Insights into the Role of Rifampicin Exposure and Clinical Baseline Covariates on the Response to Pulmonary Tuberculosis Treatment

Abstract

Background

Quantitative markers of tuberculosis disease burden are essential for assessing treatment response and optimizing therapeutic strategies. This study evaluated the impact of antimicrobial plasma drug levels on time-to-positivity (TTP) trajectories in patients with drug-susceptible tuberculosis and explored the relationship between bacillary clearance and treatment outcomes.

Methods

Patients with drug susceptible pulmonary tuberculosis initiating treatment were recruited in Worcester, South Africa. Weekly sputum samples were collected for 12 weeks, with bacterial load quantified using Mycobacterial Growth Indicator Tubes, yielding TTP. Nonlinear mixed-effects modeling was used to empirically describe longitudinal TTP, assessing rifampicin exposure and other participant characteristics as covariates. TTP trajectories were stratified by treatment outcomes to identify trends between change in TTP and outcome.

Results

A total of 402 participants were included, of whom 60% were male. Median age was 37 years (interquartile ratio 26–48) and 90% had successful treatment outcomes (were cured or completed treatment) at 6 months with 3% who experienced treatment failure, 2% died, 1% transferred, and 4% were lost to follow-up. Baseline smear grade and lung cavitation decreased baseline TTP. Higher rifampicin concentration area under the curve was associated with faster longitudinal change in TTP, whereas lung cavitation and older age were associated with slower longitudinal change. Furthermore, slower change in TTP during the first 12 weeks of treatment was linked to treatment failure at 6 months.

Conclusions

Our findings highlight the role of rifampicin plasma exposure in optimizing bacillary clearance and improving treatment outcomes, even within standard dosing regimens for drug-susceptible tuberculosis.

We modeled longitudinal Mycobacteria Growth Indicator Tube TTP data from drug-susceptible patients with tuberculosis. We found higher rifampicin exposure to result in faster bacillary clearance, as measured by longitudinal change in TTP, which in turn displayed a trend indicating correlation with microbiological treatment success.

Tuberculosis remains the leading cause of death from a single bacterial infection globally [1]. Despite relatively high treatment success rates, the burden of tuberculosis-related mortality and long-term post-treatment complications underscores the urgency of identifying determinants of treatment outcomes that can be modified to increase success. Sterilization of sputa correlates with end of treatment outcome as well as risk of tuberculosis relapse, making it a critical marker of treatment efficacy [2–4]. Early achievement of bacterial sterilization may indicate improved outcomes [4, 5]. Hence, understanding what factors predict sterilization is imperative.

Mycobacteria Growth Indicator Tube (MGIT) time-to-positivity (TTP) has gained prominence as a marker of sterilization, offering a quantitative inverse measure of bacterial load reduction during treatment. Traditionally, TTP is used to determine the time to culture conversion (ie, 2 consecutive tuberculosis-negative cultures are one of the outcome measures in clinical trials). However, analysis of the entire longitudinal profile of TTP across treatment provides further insight into treatment dynamics and patient response [6, 7].

Several studies have shown that higher doses of rifampicin and isoniazid increase the bacillary clearance [7–9]. This relationship has also been investigated in standard treatment for drug-susceptible tuberculosis [3, 10]. Chigutsa et al [11] found rifampicin exposure to predict sterilization activity measured by TTP in patients receiving first-line treatment. McCallum et al reported that only intrapulmonary, not plasma, exposure of rifampicin and isoniazid affected longitudinal TTP [10]. Kloprogge et al developed a pharmacokinetic–pharmacodynamic model to identify predictors of colony forming units (CFU) decline and found that higher rifampicin and isoniazid exposure correlated with greater bacillary killing [3]. The effect of several other factors and patient characteristics on TTP during tuberculosis treatment have also been investigated. People with lung cavitations have been shown to have higher bacterial burden [6, 10, 12]. It has also been reported that baseline acid-fast bacilli smear grade correlates with TTP, with smear-positive individuals having higher bacterial load and slower bacillary clearance [13, 14].

