Xpert MTB/Rif, a New Tool for the Diagnosis of Pulmonary Tuberculosis in Two Municipalities in Brazil

Diagnosis of tuberculosis (TB) is a challenge because sputum smear, the most rapid and inexpensive test, often fails to detect the disease, in around 20 to 30% of cases. Culture of sputum yields a correct diagnosis in up to 90% of cases, but results are only available in 4 to 8 weeks, depending on the method. A new test (Xpert MTB/Rif) based on a rapid technique, named polymerase chain reaction (PCR), detects TB in less than 2 hours over 95% of cases, in addition to identification of cases resistant to certain drugs used to treat TB. The test is expensive, but several studies have demonstrated its accuracy, and since most steps are automatized, savings can be expected from human resources work. The aims of our study are (1) to evaluate this tool as a substitute test for sputum smears in routine conditions; (2) evaluate if it is cost-effective, meaning that effectiveness of the test may outweigh the extra cost, and (3) evaluate the acceptability of the test among patients and health care workers.

The main study consisted of a group-randomized pragmatic trial following a stepped-wedge design. The GeneXpert machines will be installed in laboratories of the basic health units in Rio de Janeiro and Manaus, two cities with a high burden of the disease in Brazil. Doctors will prescribe routine tests in the clinics. When sputum specimens arrive in the laboratory, instead of performing smears, the Xpert MTB/Rif will be performed during the intervention period. If positive, a sputum smear will also be performed. The intervention period will be preceded by an observation period during which routine will be kept unchanged, i.e., only smears will be performed. The observation period will serve as a comparator (control). The investigators will then see how many more cases were detected during the intervention period, and in particular, how many smear-negative cases were detected, compared to the observation period.

A total of 35,000 specimens are expected to be tested with the new technology in 14 laboratories. Laboratories, not individuals, will be randomized.