AI for Detection of Tuberculosis: Implications for Global Health.

Tuberculosis, which primarily affects developing countries, remains a significant global health concern. Since the 2010s, the role of chest radiography has expanded in tuberculosis triage and screening beyond its traditional complementary role in the diagnosis of tuberculosis. Computer-aided diagnosis (CAD) systems for tuberculosis detection on chest radiographs have recently made substantial progress in diagnostic performance, thanks to deep learning technologies. The current performance of CAD systems for tuberculosis has approximated that of human experts, presenting a potential solution to the shortage of human readers to interpret chest radiographs in low- or middle-income, high-tuberculosis-burden countries. This article provides a critical appraisal of developmental process reporting in extant CAD software for tuberculosis, based on the Checklist for Artificial Intelligence in Medical Imaging. It also explores several considerations to scale up CAD solutions, encompassing manufacturer-independent CAD validation, economic and political aspects, and ethical concerns, as well as the potential for broadening radiography-based diagnosis to other nontuberculosis diseases. Collectively, CAD for tuberculosis will emerge as a representative deep learning application, catalyzing advances in global health and health equity. Keywords: Computer-aided Diagnosis (CAD), Conventional Radiography, Thorax, Lung, Machine Learning Supplemental material is available for this article. © RSNA, 2024.

T-cell deficiency and hyperinflammatory monocyte responses associate with Mycobacterium avium complex lung disease.

Immunological mechanisms of susceptibility to nontuberculous mycobacterial (NTM) disease are poorly understood. To understand NTM pathogenesis, we evaluated innate and antigen-specific adaptive immune responses to Mycobacterium avium complex (MAC) in asymptomatic individuals with a previous history of MAC lung disease (MACDZ). We hypothesized that Mav-specific immune responses are associated with susceptibility to MAC lung disease. We measured MAC-, NTM-, or MAC/Mtb-specific T-cell responses by cytokine production, expression of surface markers, and analysis of global gene expression in 27 MACDZ individuals and 32 healthy controls. We also analyzed global gene expression in Mycobacterium avium-infected and uninfected peripheral blood monocytes from 17 MACDZ and 17 healthy controls. We were unable to detect increased T-cell responses against MAC-specific reagents in MACDZ compared to controls, while the responses to non-mycobacteria derived antigens were preserved. MACDZ individuals had a lower frequency of Th1 and Th1* T-cell populations. In addition, MACDZ subjects had lower transcriptional responses in PBMCs stimulated with a mycobacterial peptide pool (MTB300). By contrast, global gene expression analysis demonstrated upregulation of proinflammatory pathways in uninfected and M. avium-infected monocytes, i.e. a hyperinflammatory in vitro response, derived from MACDZ subjects compared to controls. Together, these data suggest a novel immunologic defect which underlies MAC pathogenesis and includes concurrent innate and adaptive dysregulation which persists years after completion of treatment.

The impact of the COVID-19 pandemic and associated suppression measures on the burden of tuberculosis in India.

BACKGROUND: Tuberculosis (TB) is a major cause of death globally. India carries the highest share of the global TB burden. The COVID-19 pandemic has severely impacted diagnosis of TB in India, yet there is limited data on how TB case reporting has changed since the pandemic began and which factors determine differences in case notification. METHODS: We utilized publicly available data on TB case reporting through the Indian Central TB Division from January 2017 through April of 2021 (prior to the first COVID-19 related lockdown). Using a Poisson model, we estimated seasonal and yearly patterns in TB case notification in India from January 2017 through February 2020 and extended this estimate as the counterfactual expected TB cases notified from March 2020 through April 2021. We characterized the differences in case notification observed and those expected in the absence of the pandemic by State and Territory. We then performed a linear regression to examine the relationship between the logit ratio of reported TB to counterfactual cases and mask use, mobility, daily hospitalizations/100,000 population, and public/total TB case reporting. RESULTS: We found 1,320,203 expected cases of TB (95% uncertainty interval (UI) 1,309,612 to 1,330,693) were not reported during the period from March 2020 through April 2021. This represents a 63.3% difference (95% UI 62.8 to 63.8) in reporting. We found that mobility data and average hospital admissions per month per population were correlated with differences in TB case notification, compared to the counterfactual in the absence of the pandemic (p > 0.001). CONCLUSION: There was a large difference between reported TB cases in India and those expected in the absence of the pandemic. This information can help inform the Indian TB program as they consider interventions to accelerate case finding and notification once the pandemic related TB service disruptions improve. Mobility data and hospital admissions are surrogate measures that correlate with a greater difference in reported/expected TB cases and may correlate with a disruption in TB diagnostic services. However, further research is needed to clarify this association and identify other key contributors to gaps in TB case notifications in India.

