstrated activity against Mtb, is recommended as one of the primary drugs for treatment of MDR-TB in shorter and longer regimens. (Schecter et al., 2010; World Health Organization, 2020b).

Linezolid has a rapid and extensive absorption after oral administration with high bioavailability (∼100%). (Qin et al., 2022). Approximately 30% is excreted unchanged through the kidneys, with non-renal clearance accounting for 65% of total clearance, resulting in a terminal half-life of 3.5–7 h. (Macgowan, 2003; Stalker and Jungbluth, 2003). Linezolid works by binding the Mtb 23S ribosomal subunit and preventing protein synthesis, (Livermore, 2003), but, unfortunately, it also affects the human mitochondrial 16 S rRNA subunit leading to mitochondrial toxicity, causing dose reductions or interruptions of intended linezolid treatment course. (Wasserman et al., 2016; Conradie et al., 2020). The significant toxicity profile and ongoing uncertainty about optimal linezolid dosing for MDR-TB confounds clinical efforts to balance efficacy, resistance suppression, and adverse events. (Wasserman et al., 2016). Previous studies have proposed a minimum concentration (C_min) threshold of 2 mg/L as a marker of adverse effects related to mitochondrial toxicity from linezolid. (Song et al., 2015).

The most widely-used linezolid dose for MDR-TB supported by trial evidence is 600 mg daily, (Conradie et al., 2022), but this frequently is reduced to 300 mg daily because of toxicity, which may fail to achieve the therapeutic target for Mtb. (Maartens and Benson, 2015). Previous studies suggest that efficacy is driven by the ratio of the area under the curve of unbound linezolid divided by minimal inhibitory concentration (ƒAUC/MIC), with a target ratio of >100 likely reach an appropriate exposure and minimize toxicity. (Bolhuis et al., 2018). Population-specific pharmacokinetic (PK) models are paramount to ensuring that patients receive the best-possible dosing to achieve the efficacy target. Previous population PK studies of linezolid for MDR-TB have been performed in patients in South Africa, (Garcia-Prats et al., 2019; Abdelwahab et al., 2021; Padmapriyadarsini et al., 2022), Italy, (Tietjen et al., 2022), Brazil, (Alghamdi et al., 2020), United States, (Alghamdi et al., 2020), and China, (Zhou et al., 2022), but despite the global burden, there are limited population PK data from India. Given possible differences in comorbidities, coadministered drugs, diet, and other unmeasured characteristics, we sought to better understand the PK of linezolid in an Indian population. Our study developed a population PK model using linezolid blood levels collected through a cohort study of Indian adolescents and adults with MDR-TB and explored the probability of target attainment (PTA) with daily linezolid treatment at doses of 600 and 300 mg.

In this study, we describe the population PK of linezolid in a cohort of Indian adults and adolescents treated for MDR-TB and evaluate dosing strategies to balance treatment efficacy and toxicity. We found that while 300 mg daily is expected to be effective against strains with MICs 0.25 mg/L or lower, 600 mg daily may be more appropriate for MIC levels of 0.5–1 mg/L, albeit with a higher likelihood of treatment-associated toxicity.

Our PK model identified one-compartment disposition with first-order elimination, and first-order absorption with transit compartments. This is consistent with previous studies on linezolid PK, which have primarily employed similar one-compartment models for both drug-susceptible TB and MDR-TB. (Alghamdi et al., 2020; Abdelwahab et al., 2021; Tietjen et al., 2022). Other studies have reported two-compartment disposition with Michaelis-Menten elimination when linezolid is dosed at 1,200 mg daily, (Imperial et al., 2022), or linezolid inhibiting its own clearance with repeated administration. (Plock et al., 2007; Mockeliunas et al., 2022). In our data, testing two-compartments disposition and saturation did not improve model fit, possibly due to differences in the sampling schedule and the fact that non-linear kinetics may not be evident at the lower doses prescribed to participants in our study. Due to the limitations of our overall study design, it was not possible to meaningfully evaluate autoinhibition, since that would require PK data from the first days of linezolid administration.

Interestingly, reports for linezolid PK in different populations show different values of clearance. The typical clearance estimated in our population, 3.81 L/h, is similar to the clearance of 3.57 l/h reported in South African patients from two separate clinical trials. (Abdelwahab et al., 2021). Higher clearance has been reported in other publications. For example, clearance was found to be 6.06 l/h in study participants from Brazil and the USA, (Alghamdi et al., 2020), 7.69 l/h in an Italian population, (Tietjen et al., 2022), and 4.59 l/h in a Chinese population. (Zhou et al., 2022). While part of the difference could be ascribed to the larger weight of the participants in the other cohorts, who were slightly heavier than those in our study and in South African studies, the difference persisted despite allometric scaling in our model. Other differences between studies that could have accounted for the variability in the parameters are the linezolid dose ranging from 300 to 600 mg once or twice daily, and the prolonged time on treatment for participants in this study. However, we did not find a significant effect of these covariates in the final model. New evidence has shown that linezolid is mainly metabolized to its inactive metabolites in the liver through oxidation of the morpholine ring by distinct isoforms CYP2J2, CYP4F2, and CYP1B1 of the Cytochrome P450 enzyme family. (Obach, 2022). This could also contribute to differences in linezolid metabolism based on differences in allele distribution frequencies between distinct populations. However, the presence of polymorphisms on these enzymes and their influence on linezolid metabolism needs to be evaluated in future studies. Importantly, we found high between-subject variability for clearance and between-occasion variability for absorption process within the Indian population studied in this cohort, suggesting a potential role for therapeutic drug monitoring to optimize linezolid treatment, though PK laboratory capacity is limited, particularly in high incidence settings.