Variability in bacillary clearance has been linked to treatment outcome. Ntinginya et al [15] investigated correlations between TTP or the tuberculosis molecular bacterial load (MBL) assay and treatment outcome in participants receiving standard treatment for drug-susceptible tuberculosis or 1 of 4 experimental treatment arms [15]. They compared TTP and MBL between outcome groups by a pairwise test at each sampling time point, by time to zero bacterial load, and analyzed longitudinal TTP and MBL and investigated the effects of outcome. They found treatment outcome to have a significant effect on both measures of bacterial load in all 3 analyses, except for the modeling of longitudinal TTP. Kloprogge et al, in the study mentioned previously, found lower bacillary clearance rate, as measured by CFU, to be correlated with treatment failure [3]. Sloan et al [5] investigated the association between bacillary elimination rates, as measured by longitudinal TTP and CFU, and outcome and found slower elimination rates to correlate with unfavorable outcome.

In this study, we investigate the effect of drug exposure and microbiological and patient factors, such as age, sex, body mass index (BMI), human immunodeficiency virus (HIV) status, and alcohol use, on bacillary clearance using longitudinal TTP in patients with drug-susceptible tuberculosis. Furthermore, we explore the correlation between bacillary clearance and end of treatment outcomes.

METHODS

Study Population and Study Design

The Tuberculosis Treatment and Alcohol Use Study (NIAID R01AI119037) was a prospective cohort study recruiting patients with drug-susceptible tuberculosis in Worcester, South Africa [16]. The study was conducted in accordance with the Declaration of Helsinki and was approved by the South African Medical Research Council Human Research Ethics Committee (EC011–5/2016), University of Cape Town Human Research Ethics Committee (497/216), Stellenbosch University Human Research Ethics Committee (N17/04/039 and N18/02/024), and Boston University Medical Campus Institutional Review Board (H-34970).

Eligible participants were ≥15 years old, initiating treatment for Xpert-, culture-, and/or smear-positive tuberculosis with rifampicin susceptibility confirmed by Xpert, but without confirmed susceptibility to the remaining 3 first-line drugs. Exclusion criteria included rifampicin resistance, extrapulmonary tuberculosis, pregnancy at enrollment, or prior tuberculosis treatment within 2 years. Participants received weight-based, first-line therapy per World Health Organization guidelines [17]: 2 months of rifampicin, isoniazid, pyrazinamide, and ethambutol, followed by 4 months of rifampicin and isoniazid. Adherence was monitored by directly observed therapy (DOT) workers 5 weekdays per week [18].

Data Collection

Participant demographics were collected at baseline, including age, sex, BMI, HIV status, alcohol use measures (categorized into low, moderate, or high alcohol use as previously reported [19]), and presence of lung cavitation as well as percent of lung involved (PLI) [20] from chest radiographs.

Expectorated sputum samples were collected at baseline and first in the morning weekly until week 12 into treatment. To obtain TTP, sputum was inoculated on liquid media using BACTEC MGIT (Supplementary A). The time from inoculation to detection of tuberculosis (ie, positivity) was recorded, yielding TTP. Samples were considered tuberculosis negative if TTP was not reported within 42 days (culture negative).

Acid-fast bacilli smear grading (Ziehl Neelsen) and minimum inhibitory concentration (MIC) determination for isoniazid, rifampicin, ethambutol, and pyrazinamide were performed using baseline sputum samples. The degree of smear positivity was graded as negative, scanty, 1+, 2+, or 3+. MICs were measured using the 1% proportion method in the BACTEC MGIT 960 instrument (Supplementary B) and predicted using convolutional neural network models [21, 22] (Supplementary C). Additionally, isoniazid-resistant variants were annotated based on the second edition of the World Health Organization mutation catalog [21, 23] (Supplementary D).

Pharmacokinetic Data

A subset of 104 participants without HIV underwent pharmacokinetic evaluation 4–8 weeks into treatment, as previously reported [19]. Steady-state area under the concentration-time curve (AUC) over 24 hours (AUC_0-24) were generated for each drug based on individual Bayesian estimates of the pharmacokinetic parameters for all subjects in the pharmacokinetic substudy. For participants with no pharmacokinetic sampling, the final pharmacokinetic models were used to predict AUC_0-24 using each individual's dose and predictive covariates [19].

Correlation Between Longitudinal TTP and 6-month Clinical Outcomes

To investigate if the change in longitudinal TTP during the first 12 weeks was correlated with final treatment outcome, the raw TTP data were stratified by the treatment outcome categories defined as per the study protocol. All participants who were smear- or culture-negative in the last month of treatment and on at least 1 previous occasion, or had completed treatment but did not have bacteriologic proof of cure were classified as “cured or completed treatment.” All participants who were smear or culture positive at treatment month 5 or later, or had a change of regimen were classified as “treatment failure.” Participants who had died regardless of the reason during the treatment were classified as “died,” participants lost to follow-up were classified as “lost to follow-up,” and participants who moved to another healthcare facility during treatment were classified as “transferred.”