Machine Learning for Health: Algorithm Auditing & Quality Control.

Developers proposing new machine learning for health (ML4H) tools often pledge to match or even surpass the performance of existing tools, yet the reality is usually more complicated. Reliable deployment of ML4H to the real world is challenging as examples from diabetic retinopathy or Covid-19 screening show. We envision an integrated framework of algorithm auditing and quality control that provides a path towards the effective and reliable application of ML systems in healthcare. In this editorial, we give a summary of ongoing work towards that vision and announce a call for participation to the special issue  Machine Learning for Health: Algorithm Auditing & Quality Control in this journal to advance the practice of ML4H auditing.

Systematic review of the performance of rapid rifampicin resistance testing for drug-resistant tuberculosis.

INTRODUCTION: Rapid tests for rifampicin resistance may be useful for identifying isolates at high risk of drug resistance, including multidrug-resistant TB (MDR-TB). However, choice of diagnostic test and prevalence of rifampicin resistance may both impact a diagnostic strategy for identifying drug resistant-TB. We performed a systematic review to evaluate the performance of WHO-endorsed rapid tests for rifampicin resistance detection. METHODS: We searched MEDLINE, Embase and the Cochrane Library through January 1, 2012. For each rapid test, we determined pooled sensitivity and specificity estimates using a hierarchical random effects model. Predictive values of the tests were determined at different prevalence rates of rifampicin resistance and MDR-TB. RESULTS: We identified 60 publications involving six different tests (INNO-LiPA Rif. TB assay, Genotype MTBDR assay, Genotype MTBDRplus assay, Colorimetric Redox Indicator (CRI) assay, Nitrate Reductase Assay (NRA) and MODS tests): for all tests, negative predictive values were high when rifampicin resistance prevalence was ≤ 30%. However, positive predictive values were considerably reduced for the INNO-LiPA Rif. TB assay, the MTBDRplus assay and MODS when rifampicin resistance prevalence was < 5%. LIMITATIONS: In many studies, it was unclear whether patient selection or index test performance could have introduced bias. In addition, we were unable to evaluate critical concentration thresholds for the colorimetric tests. DISCUSSION: Rapid tests for rifampicin resistance alone cannot accurately predict rifampicin resistance or MDR-TB in areas with a low prevalence of rifampicin resistance. However, in areas with a high prevalence of rifampicin resistance and MDR-TB, these tests may be a valuable component of an MDR-TB management strategy.

Diagnostic accuracy and reproducibility of WHO-endorsed phenotypic drug susceptibility testing methods for first-line and second-line antituberculosis drugs.

In an effort to update and clarify policies on tuberculosis drug susceptibility testing (DST), the World Health Organization (WHO) commissioned a systematic review evaluating WHO-endorsed diagnostic tests. We report the results of this systematic review and meta-analysis of the diagnostic accuracy and reproducibility of phenotypic DST for first-line and second-line antituberculosis drugs. This review provides support for recommended critical concentrations for isoniazid and rifampin in commercial broth-based systems. Further studies are needed to evaluate critical concentrations for ethambutol and streptomycin that accurately detect susceptibility to these drugs. Evidence is limited on the performance of DST for pyrazinamide and second-line drugs.

