We described the pharmacokinetics and safety of levofloxacin among 11 women receiving treatment for RR-TB during the second and third trimesters of pregnancy and the postpartum period. Other published pharmacokinetic data on the use of the newer-generation fluoroquinolones during pregnancy are limited to one case report of a woman with RR-TB and three single-dose exposure studies (total of 48 women without tuberculosis or HIV) [13]. Two of the studies measured levofloxacin concentrations after a single oral dose of levofloxacin 500 mg given 2 h before amniocentesis (n = 10) [22] and after an intravenous infusion of levofloxacin 500 mg given 25 min before planned Caesarean delivery (n = 12) [23]. The mean ± standard deviation plasma concentration of levofloxacin was 3.95 ± 0.77 µg/mL when measured at a single timepoint during amniocentesis (at 16–20 weeks’ gestation) and 8.18 ± 1.68 µg/mL at the time of clamping the umbilical cord during Caesarean delivery (median gestational age 38 weeks). Fluoroquinolones exhibit concentration-dependent bactericidal activity, and these single-point levofloxacin concentrations, which are lower than the median levofloxacin Cmax measured at steady state in our cohort, reflect the limitations of single-dose pharmacokinetic studies where steady state is not achieved.

Early pharmacokinetic data from 60 pregnant and 30 lactating women receiving either ciprofloxacin, pefloxacin, or ofloxacin found lower serum quinolone concentrations among women in the second trimester of pregnancy than among lactating women [24]. More recently, a population pharmacokinetic model based on pharmacokinetic sampling over 24 h from 47 South African women with RR-TB demonstrated that renal clearance of levofloxacin during pregnancy may be increased by up to 38.1% above non-pregnant concentrations [25]. Levofloxacin is cleared by renal elimination [7], so the physiological changes of pregnancy, particularly higher blood flow to the kidneys resulting in increased glomerular filtration, may explain this variation. Levofloxacin AUC0–12 and AUC0–24 in our cohort were lower, with higher drug clearance, during both trimesters of pregnancy, with a more notable difference in the second trimester, than at 2–8 weeks postpartum. These differences may be due in part to variations in glomerular filtration, which peaks in mid-pregnancy [26]; this is supported by the observed renal clearance estimates at the different sampling points in our study.

In our cohort, the median (Q1–Q3) levofloxacin AUC0–24 of 90.5 µg·h/mL (82.3–98.6) and 111.9 µg·h/mL (90.8–121.5) in the second and third trimesters, respectively, were comparable to the AUC0–24 range of 77.3–295.9 µg·h/mL reported in a pharmacokinetic study of 10 non-pregnant adults with tuberculosis who had received levofloxacin 1000 mg daily monotherapy over 7 days [27]. Other pharmacokinetic parameters from the same study by Peloquin et al. [27], such as tmax of 1 h, a median t½ of 7.37 h, and a Cmax range of 9–43 µg/mL, were also similar to those observed in our cohort; however, we reported a shorter half-life for levofloxacin among participants sampled during the second trimester of pregnancy.

The minimum Cmax target we used for this study was based on results of non-compartmental pharmacokinetic analyses from the Tuberculosis Trials Consortium/National Institute of Allergy and Infectious Diseases Opti-Q trial (a phase II randomised dose-finding trial), in which adults receiving therapy for RR-TB were randomised to four levofloxacin doses of 11, 14, 17, or 20 mg/kg/day with an optimised background regimen [18]. We opted for the lower limit of the Cmax range (6.90–21.03 µg/mL) observed in 26 non-pregnant patients receiving levofloxacin doses of either 750 mg or 1000 mg daily (i.e., 14 mg/kg/day) [18], as this reflected the range of doses received by participants in our cohort. Cmax concentrations in our cohort were mostly above the minimum target of 7 µg/mL, except for one patient, who received a levofloxacin dose of > 15 mg/kg and the measured Cmax during the third trimester (4.3 µg/mL) and postpartum (4.7 µg/mL) were considerably lower than expected. This discrepancy was attributed to potential poor adherence to treatment and lower pre-dose drug concentrations on intensive sampling days, highlighting the study limitation that direct observation of study drug administration and adherence support were not required to be implemented by the research teams. Nevertheless, median Cmax and AUC0–24 values measured in pregnancy and postpartum in our study were within the ranges reported in the Opti-Q trial (AUC0–24 70–248 µg·h/mL for the group receiving 14 mg/kg/day) [18], despite the demonstrated interindividual variability in levofloxacin exposures in our cohort.

