Open Air Scholar

Design:

The study employed a combination of in vivo and in vitro approaches. In vivo experiments used a rat model of HAPE induced by simulated high-altitude hypoxia (equivalent to 6,000 meters). Rats were treated with either control sEVs (sEVNull) or miR-486-overexpressing sEVs (sEVmiR-486). In vitro studies utilized pulmonary microvascular endothelial cells (PMVECs) exposed to hypoxia (CoCl₂) to assess molecular mechanisms. The study design included comparative analyses of treatment efficacy, oxidative stress markers, inflammatory cytokines, and endothelial barrier function.

Procedures:

DPSCs were engineered to overexpress miR-486 using adenoviral vectors (Ad.miR-486). sEVs were isolated via differential ultracentrifugation and characterized by nanoparticle tracking analysis (size: 70–150 nm), Western blotting for sEV markers (Tsg101, CD63, CD9), and DPSC-specific CD90. Uptake studies confirmed sEV internalization by PMVECs and localization in rat lungs post-intravenous injection. HAPE was modeled by exposing rats to low-pressure hypoxia for 72 hours. Outcomes included pulmonary artery pressure (PAP), lung water content, bronchoalveolar lavage fluid (BALF) protein leakage, histopathology, oxidative stress markers (MDA, SOD), cytokine profiling (IL-1β, IL-6, TNF-α, IL-10), tight junction proteins (Occludin, ZO-1), and vasoactive mediators (ET-1, NO). Molecular mechanisms were explored via qPCR, Western blotting, and pathway inhibitors (e.g., Ly294002 for PI3K).

Primary Findings:

Both sEVNull and sEVmiR-486 reduced HAPE-induced pulmonary hypertension (PAP), lung edema (lung water content), and protein leakage in BALF. However, sEVmiR-486 demonstrated superior efficacy, particularly in restoring tight junction integrity (increased Occludin and ZO-1 expression) and reducing histopathological lung damage. Oxidative stress was mitigated in both treatment groups, with sEVmiR-486 further enhancing SOD activity and Nrf2/HO-1 pathway activation. Pro-inflammatory cytokines (IL-1β, IL-6, TNF-α) were significantly reduced, while anti-inflammatory IL-10 was elevated in sEVmiR-486-treated rats. Vasoactive mediators showed improved balance, with reduced vasoconstrictors (ET-1) and increased vasodilators (NO) in sEVmiR-486 groups.

Secondary Findings:

Mechanistically, sEVmiR-486 inhibited PTEN expression, enhancing PI3K/AKT phosphorylation and eNOS activity in PMVECs, which promotes endothelial repair and vasodilation. Blocking PI3K with Ly294002 abolished these effects, confirming pathway dependency. miRNA sequencing confirmed miR-486 enrichment in sEVmiR-486, which directly targeted PTEN mRNA. Additionally, sEVmiR-486 downregulated TRPV4 (a channel linked to permeability) and upregulated aquaporins (AQP-1, AQP-5) and Na⁺/K⁺-ATPase, critical for fluid transport. These findings highlight dual actions on endothelial barrier function and vascular reactivity.

Interpretation:

The study links HAPE pathogenesis to oxidative stress, inflammation, and endothelial dysfunction. sEVmiR-486’s efficacy stems from its dual targeting of PTEN (activating PI3K/AKT/eNOS) and Nrf2/HO-1 pathways, which synergistically combat oxidative damage and restore endothelial integrity. The PI3K/AKT pathway’s central role in endothelial survival underscores its therapeutic potential. The superiority of sEVmiR-486 over sEVNull suggests miR-486’s cargo enhances specificity and efficacy. These findings align with prior studies linking miR-486 to endothelial protection and Nrf2 activation.

Strengths:

The study combines robust in vivo and in vitro models, providing mechanistic insights into HAPE pathophysiology. The use of miRNA-modified sEVs addresses limitations of traditional stem cell therapies, such as immunogenicity and delivery challenges. Clear dose-response relationships and pathway validation via inhibitors strengthen the conclusions.

Limitations:

The animal model may not fully replicate human HAPE’s genetic and environmental complexity. Long-term safety and efficacy of sEVmiR-486 in humans remain untested. Additionally, while miR-486’s role in PTEN inhibition is clear, other potential targets of miR-486 warrant further exploration.

