Early Microbiologic Markers of Pulmonary Tuberculosis Treatment Outcomes

Abstract

Rationale

Earlier biomarkers of pulmonary tuberculosis (PTB) treatment outcomes are critical to monitor shortened anti-TB treatment (ATT).

Objectives

To identify early microbiologic markers of unfavorable TB treatment outcomes.

Methods

We performed a subanalysis of 2 prospective TB cohort studies conducted from 2013 to 2019 in India. We included participants aged ⩾18 years who initiated 6-month ATT for clinically or microbiologically diagnosed drug-sensitive PTB and completed at least one follow-up visit. Sputum specimens were subjected to a baseline Xpert Mycobacterium tuberculosis/rifampin (MTB/RIF) assay, acid-fast bacilli (AFB) microscopy and liquid and solid cultures, and serial AFB microscopy and liquid and solid cultures at weeks 2, 4, and 8. Poisson regression was used to assess the impact of available microbiologic markers (test positivity, smear grade, time to detection, and time to conversion) on a composite outcome of failure, recurrence, or death by 18 months after the end of treatment. Models were adjusted for age, sex, nutritional status, diabetes, smoking, alcohol consumption, and regimen type.

Results

Among 1,098 eligible cases, there were 251 (22%) adverse TB treatment outcomes: 127 (51%) treatment failures, 73 (29%) recurrences, and 51 (20%) deaths. The primary outcome was independently associated with the Xpert MTB/RIF assay (medium-positive adjusted incidence rate ratio [aIRR], 1.91; 95% confidence interval [CI], 1.07–3.40; high-positive aIRR, 2.51; 95% CI, 1.41–4.46), positive AFB smear (aIRR, 1.48; 95% CI, 1.06–2.06), and positive liquid culture (aIRR, 1.98; 95% CI, 1.21–3.23) at baseline; Week 2 positive liquid culture (aIRR, 1.47; 95% CI, 1.04–2.09); and Week 8 positive AFB smear (aIRR, 1.63; 95% CI, 1.06–2.50) and positive liquid culture (aIRR, 1.54; 95% CI, 1.07–2.22). There was no evidence of Mycobacterium tuberculosis growth in the Mycobacterium Growth Indicator Tube at Week 4 conferring a higher risk of adverse outcomes (aIRR, 1.25; 95% CI, 0.89–1.75).

Conclusions

Our analysis identifies Week 2 respiratory mycobacterial culture as the earliest microbiologic marker of unfavorable PTB treatment outcomes.

Tuberculosis (TB), caused by Mycobacterium tuberculosis (Mtb), affected an estimated 10.6 million people worldwide in 2021 and caused 1.6 million deaths (1). Optimizing prevention, early microbiologic detection, and prompt anti-TB treatment (ATT) are key components of the global TB control strategy (2–4). Despite these efforts, decreases in TB incidence and mortality have fallen short of global milestones to end the TB epidemic by 2030 (2, 4–7). The reasons are multifactorial. With respect to TB treatment, long and complicated traditional 6-month regimens are linked to poor patient compliance, which may contribute to increased adverse treatment outcomes, infectiousness, transmission, and development of drug resistance (8, 9).

To improve the real-world impact of ATT, multiple global clinical trials are investigating 3- to 4-month courses of potent antimycobacterial agents (10–14). If these are successful, shortened regimens could improve compliance and ease the logistics of directly observed therapy programs, which are inconsistently implemented (8, 9, 15). However, updated treatment monitoring strategies that reliably predict TB treatment outcomes are critical to advance this research (16).

For years, clinical trials and national TB programs have used acid-fast bacilli (AFB) microscopy and mycobacterial culture conversion at the end of the intensive phase (Week 8 of ATT) to predict relapse-free TB cure (17, 18). These respiratory microbiologic markers leverage widely available technology and require minimally invasive specimens, and culture remains the gold standard for Mtb detection. However, a period of 6–8 weeks is required to confirm Mtb growth, rendering culture conversion at Week 8 an outdated marker for shortened regimens (18, 19). Furthermore, even though AFB microscopy provides rapid results, evidence suggests that neither marker may consistently predict TB treatment outcomes (20, 21).

