Diagnosis and Screening

SAIA-TB: Using the Systems Analysis and Improvement Approach (SAIA) to prevent TB in rural South Africa

  • Investigator: Meredith Brooks
  • Funding Source: National Institute of Nursing Research

South Africa estimates 80% of their population has TB infection, and 14% of the population lives with HIV, with an estimated 5-15% of South Africans at high risk of developing TB disease from recent infection or immunocompromised status. Therefore, utilization of routinely collected data to optimize the comprehensive TB care cascade – screening, evaluation, diagnosing, linkage to care, treatment, and TB-free survival – is important to assess at the clinic level to improve clinic flow and patient outcomes. The proposed study will leverage an evidence-based implementation science strategy, the Systems Analysis and Improvement Approach (SAIA), and recent TB cascade analyses piloted in the proposed site, to adapt and evaluate the effectiveness of SAIA-TB using a stepped wedge crossover cluster randomized trial across 12 clinics in rural Eastern Cape, South Africa.

Accuracy of the Xpert Ultra for Diagnosis of Pulmonary Tuberculosis

  • Investigator: Carlos Acuna-Villaorduna
  • Funding Source: National Institute of Allergy and Infectious Diseases

The Xpert MTB/RIF assay, an automated, integrated nucleic acid amplification diagnostic test, has been a major breakthrough in tuberculosis (TB) diagnostics. Yet, large proportions of TB suspects are placed on TB treatment based on empirical decision-making even in settings where Xpert MTB/RIF is available. A major driver of empirical TB treatment is Xpert MTB/RIF’s modest sensitivity in smear-negative pulmonary TB patients – Xpert MTB/RIF detects only about one-half to two-thirds of such patients. To maximize the individual and population level impact of rapid TB diagnostics, the Xpert MTB/RIF assay has been re-engineered to substantially increase test sensitivity. The proposed research will establish the clinical diagnostic accuracy of the new Ultra test in adults with signs/symptoms of pulmonary TB, and will provide information about the potential impact of Ultrabased diagnostic algorithms on individual and public health. Project aims are 1) to determine the clinical diagnostic accuracy of the Ultra test for detection of M. tuberculosis in sputum; 2) to refine estimates of Ultra specificity for detection of rifampin resistance; and 3) to characterize short-term participant outcomes including initiation of treatment and to model the potential health impact and cost-effectiveness of TB diagnosis using Ultra. Two hypotheses will be tested with regard to TB case detection – first, that Ultra sensitivity is superior to that of Xpert MTB/RIF in adults with smear-negative/culture-confirmed pulmonary TB; and second, that the sensitivity of a single Ultra test is non-inferior to that of a single liquid culture in adults with pulmonary TB. To accomplish these aims, a multi-site clinical diagnostic accuracy study will be conducted in Uganda, Kenya, South Africa, and Brazil.


    Artificial Intelligence to Improve the Diagnosis of Pediatric Tuberculosis

    • Investigator: Meredith Brooks
    • Funding Source: William F. Milton Fund, Harvard University

    Tuberculosis is a preventable infectious disease, yet more than one million children fall sick with TB every year. Under-diagnosis remains a major challenge, because (1) children present with diverse and non-specific symptoms and (2) tests that diagnose TB in adults have low sensitivity in children, who often have paucibacillary TB or cannot produce sputum. Other challenges include complex diagnostic algorithms. Coupled with limited laboratory capacity in high Tb-burden settings and costly tests, these factors often preclude microbiological confirmation of TB in children. As a result, each year millions of children with Tb or sub-clinical TB infection are missed by TB services and never receive life-saving treatment. Effective diagnostic algorithms tailored for children need to be developed, ideally using large datasets to ensure validity. To address this knowledge gap, we leverage an existing dataset of children screened for TB in Kotri, Pakistan, providing a unique opportunity to assess age-specific barriers to TB diagnosis in children. We aim to: (1) identify age-specific gaps in the sequence of steps required for screening children for TB disease; and (2) refine diagnostic algorithms for classification of TB using age-specific predictors through machine-learning methods.


    Development of a Costing and Cost-Effectiveness Framework to Compare Tuberculosis Infection Tests in a High-Burden Setting

    • Investigator: Meredith Brooks
    • Funding Source: QIAGEN

    We aim to develop a framework to estimate the practical costs incurred from, and programmatic impact related to, tuberculosis (TB) infection testing—tuberculin skin tests (TST) versus interferon gamma release assay (IGRA)—in a densely populated high-burden TB area. We will develop a seven-step framework that can be tailored to individual TB programs to inform decision-making around costing for comparing different TB infection diagnostics. We will present methodology to calculate (1) the prevalence of TB infection, (2) the rate of observed, true, and false positives and negatives for each test, (3) the cost of test administration, (4) the cost of false negatives, (5) the cost of treating all that test positive, (6) the per-test cost incurred due to treatment and misdiagnosis, and (7) the threshold at which laboratory infrastructure investments for IGRA become cost effective over TST. We will then applied this framework in a densely populated, peri-urban district in Lima, Peru with high rates of BCG vaccination.


    Identifying Effective and Efficient Approaches to Tuberculosis Screening in Brazilian Prisons

    • Investigator: Leonardo Martinez
    • Funding Source: National Institute of Allergy and Infectious Diseases

    Point-of-care Questionnaire and mHealth Assisted Diagnosis of Post-TB Lung Disease

    • Investigator: Akshay Gupte
    • Funding Source: National Institutes of Health

    Pulmonary tuberculosis (PTB) is associated with lung injury which can persist despite successful therapy. Lung sequelae of treated PTB are increasingly recognized as an independent risk factor for chronic obstructive pulmonary disease (COPD) and, an important contributor of excess morbidity and mortality. Spirometry is the gold standard for diagnosing lung function defects, however it is technically challenging and expensive to perform, and may not be available at the point-of-care in many TB-endemic settings. The overall goal of this project is to develop and validate a questionnaire-based screening algorithm, assisted by machine learning analysis of cough sounds and lung auscultation data, to identify individuals with a high probability of having Post-TB Lung Disease for referral and confirmatory testing.