The role of new T-cell–based blood tests for tuberculosis in the diagnosis of active tuberculosis is unclear.

To compare the performance of 2 interferon-γ assays and tuberculin skin testing in adults with suspected tuberculosis.

Prospective study conducted in routine practice.

2 urban hospitals in the United Kingdom.

389 adults, predominantly of South Asian and black ethnicity, with moderate to high clinical suspicion of active tuberculosis.

Tuberculin skin testing, the enzyme-linked immunospot assay (ELISpot) incorporating early secretory antigenic target-6 and culture filtrate protein-10 (standard ELISpot), and ELISpot incorporating a novel antigen, Rv3879c (ELISpot^PLUS) were performed during diagnostic assessment by independent persons who were blinded to results of the other test.

Sensitivity, specificity, predictive values, and likelihood ratios.

194 patients had a final diagnosis of active tuberculosis, of which 79% were culture-confirmed. Sensitivity for culture confirmed and highly probable tuberculosis was 89% (95% CI, 84% to 93%) with ELISpot^PLUS, 85% (CI, 79% to 90%) with standard ELISpot, 79% (CI, 72% to 85%) with 15-mm threshold tuberculin skin testing, and 83% (CI, 77% to 89%) with stratified thresholds of 15 and 10 mm in vaccinated and unvaccinated patients, respectively. The ELISpot^PLUS assay was more sensitive than tuberculin skin testing with 15-mm cutoff points (P = 0.01) but not with stratified cutoff points (P = 0.10). The ELISpot^PLUS assay had 4% higher diagnostic sensitivity than standard ELISpot (P = 0.02). Combined sensitivity of ELISpot^PLUS and tuberculin skin testing was 99% (CI, 95% to 100%), conferring a negative likelihood ratio of 0.02 (CI, 0 to 0.06) when both test results were negative.

Local standards for tuberculin skin testing differed from others used internationally. The study sample included few immunosuppressed patients.

The ELISpot^PLUS assay is more sensitive than standard ELISpot and, when used in combination with tuberculin skin testing, enables rapid exclusion of active infection in patients with moderate to high pretest probability of tuberculosis.