The research, led by Neha Janrode and Molebogeng Rangaka together with colleagues at institutions including the University of Cape Town, set out to fill a conspicuous evidence gap. Although international guidelines endorse symptom-based and chest X-ray screening for people with diabetes, the accuracy of these tools in this specific population had never been rigorously quantified. That omission matters because diabetes does not merely raise tuberculosis risk; it can also alter how the disease presents, how the immune system responds, and how confidently clinicians can interpret screening results. Screening strategies validated in the general population, or in people living with HIV, cannot simply be assumed to transfer to a metabolic disease population.

To measure performance directly, the team consecutively enrolled 673 adults with diabetes attending routine care at clinics in South Africa, a country that carries one of the heaviest tuberculosis burdens in the world. Every participant, regardless of whether they felt ill, underwent the same battery of assessments: a structured symptom screen, a chest X-ray, and collection of a single sputum specimen. The sputum was tested with Xpert Ultra, the molecular diagnostic platform that serves as the backbone of tuberculosis testing in high-burden settings. A positive Xpert Ultra result served as the reference standard against which the screening tools were judged, allowing the researchers to calculate sensitivity, the proportion of true cases caught, and specificity, the proportion of disease-free people correctly cleared.

The enrolled cohort reflected the demographic reality of diabetes care in urban South Africa. The median age was 54 years, with an interquartile range of 47 to 60, and 63.8 percent of participants were female. HIV prevalence, a critical consideration in any South African tuberculosis study, stood at 17.2 percent. Most strikingly, the yield of screening was low: only nine participants, or 1.33 percent, were diagnosed with prevalent tuberculosis. That modest number of cases is itself informative, because it illustrates how difficult it is to study tuberculosis in diabetes populations even in high-burden countries, and it frames the central statistical challenge of the analysis, since sensitivity estimates rest on a very small numerator.

The symptom screen performed poorly. Any cough, the anchor question of most national screening algorithms, detected just 22.2 percent of the tuberculosis cases, with a wide 95 percent confidence interval stretching from 2.1 to 60.0 percent, a consequence of the small number of cases. Its specificity was high at 97.5 percent, meaning few people without tuberculosis were falsely flagged, but high specificity is cold comfort when the test misses more than three quarters of true cases. The researchers also tested whether broadening the symptom definition would help, expanding it to include any of cough, fever, weight loss, or night sweats, the classic constitutional quartet of tuberculosis. It did not. Sensitivity remained exactly 22.2 percent, while specificity slipped slightly to 96.0 percent. Adding more symptoms, in other words, caught no additional cases while generating more false alarms.

This finding strikes at the heart of a widely held assumption. Symptom screening is attractive precisely because it is cheap, fast, and requires no equipment, and it underpins tuberculosis case-finding programs across the globe. But the classic symptoms of tuberculosis are also nonspecific, and in a population with diabetes the overlap with ordinary morbidity may be even greater. People with diabetes attend clinics frequently, report a range of chronic complaints, and may have pulmonary pathology unrelated to tuberculosis. The study’s results suggest that in this group, the presence or absence of symptoms is a weak signal indeed, and that a symptom-driven pathway will systematically overlook infectious individuals who could then continue transmitting Mycobacterium tuberculosis in households and communities.

Chest X-ray screening fared better, but not well enough. Abnormalities on chest radiography suggestive of tuberculosis detected 55.6 percent of cases, with a 95 percent confidence interval of 21.2 to 86.3 percent, and a specificity of 95.4 percent. In practical terms, even the more technologically demanding tool missed nearly half of the tuberculosis cases circulating in the cohort. For a screening strategy to be effective at the population level, sensitivity needs to be substantially higher, because every missed case represents a missed opportunity for treatment and a continued risk of onward transmission. The authors conclude that chest X-ray sensitivity, while an improvement over symptoms, remained suboptimal in people with diabetes.

The study also uncovered a clinically important nuance: the accuracy of chest X-ray depends heavily on a patient’s tuberculosis history. Among participants who had experienced tuberculosis in the past, the specificity of chest X-ray fell to 87.6 percent, significantly lower than the 97.2 percent observed in those with no prior disease, a difference the authors report as statistically significant. The explanation is almost certainly scarring, cavitation, and other residual abnormalities left behind by previous episodes of disease, which radiologists cannot reliably distinguish from active infection. In high-burden settings, where many adults carry the imprint of past tuberculosis, this loss of specificity could translate into large numbers of unnecessary confirmatory workups, straining laboratory capacity and burdening patients with anxiety and repeat visits.

The implications for policy are sobering. The World Health Organization has positioned systematic screening as a pillar of the End Tuberculosis Strategy, and people with diabetes are explicitly named as a priority group. Yet if the standard screening instruments miss most cases, as this study suggests, then programs built on those instruments will deliver far less benefit than models predict. The low prevalence observed in the cohort, 1.33 percent, also raises a question about resource allocation: screening hundreds of people with diabetes to find a handful of cases is only worthwhile if the tools used are accurate enough to justify the effort. Neither symptom screening nor chest radiography, on this evidence, clears that bar on its own.

What the study points toward, implicitly, is the need for better screening approaches tailored to diabetes populations. The authors’ use of a single sputum specimen tested by Xpert Ultra hints at one possible direction, since molecular testing on routinely collected samples could bypass the weak signal of symptoms altogether, though cost and logistics remain obstacles in many settings. More research will also be needed to define optimal screening intervals, to test combinations of tools, and to evaluate newer technologies such as computer-aided chest X-ray interpretation. For now, the message for clinicians and program managers is clear: in people with diabetes, absence of symptoms and a clean-looking chest X-ray cannot be taken as reassurance, and the search for screening strategies that actually work in this growing, high-risk population is far from over.

Subject of Research: Accuracy of symptom and chest X-ray screening for tuberculosis in adults with diabetes

Article Title: Sensitivity and specificity of tuberculosis screening tools in people with diabetes

Article References: Janrode, N., Hamada, Y., Taliep, A., Barron, L., Chabaya, W., Goliath, R. T., Duong, T., Jackson, A., Galant, S., Omar-Davies, N., Sai, L. L., Twentiey, L., Wilkinson, R. J., & Rangaka, M. X. (2026). Sensitivity and specificity of tuberculosis screening tools in people with diabetes. PLOS Global Public Health, 6(9), e0006479. https://doi.org/10.1371/journal.pgph.0006479

Image Credits: AI Generated

DOI: 10.1371/journal.pgph.0006479

Keywords: tuberculosis, diabetes, screening, chest X-ray, Xpert Ultra, sensitivity, specificity, South Africa, public health, symptom screening, PLOS Global Public Health, diagnostic accuracy