Background: Culture-negative TB comprises 20% of TB cases in settings where TB cultures are routinely performed. In resource-limited settings, cultures are obtained less frequently, and the proportion of culture-negative TB is often much higher. Although acid fast bacillus (AFB) smear-negative TB is associated with increased mortality in HIV+ persons, there are few data on mortality risk of culture-negative TB.
Methods: We performed an observational cohort study of HIV+ adults treated for TB with standard therapy (2-month initiation phase of isoniazid, rifampin, pyrazinamide +/- ethambutol + continuation phase of isoniazid + rifampin) at or after their first HIV clinic visit. Persons were excluded if date of TB treatment relative to HAART initiation was unknown. Patients were enrolled in 2000-2013 from Brazil, Peru, Argentina, Chile, Honduras, and Mexico. Kaplan-Meier curves and Cox proportional hazards models of time from TB diagnosis to death stratified by study site were fit. For the Cox model, missing data were multiply imputed.
Results: 635 TB patients met inclusion criteria, of whom 535 had known AFB smear status (265 (50%) smear-negative) and 428 had known culture status (137 (32%) culture-negative). Median age was 36 years; 76% were male, 71% had any pulmonary TB, 56% had any extrapulmonary TB. Median CD4 count at TB diagnosis was 107 (IQR: 41-235) and 526 (83%) received concurrent HAART and TB treatment. Of the 635 patients, 139 (22%) died: 36/137 (26%) culture-negative vs. 47/291 (16%) culture-positive. The Kaplan-Meier curve of time to death by culture status is in the Figure. There was no significant difference in time to death according to AFB smear status (P=0.64). In a multivariable Cox model of all 635 patients adjusted for age, sex, site of TB disease, CD4 count, and timing of HAART initiation relative to TB treatment, persons who were culture-negative had a significantly increased risk of death (HR=1.61; 95% CI: 1.09,2.38; P=0.02). There were 12 episodes of TB recurrence occurring >180 days after initiation of TB treatment; recurrence occurred more frequently in culture-negative compared to culture-positive persons (log-rank P=0.05).
Conclusions: In this cohort, culture-negative TB was associated with a 61% increased risk of death compared to those with culture-confirmed TB. These findings raise the possibility that persons diagnosed with culture-negative TB may not have had TB, and died of other causes. This underscores the importance of accurate TB diagnosis in HIV+ persons.