Predictors of treatment outcomes in adults with drug-sensitive Tuberculosis in Maharashtra, India: A retrospective study

This retrospective study of over 320,000 adult drug-sensitive tuberculosis patients in Maharashtra, India, utilized programmatic data to identify that increasing age, male gender, lower body weight, comorbidities (HIV and diabetes), and substance use are significant predictors of both unfavorable treatment outcomes and mortality.

The Gist

Imagine the fight against Tuberculosis (TB) in India as a massive, complex marathon. The country has a digital "scoreboard" called Ni-kshay, which records every runner (patient), their starting line, their pace, and whether they finished the race.

This study is like a team of data detectives who took a giant slice of that scoreboard—specifically looking at runners in the state of Maharashtra during 2021 and 2022. Their goal wasn't to treat the runners, but to look at the data and answer a simple question: "What clues in the runner's profile tell us they might stumble, drop out, or not finish the race?"

Here is what they found, broken down into everyday concepts:

1. The "Finish Line" Definitions

The researchers looked at two different ways a race can go wrong:

• The "Unfavourable Outcome" Race: This is like a runner who doesn't cross the finish line successfully. They might have died, given up (lost to follow-up), failed the treatment, or had to change their running shoes (regimen change) mid-race.
• The "Mortality" Race: This is a stricter look, focusing only on the runners who didn't make it because they passed away.

2. The "Risk Factors" (The Clues)

The detectives found that certain "backpacks" or "handicaps" made it much harder for runners to succeed. If a runner had these items in their backpack, they were statistically more likely to have a bad outcome:

• The "Age" Backpack: Being older was a heavy burden. Runners over 60 were significantly more likely to struggle. In fact, for the "Mortality" race, being over 60 made the risk of not finishing eight times higher than for younger runners.
• The "Gender" Backpack: Men were more likely to have trouble finishing than women. Interestingly, the transgender community faced an even steeper hill, with risks more than double that of the general population.
• The "Weight" Backpack: This was a major finding. Think of body weight as the runner's fuel tank. If a runner was very light (under 40 kg), their tank was nearly empty. The lighter they were, the higher the risk. The lightest runners had the highest odds of not finishing.
• The "Comorbidity" Backpack (HIV & Diabetes): Carrying these extra conditions was like running with a heavy anchor. Having HIV or Diabetes significantly increased the chance of a bad outcome.
• The "Habits" Backpack (Tobacco & Alcohol): Smoking and drinking were like running with a limp. Both habits increased the risk of failure.
• The "Unknown" Backpack (The Mystery): This was a surprising discovery. Runners where the data said "We don't know if they smoke/drink/have diabetes" were actually more at risk than those where we knew the answer. The researchers suggest this isn't because "unknown" is dangerous, but because it means the medical team missed a chance to check. It's like a runner showing up to the start line without a medical checkup; the system didn't catch the problem early.

3. The "Extra-Pulmonary" Twist

The study looked at where the TB was hiding in the body.

• Lungs (Pulmonary): The most common spot.
• Elsewhere (Extra-pulmonary): When TB hides in other organs.
• The Twist: Usually, runners with TB in other organs seemed to do better at finishing the race generally. However, if they did die, they were slightly more likely to die than those with lung TB. It's a complex mix: they often survive, but if the situation turns fatal, it's very serious.

4. The "Data Quality" Lesson

The paper emphasizes that the "Unknown" category is a warning sign. It's like a coach saying, "If we don't know if our runner has a heart condition, we can't help them properly." The study suggests that filling in these blanks is just as important as the treatment itself.

5. The "Map" for the Future

Finally, the authors didn't just list the problems; they drew a new map for how to read the scoreboard in the future. They created a Standardized Reporting Framework.

• Analogy: Before, everyone might have read the scoreboard differently (some counting runners by name, others by ID, some ignoring missing data). This paper says, "Let's all agree to use the same ruler and the same counting method."
• They propose a checklist for anyone else who wants to analyze this data, ensuring that future studies are comparing apples to apples, not apples to oranges.

Summary

In short, this paper used a massive digital record book to find that older age, being male, having low body weight, and having HIV or diabetes are the biggest red flags for TB treatment failure in Maharashtra. It also highlighted that missing information (not knowing a patient's habits or health status) is itself a huge risk factor, likely because it means the patient wasn't screened properly. The study provides a clear, repeatable recipe for how to analyze these records so health officials can spot the runners who need extra help before they stumble.

Technical Summary

Technical Summary: Predictors of Treatment Outcomes in Drug-Sensitive Tuberculosis in Maharashtra, India

Problem Statement
Tuberculosis (TB) remains the leading infectious cause of death globally, with India accounting for approximately one-fourth of global cases. Despite the National Strategic Plan (NSP) and the National TB Elimination Programme (NTEP), a significant gap remains in achieving TB elimination targets, evidenced by a case fatality ratio (CFR) of 11% in India in 2024. While the Ni-kshay digital platform provides rich programmatic data on the continuum of care, there is a need for standardized analytical approaches to utilize this data for identifying predictors of unfavourable treatment outcomes (death, failure, loss to follow-up, regimen change, or not evaluated) and mortality. This study addresses the lack of a reproducible framework for analyzing routine programmatic data to stratify risk among adult drug-sensitive TB (DS-TB) patients.