Numerous blood-based biomarker discovery studies are under way, but none are ready for implementation, and data are lacking on respiratory microbiologic markers before Week 8. To address this data gap, we conducted a subanalysis among adult participants of 2 longitudinal TB cohort studies in India (22–24). The parent studies enrolled new clinically or microbiologically diagnosed drug-susceptible PTB cases and assessed factors associated with a composite unfavorable TB treatment outcome (failure, recurrence, or death). We aimed to evaluate all potential respiratory microbiologic markers (AFB smear, culture, and polymerase chain reaction–based tests assessed at baseline and serially during the intensive phase of ATT) of adverse treatment outcomes. Identifying an earlier marker of poor treatment outcome would define a primary endpoint for treatment-shortening trials and provide a treatment monitoring strategy to potentially optimize outcomes (16, 19, 25).

Methods

The Cohort for TB Research by the Indo-U.S. Medical Partnership (CTRIUMPh) and the Impact of Diabetes on TB Treatment Outcomes (TBDM) study are prospective TB cohort studies conducted in western India between 2013 and 2019 at Byramjee Jeejeebhoy Government Medical College, Dr. D.Y. Patil University, and the National Institute for Research in Tuberculosis (22–24). The primary objective was to assess host and mycobacterial factors associated with a composite unfavorable TB treatment outcome (failure, recurrence, or death); this subanalysis aimed to assess all available respiratory microbiologic markers from baseline through Week 8 of ATT.

CTRIUMPh and TBDM followed similar protocols with respect to TB case recruitment, follow-up, and outcomes assessed. Both recruited patients with new clinically diagnosed (based on clinical or radiologic features) or microbiologically confirmed (Xpert MTB/RIF assay, AFB microscopy, or culture) drug-sensitive PTB; TBDM excluded people living with HIV. Participants initiated 6-month ATT (thrice-weekly directly observed therapy before 2017 and daily regimens after 2017) within 1 week of enrollment. Follow-up during ATT was at Weeks 2, 4, 8, 20, and 24 (TBDM included additional visits at Weeks 6, 12, and 16) and continued for as long as 18 months after the end of treatment; a TB event visit was scheduled if treatment failure or recurrence was suspected. Relevant sociodemographic, epidemiologic, clinical, and laboratory data were collected at baseline and follow-up visits. The participant-reported adherence to ATT during the follow-up visits was evaluated in the CTRIUMPh cohort as the total number of doses missed among the total doses expected to be taken. At each visit, 2 sputum specimens were collected for microbiologic analysis.

This subanalysis was restricted to adult participants ⩾18 years of age with at least one follow-up visit. We pooled data for all respiratory microbiologic markers assessed at baseline (Xpert MTB/RIF assay, AFB microscopy, and liquid and solid culture) and serially during the intensive phase of ATT (AFB microscopy and liquid and solid culture at Weeks 2, 4, and 8). These data included Xpert positivity grade (high, medium, or low positivity or negative) and cycle threshold; qualitative AFB smear, World Health Organization (WHO) smear grade (26), and time to conversion (TTC) from positive to negative AFB smear; and qualitative culture, colony count (solid culture only), time to detection of growth (TTD), and TTC from positive to negative culture.