Modeling of TTP Data

We used nonlinear mixed-effects modeling in NONMEM version 7.5.0 to characterize the longitudinal change in TTP versus time on treatment [24]. Different empirical models were tested to fit the data. Contaminated samples were excluded from the analysis, as well as individuals who never had a culture that grew Mycobacterium tuberculosis. Participants with missing covariate data were excluded, unless a substantial proportion of the cohort lacked a specific covariate, in which case the missing values were imputed using mixture modeling. More details can be found in Supplementary E and G, Supplementary Figure 3. Baseline smear grade, lung cavitation, PLI, HIV status, age, sex, BMI, and alcohol use were tested as covariates on baseline TTP and the longitudinal change in TTP, whereas drug exposure, measured MIC, predicted MIC, and overall adherence rate were tested on the longitudinal change in TTP only.

RESULTS

Study Population and Data

Of 458 enrolled participants (May 2017–March 2024), 53 were excluded due to having contaminated or no tuberculosis growth on any samples (Supplementary G, Supplementary Figure 1). Furthermore, 26 participants were missing lung cavitation status and were imputed for the analysis. Two participants were missing baseline smear and one was missing HIV status. The median age was 37 years (interquartile range 26–48), 60% were male, and 27% had HIV (Table 1). The overall adherence rate was 84%. After excluding contaminated samples (22%), 3348 remained, of which 57% grew M. tuberculosis (Figure 1, Supplementary A).

Table 1.

Participant Characteristics (n = 402)

Characteristic	Median (IQR) or n (%)
Male	241 (60)
Age (years)	37 (26–48)
Weight (kg)	47.3 (41.6–53.7)
BMI (kg/m2)	17.7 (16.2–19.4)
People with HIV	110 (27)
People on ART at baseline yes/no/missing	34 (8)/76 (19)/292 (73)
CD4 count	239 (113–477)
Alcohol use low/moderate/high/missing	144 (36)/183 (45)/36 (9)/39 (10)
Overall adherence rate (%)	84 (63–91)
Lung cavitation yes/no/inconclusive/missing	268 (67)/99 (25)/9 (2)/26 (6)
Baseline acid-fast bacilli smear none/scanty/+/++/+++	108 (27)/52 (13)/75 (19)/61 (15)/106 (26)
Rifampicin AUC0–24 (mg/h/L)	24.1 (18.6–30.9)
Isoniazid AUC0–24 (mg/h/L)	18.6 (15.8–21.8)
Pyrazinamide AUC0–24 (mg/h/L)	304 (272–350)
Ethambutol AUC0–24 (mg/h/L)	18.7 (16.0–21.8)
Rifampicin MIC (µg/mL) 0.06/0.125/0.25/0.5/missing	109 (27)/167 (42)/78 (19)/31 (8)/17 (4)
Isoniazid MIC (µg/mL) 0.025/0.05/0.1/0.2/>0.2/missing	16 (4)/236 (59)/98 (24)/28 (7)/7 (2)/17 (4)
Pyrazinamide MIC (µg/mL) 25/50/75/100/>100/missing	171 (43)/78 (19)/23 (6)/5 (1)/2 (1)/123 (30)
Ethambutol MIC (µg/mL) 0.6/1.25/2.5/>5/missing	88 (22)/259 (64)/36 (9)/2 (1)/17 (4)
Predicted rifampicin MIC (µg/mL)	0.06 (0.06–0.06)
Predicted isoniazid MIC (µg/mL)	0.05 (0.05–0.05)
Predicted pyrazinamide MIC (µg/mL)	24.9 (24.0–26.2)
Predicted ethambutol MIC (µg/mL)	1.21 (1.17–1.24)
Percent lung involved (%)	21.6 (14.0–30.9)
Isoniazid resistance-associated gene variant yes/no/missinga	14 (3)/381 (95)/7 (2)

Abbreviations: ART, antiretroviral therapy; AUC, area under the curve; BMI, body mass index; IQR, interquartile range; MIC, minimum inhibitory concentration; WGS, whole-genome sequencing.