Methodology
This retrospective observational study utilized routinely collected programmatic data from the Ni-kshay platform, specifically the Notification and Comorbidity registers, for the state of Maharashtra, India, covering the years 2021 and 2022.

• Study Population: The analysis focused on adult patients (≥15 years) diagnosed with drug-sensitive TB (DS-TB) who had a documented treatment initiation date. Drug-resistant TB cases and those without treatment initiation were excluded.
• Data Processing: Records were merged using the "instance" variable. Cases with missing weight data were excluded from the primary analysis, while other missing variables were categorized as "unknown" to maintain representativeness.
• Outcome Definitions:
  • Treatment Success: Defined as "cured" or "treatment completed."
  • Unfavourable Outcomes: Defined as "died," "loss to follow-up," "failure," "treatment regimen changed," or "not evaluated."
  • Mortality: Defined strictly as "died."
• Statistical Analysis: Univariate and multivariate logistic regression models were employed using STATA version 19. Adjusted Odds Ratios (aOR) were calculated for key risk factors, controlling for age, gender, and weight.
• Risk Factors Analyzed: Demographic variables (age, gender), anthropometric data (weight), disease site (pulmonary vs. extra-pulmonary), clinical comorbidities (HIV, diabetes), and behavioral risks (tobacco and alcohol consumption).
• Robustness Checks: Four sensitivity analyses were conducted, including the creation of a composite substance use variable, comparison of complete vs. missing weight cases, multiple imputation for missing weight data, and exclusion of "not evaluated" cases from the unfavourable outcome cohort.

Key Results
The final cohort included 323,124 cases for unfavourable outcome analysis and 315,579 cases for mortality analysis.

• Demographic Predictors:
  • Age: Advanced age was a strong predictor. Patients aged ≥60 years had the highest risk, with an aOR of 3.85 for unfavourable outcomes and 8.06 for mortality compared to younger cohorts.
  • Gender: Males had elevated risks (aOR 1.50 for unfavourable outcomes; 1.47 for mortality). Transgender individuals demonstrated substantially higher risks (aOR 2.52 for unfavourable outcomes; 2.63 for mortality).
• Comorbidities and Behavioral Risks:
  • HIV and Diabetes: Known HIV status (aOR 2.84 for unfavourable outcomes; 3.40 for mortality) and diabetes (aOR 1.43 for unfavourable outcomes; 1.56 for mortality) were significant predictors.
  • Substance Use: Both tobacco and alcohol consumption were associated with increased risks. Notably, "unknown" status for these factors also emerged as a significant risk category (e.g., unknown HIV status aOR 3.48 for mortality), suggesting data gaps correlate with poor outcomes.
  • Weight: A clear dose-response relationship was observed between low body weight and adverse outcomes. Patients weighing <20 kg had the highest risk (aOR 3.29 for unfavourable outcomes; 5.34 for mortality).
• Disease Site: Extra-pulmonary TB (EPTB) was associated with lower odds of unfavourable outcomes overall (aOR 0.95) but conferred a slightly increased risk of mortality (aOR 1.10).
• Sensitivity Analysis: Results remained consistent across multiple imputation methods and when "not evaluated" cases were excluded, confirming that missing weight data did not substantially bias the findings.

Key Contributions

1. Standardized Analytical Framework: The study proposes a reproducible protocol for merging and analyzing Ni-kshay data, specifically detailing the use of the "instance" variable for register merging and handling of missing data.
2. Identification of "Unknown" as a Risk Factor: The research highlights that the "unknown" status for comorbidities and behavioral risks is itself a significant predictor of poor outcomes, pointing to systemic data entry gaps or missed screening opportunities rather than just clinical absence.
3. Weight and Age Synergy: The study quantifies the compounding risk of advanced age and low body weight, reinforcing the need for differentiated care strategies that prioritize nutritional assessment and monitoring for these high-risk groups.
4. Reporting Framework: A standardized reporting framework (Table 3) is proposed to guide future researchers and program managers in conducting inferential analyses using TB programmatic data, addressing inconsistencies in previous studies regarding case counting and identifier usage.

Significance and Claims
The authors claim that this study represents one of the first large-scale attempts to apply a standardized analytical protocol to programmatic data for retrospective analysis of DS-TB outcomes in a high-burden setting. The significance lies in demonstrating the utility of routine programmatic data not just for surveillance, but as a decision-support tool for identifying high-risk patients.

The paper asserts that the findings support the integration of regular anthropometric assessments and risk stratification into the National Guidance on Differentiated TB Care. It suggests that patients with substance use, low weight, or "unknown" comorbidity status could be flagged within the Ni-kshay system for enhanced adherence support and clinical review. The authors maintain a modest tone, noting that while the study confirms known risk factors, the specific contribution is the methodological rigor and the identification of data completeness itself as a critical determinant of patient outcomes. The study concludes that future work should expand this framework to other states and incorporate advanced predictive modeling to enable real-time risk stratification.