The established pharmacokinetic–pharmacodynamic relationship between the AUC, minimum inhibitory concentration (MIC), and anti-mycobacterial response of levofloxacin must be considered when determining the clinical relevance of lower drug exposures during pregnancy. The AUC0–24/MIC ratio is a robust predictor of fluoroquinolone efficacy, favourable clinical outcomes, and acquisition of mycobacterial drug resistance [28, 29]. In a hollow fibre system model described by Deshpande et al. [28], an AUC0–24/MIC ratio of 146 was identified as the optimal target exposure for achieving at least 80% maximal microbial kill, and a higher dose of levofloxacin ≥ 20 mg/kg/day was considered necessary to achieve suppression of acquired levofloxacin resistance. Similarly, a prospective pharmacokinetic study of 23 patients with RR-TB who received levofloxacin 750–1000 mg daily (equivalent to 11–14 mg/kg/day) found that one-third of patients did not attain target drug exposures at these doses. Only one-half of the patients with MICs ≥ 1 mg/L achieved optimal target exposures, prompting suggestions for higher levofloxacin doses (17–20 mg/kg/day) for these individuals [29]. Furthermore, pharmacokinetic modelling studies have consistently suggested that adults with RR-TB may require higher levofloxacin doses for maximum M. tuberculosis kill [30,31,32,33]. These findings underscore concerns regarding potentially suboptimal treatment efficacy and increased risk for M. tuberculosis isolates to acquire further drug resistance [34], especially among people with lower than expected drug exposures with currently recommended fluoroquinolone doses. This scenario is particularly relevant during pregnancy, when drug exposures may be lower because of the increased volume of distribution and drug clearance. Our study was not designed to assess the MIC of levofloxacin and RR-TB treatment outcomes. However, our findings of lower levofloxacin exposures in mid-pregnancy, although modest and among only four women, still warrant further research into the optimal dosing of this drug and clinical implications of lower exposures during pregnancy.

The limited sample size of our cohort meant that a direct comparison of levofloxacin exposures between participants with or without HIV (all on protease inhibitor-based ART regimens) was not feasible. The presence and magnitude of potential drug–drug interactions between fluoroquinolones and antiretrovirals remain poorly understood, and our study did not address these. Although ritonavir may potentially diminish moxifloxacin concentrations, a clinically significant interaction with levofloxacin is considered unlikely [35]. Nonetheless, the clinical implications of potential differences in second-line tuberculosis drug exposures between people without HIV and those with HIV and on ART can change during pregnancy [36] and remains a topic for further research.

The Opti-Q study findings demonstrated that escalating levofloxacin doses increased drug exposures and increase the likelihood of successful RR-TB treatment outcomes in adults [18]; this benefit must be weighed against the potential for an increased risk of adverse effects. Fluoroquinolones generally exhibit good tolerability over short dosing durations in pregnancy among people without RR-TB. Levofloxacin use has been linked to gastrointestinal disturbances, arthropathies, tendonitis, photosensitivity, and central nervous system toxicity, including seizures, headaches, dizziness, sleep disorders, and hallucinations [37, 38]. AEs are common among adults receiving treatment for RR-TB [39], and establishing causality to a specific drug within a multidrug treatment regimen can be challenging, particularly when such events may also be attributed to HIV, pregnancy complications, and other comorbid conditions or concomitant medications. However, levofloxacin is relatively well tolerated compared with most other second-line antituberculosis drugs in adults [40], and none of the severe or life-threatening maternal AEs observed in our cohort were assessed by the site investigators and protocol team as attributable to levofloxacin during or after RR-TB treatment.