^aParticipants did not have WGS or WGS was of too poor quality.

Figure 1.

Proportion of tuberculosis positive, tuberculosis negative, tuberculosis positive contaminated, tuberculosis negative contaminated, and missing samples at each sampling time point. All contaminated samples were excluded for the analysis. The numbers above the x-axis represent the number of samples included in the analysis at each time point. Week 1 represents baseline TTP (ie, pretreatment). TTP, time-to-positivity.

Correlation Between Longitudinal TTP and 6-month Clinical Outcome

Of the 402 participants, 360 (90%) were cured or completed treatment, 12 (3%) had treatment failure, 8 (2%) died, 5 (1%) transferred, and 17 (4%) were lost to follow-up at the end of 6 months of treatment. Figure 2 and Supplementary Figure 7 show that 10 of 12 participants experiencing treatment failure had TTP <12.5 days at baseline and suggest that they have slower bacillary clearance compared to those who were cured or completed treatment. Individual model-predicted longitudinal change in TTP among patients with baseline TTP <12.5 days was significantly slower for participants experiencing treatment failure compared to individuals who were cured or completed treatment (P = .03) (Supplementary F, G, Supplementary Figure 8).

Figure 2.

Raw TTP data in days versus time, stratified by treatment outcome category. The category “other” includes those lost to follow-up, who died, or were transferred. The smoothed lines represent the local polynomial regression line in each category, with the shaded areas representing their standard error. The dots represent individual observations. Tuberculosis negative samples have been imputed as 42 days. TTP, time-to-positivity.

Longitudinal TTP Model and Predictors

An exponential model provided the best fit for the data (Table 2), as per Equation 1:

Table 2.

Final Parameter Estimates

Parameter	Typical Value (95% CIa)
B0TTP (days)	12.1 (11.1–13.0)
kTTP (1/week)	0.389 (0.337–0.447)
Proportional error (%)	28.8 (27.8–30.0)
Effect of lung cavitation on B0TTP (%)	−18.7 (−23.5 to −13.3)
Effect of lung cavitation on kTTP (%)	−33.2 (−38.3 to −28.6)
Effect of age on kTTP (%/10 y)	−10.2 (−14.4 to −6.28)
Effect of rifampicin AUC on kTTP (%/doubling)b	+15.6 (4.38–25.2)
Effect of smear grade 1+ on B0TTP (%)	−30.2 (−35.8 to −22.6)
Effect of smear grade 2+/3+ on B0TTP (%)	−41.6 (−45.9 to −36.9)
Between-subject variability in B0TTPc (%CV)	24.1 (21.6–27.2)
Between-subject variability in kTTPc (%CV)	54.0 (48.8–60.4)

Abbreviations: CI, confidence interval; CV, coefficient of variation.

^a95% CI obtained by sampling importance resampling using PsN software.

^bThe effect was found on log2 scale.

^cBetween-subject variability was assumed to be log-normally distributed.

$$TTP=B{0}_{TTP}\cdot e^{k_{TTP}\cdot Time}$$ (1)

where B0_TTP is baseline TTP and k_TTP is the longitudinal change in TTP. Details of modeling results (including part of the data and model code) can be found in Supplementary E, G, Supplementary Figures 2–6, and Supplementary H.

Covariates affecting B0_TTP included baseline smear grade and lung cavitation. Individuals with smear grade 1+ and 2+/3+ had 30.2% and 41.6% lower B0_TTP (ie, higher initial bacterial load) compared to individuals with scanty or no baseline smear, respectively (P < .001). A visual predictive check of the model, stratified by baseline smear grade, can be found in Figure 3. Individuals with lung cavitations had 18.7% lower B0_TTP (ie, higher initial bacterial load) (P < .001). When comparing lung cavitation and PLI as predictors of B0_TTP and k_TTP, lung cavitation performed better and was kept in the model (Supplementary E, Supplementary Table 1).

Figure 3.

Visual predictive check (VPC) (n = 1000) showing TTP in days versus time, stratified by baseline smear grade. The solid and dashed lines represent the 5th, 50th, and 95th percentiles of the observations, whereas the shaded areas represent the 95% model predicted confidence intervals for the same percentiles. The red solid lines in the upper panels represent the median predicted proportion of samples above the upper limit of quantification. The circles represent observations. TTP, time-to-positivity.

