Uneven testing hides drug-resistant tuberculosis burden

Drug resistance is one of the biggest obstacles to ending tuberculosis, but how often resistant cases are actually found depends on who gets the right tests. Jhancy Rocío Aguilar‐Jiménez led a study that examined whether incomplete access to modern diagnostics leads surveillance systems to underestimate drug-resistant TB (DR-TB). The team focused on Brazil’s national notification database, using records for new TB cases in people older than 18 collected between 2015 and 2020 in the Brazilian Information System on Notifiable Diseases (SINAN). They began from a clear idea: patients who received the rapid molecular test Xpert ® MTB/RIF tended to have higher detection of resistance overall, not only resistance to rifampicin. That pattern suggested that access to Xpert ® MTB/RIF could be a useful proxy for broader diagnostic capacity. The researchers set out to correct reported DR-TB prevalence by accounting for uneven access to this test, aiming to produce a more accurate picture of how common drug-resistant TB really is across different municipalities and patient groups.

To estimate and correct for disparities in testing, the team carried out a cross-sectional analysis of 406,331 new adult TB cases recorded in SINAN from 2015 to 2020. They assumed access to Xpert ® MTB/RIF indicated better overall conditions for detecting drug resistance and used that as the target for prediction. A multilevel mixed-effects logistic regression model that included both individual- and municipality-level variables was developed on a random sample of 81,027 observations. The model was validated using out-of-sample observations across geographic areas and showed strong discrimination (AUC = 80·93%, 95%CI: 80·71%–81·33%). The researchers then applied inverse probability weighting based on the model’s predicted access to Xpert ® MTB/RIF to adjust DR-TB prevalence estimates. Reported DR-TB prevalence in the surveillance system was 1·54% (95%CI: 1·50%–1·57%), among patients tested with Xpert ® MTB/RIF it was 3·86% (95%CI: 3·75%–3·96%), and the corrected weighted prevalence rose to 5·95% (95%CI: 5·54%–6·38%).

The findings suggest that routine surveillance figures may substantially understate the true burden of drug-resistant TB when access to high-quality diagnostics is uneven. By using predictions of who had access to Xpert ® MTB/RIF as a stand-in for diagnostic capacity, and then reweighting cases to account for that, the study produced a notably higher estimate of DR-TB prevalence. This doesn’t change laboratory results themselves, but it changes how we interpret surveillance data: areas or groups with less access to rapid molecular testing may look like they have lower resistance simply because resistant cases are missed. The authors argue that correcting surveillance indicators for diagnostic access can make public health records more accurate and help target resources, testing, and treatment where they’re most needed. Improving equity in access to Xpert ® MTB/RIF and similar diagnostics is therefore central to finding and treating more people with DR-TB and to tracking progress toward TB control goals.