Fluoroquinolones can cross the foetal–placental barrier and reach the foetal compartment, albeit with low observed levels of in utero foetal drug exposure [23]. Case reports have raised concerns regarding teratogenicity in animals and potential risks of impaired foetal cartilage development and erosion with prolonged in utero levofloxacin exposure [37, 41,42,43]. However, these concerns have not been supported by subsequent clinical studies of children receiving fluoroquinolones for extended periods [44]. A systematic review and meta-analysis involving 2821 pregnant women exposed to fluoroquinolones at any stage of pregnancy for any indication found no significant increase in major anomalies or adverse pregnancy outcomes compared with unexposed pregnant women [42]. However, a notable decrease in live birth rates and an increased rate of elective terminations, possibly influenced by perceived teratogenicity risks, was observed. Another meta-analysis corroborated these findings, revealing no significant association between quinolone use and risk of birth defects, stillbirths, preterm births, or low birth weight [43]. Many of the studies that met inclusion criteria in both meta-analyses focused on the early stages of pregnancy, with relatively short durations of drug exposure compared with the minimum 6-month treatment duration of therapy for RR-TB. The observation of low birth weight in over half of the infants in our cohort is consistent with findings from an observational study among 108 pregnant women treated for RR-TB in a South African province, where in utero exposure to levofloxacin and bedaquiline emerged as significant, independent predictors of low birth weight [45]. These findings underscore the need for integrated pregnancy and tuberculosis treatment registries, alongside coordinated surveillance systems to identify and address adverse pregnancy outcomes among such cohorts, particularly in South African and analogous settings.

Our study has several strengths and limitations. This was one of very few pharmacokinetic studies assessing levofloxacin for the treatment of RR-TB during pregnancy and the postpartum period. Pregnant women enrolled in the RR-TB treatment arm of the IMPAACT P1026s study were monitored throughout pregnancy and postpartum, with regular clinical evaluation of pregnant and postpartum women and their infants during and after exposure to levofloxacin. However, enrolment of participants with RR-TB was slower than for other arms of the study (which enrolled pregnant women with drug-susceptible tuberculosis and/or HIV) because of the relatively lower disease prevalence and the inherent challenges in timing enrolment with the limited duration of drug exposure during pregnancy. At the time of study closure, the enrolment target for this arm had not been reached, and the modest number of participants reduced the study’s power to detect significant differences in pharmacokinetic parameters during pregnancy and postpartum. Additionally, incomplete pharmacokinetic sampling further decreased the sample size for within-participant comparison between trimesters. In addition, 24-h sampling was not carried out despite once-daily dosing of levofloxacin. All participants received generic versions of levofloxacin from different suppliers over different time periods, and inconsistencies in drug formulations could theoretically have affected pharmacokinetic findings. However, this is unlikely as 10 of 11 participants were enrolled in South Africa, where all generic medicines are required to meet strict regulatory standards to ensure bioequivalence to their brand name counterparts.

Selection of participants was biased towards those who could tolerate levofloxacin, and drug dosages were prescribed by routine healthcare providers and therefore not standardised to a specific mg/kg dose for all participants. Our study was not designed to assess pharmacodynamic parameters because levofloxacin MICs were not available, and RR-TB treatment outcomes were not recorded. Infant and breastmilk pharmacokinetic sampling were not conducted to assess levofloxacin passage to neonates, and drug–drug interactions between levofloxacin and other concomitant medications were not assessed in this cohort. Safety assessment was limited by the extended intervals between study visits and reliance on clinical information from routine medical records. Determining the causality of AEs was challenging because of the multiple comorbidities and many concomitant antituberculosis and antiretroviral medications. We assessed potential relationships of AEs to specific study drugs but did not actively collect comprehensive data on other attributable causes or on the discontinuation or reintroduction of antituberculosis drugs in response to AEs.

In conclusion, levofloxacin appeared well tolerated despite the high number of unrelated AEs in this small cohort of women treated for RR-TB during pregnancy. Although levofloxacin exposures were lower in the second trimester than in the postpartum period, Cmax and AUC were within the ranges reported among non-pregnant adults with RR-TB. Registration and monitoring of pregnant women with RR-TB, and regular reporting of pregnancy outcomes and neonatal and infant health, alongside RR-TB treatment outcomes, must be implemented through routine tuberculosis surveillance and integrated reporting systems. Further exploration of levofloxacin exposures during pregnancy, as well as maternal–infant drug transfer of levofloxacin, both in utero and through breastfeeding, is warranted.