Daily Respiratory Research Analysis
Three impactful respiratory studies stand out today: a nationwide French cohort characterizes the post-pandemic resurgence of pediatric Mycoplasma pneumoniae and identifies risk factors for critical care; a multi-centre analysis demonstrates that an AI model (TabPFN) robustly predicts high-flow nasal cannula (HFNC) failure despite realistic vital sign measurement errors; and a translational study reveals circulating CD8+CX3CR1+ T cells as predictive precursors of response to neoadjuvant immune c
Summary
Three impactful respiratory studies stand out today: a nationwide French cohort characterizes the post-pandemic resurgence of pediatric Mycoplasma pneumoniae and identifies risk factors for critical care; a multi-centre analysis demonstrates that an AI model (TabPFN) robustly predicts high-flow nasal cannula (HFNC) failure despite realistic vital sign measurement errors; and a translational study reveals circulating CD8+CX3CR1+ T cells as predictive precursors of response to neoadjuvant immune checkpoint blockade in NSCLC.
Research Themes
- Post-pandemic resurgence and risk stratification in pediatric Mycoplasma pneumoniae
- Robust AI prediction of HFNC outcomes under measurement error
- Immune correlates and biomarkers of neoadjuvant ICB response in NSCLC
Selected Articles
1. Post-COVID-19 resurgence of Mycoplasma pneumoniae infections in French children (ORIGAMI): a retrospective and prospective multicentre cohort study.
In a 37-centre French cohort of 969 hospitalized children during the 2023–2024 resurgence, PCR-confirmed Mycoplasma pneumoniae frequently caused pneumonia and occasionally cutaneous disease. PICU admission (6%) was associated with age >11 years, asthma, comorbidities, and erythema multiforme, while macrolide resistance was uncommon among tested samples.
Impact: This nationwide cohort provides timely risk stratification for pediatric M. pneumoniae admissions post-pandemic and supports current macrolide-based management while underscoring surveillance needs.
Clinical Implications: Prioritize monitoring and early escalation for older children, asthmatics, and those with erythema multiforme or comorbidities; macrolides remain appropriate first-line therapy with continued resistance surveillance.
Key Findings
- Among 969 hospitalized children, 97% were PCR-positive and 87% of those with respiratory involvement had pneumonia.
- Cutaneous manifestations occurred in 14%, with erythema multiforme comprising 42% of cutaneous cases.
- Macrolides were prescribed in 95% of antibiotic-treated cases; macrolide resistance was detected in 1/21 (5%) tested samples.
- PICU admission occurred in 6%; risk factors included age >11 years (aOR 2.0), asthma (aOR 2.2), other comorbidities (aOR 2.1), and erythema multiforme (aOR 3.7).
Methodological Strengths
- Nationwide, multicentre design with large sample size and standardized laboratory confirmation (PCR/serology).
- Pre-registered observational protocol and multivariable modeling for PICU risk factors.
Limitations
- Antimicrobial resistance testing was performed in a small subset (21 samples), limiting precision of resistance estimates.
- Hospital-based cohort may over-represent severe cases; mix of retrospective and prospective data collection.
Future Directions: Develop and validate clinical risk scores incorporating age, asthma, comorbidities, and erythema multiforme; expand macrolide resistance surveillance and assess outcomes by timely macrolide initiation.
2. Evaluating the effect of heart and respiratory rate measurement errors on the ability to predict the outcome of high flow nasal cannula therapy: a multi-centre study.
Using first 2-hour data from 596 training and 241 external validation HFNC episodes, an ML model (TabPFN) outperformed ROX-based indices for predicting HFNC failure despite simulated RR/HR counting errors. Adding arterial blood gas data further improved performance and robustness.
Impact: Demonstrates clinically robust AI that accounts for common bedside measurement errors, addressing a key barrier to real-world deployment of HFNC decision support.
Clinical Implications: Adopting robust ML models can improve early identification of HFNC failure risk; accuracy could be enhanced by reducing RR/HR measurement error via better monitoring and incorporating arterial blood gas values.
Key Findings
- TabPFN achieved external validation Accuracy 0.79 (95% CI 0.73–0.84) and AUC 0.86 (0.82–0.89), outperforming ROX variants (Accuracy 0.71; AUC 0.78) despite simulated RR/HR counting errors.
- Simulated 30-second and 15-second manual counting windows for RR and HR degraded performance more for TabPFN than ROX, but TabPFN remained superior overall.
- Adding arterial blood gas measurements improved TabPFN performance and robustness, with no improvement for ROX-based indices.
Methodological Strengths
- External validation across two countries with bootstrapping and Monte Carlo error simulations.
- Use of trial-derived training data (RENOVATE) with standardized HFNC protocols and early time-window features.
Limitations
- Retrospective modeling without prospective interventional testing; generalizability to other settings and sensors needs confirmation.
- Measurement error was simulated from literature rather than measured contemporaneously at the bedside.
Future Directions: Prospective impact trials integrating real-time data capture and automated RR/HR measurement; evaluation of clinical outcomes (escalation timing, intubation, mortality) and fairness across subgroups.
3. Systemic activation and tissue infiltration of CD8 + CX3CR1 + T cells in non-small cell lung cancer treated with neoadjuvant immune checkpoint blockade.
In NA-ICB–treated NSCLC, CD8+CX3CR1+ T cells expanded systemically and within tumors in responders, with evidence of clonal sharing across blood and tumor and differentiation toward exhausted and NK-like cytotoxic states. These cells represent circulating cytotoxic precursors linked to effective responses and may serve as predictive biomarkers and therapeutic targets.
Impact: Provides mechanistic, multi-omic evidence linking a specific circulating CD8+ subset to NA-ICB response, informing biomarker development and next-generation adoptive cell strategies in lung cancer.
Clinical Implications: CD8+CX3CR1+ T-cell frequencies and clonotypes in blood could serve as minimally invasive biomarkers to predict NA-ICB response; enrichment strategies for these cells may augment adoptive therapies.
Key Findings
- CD8+CX3CR1+ T cells were significantly enriched in tumors of responders (scRNA-seq proportion p=0.0027; flow cytometry frequency p=0.021) with clonal expansion.
- Longitudinal blood analyses showed post-treatment proliferation and shared TCR clonotypes across blood and tumor, suggesting trafficking.
- Pseudotime trajectories indicated differentiation into exhausted and NK-like cytotoxic states upon tumor infiltration.
Methodological Strengths
- Integrated single-cell transcriptomics, TCR repertoire mapping, and flow cytometry across paired tissues and blood.
- Longitudinal sampling enabling assessment of systemic proliferation and trafficking.
Limitations
- Modest sample size and observational design limit causal inference and generalizability.
- Functional validation in vivo and prospective predictive performance testing were not performed.
Future Directions: Prospective validation of blood-based CD8+CX3CR1+ biomarkers for NA-ICB response and development of enrichment/engineering strategies to enhance their cytotoxic potential in adoptive therapies.