Prediction of the risk of active tuberculosis in HIV-infection with an interferon-γ release assay

Abstract Background Globally, tuberculosis remains the main cause of death in HIV-infected people. The spread of HIV/AIDS cannot be curbed without effective tuberculosis control strategies. Targeted treatment of latent tuberculosis infection is cost effective, but requires accurate diagnostic methods. QuantiFERON-TB Gold (QFT), an interferon-γ release assay, is promising, but its predictive value in HIV-infected patients is uncertain. We followed up a cohort of HIV-infected people to establish the incidence of tuberculosis by linking to a national tuberculosis database, and determined the predictive value of QFT. Methods We did a prospective 5-year cohort study in one medical centre and one regional hospital in Taiwan. HIV-infected adults attending outpatient clinics, from Jan 1, 2006, to Jan 31, 2010, were invited to participate after excluding active tuberculosis through absence of symptoms and a normal result of chest radiography. We used a questionnaire to obtain demographic information and information on past exposure to tuberculosis, previous tuberculosis disease, previous vaccination, HIV risk factors, CD4 cell counts, and HIV viral load testing within the past 3 months. At study entry, blood was taken for the QFT test. The cohort was followed up every 3 months, until development of active tuberculosis disease, death, or censoring on Dec 31, 2012. Cases of incident active tuberculosis were ascertained by linking participants to the national tuberculosis database registry. Patients were offered treatment if they tested positive. We used Kaplan-Meier survival analysis to explore risk factors. Multivariate Cox proportional hazards regression models were used to estimate hazard ratios for development of incident active tuberculosis, and were adjusted for age and baseline CD4 cell count. Findings We enrolled 772 HIV-infected adults with a mean age of 36·8 years (SD 9·0), mostly men (744 [96·4%]), with a median CD4 cell count of 460 cells per μL (IQR 329–634), and a median log plasma HIV viral load of 3·40 copies per mL (IQR 1·74–3·97). HIV risk factors included being injecting drug users (517 [67%]), men who have sex with men (187 [24·2%]), and heterosexuals (62 [8%]). Almost a third (254 [32·9%]) were receiving antiretroviral treatment at study entry. Subsequently, 431 (60·3%) were placed on treatment during the study. The QFT was positive in 90 participants (11·7% [95% CI 9·5–14·0]) and indeterminate in 31 (4·0% [2·7–5·7]). 17 incident active tuberculosis cases (incidence rate 0·44 per 100 person-years) were recorded during 3843 person-years of follow-up (median 5·21 person-years), with two (11·8%) of 17 participants dying during the study. The median CD4 cell count at diagnosis of tuberculosis was 308 cells per μL (IQR 56–531). One patient had extrapulmonary tuberculosis and tested QFT negative. We found that people with a positive QFT had a five-fold risk (adjusted hazard ratio [HR] 5·02 [95% CI 1·75–14·42]; p=0·003) of developing active tuberculosis after ad justment for age, CD4 cell count, and HIV viral load. Individuals with an HIV viral load greater than or equal to 100 000 copies per mL had an eight-fold risk of developing active tuberculosis (adjusted HR 8·69 [1·21–62·23]; p=0·03). The negative predictive value of the QFT was 98·5% (95% CI 97·2–99·3). Interpretation A high negative predictive value of QFT for incident tuberculosis confirmed the safety of not treating those with negative tests. QFT accurately identifies patients who might benefit from treatment of latent tuberculosis infection, which is an important step towards halting the dual epidemic. Funding This study was supported by grants from Kaohsiung Veterans General Hospital, Taiwan (VGHKS97-066); the National Science Council of Taiwan (NSC 100-2314-B-010-039, NSC102-2314-B-010-017); and the National Health Research Institute, Department of Health, Taiwan (NHRI-99A1-PDCO-0710101).