Bulletin of the World Health Organization

Operational lessons drawn from pilot implementation of Xpert MTB/Rif in Brazil

Betina Durovni a, Valeria Saraceni a, Marcelo Cordeiro-Santos b, Solange Cavalcante a, Elizabeth Soares a, Cristina Lourenço c, Alexandre Menezes d, Susan van den Hof e, Frank Cobelens f & Anete Trajman g

a. Rio de Janeiro Municipal Health Secretariat, Rio de Janeiro, Brazil.
b. Infectious Diseases Graduate Programme, Tropical Medicine Foundation Dr Heitor Vieira Dourado, Manaus, Brazil.
c. Evandro Chagas Research Institute, Oswaldo Cruz Foundation, Rio de Janeiro, Brazil.
d. Global Health Strategies, Rio de Janeiro, Brazil.
e. KNCV Tuberculosis Foundation, The Hague, Netherlands.
f. Department of Global Health, Amsterdam Institute of Global Health and Development, Amsterdam, Netherlands.
g. Internal Medicine Graduate Programme, Rio de Janeiro Federal University, Av Brigadeiro Trompowsky s/nº,11th floor, Ilha do Fundão, Rio de Janeiro, 21941-50, Brazil.

Correspondence to Anete Trajman (email: atrajman@gmail.com).

(Submitted: 08 October 2013 – Revised version received: 20 January 2014 – Accepted: 21 January 2014 – Published online: 01 May 2014.)

Bulletin of the World Health Organization 2014;92:613-617. doi: http://dx.doi.org/10.2471/BLT.13.131409


In health services in areas with a high burden of tuberculosis, diagnosis can be especially difficult because it takes several weeks or months to obtain the results of mycobacterial culture, which is the gold standard test for the diagnosis of tuberculosis. In such settings, diagnosis is usually based on microscopic examination of at least two sputum smears,1 but because sputum smear microscopy has low sensitivity, patients are often started on antituberculous therapy based on clinical evidence, without bacteriological confirmation.2 This leads to the underreporting of tuberculosis cases, unnecessary exposure to therapy with potential toxicity, and delay in the correct diagnosis and treatment of patients. In Brazil, around 26% of new tuberculosis cases are not confirmed with any bacteriological test and culture is not performed in 73% of re-treatment cases.3

Xpert MTB/RIF (Xpert), a new nucleic acid amplification test based on polymerase chain reaction (PCR), has been recently developed to detect Mycobacterium tuberculosis DNA and genetic sequences indicative of rifampicin resistance (i.e. mutations of rpoB).4 The entire PCR assay is performed automatically within a cartridge, where the sample and reagents are mixed. The test, which takes less than two hours, has 88% sensitivity for tuberculosis and 94% sensitivity for rifampicin resistance, as well as 98% specificity for both.5,6 In light of these advantages, the World Health Organization has recommended the use of Xpert for the diagnosis of tuberculosis in countries with high prevalences of human immunodeficiency virus (HIV) infection and multidrug-resistant tuberculosis (MDR-TB).7

Brazil is considered a high-burden tuberculosis country (90 cases per 100 000 population). However, it is unusual among countries with a high tuberculosis burden in that primary MDR-TB is relatively uncommon (< 2%).3 Yet despite low rates of HIV co-infection (10%)3 and MDR-TB, the Brazilian National Tuberculosis Programme has recommended the incorporation of Xpert for the routine diagnosis of pulmonary tuberculosis in the public health system in an effort to increase the notification of cases with bacteriologically confirmed tuberculosis. To monitor the implementation of this new diagnostic test in the routine work of public health services, a roll-out pilot study was conducted. In this paper we report the lessons learnt during the introduction of Xpert in two cities. The results of the pilot study will be used to plan for the national scale-up of Xpert in Brazil and may be useful to other countries that are trying to incorporate this new technology.

