Role of empiric treatment in hospitalized patients with Xpert MTB/RIF-negative presumptive pulmonary tuberculosis: A prospective cohort study.

OBJECTIVES The ability of clinical algorithms to identify tuberculosis disease and the impact of empiric treatment on survival in people with a negative Xpert MTB/RIF (Xpert) result remains poorly documented. METHODS Hospitalized Xpert-negative patients (125 initiated on empiric tuberculosis treatment based on a clinical algorithm and 125 in whom tuberculosis treatment was not started) were enrolled. Sputum samples were evaluated for Mycobacterium tuberculosis by culture. All study participants were followed up for six months. RESULTS Xpert-negative inpatients in whom empiric tuberculosis treatment was initiated were more likely to have microbiological confirmed tuberculosis compared to those in whom empiric tuberculosis treatment was not started (24.8% vs 6.4%, p=0.0001). Six-month risk of death was 5.2%, but the risk was twice as high in people with bacteriological confirmation of TB (10.3% vs 4.3%, p=0.12). Cardinal symptoms of TB were associated with bacteriological confirmation and a decision to start empiric treatment. The positive predictive value of the clinical algorithm was 24.8% and empiric treatment did not affect six-month risk of death (5.6% vs 4.8%, p=0.78). CONCLUSIONS Clinical algorithm identifies the majority of confirmed tuberculosis cases among Xpert-negative inpatients. Empiric treatment did not impact survival and resulted in substantial overtreatment. The more sensitive Xpert Ultra assay should be used to eliminate the need for empiric tuberculosis treatment.