Outcomes and Definitions

The primary outcome was a composite of treatment failure (Mtb growth on liquid or solid culture or clinical TB diagnosis from Week 20 to Week 24), recurrence (Mtb growth on liquid and/or solid culture or clinical TB diagnosis after successful ATT completion, defined as no TB symptoms and no microbiologic evidence of Mtb at Week 24), or all-cause mortality by 18 months after the end of treatment (23, 24, 27). If treatment failure or recurrence preceded death, the former was assigned as the treatment outcome. Mortality was further classified as TB-related or unlikely due to TB based on independent review by at least 2 clinician reviewers, and discordant cases were adjudicated by a third clinician. The risk factors known to be associated with the negative outcomes measured were baseline age, sex, body mass index (BMI) (28, 29), diabetes mellitus (DM), tobacco use, alcohol use, and treatment regimen type (daily or thrice weekly).

Laboratory Methods

One sputum specimen was directly analyzed using the Xpert MTB/RIF assay (Cepheid) to detect and quantify Mtb DNA (30) and Ziehl-Neelsen smear microscopy to detect and quantify AFB (26). The second specimen was decontaminated and homogenized using the N-acetyl-L-homocysteine sodium hydroxide method, and the sediment was inoculated on liquid Mycobacterium Growth Indicator Tube (MGIT) and solid Löwenstein-Jensen (LJ) culture media (the culture contamination rate did not exceed 6% at any study visit); the Global Laboratory Initiative protocol was followed for MGIT cultures, and LJ cultures were followed for 8 weeks before they were considered negative (26). Drug susceptibility testing for first-line anti-TB drugs was performed when a culture was positive at baseline or a TB event visit (31).

Statistical Analysis

The sample size for the study was adjusted for a 10% loss to follow-up at the design stage. The participants who were lost were considered as censored, and their follow-up data for as long as they participated in the study were used in the analysis. Participants lost to follow-up during TB treatment (incomplete treatment, two consecutive missed follow-up visits, or not reachable [22–24]) were censored for the duration of follow-up; those lost to follow-up after the end of treatment were included if a primary outcome was assigned. Covariates were summarized as proportions or medians with IQRs and compared by primary outcome status using the Fisher’s exact test and Wilcoxon rank-sum test, respectively. The relative change in baseline TTD was calculated at Weeks 2, 4, and 8 as (TTD_{time point} – TTD_baseline) / TTD_baseline. TTC by AFB smear and culture was estimated using the Kaplan-Meier product limit estimator and compared using the log-rank test.

Poisson regression analysis was used to evaluate baseline and longitudinal markers associated with the primary outcome, and random-effects Poisson regression was used to evaluate the serial change in TTD. Models adjusted for measured confounders; Week 2 cultures included additional adjustment for initial bacterial burden (i.e., baseline AFB smear grade). Receiver operating characteristic (ROC) curves were used to assess the accuracy of baseline and early longitudinal markers (AFB smear and culture at Weeks 2 and 4) as predictors of the primary outcome; areas under the ROC curve (AUCs) were evaluated for each marker. All adjusted models included variables that were known to have association with, or showed an increased risk of, unfavorable TB treatment outcomes.

The primary analysis was repeated with TB-related death substituted for all-cause mortality in the primary outcome definition. Additionally, we performed subgroup analyses restricted to: 1) MGIT culture–confirmed, isoniazid-sensitive Mtb; 2) microbiologically confirmed drug-sensitive PTB (positive AFB smear, Xpert MTB/RIF assay, or culture); and 3) MGIT culture–confirmed Mtb irrespective of other drug resistance status.

Statistical methods for these analyses are tabulated in Table E1 in the data supplement. All P values were 2-sided, and statistical significance was evaluated at the 0.05 α-level. Data were analyzed using Stata version 14.2 (StataCorp).

We performed predictive modeling at each predetermined follow-up time (Weeks 2, 4, 8, 12, 16, 20, and 24) after the initiation of treatment. In the modeling, each microbiological marker was examined independently to assess sensitivity and specificity of predicting an unfavorable TB treatment outcome. The original data of 1,098 participants were randomly divided into 2 parts: the first two thirds constituted the discovery cohort and the remaining third was the validation cohort. All models were adjusted for age, sex, alcohol use, BMI, daily versus thrice-weekly AKT, and lung cavitation.

