Daily Respiratory Research Analysis
Three studies advance respiratory medicine across diagnosis, mechanisms, and data science: a multicenter trial shows stool Xpert Ultra can augment tuberculosis detection in people with HIV, particularly with low CD4 counts; a mechanistic investigation identifies ICOS+ CD4+ T cells in aged hosts as drivers of anti–PD-1-related lung toxicity, suggesting actionable biomarkers and targets; and an externally validated EHR/radiology machine-learning model accurately identifies ARDS across institutions
Summary
Three studies advance respiratory medicine across diagnosis, mechanisms, and data science: a multicenter trial shows stool Xpert Ultra can augment tuberculosis detection in people with HIV, particularly with low CD4 counts; a mechanistic investigation identifies ICOS+ CD4+ T cells in aged hosts as drivers of anti–PD-1-related lung toxicity, suggesting actionable biomarkers and targets; and an externally validated EHR/radiology machine-learning model accurately identifies ARDS across institutions.
Research Themes
- Stool-based molecular testing to enhance TB diagnosis in people with HIV
- Mechanisms and biomarkers of immunotherapy-related lung toxicity in aging
- Cross-institutional machine learning for ARDS phenotyping using EHR and radiology
Selected Articles
1. ICOS+CD4+ T cells define a high susceptibility to anti-PD-1 therapy-induced lung pathogenesis.
In aged tumor-bearing mice, anti–PD-(L)1 therapy induces ICOS+ CD4+ T cell activation that drives germinal center B cell responses and lung injury mimicking irAEs; blocking ICOS–ICOSL attenuates damage, while local IL-21 restores it. Adoptive transfer and single-cell data show both aged host milieu and pathogenic CD4+ T cells are required. In patients, CD4+ T-cell ICOS upregulation correlates with later irAE incidence.
Impact: This study uncovers an age-amplified, ICOS+ CD4+ T cell–centric mechanism for anti–PD-1 lung toxicity and validates a translational biomarker/target, informing risk stratification and preventive strategies for irAEs in the elderly.
Clinical Implications: Monitoring ICOS expression on CD4+ T cells could anticipate lung irAEs in patients on anti–PD-1 therapy, especially the elderly. Therapeutically, modulating the ICOS–ICOSL axis or downstream IL-21 signaling may mitigate lung toxicity without broadly suppressing antitumor immunity.
Key Findings
- Anti–PD-(L)1 therapy induced ICOS+ CD4+ T cell activation and ectopic T/B cell infiltration with antibody deposition in aged, not young, mouse lungs.
- Blocking ICOS–ICOSL reduced germinal center B cell differentiation and lung injury; local IL-21 administration reversed protection.
- Adoptive transfer showed both pathogenic aged lung CD4+ T cells and an aged host environment were required for irAE-like responses.
- In cancer patients, increased ICOS expression on CD4+ T cells associated with later irAE incidence.
Methodological Strengths
- Integrated in vivo aging models with adoptive transfer and single-cell transcriptomics to dissect mechanisms.
- Translational validation in patient samples linking ICOS upregulation to irAE incidence.
Limitations
- Predominantly preclinical murine data; generalizability across tumor types and human heterogeneity remains to be established.
- Clinical cohort details (sample size, timing, confounders) for ICOS association were limited in the abstract.
Future Directions: Prospective studies to validate ICOS+ CD4+ T cells as predictive biomarkers and trials testing ICOS–ICOSL modulation or IL-21 pathway interventions to prevent/manage lung irAEs, with stratification by age.
Managing immune-related adverse events (irAEs) caused by cancer immunotherapy is essential for developing effective and safer therapies. However, cellular mechanism(s) underlying organ toxicity during anti-PD-(L)1 therapy remain unclear. Here, we investigated the effect of chronological aging on anti-PD-(L)1 therapy-induced irAE-like lung toxicity, utilizing tumor-bearing aged mice. Anti-PD-(L)1 therapy facilitated ectopic infiltration of T and B cells, and antibody deposition in lungs of aged but not young mice. Adoptive transfer of aged lung-derived CD4+ T cells into TCR-deficient mice revealed that both pathogenic CD4+ T cells and an aged host environment were necessary for the irAE-inducible responses. Single-cell transcriptomics of lung-infiltrating cells in aged mice demonstrated that anti-PD-(L)1 therapy elicited ICOS+CD4+ T cell activation. Disruption of the ICOS-ICOSL interaction attenuated germinal center B cell differentiation and subsequent lung damage, which were overcome by local administration of IL-21 in the lungs of anti-PD-1 therapy-treated aged mice. Therefore, ICOS+CD4+ T cells elicited under an aged environment exacerbated aberrant immune responses and the subsequent lung dysfunction. Consistent with the findings from the mouse model, ICOS upregulation in CD4+ T cells was associated with later irAE incidence in patients with cancer. These finding will help development of useful strategies for irAE management in patients with cancer, many of whom are elderly.
