Lab-Based Frailty Score Predicts Poor TB Outcomes in Older Adults
Key Takeaways
- Using the study’s prespecified outcome definition, 119 of 627 older adults with DS-PTB treated at a tertiary academic medical center in the Republic of Korea (19%) had an unfavorable outcome; death counted only during treatment or within 30 days after it, and 30 later deaths were not included in either unfavorable outcome or treatment success. Within that definition, unfavorable outcomes became more common with higher FI-Lab.
- Higher baseline FI-Lab in the highest quartile was significantly associated with greater adjusted odds of death during treatment, treatment failure, or treatment discontinuation than the lowest quartile.
- Treatment discontinuation accounted for much of the FI-Lab gradient, while treatment failure was rare.
- FI-Lab remained associated with unfavorable outcomes in time-to-event and sensitivity analyses and improved risk classification beyond age, sex, and comorbidity burden.
In Park and colleagues’ retrospective cohort of FI-Lab in older adults with drug-susceptible pulmonary tuberculosis at Jeonbuk National University Hospital in the Republic of Korea, investigators used de-identified electronic health record data to study adults aged 65 years or older with drug-susceptible pulmonary tuberculosis (DS-PTB) treated between January 1, 2015, and December 31, 2024, who started anti-tuberculosis therapy with at least three agents including isoniazid and rifampicin within 14 days of a pulmonary tuberculosis diagnostic code entry and had at least 180 days of follow-up. Using the FI-Lab methods and composite outcome definition, they calculated FI-Lab from 23 laboratory variables collected within 60 days before to 7 days after treatment initiation by dividing out-of-range deficits by the number of measured items; at least 16 of 23 items were required, and the value nearest the index date was used when more than one result was available.
The primary outcome was the authors’ composite of all-cause death during treatment or within 30 days after it ended, treatment failure (>360 days on at least 2 anti-tuberculosis drugs without a 60-day interruption), or treatment discontinuation (<150 concurrent dual-drug days or a treatment gap of at least 60 days). Patients whose treatment was still ongoing at the study end date were excluded from the primary analysis. Models adjusted for age, sex, body mass index (BMI), Rx-Risk comorbidity score, insurance type, residence area, treatment delay, selected baseline medications, cavitary disease, and acid-fast bacillus (AFB) smear positivity, with missing data handled by multiple imputation by chained equations (MICE), alongside Cox, discrimination, sensitivity, and subgroup analyses.
Unfavorable outcomes became more common across FI-Lab quartiles, with treatment discontinuation driving much of the gradient and treatment failure remaining uncommon. In fully adjusted analysis, the highest quartile was associated with greater odds of the composite outcome than the lowest quartile, with OR 3.42 (95% CI 1.75–6.69; p<0.001; p-trend=0.0002). FI-Lab also retained a continuous association with risk, with OR 1.30 (95% CI 1.12–1.50; p<0.001) per 0.1-unit increase, and time-to-event modeling showed a similar pattern with HR 3.28 (95% CI 1.60–6.73; p=0.001). Spline modeling supported a linear dose-response relationship across the observed FI-Lab range.
Adding FI-Lab to age, sex, and Rx-Risk score improved model performance with ΔAUC +0.047, NRI 0.342, and IDI 0.026. Results were also consistent across prespecified sensitivity analyses, no significant subgroup interaction was detected, and estimates were less precise in smaller strata. Because this was a retrospective single-center study at a tertiary academic medical center in the Republic of Korea, the findings are associative rather than causal and may not generalize beyond similar Korean referral settings or beyond adults aged 65 years or older. Possible incident-versus-retreatment misclassification remained because records before 2015 and care delivered outside the institution were unavailable, deaths were counted only when captured by or reported to the hospital, and isoniazid mono-resistant cases could not be fully excluded. Residual confounding from frailty trajectory, nutritional status, and functional status also remains possible, and the laboratory data could not fully distinguish pre-existing frailty from tuberculosis-related abnormalities that may change after treatment begins.
The authors concluded that higher baseline FI-Lab was independently associated with unfavorable treatment outcomes in older adults with DS-PTB and may offer a practical way to quantify frailty from routine laboratory data around treatment initiation. Prospective studies with serial FI-Lab measurement are still needed to clarify how this signal behaves over the course of tuberculosis therapy.
Clinician Questions
Which older adults with drug-susceptible pulmonary tuberculosis were included in the FI-Lab cohort?
Adults aged 65 years or older at Jeonbuk National University Hospital in the Republic of Korea were included if they began at least three anti-tuberculosis drugs, including isoniazid and rifampicin, within 14 days of a pulmonary tuberculosis diagnostic code and had at least 180 days of follow-up. The analysis excluded drug-resistant tuberculosis, nontuberculous mycobacterial disease, insufficient baseline laboratory data to calculate FI-Lab, and prior tuberculosis treatment identified during the washout period.
What did FI-Lab add beyond age, sex, and Rx-Risk score in older adults with drug-susceptible pulmonary tuberculosis?
Adding FI-Lab improved risk discrimination and reclassification beyond age, sex, and prescription-based comorbidity burden measured by the Rx-Risk score. The authors interpreted that increment as evidence that FI-Lab may capture physiologic reserve not fully reflected by medication-derived comorbidity measures.
Could acute tuberculosis-related laboratory abnormalities explain the FI-Lab signal?
The association persisted when investigators removed five inflammation-sensitive items from FI-Lab and after adjustment for cavitary disease and AFB smear positivity, which argued against the signal being driven only by acute inflammatory burden. Even so, the retrospective dataset could not fully separate long-standing frailty from reversible laboratory changes related to active tuberculosis or its early treatment course.
How far do these FI-Lab findings extend beyond the Korean tertiary-center cohort?
The analysis was limited to adults aged 65 years or older treated at a single tertiary academic medical center in the Republic of Korea, so applicability to younger patients, primary care or community settings, and regions with different tuberculosis burden or demographic profiles remains uncertain. The authors called for prospective multicenter validation before extending the findings more broadly.