According to our simulations, 300 mg daily would achieve a ƒAUC/MIC ≥100 in at least 97% of the simulated individuals when the MIC is ≤ 0.25 mg/L, while a dose of 600 mg daily would be more suitable for isolates with higher MICs. This dosing suggestion is similar to the one published for the South African cohort participants, (Abdelwahab et al., 2021), while higher doses have been proposed in studies finding higher clearance values. (Alghamdi et al., 2020; Mockeliunas et al., 2022). It is also important to highlight that the MIC distribution may vary between countries, and indeed we report lower MICs in this study than reported in South Africa. (Abdelwahab et al., 2021). From a toxicity perspective, our simulations suggest that 31% of people receiving 600 mg daily would exceed the literature-derived C_min threshold of 2 mg/L, (Song et al., 2015), compared to 9.6% of those receiving 300 mg daily While more data are needed to confirm the predictive value of this threshold for clinical application, particularly as competing toxicity thresholds are considered, (Wasserman et al., 2019a), the high proportion of people affected at either dose and the large between-subject variability we observed offers an argument in support of individualized therapy that considers clinical features such as fat-free mass and the extent of microbiological resistance to better identify the 31% at risk of toxicity as well as to ensure efficacy among those given lower or less frequent doses.

As TB control programs worldwide move to adopt linezolid-based treatments for MDR-TB, (Padmapriyadarsini et al., 2022; World Health Organization, 2022), it will be important to recognize that despite evidence of resistance in high-burden settings, (Tornheim et al., 2020), the majority of such treatment is prescribed in the absence of susceptibility testing for linezolid, let alone MIC testing. While infrequently discussed in the literature and rarely tested in clinical practice, linezolid resistance is identified with increasing frequency in high-burden settings where testing is performed. Previously considered to be rare, (Bharadwaj, 2021), resistance was noted to affect 1% of isolates tested in Mumbai, India in 2017, (Tornheim et al., 2017), with more recent publications documenting resistance among 6.7%, (Tornheim et al., 2020), similar to rates documented in China (6.9%). (Du et al., 2021). Among South African patients with MDR-TB and treatment failure, linezolid resistance has been reported among 33% of isolates. (Wasserman et al., 2019b). Given the increasing prescription of linezolid in shorter regimens with bedaquiline and pretomanid, with or without moxifloxacin, (World Health Organization, 2022), increased vigilance for emerging linezolid resistance is crucial to secure the success of TB elimination programs. (Bharadwaj, 2021).

Because the efficacy target for linezolid is normalized to MIC, improving local knowledge of drug resistance and MICs can help TB providers determine the MIC distributions to target with different linezolid dosing strategies, with or without assistance from PK models such as this one. Given the WHO-endorsed critical concentration of 1 mg/L, (World Health Organization, 2018), in the absence of local MIC knowledge, a dose of 600 mg would ensure a PTA ≥97% against MICs ≤0.5 mg/L, and which represented 90.5% of isolates in this study. This finding supports the results of the recently published ZeNIX trial demonstrating improved efficacy and toxicity of a 600 mg daily dose compared to other dosing strategies. (Conradie et al., 2022). While the use of higher doses such as 1,200 mg daily in the NIX-TB study may increase drug exposure, (Conradie et al., 2020), the high rates of toxicity-associated linezolid dose reduction or treatment interruption during 6 months of therapy (85%) leaves much room for improvement. Similarly, a dose of 600 mg twice-daily achieves a target attainment of 100% in simulation studies, but with >99% of the simulated individuals exceeding the safety levels. (Millard et al., 2018).

Our study had several limitations. This relatively small, single-site study may not be generalizable to non-Indian populations, populations with different rates of comorbid diseases such as HIV, diabetes, or malnutrition that affect absorption, children, or those with drug-susceptible tuberculosis with different concomitant treatments. Due to the small number of participants coinfected with HIV, we could not assess drug-drug interactions with antiretroviral therapy, which is an important factor that may influence the treatment in people living with HIV. Given that the majority of MDR-TB globally is not associated with HIV, however, our data are relevant to a large proportion of global cases. Additionally, model development was better informed by intensive than sparse PK data, which is easier to obtain in clinical settings, but is simultaneously less-informative because parameters estimation relies on a reduced number of observations within a dosing interval. For this reason, model building and assessment of PK parameters relied in an initial stage only on the intensive data, which was available for only 26% of participants. Additionally, the pre-dose concentrations from both intensive and sparse data were affected by larger variability than post-dose concentrations, likely due to self-reported dosing history. We tried to mitigate the effect of this uncertain information by allowing larger between-occasion variability for absorption parameters and bioavailability in the final model. As shown in the final visual predictive check, the model prioritized intensive data, but still described sparse data adequately. Finally, this non-interventional cohort study assessed the extent to which participants achieved literature-derived efficacy and toxicity targets. Future studies will need to evaluate the impact of model-derived treatment decisions on improved treatment outcomes and frequency of treatment-associated side effects.

To the best of our knowledge, this is the first PK model for linezolid developed in a population of Indian patients with MDR-TB. We report values of clearance similar to those reported in South African patients. These results suggest that while linezolid dosed at 300 mg daily may be effective against isolates with MICs ≤0.25 mg/L, a dose of 600 mg is more likely to achieve the efficacy target for isolates with higher or unknown MICs and improved the PTA from 60% to 97% against such isolates, representing nearly a third of samples tested in this study. The high variability in multiple important PK parameters demonstrates a role for model-based, individualized therapy to optimize linezolid exposure. Additional prospective studies are needed to confirm these findings and evaluate their role in improving both efficacy and toxicity thresholds in clinical settings.