Ethical Approval

The CTRIUMPh and TBDM studies were approved by the institutional review board or ethics committee at each participating institution: Johns Hopkins University (Baltimore, Maryland), Byramjee Jeejeebhoy Government Medical College (Pune, India), Dr. D.Y. Patil University (Pune, India), and the National Institute for Research in Tuberculosis (Chennai, India). All participants provided written informed consent.

Results

Study Population and Characteristics

Of 1,143 adult PTB cases enrolled in the parent studies, we excluded 45 for the following reasons: multidrug-resistant TB (n = 24), no follow-up data (n = 20), and no sputum microbiology results available (n = 1). The remaining 1,098 (96%) were included in this analysis; 1,003 and 95 had microbiologically confirmed and clinically diagnosed PTB, respectively (Figure 1). Baseline median age was 35 years (IQR, 25–48 yr), the prevalence of DM was 32% (50% treated with metformin), the prevalence of HIV was 3%, and there were 251 (23%) primary outcomes, including 127 (51%) treatment failures, 73 (29%) TB recurrence events, and 51 (20%) deaths (35 considered TB-related). Compared with those with a favorable outcome, cases with the primary outcome were more likely to be in male participants (75% vs. 64%; P = 0.002), undernourished participants (69% vs. 52%; P < 0.001), ever-smokers (28% vs. 19%; P = 0.004), those who reported alcohol use (48% vs. 31%; P < 0.0001), and those who received intermittent ATT (90% vs. 80%; P < 0.0001), but were less likely in participants with DM (27% vs. 33%; P = 0.01) (Table 1). The overall participant-reported adherence assessed in one of our cohorts was >91% (n = 433).

Figure 1.

Study flow of adults with pulmonary tuberculosis enrolled in the parent cohort studies conducted in India from 2013 to 2019. The subanalysis included 1,098 participants with drug-sensitive microbiologically confirmed or clinically diagnosed pulmonary tuberculosis who completed at least one follow-up visit. Microbiologic confirmation was defined by an Xpert MTB/RIF assay, acid-fast bacilli microscopy, liquid Mycobacterium Growth Indicator Tube culture, or solid Löwenstein-Jensen culture positivity. AFB = acid-fast bacilli; DSTB = drug-sensitive tuberculosis; LJ = Löwenstein-Jensen; MDR = multidrug-resistant; MGIT = Mycobacterium Growth Indicator Tube; MTB = Mycobacterium tuberculosis; RIF = rifampicin; TB = tuberculosis.

Table 1.

Baseline characteristics of adult pulmonary TB cases enrolled in 2 prospective cohort studies from 2013 to 2019 in India

Characteristic	Overall (N = 1,098)	Unfavorable Treatment Outcome*
		Yes (n = 251)	No (n = 847)
Median age, yr (IQR)	35 (25–48)	36 (25–49)	34 (25–48)
Age group
18–44 yr	730 (66)	164 (22)	566 (78)
⩾45 yr	368 (33)	87 (24)	281 (76)
Sex
Male	732 (67)	188 (75)	544 (64)
Female	366 (33)	63 (25)	303 (36)
Body mass index†
<18.5 kg/m2	615 (56)	172 (69)	443 (52)
18.5–25 kg/m2	410 (37)	68 (27)	342 (40)
>25 kg/m2	70 (6)	10 (14)	60 (7)
HIV status†
Negative	1,050 (97)	239 (97)	811 (97)
Positive	29 (3)	8 (3)	21 (3)
Diabetes mellitus‡
No	748 (68)	184 (73)	564 (67)
Yes	350 (32)	67 (27)	273 (33)
Ever smoker†
No	829 (79)	175 (72)	654 (81)
Yes	218 (21)	67 (28)	151 (9)
Alcohol consumption†
No	713 (65)	130 (52)	583 (69)
Yes	383 (35)	121 (48)	262 (31)
TB regimen
Thrice weekly	902 (82)	227 (90)	675 (80)
Daily	196 (18)	24 (10)	172 (20)

Definition of abbreviation: TB = tuberculosis.

