Description
Background: Although drug-resistant tuberculosis (DR-TB) exerts heavy disease burden worldwide, especially in developing countries, it still makes up a minority of all TB cases. To efficiently assess the performance of various rapid drug susceptibility tests, selecting patients at increased risk for DR-TB reduces the cost by testing fewer drug- susceptible patients. Therefore, simple screening tools are needed to identify TB cases most likely to be drug-resistant. Methods: Patients with TB presenting at collaborating sites in India, Moldova, and South Africa who were suspected, but not yet diagnosed with DR-TB were prospectively screened and enrolled into a study comparing four assays for detecting extensively DR-TB (XDR-TB). Prior to enrollment, patients were administered a five-question screener. Patients were eligible if they had previous treatment for TB, close contact with someone known to have DR-TB, been diagnosed with multi-drug resistant TB (MDR-TB), or failed standard or MDR-TB treatment. Drug resistance was ascertained by standardized liquid-culture methods. Chi-square and logistic regression analyses were used to assess the ability of the screening questions to detect DR-TB. Area-under-the-curve analysis was used to determine which combinations of questions worked well in each country. Results: Of 878 patients included in the analysis, 52.9%, 25.4%, and 21.7% came from India, Moldova, and South Africa, respectively. Overall, 66.0% had DR-TB. History of TB treatment (OR 1.46, CI 1.05–2.02), diagnosis with MDR-TB (OR 4.13, CI 2.76–6.16), failure of standard treatment (OR 5.47, CI 3.95–7.58), and failure of MDR-TB treatment (OR 3.27, CI 2.16–4.95) were individually associated with DR-TB. The number of affirmative answers given by each patient was also associated with DR-TB. For each additional affirmative answer, the odds of DR-TB increased by a factor of 1.86 (CI 1.61–2.15). The strength of the association differed by country in three of the screening questions (contact, failure of standard treatment, and diagnosis with MDR-TB). Conclusions: Screening questions were effective in identifying patients with DR-TB, although the ability of the screening questions to discriminate varied by location. The choice of screening questions to employ may need to be altered in each country and be further tested for clinical applicability.