2. Performance of stool Xpert MTB/RIF Ultra for detection of Mycobacterium tuberculosis among adults living with HIV: a multicentre, prospective diagnostic study.
In 677 adults with HIV across three African countries, stool Xpert Ultra showed overall sensitivity of 23.7% and specificity of 94.0% versus a composite reference; sensitivity rose to 45.5% when CD4 ≤200 cells/μL. Stool Ultra added 23–33% extra cases compared with sputum Ultra, sputum culture, or urine TB-LAM, supporting its role as an adjunct diagnostic in PLHIV.
Impact: Provides prospective, multicountry evidence that stool Ultra augments TB detection in PLHIV—especially the immunosuppressed—filling a critical gap where sputum is unobtainable or paucibacillary.
Clinical Implications: Incorporate stool Xpert Ultra as an adjunct test in TB diagnostic algorithms for PLHIV—prioritizing those with CD4 ≤200 cells/μL—and combine with urine LAM and sputum testing to maximize yield.
Key Findings
- Overall stool Ultra sensitivity 23.7% (95% CI 16.4–32.4) and specificity 94.0% (95% CI 91.7–95.9) versus composite reference.
- Higher sensitivity in advanced immunosuppression: 45.5% with CD4 ≤200 cells/μL vs 21.3% with CD4 >200.
- Stool Ultra added 23–33% additional cases compared with sputum Ultra, sputum culture, and urine TB-LAM across all tested.
- Diagnostic yield among all treated: stool Ultra 9%, urine TB-LAM 12%, sputum Ultra 6%, sputum culture 4%.
Methodological Strengths
- Prospective, multicenter design across three countries with standardized processing.
- Use of composite microbiologic reference including WHO-recommended tests; CD4-stratified analyses and yield comparisons.
Limitations
- Overall sensitivity was modest; performance may vary with stool processing and bacillary load.
- Study population limited to adults with HIV; generalizability to HIV-negative populations not assessed.
Future Directions: Optimize stool processing workflows and evaluate integration with triage tools (e.g., clinical scores, digital CXR) in implementation studies; assess cost-effectiveness and impact on time-to-treatment.
BACKGROUND: When people living with HIV develop pulmonary tuberculosis, it often manifests without detectable acid-fast bacilli on sputum microscopy. We aimed to assess the diagnostic accuracy of stool Xpert MTB/RIF Ultra (hereafter, Ultra) for Mycobacterium tuberculosis detection among adults with HIV. METHODS: This multicentre, prospective diagnostic accuracy study was done in outpatient and inpatient health centres in Eswatini, Mozambique, and Uganda. We enrolled adults aged 15 years and older with HIV with presumptive tuberculosis. We evaluated the diagnostic accuracy of stool Ultra using the simple one-step processing method against a composite microbiological reference standard (CMRS) including three WHO-recommended tuberculosis diagnostic tests (urine tuberculosis biomarker-based lateral-flow lipoarabinomannan [TB-LAM], sputum Ultra, and sputum culture), and stratified by CD4 cell count. We compared sputum versus stool Ultra performance against a composite reference standard of TB-LAM and sputum culture (CMRS2). We also calculated the diagnostic yield among all tested. This study is registered with ClinicalTrials.gov, NCT05047315. FINDINGS: Between Dec 2, 2021, and Aug 14, 2024, 677 participants were enrolled (247 [36%] men and 430 [64%] women). Tuberculosis was microbiologically confirmed in 119 participants by the CMRS: 39 (33%) had a positive test with sputum Ultra, 30 (25%) had a positive test with culture, and 84 (71%) had a positive test with urine TB-LAM. The sensitivity of stool Ultra compared with CMRS was 23·7% (28/118 [95% CI 16·4-32·4]) and the specificity was 94·0% (504/536 [91·7-95·9]). The sensitivity of stool Ultra in participants with CD4 counts less than or equal to 200 cells per μL was 45·5% (10/22 [24·4-67·8]) compared with 21·3% (17/80 [12·9-31·8]) in those with CD4 counts greater than 200 cells per μL. Against the CMRS2, we observed no differences in sensitivity between sputum and stool Ultra on the basis of CD4 cell count. Stool Ultra resulted in additional cases detected of 23% (30/133) compared with sputum Ultra, 29% (38/133) compared with sputum culture, and 33% (44/133) compared with TB-LAM. The overall diagnostic yield for all treated for stool Ultra was 9% (60/677), for TB-LAM was 12% (84/677), for sputum Ultra was 6% (39/677), and for sputum culture was 4% (30/677). INTERPRETATION: These results suggest stool Ultra could be used as an additional test for tuberculosis diagnosis among people with HIV, particularly among those with CD4 counts less than 200 cells per μL. FUNDING: EDCTP2 and EDCTP3 Programmes supported by the EU.