Data presented as n (%) unless otherwise indicated.

^*Composite of treatment failure, recurrence or all-cause mortality by 18 mo after the end of treatment.

^† Summarized among cases for which data were available: body mass index (n = 1,095), HIV status (n = 1,079), ever smoker (n = 1,047), and alcohol consumption (n = 1,096).

^‡ Defined as prior diagnosis, glycated hemoglobin ⩾6.5%, fasting blood glucose ⩾126 mg/dl, or random blood glucose ⩾200 mg/dl.

Primary Outcome

Baseline respiratory microbiologic markers

Most participants had Xpert assay (93%), AFB microscopy (99%), and liquid (95%) and solid (98%) culture data available at baseline (Figure 1). In multivariable analyses, each adjusted for baseline age, sex, BMI, DM, smoking, alcohol consumption, and regimen type (daily or thrice weekly), the primary outcome was associated separately with Xpert positivity grade (medium-positive adjusted incidence rate ratio [aIRR], 1.91; 95% confidence interval [CI], 1.07–3.40; high-positive aIRR, 2.51; 95% CI, 1.41–4.46), Xpert cycle threshold (aIRR, 0.96; 95% CI, 0.93–0.99), positive AFB smear (aIRR, 1.48; 95% CI, 1.06–2.06), and positive liquid culture (aIRR, 1.98, 95% CI, 1.21–3.23). Median TTD was 8.5 (IQR, 6.3–12.8) days on MGIT; a TTD <8 days was not associated with the primary outcome (Figure 2A and Table E2).

Figure 2.

(A) Forest plot summarizing associations between baseline respiratory microbiologic markers and the composite unfavorable tuberculosis (TB) treatment outcome. Independent predictors were baseline Xpert MTB/RIF assay, acid-fast bacilli smear, and liquid MGIT culture positivity. Results are presented as adjusted incidence rate ratios from Poisson regression models adjusted for baseline age, sex, body mass index, diabetes, smoking, alcohol consumption, and regimen type (daily or thrice-weekly anti-TB treatment). (B) Forest plot summarizing associations between longitudinal respiratory microbiologic markers and the composite unfavorable TB treatment outcome. Independent predictors were Week 2 liquid culture positivity, Week 8 acid-fast bacilli smear and liquid and solid Löwenstein-Jensen culture positivity, and change in baseline time to detection on liquid culture at Week 8. Results are presented as adjusted incidence rate ratios from Poisson regression models adjusted for baseline age, sex, body mass index, diabetes, smoking, alcohol consumption, and regimen type (daily or thrice-weekly anti-TB treatment). aIRR = adjusted incidence rate ratio; CI = confidence interval; LJ = Löwenstein-Jensen; MGIT = Mycobacterium Growth Indicator Tube; MTB = Mycobacterium tuberculosis; RIF = rifampicin.

Longitudinal microbiologic markers

Availability of AFB smear and culture data ranged from 86% to 95% at Week 2, from 79% to 87% at Week 4, and from 76% to 86% at Week 8 (Table E3). Multivariable analysis identified Week 2 MGIT culture as the only early marker associated with the primary outcome (aIRR, 1.47; 95% CI, 1.04–2.09). There was no evidence that Mtb growth at Week 4 conferred a higher risk (aIRR, 1.25; 95% CI, 0.89–1.75) (Figure 2B); however, on univariate analysis, Week 4 MGIT culture was associated with adverse outcomes (adjusted IRR, 1.42; 95% CI, 1.06–1.89). At Week 8, the primary outcome was associated with a positive AFB smear result (aIRR, 1.61; 95% CI, 1.05–2.46), positive liquid culture (aIRR, 1.54; 95% CI, 1.07–2.22), and positive solid culture (aIRR, 1.75; 95% CI, 1.10–2.76) (Figure 2B).