Design:

The experimental study applied lignin-based polyurethane (PU) coatings to cotton and polyester fabrics. Unmodified kraft lignin (provided by Suzano) was incorporated into PU formulations at five concentrations. Coated fabrics underwent thermal curing (110°C for 10 minutes) and were tested for surface morphology, chemical composition, thermal stability, air/water vapor permeability, hydrophobicity, and antibacterial activity. Durability assessments included domestic washing cycles (1, 5, 10, and 20 cycles) and hot pressing at 110°C and 200°C. Antibacterial efficacy was evaluated against Staphylococcus aureus (Gram-positive) and Klebsiella pneumoniae (Gram-negative) using ASTM E2149. Material characterization employed FTIR spectroscopy, TGA, and microscopy, while permeability and wettability were measured via ISO 9237, BS 7209, and contact angle analysis.

Procedures:

Lignin-PU coatings were applied to cotton and polyester substrates using a knife-coating method. Antibacterial testing quantified bacterial reduction percentages using agar diffusion assays. Thermal stability was assessed via TGA, and surface modifications were analyzed via SEM and FTIR. Hydrophobicity was measured via water contact angles, while breathability metrics (air permeability, water vapor permeability index) were evaluated using standardized protocols. Durability tests simulated real-world conditions, with washing cycles conducted in a domestic machine and thermal stress applied via hot pressing. Crosslinking mechanisms and coating adhesion were inferred from microscopic and spectroscopic data.

Primary Findings:

Cotton fabrics coated with ≥10% lignin achieved over 96% antibacterial efficacy against S. aureus, while polyester showed reduced activity (84.3% at 10% lignin). Hydrophobicity improved significantly: cotton reached 112.6° and polyester 110.3° contact angles at 30% lignin. Thermal stability increased with lignin content, delaying cotton’s thermal degradation onset to 250°C (vs. 300°C for uncoated cotton) and enhancing char formation. Coated fabrics exhibited drastically reduced air permeability (cotton: <3 L/m²/s; polyester: <8 L/m²/s) and water vapor permeability (WVP index <45% for cotton, <12% for polyester), compromising breathability. Durability tests revealed rapid antibacterial degradation after five washing cycles, though hydrophobicity and antibacterial activity were preserved under hot pressing (up to 200°C).

Secondary Findings:

SEM analysis showed lignin accumulation at cotton yarn intersections, reducing surface air gaps, while polyester exhibited less pronounced structural masking. FTIR confirmed lignin integration via hydrogen bonding. Non-linear effects were observed in polyester’s air permeability at 30% lignin (7.8 L/m²/s), attributed to microfissures. Antibacterial activity against K. pneumoniae was negligible for both fabrics. Washing induced coating detachment and fiber protrusion, whereas hot pressing improved hydrophobicity (e.g., cotton reached 136.2° at 200°C). Crosslinking agents were proposed to enhance wash resistance.

Interpretation:

Lignin’s polyphenolic groups likely disrupted S. aureus cell membranes, explaining its efficacy against Gram-positive bacteria. Hydrophobicity arose from lignin’s aromatic structure and surface roughness. Thermal stability improvements stemmed from lignin’s char-forming aromatic rings. Reduced breathability reflected coating thickness, while polyester’s lower antibacterial activity was linked to its hydrophobicity limiting microbial adhesion. Durability issues under washing highlight the need for improved adhesion or crosslinking. The findings underscore lignin’s potential for niche applications where breathability is non-essential, such as disposable medical textiles or protective gear.

Strengths:

The study demonstrates lignin’s multifunctional benefits (antibacterial, hydrophobic, thermal stability) using unmodified waste material. It provides a systematic evaluation of performance trade-offs and durability under real-world conditions. Cross-disciplinary methods (microscopy, spectroscopy, mechanical testing) strengthened the analysis.

Limitations:

The lack of antibacterial activity against K. pneumoniae limits applicability. Wash resistance remains a critical barrier for long-term use. The study did not explore lignin modification or blending strategies to address these limitations. Standardized breathability thresholds (e.g., WVP >50%) were not met, restricting use in apparel requiring comfort.