The k_TTP was significantly influenced by lung cavitation, age, and rifampicin AUC_0–24. Individuals with lung cavitation had 33.2% slower k_TTP (ie, slower decrease in bacterial load) (P < .001). Furthermore, each 10-year increase from the median age of 38 years resulted in a 10.2% slower k_TTP (P < .001). As for rifampicin AUC_0–24, each doubling of AUC_0–24 from the median of 25 mg/h/L resulted in 15.6% faster k_TTP (ie, faster decrease in bacterial load) (P = .005). This was also tested excluding the participants with imputed rifampicin AUC_0–24 and remained significant (P = .01). Adjusting rifampicin AUC_0–24 by measured or predicted MIC (AUC/MIC) rendered the effect not significant (P = .74). Although measured MIC levels did not significantly affect k_TTP for any of the drugs, each unit increase in predicted isoniazid MIC from the median of 0.05 µg/mL resulted in 13.3% slower k_TTP (P = .03). However, this effect was driven by 3 outliers with predicted MIC ≥3.29 µg/mL and was no longer significant after exclusion of these individuals. Furthermore, a trend was observed with individuals with isoniazid resistant-associated genetic variants having slower k_TTP, but the effect was not statistically significant (P = .25, Supplementary G, Supplementary Figure 4). No significant trend was found between alcohol use, sex, HIV status, and BMI or adherence rate and B0_TTP or k_TTP. Figure 4 presents model-predicted profiles for 2 participant groups: 1 with favorable baseline covariates (negative smear, absence of lung cavitation, and younger age) and 1 with unfavorable baseline covariates (positive smear, presence of lung cavitation, and older age). Rifampicin exposure varied within each group and suggests the impact of low rifampicin exposure is most pronounced among participants with unfavorable baseline covariates. The latter results in a typical predicted time to sputum conversion beyond the duration of our study, which was 12 weeks.

Figure 4.

Model-predicted profiles for 6 representative participants. The participants with favorable covariates (negative baseline smear, absence of lung cavitation, and younger age) are shown in green, whereas the participants with unfavorable covariates (high baseline smear, presence of lung cavitation, and older age) are shown in red. The solid, dotted, and dot-dashed lines represent high, low, and median rifampicin exposure, respectively. The black dashed line represents a TTP of 42 days. TTP, time-to-positivity.

DISCUSSION

We characterized the changes in TTP over the first 12 weeks of treatment in a cohort of patients with drug-susceptible tuberculosis using an empirical nonlinear mixed-effects model. Our analysis demonstrated that rifampicin plasma exposure significantly impacts the rate of bacillary clearance, as measured by longitudinal change in TTP, even within the standard dosing regimen for rifampicin-susceptible tuberculosis. Our findings also suggest that poor bacillary clearance is linked to worse treatment outcomes, even among a cohort with high treatment adherence, thus underscoring the importance of optimizing rifampicin dosing in first-line tuberculosis treatment to enhance bacillary clearance and improve clinical outcomes. Furthermore, the significant effect of baseline smear grade, age, and lung cavitation on baseline TTP and longitudinal change in TTP highlights the need for tailored monitoring in individuals with increased risk for poor treatment outcome.

There have been major efforts to identify factors that might allow for shortened treatment regimens or need for more follow-up in persons with tuberculosis disease [25, 26]. Individuals with unfavorable baseline covariates may particularly benefit from identification of low rifampicin exposure leading to slow bacillary clearance and delayed sputum conversion, which in turn associates with poor overall treatment outcome. Ongoing investigations into increasing rifampicin dosing from the currently recommendation of 10 mg/kg are expected to contribute to improved bacillary clearance rates and earlier sputum conversion amongst patients with unfavorable baseline covariates [8, 9, 27, 28].

