Interpretation of serial interferon-gamma test results to measure new tuberculosis infection among household contacts in Zambia and South Africa

BACKGROUND: A more stringent QuantiFERON-TB Gold In-Tube (QFT) conversion (from negative to positive) definition has been proposed to allow more definite detection of recent tuberculosis (TB) infection. We explored alternative conversion definitions to assist the interpretation of serial QFT results and estimate incidence of TB infection in a large cohort study. METHODS: We used QFT serial results from TB household contacts aged ≥15 years, collected at baseline and during two follow-up visits (2006–2011) as part of a cohort study in 24 communities in Zambia and South Africa (SA). Conversion rates using the manufacturers’ definition (interferon-gamma (IFN-g) < 0.35 to ≥0.35, ‘def1’) were compared with stricter definitions (IFN-g < 0.2 to ≥0.7 IU/ml, ‘def2’; IFN-g < 0.2 to ≥1.05 IU/ml, ‘def3’; IFN-g < 0.2 to ≥1.4 IU/ml, ‘def4’). Poisson regression was used for analysis. RESULTS: One thousand three hundred sixty-five individuals in Zambia and 822 in SA had QFT results available. Among HIV-negative individuals, the QFT conversion rate was 27.4 per 100 person-years (CI:22.9–32.6) using def1, 19.0 using def2 (CI:15.2–23.7), 14.7 using def3 (CI:11.5–18.8), and 12.0 using def4 (CI:9.2–15.7). Relative differences across def1-def4 were similar in Zambia and SA. Using def1, conversion was less likely if HIV positive not on antiretroviral treatment compared to HIV negative (aRR = 0.7, 95%CI = 0.4–0.9), in analysis including both countries. The same direction of associations were found using def 2–4. CONCLUSION: High conversion rates were found even with the strictest definition, indicating high incidence of TB infection among household contacts of TB patients in these communities. The trade-off between sensitivity and specificity using different thresholds of QFT conversion remains unknown due to the absence of a reference standard. However, we identified boundaries within which an appropriate definition might fall, and our strictest definition plausibly has high specificity.