The field experience

Study design

The roll-out pilot study, registered at www.clinicaltrials.gov (NCT01363765) and approved by the National Ethics Board-CONEP (#494/2011), was a stepped wedge randomized clinical trial. It was conducted between February and October 2012 in Rio de Janeiro and Manaus, two large state capitals with high tuberculosis incidence rates (94.4 and 89.3 per 100 000 population, respectively).3

In total, 15 four-slot Xpert systems were implemented in 14 laboratories covering 70% of the TB diagnosis in Rio de Janeiro ( = 11) and Manaus (= 3). The laboratories were chosen because they had different workloads and their population catchment areas did not overlap. Every month, two laboratories migrated overnight from two-sample smear microscopy diagnostics to a one-sample Xpert test. There were no changes to the patient’s routine management and standard clinical guidelines; two sputum samples continued to be collected from each patient (despite the switch to Xpert) and the same information systems remained in place. If an Xpert result was positive, regardless of a rifampicin resistance signal, the Xpert assay was repeated on a new sputum sample. If a positive rifampicin resistance was detected, culture and drug-susceptibility tests were performed. If resistance was confirmed, the patient was referred to the MDR centre for further evaluation and treatment.

Preparing for implementation of Xpert

For reporting purposes, the existing national electronic laboratory information system, Gerenciador de Ambiente Laboratorial (GAL), had to be modified to include tuberculosis cases diagnosed with the new technology.8 A new item was added to the notification forms (PCR results) of both GAL and the national reporting system. Training of laboratory technicians on the use of Xpert was conducted two weeks before the laboratory entered the intervention phase.9 The training, which lasted one day, was carried out by a representative of the manufacturer and the municipal tuberculosis programme team. In addition, a four-hour training session was organized by the National Tuberculosis Programme and the municipal tuberculosis programme to teach physicians and nurses how to interpret the results of a resistance signal in a country with a low prevalence of MDR-TB such as Brazil. The recommendation given was to put patients on four-drug combination therapy – i.e. rifampicin, isoniazid, pyrazinamide and ethambutol – until the results of conventional drug susceptibility tests were available.

Operational findings

During the study period, several bottlenecks and opportunities for strengthening the health-care system were identified. In informal discussions, the health staff said that the GAL system was a very useful tool but pointed out that the lack of a unique patient identifier number in the Brazilian health system resulted in the need to manually enter each patient’s information when requesting a test or a laboratory result – a problem identified in an earlier study.10 Apart from being time consuming, this practice often led to minor errors that resulted in truncated information and misidentification of samples or double entries for the same patient. It also led to the misuse of PCR cartridges. In some cases two diagnostic samples from the same patient or follow-up samples were processed by mistake.

Certain problems came up, in addition to errors involving Xpert syringes, probes and signals previously described by the manufacturer of Xpert.11 Such problems were attributable to the characteristics of the sputum samples. Among 15 701 samples analysed by Xpert, 1151 (7.3%) had insufficient volume (less than 1 ml) and 200 (1.3%) had heavy traces of blood or food residuals. Thus, 8.6% of the Xpert samples had to be examined by smear microscopy.

The training of laboratory technicians without computer skills was straightforward and the learning curve was quick. A previous study has shown that Xpert performance is less subject to the influence of user skills, motivation or workload than sputum smear microscopy.12 In one laboratory with a very high volume of samples, a change in the working shift of one employee was sufficient to enable the processing of all samples.

To confirm rifampicin resistance and investigate resistance to other drugs, it became necessary to expand reference laboratory capacity, especially in performing mycobacterial culture and drug-susceptibility tests, which are not routinely performed in Brazil.

Despite frequent energy shortages in Manaus, Xpert machines were able to complete the PCR cycles with support from additional, uninterrupted power supplies. Nonetheless, the maintenance of Xpert equipment consumed much time and energy because spare modules and replacement parts were not immediately available in Brazil. We were able to use parts from Xpert modules that weren’t yet being used in the study, but this is not a good solution in a routine laboratory. During the short pilot study, six of the 60 modules had to be replaced, along with a defective computer that could not be replaced locally on account of software requirements.