The association between rifampicin exposure and longitudinal change in TTP aligns with prior studies suggesting a dose-dependent response to rifampicin. This strengthens the findings by Chigutsa et al in a study within standard treatment for drug-susceptible tuberculosis using TTP as the pharmacodynamic endpoint [6]. McCallum et al did not find rifampicin plasma concentrations to affect longitudinal TTP in drug-susceptible patients with tuberculosis, possibly because of the sparser sampling schedule used in their study, both for sputum and pharmacokinetic plasma sampling [10]. Although AUC/MIC has been suggested as the primary endpoint for the relationship between antituberculosis drug exposure and bacillary killing [29, 30], recent studies found AUC alone to be a better predictor of bacillary clearance [3, 9, 10]. Consistent with these findings, our results suggest that rifampicin AUC is a better predictor of bacillary clearance than AUC/MIC. This calls into question the long-standing assumption that AUC/MIC serves as the primary pharmacokinetic–pharmacodynamic endpoint for rifampicin [9, 10]. Furthermore, no significant relationships were found between measured MIC and longitudinal change in TTP. In contrast, predicted isoniazid MIC showed a significant effect on the longitudinal change in TTP, although this effect was driven by individuals with predicted MIC values greater than the cutoff for resistance. The validity of this trend remains unproven because these resistance patterns could not be experimentally confirmed.

Baseline smear grade and lung cavitation were significant predictors of baseline TTP and longitudinal change in TTP, supporting the association of these parameters with an increased bacterial load found in previous studies [6, 10, 12–14]. The slower longitudinal change in TTP observed in older participants highlights the importance of age as a determinant of treatment dynamics. Although it is well known that treatment outcome is worsened with age, previous studies investigating the effect of age on TTP trajectories during different treatment regimens found no significant trends, possibly from the less dense sputum sampling schedules used in those studies [12, 13, 31].

The observed trend of individuals with isoniazid resistant-associated gene variants having slower bacillary clearance is expected because these individuals would have lower efficacy of isoniazid, 1 of the cornerstones of drug-susceptible tuberculosis treatment. However, because of the low number of individuals with such gene variants in this cohort, no conclusion can be drawn. Furthermore, HIV status was not found to affect baseline TTP or longitudinal change in TTP, potentially reflecting limited statistical power from the relatively small number of participants with HIV. However, consistent with our findings, other studies have also reported no significant effect of HIV status on longitudinal TTP [6, 13]. Finally, we did not find any effect of alcohol use in our study, possibly because few individuals have high alcohol use in this cohort. Furthermore, the poor outcomes observed in other studies in individuals consuming alcohol could be due to poor adherence to treatment [32, 33]. This was challenging to investigate in our cohort, as adherence was recorded by DOTS workers, resulting in generally high adherence rates among participants.

Pharmacodynamic modeling is a powerful tool for achieving deeper insights from bacterial load data in tuberculosis treatment. Unlike traditional endpoints, such as time to culture conversion, which provides a binary outcome at a single time point, longitudinal modeling of TTP enables more detailed characterization of treatment response dynamics. It offers greater statistical power compared to modeling time to culture conversion, which discards most of the data, and more rigorous testing of predictors of treatment response. This is particularly valuable for optimizing dosing regimens, as demonstrated by studies showing the impact of drug exposure on bacillary clearance.

A key limitation of our study is using an empirical model, which characterizes TTP data without incorporating mechanistic interpretation. Consequently, while we infer associations between TTP, bacterial load, and bacillary clearance, these interpretations rely on assumptions rather than direct mechanistic evidence. The relationship between TTP and bacterial load is complex, as highlighted by Maitra et al [34], and should be considered when interpreting our findings. Another limitation is the exclusion of tuberculosis-negative samples that were followed by tuberculosis-positive samples. A potential alternative approach to address these data could involve incorporating a component describing the probability of bacterial presence in a sample, as demonstrated by Svensson et al [35]. Additionally, although we observed a trend between bacillary clearance and treatment outcomes, predicting individualized treatment durations was limited by the uniform 6-month regimen in our study. The low prevalence of treatment failure in our study further restricted robust quantification of correlations between baseline factors, rifampicin exposure, and treatment outcomes. Finally, we did not have pharmacokinetic data for all participants included in the analysis and thus imputed drug exposures for those individuals. However, we used individual characteristics that significantly impact drug exposure in our imputation, and the resulting AUCs were similar to those of the cohort for which we had pharmacokinetic data.

In conclusion, we show that rifampicin plasma exposure, along with other previously confirmed predictors, impacts bacillary clearance in drug-susceptible patients with tuberculosis as measured by changes in MGIT TTP. We also identified a trend between longitudinal change in TTP and final treatment outcome, highlighting the importance of early and dynamic monitoring of bacterial load. These findings support the value of individualized treatment approaches, such as dose optimization of rifampicin, to enhance treatment efficacy and potentially reduce the duration of therapy. Further research is warranted to explore their implications for global tuberculosis treatment guidelines.

Supplementary Data

Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.