A comparative study of diagnostic performance for tuberculosis and rifampicin- resistant tuberculosis between standard and pooled sputum methods using GeneXpert Ultra at Lampang Hospital, Thailand

Abstract

Background: Molecular diagnosis using the Xpert MTB/RIF Ultra assay enables rapid detection of Mycobacterium tuberculosis and rifampicin resistance.

Objectives: This study evaluated the diagnostic performance and cost efficiency of a two-sample pooled testing approach compared with standard individual testing.

Materials and methods: A cross-sectional diagnostic study was conducted among 3,504 presumptive TB patients. Sputum specimens were tested by Xpert Ultra individually, with those results serving as the reference standard, and in two-sample pools. The pooling protocol utilized 2.0 mL from each specimen to maximize load volume. Deconvolution, which required retesting both individual specimens, was mandatory for all positive and invalid pooled results. Performance metrics, RIF resistance concordance, and cost-effectiveness modeled on cartridge consumption and direct cost 550 Thai Baht per unit were compared.

Results: The pooled method showed excellent concordance with the individual method, yielding 240 concordant positives and 3,249 concordant negatives, with no statistically significant difference in MTB detection (McNemar χ²=0.267, p=0.605). The pooled approach achieved a sensitivity of 96.39% (95% CI, 93.25-98.33) and a specificity of 99.82% (95% CI, 99.60-99.93). The assay maintained reliable detection of RIF resistance, indicating that pooling did not compromise molecular accuracy. In practice, 1,752 pooled runs were performed, with 105 (5.7%) error results and 246 positive pools requiring deconvolution. Including repeats and deconvolution, total cartridge use was 2,349 compared with 3,504 for individual testing, corresponding to an actual cost reduction of 32.96% (approx. 180,000 Thai Baht saved per 1,000 tests).

Conclusion: Two-sample pooled Xpert Ultra testing demonstrated high diagnostic accuracy and doubled analytical throughput. Although deconvolution limited cost savings to approximately 33%, the strategy proved highly cost-effective and operationally feasible. This method offers a practical, scalable approach for optimizing molecular TB diagnostics and resource utilization, especially in low- to moderate-prevalence settings.