Lessons learnt

As newer and more accurate technologies for tuberculosis diagnosis are developed, understanding the factors that facilitate or hinder their implementation becomes important. Health system staff should carefully consider the many factors involved in incorporating a new technology when seeking to maximize clinical impact and minimize the tuberculosis burden in their settings (Box 1).

Box 1. Summary of main lessons learnt

  • To minimize errors when using Xpert, notification forms have to be adapted and an efficient communication system has to be in place.
  • A regular supply of PCR cartridges and the ready availability of spare parts must be negotiated with the manufacturer of Xpert.
  • The capacity to perform sputum smear microscopy should be maintained for follow-up tests and for the testing of sputum samples too scanty to run Xpert.

Ensuring an efficient laboratory information system is essential. Such a system can lead to improved health outcomes through several mechanisms, including increased adherence to guideline-based care, enhanced disease surveillance and monitoring and fewer medication errors.13,14 Above all, laboratory information systems can expedite the reporting of test results. A patient identifier number would reduce the probability of errors.

The eventual need for changes in national notification systems when incorporating a new technology should also be considered in advance. For instance, it will be necessary to modify reporting forms to ensure that bacteriologically confirmed TB cases are accurately captured.

Training also needs careful planning. In the case of Xpert, it is particularly important to make physicians understand that the test’s negative and positive predictive values for tuberculosis and rifampicin resistance depend on the local prevalences of both and should be adapted accordingly.6 Before adopting Xpert, national tuberculosis programmes might want to prepare algorithms and technical notes on how to investigate suspected tuberculosis cases and how to handle positive rifampicin resistance. In addition, since Xpert is not currently recommended for testing follow-up samples, to avoid wasting PCR cartridges health-care workers should be trained in how and when to request the test and laboratory staff should be shown how to manage the flow of samples obtained for different purposes (diagnosis versus follow-up).

It is important to ensure that laboratory technicians feel comfortable with the use of new technologies, including the simple computer-based tasks they require. A short adaptation period should be planned in advance to allow them time to learn to use the equipment and to gradually shift from paper-based to electronic record keeping. With advances in technology, it will be necessary to ensure that health workers, especially in low-resource settings, have the training and support required to effectively operate computers and other equipment. Furthermore, since Xpert cannot be performed from time to time because of inadequate samples or other technical difficulties,15 it is important to retain the laboratory capacity to perform good quality sputum smears. Moreover, countries where mycobacterial culture is not routinely performed in patients suspected of having tuberculosis should be prepared to expand their laboratory capacity to process samples with a positive rifampicin resistance signal, as well as the capacity to manage patients suspected of having MDR-TB.

Plans for scaling up the use of Xpert should include negotiations with the manufacturer for the maintenance of equipment and the regular supply of cartridges, syringes and other necessities. It is critically important to have available spare parts and local technicians trained to quickly solve occasional technical problems, together with a sustainable and regular supply of cartridges. Hopefully, some of the issues identified in this pilot exercise will help Xpert manufacturers to provide users with effective technical support. Lastly, countries planning to scale up Xpert nationwide should conduct health system research, which plays an essential role in the implementation of new technologies.


We thank all directors of the clinics and laboratories, as well as laboratory staff. Without them this study would not have been possible.

Authors’ secondary affiliations: VS, Infectious Diseases Graduate Programme, Tropical Medicine Foundation, Dr Heitor Vieira Dourado, Manaus, Brazil; MCS, Amazonas State University, Manaus, Brazil; SC, Oswaldo Cruz Foundation, Rio de Janeiro, Brazil; SvdH, Department of Global Health, Academic Medical Center and Amsterdam Institute of Global Health and Development, Amsterdam, Netherlands; FC, KNCV Tuberculosis Foundation, The Hague, Netherlands; AT, McGill University, Montreal Chest Institute, Montreal, Canada. BD is a PhD student at the Internal Medicine Graduate Programme, Rio de Janeiro Federal University, Rio de Janeiro, Brazil.

Competing interests:

None declared.