Figure 3 shows serial TTD from baseline to Week 8 on liquid and solid culture media according to TB treatment outcome. The relative increase in baseline TTD on MGIT culture at Week 8 was associated with the primary outcome (aIRR, 0.71; 95% CI, 0.67–0.91) (Figure 2B). A 1-day increase in TTD reduced the risk of the primary outcome by 4% (aIRR, 0.96; 95% CI, 0.93–0.99) (see Table E3).

Figure 3.

Longitudinal time to detection of growth of Mycobacterium tuberculosis on (A) liquid MGIT and (B) solid LJ culture media. Box plots summarize time to detection of growth at each time point from baseline to the end of the intensive phase of treatment according to tuberculosis treatment outcome. LJ = Löwenstein-Jensen; MGIT = Mycobacterium Growth Indicator Tube.

Overall, median TTCs were 3.3 weeks (IQR, 2.1–5.3) for AFB smear and 5.3 weeks (IQR, 3.3–8.4) and 3.7 weeks (IQR, 2.1–6) for liquid and solid culture, respectively. TTC was significantly longer in cases with the primary outcome than without for MGIT (5.7 wk vs. 5.3 wk; log-rank P = 0.003) and LJ culture (3.9 wk vs. 3.7 wk; log-rank P = 0.03). TTC was not associated with the primary outcome in multivariable models (see Table E3).

ROC curves plotted for early longitudinal markers are shown in Figure 4. AUC values were similar between AFB smear and liquid culture at Week 2 (AFB smear, 0.68; 95% CI, 0.64–0.72; MGIT, 0.67; 95% CI, 0.63–0.71) and Week 4 (AFB smear, 0.68; 95% CI, 0.64–0.72; MGIT, 0.67; 95% CI, 0.63–0.71), and these were similar to the Week 8 AUC values. The sensitivity values of MGIT culture were 40.2% (95% CI, 33.9–46.7%) and 55.3% (95% CI, 47.9–62.6%) at Weeks 2 and 4, respectively (Table E4). Additionally, the ROC analysis for baseline covariates (including age, sex, BMI, alcohol consumption, DM, smoking, ATT regimen type, and cavitation on chest radiography) was performed without baseline AFB smear and MGIT (AUC, 0.596), with baseline smear (AUC, 0.627), and with baseline MGIT (AUC, 0.631), which demonstrated the added value of markers over and above baseline covariates.

Figure 4.

ROC curves illustrate similar performance among Week 2 (A and B) and Week 4 (C and D) acid-fast bacilli smear microscopy and liquid MGIT cultures as markers of the composite unfavorable tuberculosis treatment outcome. MGIT = Mycobacterium Growth Indicator Tube; ROC = receiver operating characteristic.

Sensitivity Analysis

In repeat analysis restricted to MGIT culture–confirmed PTB cases (n = 909) (Tables E5 and E6), the aIRR between early MGIT culture (Weeks 2 and 4) and the primary outcome still showed an increased risk, but this did not reach statistical significance in this subset (aIRRs, 1.35 and 1.20; 95% CIs, 0.93–1.95 and 0.86–1.68, respectively) (see Table E6). Additionally, Week 2 liquid culture remained independently associated with the primary outcome after excluding 95 cases without microbiologically confirmed PTB (adjusted odds ratio, 1.57; 95% CI, 1.13–2.18) and 71 cases with isoniazid-resistant Mtb (adjusted odds ratio, 1.49; 95% CI, 1.07–2.08). All results were similar when TB-related death was substituted for all-cause mortality in the primary outcome definition (data not shown). The predictive modeling showed poor sensitivity, specificity, and AUC for all baseline and early microbiological markers of adverse treatment outcomes. The Week 2 MGIT culture showed 45% (95% CI, 26–64%) sensitivity and 32% (95% CI, 23–42%) specificity, with an AUC of 61% (95% CI, 51–72%). The detailed prediction analysis is shown in Table E7.

