What Is the Optimal Community-Based Tuberculosis Screening Algorithm for People Who Inject Drugs in a High-Burden Setting?

BACKGROUND

Although people who inject drugs (PWID) are a high-risk group for tuberculosis (TB), current case-finding strategies fail to identify most TB cases. There is a need for an optimized community-based algorithm to improve TB detection in such disproportionately affected populations.

METHODS

Using respondent-driven sampling, we recruited PWID at community sites in Hai Phong, Vietnam, screening for classic TB symptoms, C-reactive protein blood measurement, portable on-site chest x-ray with CAD4TB software (Computer-Aided Detection for Tuberculosis version 7; Delft Imaging Systems BV), and Xpert MTB/RIF on sputum. Any participants suspected of TB by on-site physicians were referred to the infectious disease hospital for confirmatory testing, and external experts validated final diagnoses, which were then used as the TB gold standard. We aimed to identify the screening algorithm with the highest case detection at the lowest cost among different on-site testing combinations. Ingredients-based costing was used to evaluate the cost per test and cost per case detected for each algorithm.

RESULTS

Among the 1080 PWID enrolled, 47 (4.4%; 95% CI, 2.8%-6.4%) were diagnosed with TB disease. When compared with the current symptom-based TB screening strategy in Vietnam (double D), systematic chest x-ray with CAD4TB, Xpert MTB/RIF for those with CAD4TB ≥50, and referral to care for those with either CAD4TB ≥70 or a positive Xpert test result doubled the sensitivity (82.9% vs 43.9%) and yield (3.2% vs 1.7%) while maintaining a reasonable cost per TB case detected (US $439 vs $309 for standard of care).

CONCLUSIONS

We defined an acceptable and moderate cost algorithm that improves efficiency for community-based TB screening among PWID in Vietnam. To reflect real TB prevalence, we make the case that active case finding and systematic screening strategies should not limit testing to those with a positive symptom screen.