Design:

The study employed a cross-sectional observational design, analyzing data from 8,611 adults aged 20–80 years who underwent polysomnography (PSG) at a single sleep center in Taipei between 2015 and 2021. Participants were excluded if they used continuous positive airway pressure (CPAP) therapy, consumed alcohol, or were undergoing specific medical treatments (e.g., immunosuppressants). Environmental data (PM2.5, RH, temperature) were interpolated to participants’ residential addresses using radial basis function methods, which estimate exposure levels based on nearby monitoring stations. The analysis accounted for seasonal variations, dividing the year into cold (November–April) and hot (May–October) seasons.

Procedures:

Key measures included PSG-derived sleep parameters (sleep efficiency, duration of N1, N2, N3, and REM stages, AHI, and arousal indices). Environmental exposures were assessed using short-term (1-day, 7-day) and long-term (30-day) averages of PM2.5, RH, and temperature. Linear regression models adjusted for age, sex, and body mass index (BMI) were used to quantify associations between environmental variables and sleep outcomes. Mediation analysis further explored whether PM2.5 acted as an intermediary in the relationships between RH, temperature, and sleep parameters. This approach allowed the researchers to disentangle direct effects of meteorological factors from indirect effects mediated by PM2.5, particularly across seasonal contexts.

Primary Findings:

Short-term increases in RH (1-day or 7-day averages) were associated with a 0.04–0.06% rise in N2 sleep and a 0.02–0.03% decrease in N3 sleep, alongside higher arousal indices (0.02–0.04 events/h). Temperature rises of 1°C increased REM sleep by 0.04–0.05% and reduced AHI by 0.2–0.24 events/h but also elevated arousal. PM2.5 demonstrated a mediating role in these relationships, particularly in cold seasons. For example, cold-season PM2.5 and low temperatures amplified light sleep (N1) and arousal, while hot-season PM2.5 effects on sleep stages and arousal were strongly RH-dependent. Long-term temperature increases (7-day or 30-day averages) in cold seasons reduced AHI by 0.41–0.42 events/h and improved N1 and REM sleep. Conversely, in hot seasons, higher RH reduced N3 sleep (0.05–0.08%) but increased N1 and N2 sleep, while temperature rises elevated total arousal (0.07–0.08 events/h) and NREM arousal (0.07–0.09 events/h).

Secondary Findings:

Mediation analysis revealed that PM2.5 mediated RH’s effects on arousal indices and temperature’s effects on sleep stage transitions. Specifically, PM2.5 amplified the negative impact of low RH on sleep efficiency (0.18% reduction in cold seasons) and the positive correlation between cold temperatures and obstructive sleep apnea (OSA) severity. In contrast, higher RH in cold seasons increased total arousal (0.04–0.1 events/h) and REM arousal (0.05–0.11 events/h), while hot seasons saw RH-induced reductions in N3 sleep and temperature-driven increases in arousal. PM2.5 also mediated interactions between temperature and sleep architecture, particularly in REM and NREM stages. These findings underscore PM2.5’s dual role as both a direct pollutant and an environmental mediator influencing sleep through seasonal interactions with RH and temperature.

Interpretation:

The study highlights seasonal modulations of environmental effects on sleep. In cold seasons, lower RH and temperatures may exacerbate airway dryness and inflammation, reducing deep sleep (N3) and increasing arousal, which aligns with known mechanisms of nasal mucociliary dysfunction and osmotic stress. PM2.5 likely worsens these effects by damaging airway barriers via pyroptosis and inducing neurotoxicity through macrophage-derived glutamate, disrupting sleep regulation pathways. In hot seasons, higher RH and temperatures may trigger airway irritation via TRPV1 channels, elevating arousal, while PM2.5’s influence on sleep stages becomes more pronounced under humid conditions. The hypothalamic-pituitary-adrenal (HPA) axis is proposed as a key pathway linking PM2.5 exposure to systemic inflammation and sleep disturbances, as stress responses and cytokine release could fragment sleep. These findings suggest that environmental policies targeting PM2.5 reduction, particularly during cold months, could improve sleep quality and reduce cardiovascular risks. Clinically, optimizing bedroom humidity and temperature based on seasonal patterns may mitigate sleep disruptions.