Discussion

This subanalysis of respiratory microbiological markers from 2 prospective PTB cohorts in a setting of a high TB burden identified that early mycobacterial liquid culture is independently associated with unfavorable TB treatment outcomes. A positive culture identified as early as 2 weeks after ATT initiation was associated with a 1.5-fold increased risk of treatment failure, recurrence, or death. Week 4 culture showed a 25% higher relative association with the primary outcome, but this was not statistically significant in the multivariable analysis. However, the univariate analysis shows an association of Week 4 cultures with adverse treatment outcomes and is therefore likely to be a clinically relevant finding. Overall, our study adds to the evidence describing pretreatment respiratory microbiological markers as strong predictors of TB treatment outcomes and provides new data to support early respiratory culture during treatment as a potential marker of poor outcomes.

To our knowledge, our analysis is among the first to demonstrate the ability of early respiratory culture, as early as 2 weeks after treatment initiation, as a marker of TB treatment outcomes. Another study by Phillips and coworkers evaluated smear and culture results during treatment, but this assessment was performed for markers from Week 6 and onward (32, 33).

Key longitudinal observations support our main findings, including the consistent association between culture positivity and the primary outcome across the intensive phase of ATT and the inverse relationship between TTD and the primary outcome (34). Importantly, Week 2 and Week 4 liquid culture demonstrate high AUCs in ROC analysis, indicating their potential as independent markers of WHO-defined unfavorable outcomes. Moreover, the risk of adverse outcomes remained unchanged in the sensitivity analysis. However, the prediction modeling analysis showed poor sensitivity and specificity for all models, particularly for the early markers of interest. Taken together, these analyses imply that the presence of certain markers indicates an increased risk of adverse treatment outcomes; however, these markers cannot be used as predictors of those outcomes.

Finally, our study corroborates existing evidence that baseline respiratory microbiologic markers are strong predictors of TB treatment outcomes (17, 35, 36). Consistent with previous reports, the Xpert MTB/RIF cycle threshold was associated with TB treatment outcome in our study. Consequently, we also found that moderate and high Xpert positivity independently predicted the primary outcome, reflecting the significance of a high initial bacterial load. These findings provide evidence that pretreatment Xpert positivity grading may be useful to identify patients at high risk in whom closer monitoring is warranted during TB treatment, particularly in settings in which mycobacterial culture analysis is not routinely available (17, 37).

A major strength of our study was the availability of biweekly AFB smear and culture data across 2 multisite PTB cohorts in the country with the world’s largest absolute TB burden. The parent studies did have minor differences with respect to sample size and schedule of evaluations (23, 24). In particular, Week 6 AFB smear and culture data were available for only the TBDM cohort. Additionally, some risk of bias may exist as a result of the longitudinal cohort study design, as data were censored for loss to follow-up and/or unavailable microbiological data. Despite these limitations, CTRIUMPh and TBDM had large sample sizes; apart from Week 6, all evaluations of microbiologic markers during the intensive phase were similar, and both studies defined endpoints according to WHO guidelines.

In conclusion, this comprehensive analysis of respiratory microbiologic markers during the intensive phase of ATT suggests that liquid mycobacterial culture at Week 2 may be the earliest biomarker of PTB treatment outcomes. Although several weeks may be required to detect the growth of Mtb, collectively, these findings suggest that early respiratory culture may warrant consideration as a marker of adverse TB treatment outcomes until a more optimal biomarker emerges. Defining an early microbiologic marker of adverse treatment outcomes may help advance scientific efforts to optimize ATT by shortening TB treatment duration (38–40). In addition, our findings may inform national and international TB treatment monitoring strategies to identify individuals at risk for unfavorable outcomes and design appropriate intervention (4, 5). Further studies are warranted to confirm our findings, including evaluation among patients receiving shortened TB treatment regimens.