3. Development and External Validation of a Detection Model to Retrospectively Identify Patients With Acute Respiratory Distress Syndrome.
Using structured EHR data plus radiology reports, a regularized logistic model achieved AUROC 0.91 (internal) and 0.88 (external) with good calibration (ICI 0.13). At a set threshold, sensitivity/specificity were both 80%, PPV 64%, and cases were identified a median 2.2 hours after Berlin criteria, enabling robust retrospective ARDS phenotyping across health systems.
Impact: The model standardizes retrospective ARDS identification using routinely collected data and generalizes across systems, enabling reproducible cohort building, quality measurement, and multicenter research.
Clinical Implications: Hospitals and researchers can apply this validated EHR+radiology model to consistently identify ARDS cases for quality improvement, surveillance, and research; prospective adaptation could support earlier recognition and trial enrollment.
Key Findings
- Regularized logistic EHR+radiology model achieved AUROC 0.91 (internal) and 0.88 (external), with external ICI 0.13.
- At a prespecified threshold, sensitivity and specificity were both 80%, with PPV 64%.
- Model identified ARDS a median 2.2 hours after meeting Berlin criteria, enabling timely retrospective capture.
- Physician-adjudicated labels and cross-system validation support generalizability.
Methodological Strengths
- Physician-adjudicated ARDS labels with internal and external validation.
- Integration of structured EHR features with radiology reports; calibration assessed.
Limitations
- Retrospective design; model identifies ARDS shortly after criteria are met rather than predicting pre-onset.
- External validation limited to two health systems; performance in broader settings and languages requires testing.
Future Directions: Prospective implementation for early ARDS recognition, integration with bedside alerts, and testing across diverse health systems and international datasets.
OBJECTIVE: The aim of this study was to develop and externally validate a machine-learning model that retrospectively identifies patients with acute respiratory distress syndrome (acute respiratory distress syndrome [ARDS]) using electronic health record (EHR) data. DESIGN: In this retrospective cohort study, ARDS was identified via physician-adjudication in three cohorts of patients with hypoxemic respiratory failure (training, internal validation, and external validation). Machine-learning models were trained to classify ARDS using vital signs, respiratory support, laboratory data, medications, chest radiology reports, and clinical notes. The best-performing models were assessed and internally and externally validated using the area under receiver-operating curve (AUROC), area under precision-recall curve, integrated calibration index (ICI), sensitivity, specificity, positive predictive value (PPV), and ARDS timing. PATIENTS: Patients with hypoxemic respiratory failure undergoing mechanical ventilation within two distinct health systems. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: There were 1,845 patients in the training cohort, 556 in the internal validation cohort, and 199 in the external validation cohort. ARDS prevalence was 19%, 17%, and 31%, respectively. Regularized logistic regression models analyzing structured data (EHR model) and structured data and radiology reports (EHR-radiology model) had the best performance. During internal and external validation, the EHR-radiology model had AUROC of 0.91 (95% CI, 0.88-0.93) and 0.88 (95% CI, 0.87-0.93), respectively. Externally, the ICI was 0.13 (95% CI, 0.08-0.18). At a specified model threshold, sensitivity and specificity were 80% (95% CI, 75%-98%), PPV was 64% (95% CI, 58%-71%), and the model identified patients with a median of 2.2 hours (interquartile range 0.2-18.6) after meeting Berlin ARDS criteria. CONCLUSIONS: Machine-learning models analyzing EHR data can retrospectively identify patients with ARDS across different institutions.