Use of anti-retroviral therapy in tuberculosis patients on second-line anti-TB regimens: a systematic review.

INTRODUCTION: Use of antiretroviral therapy (ART) during treatment of drug susceptible tuberculosis (TB) improves survival. However, data from HIV infected individuals with drug resistant TB are lacking. Second line TB drugs when combined with ART may increase drug interactions and lead to higher rates of toxicity and greater noncompliance. This systematic review sought to determine the benefit of ART in the setting of second line drug therapy for drug resistant TB. METHODS: We included individual patient data from studies that evaluated treatment of drug-resistant tuberculosis in HIV-1 infected individuals published between January 1980 and December of 2009. We evaluated the effect of ART on treatment outcomes, time to smear and culture conversion, and adverse events. RESULTS: Ten observational studies, including data from 217 subjects, were analyzed. Patients using ART during TB treatment had increased likelihood of cure (hazard ratio (HR) 3.4, 95% CI 1.6-7.4) and decreased likelihood of death (HR 0.4, 95% CI 0.3-0.6) during treatment for drug resistant TB. These associations remained significant in patients with a CD4 less than 200 cells/mm(3) and less than 50 cells/mm(3), and when correcting for drug resistance pattern. LIMITATIONS: We identified only observational studies from which individual patient data could be drawn. Limitations in study design, and heterogeneity in a number of the outcomes of interest had the potential to introduce bias. DISCUSSION: While there are insufficient data to determine if ART use increases adverse drug interactions when used with second line TB drugs, ART use during treatment of drug resistant TB appears to improve cure rates and decrease risk of death. All individuals with HIV appear to benefit from ART use during treatment for TB.

Association between smoking and latent tuberculosis in the U.S. population: an analysis of the National Health and Nutrition Examination Survey.

BACKGROUND: Evidence of an association between cigarette smoking and latent tuberculosis infection (LTBI) is based on studies in special populations and/or from high prevalence settings. We sought to evaluate the association between LTBI and smoking in a low prevalence TB setting using population-based data from the National Health and Nutrition Examination Survey (NHANES). METHODS: In 1999-2000, NHANES assessed LTBI (defined as a tuberculin skin test measurement ≥10 mm) in participants, and those ≥20 years of age were queried regarding their tobacco use and serum cotinine was measured. We evaluated the association of LTBI with self-reported smoking history and smoking intensity in multivariable logistic regression models that adjusted for known confounders (gender, age, birthplace, race/ethnicity, poverty, education, history of BCG vaccination, and history of household exposure to tuberculosis disease). RESULTS: Estimated LTBI prevalence was 5.3% among those ≥20 years of age. The LTBI prevalence among never smokers, current smokers, and former smokers was 4.1%, 6.6%, and 6.2%, respectively. In a multivariable model, current smoking was associated with LTBI (OR 1.8; 95% CI, 1.1-2.9). The association between smoking and LTBI was strongest for Mexican-American and black individuals. In multivariate analysis stratified by race/ethnicity, cigarette packs per day among Mexican-American smokers and cotinine levels among black smokers, were significantly associated with LTBI. CONCLUSIONS: In the large, representative, population-based NHANES sample, smoking was independently associated with significantly increased risks of LTBI. In certain populations, a greater risk of LTBI corresponded with increased smoking exposure.

Tuberculosis Infection: Insight from Immunogenomics.

Tuberculosis continues to be one of the most important global infectious causes of morbidity and mortality. Development of a more effective vaccine is a high worldwide priority and depends on a thorough understanding of the host response to infection. In this review, we highlight recent advances in our understanding of the innate immune response to MTb infection. We also describe recent discoveries in immunogenetics that are generating insight into the potential development of immunomodulatory therapies.