Strengths:

The study’s strengths include its large sample size (8,611 participants), rigorous use of PSG for objective sleep assessment, and advanced statistical methods (linear regression and mediation analysis) to disentangle direct and indirect effects. The inclusion of seasonal variations provides context-specific insights, while the focus on PM2.5 as a mediator bridges environmental and physiological mechanisms. The use of radial basis function interpolation for environmental data ensures spatially accurate exposure estimates, enhancing ecological validity.

Limitations:

Key limitations include reliance on interpolated environmental data, which may not reflect indoor exposures or individual-level variations. The single-center design limits generalizability to other regions or populations. Additionally, unmeasured confounders—such as indoor air quality, socioeconomic status, or pre-existing respiratory conditions—could influence results. The cross-sectional nature of the study precludes causal inferences, and mediation analysis assumptions (e.g., no unmeasured confounders) require further validation. Lastly, the study did not assess long-term health outcomes directly linked to sleep changes, such as cardiovascular events.

Design:

The study used a randomized controlled design involving 60 male C57BL/6 mice divided into four groups: vehicle control, dinotefuran alone (50 or 100 mg/kg), cetylpyridinium chloride alone (0.031–0.25 mg/kg), and a mixture of both (100 mg/kg dinotefuran + 0.031–0.25 mg/kg cetylpyridinium chloride). Exposure occurred via intratracheal instillation over seven days, with cetylpyridinium chloride administered on days 1, 4, and 7.

Procedures:

Key measures included body weight changes, lung weight, bronchoalveolar lavage fluid (BALF) analysis (cell counts, cytokines), histopathology, immunohistochemistry for TLR4 and PPAR-γ expression, and qRT-PCR for fibrosis-related genes (TGF-β, NF-κB, ACTA2, TNF-α, IL-1β, MMP9). Statistical analyses evaluated dose-dependent and synergistic effects between the compounds.

Primary Findings:

The mixture group exposed to 0.062 mg/kg cetylpyridinium chloride showed significant body weight loss and elevated neutrophils in BALF. Histopathology revealed granulomatous inflammation and fibrosis in high-dose cetylpyridinium chloride groups, with synergistic effects observed in the mixture group. Immunohistochemistry confirmed increased TLR4 expression in mixture-exposed mice, while PPAR-γ was upregulated in cetylpyridinium chloride-only groups. Molecular analyses demonstrated synergistic activation of fibrotic pathways, with elevated TGF-β, NF-κB, ACTA2, TNF-α, IL-1β, and MMP9 mRNA in the mixture group.

Secondary Findings:

Dose-dependent effects were observed for cetylpyridinium chloride, with higher doses (0.25 mg/kg) causing severe fibrosis. The mixture’s synergistic effects were most pronounced at intermediate cetylpyridinium chloride doses (0.062–0.125 mg/kg), suggesting cross-talk between TLR4 and PPAR-γ pathways.

Interpretation:

The findings validate the proposed AOPs, showing that dinotefuran and cetylpyridinium chloride synergistically induce pulmonary fibrosis via TLR4/PPAR-γ cross-talk. The study highlights the importance of evaluating chemical mixtures, as individual compound effects do not fully predict toxicity. The results underscore the utility of AOP frameworks in predicting and assessing complex toxicological outcomes.

Strengths:

The study employs an AOP-based approach, in vivo validation of molecular pathways, and rigorous statistical analysis of synergistic effects. The use of multiple endpoints (histopathology, gene expression, immunohistochemistry) strengthens the evidence for mechanistic links.

Limitations:

The intratracheal instillation model may not fully replicate inhalation exposure in humans. Protein-level validation of mRNA changes and long-term fibrosis progression studies are needed. Additionally, the study does not address potential interactions with other environmental factors.

Design:

The study employed a retrospective cohort design, analyzing data from YF patients admitted to HEM between 2017 and 2018. Participants were confirmed cases through molecular (RT-qPCR, viral isolation) or serological (IgM) testing, excluding those with potential dengue cross-reactivity. The cohort included 283 patients, 87.6% of whom were male. Data were collected from medical records, focusing on demographics, clinical symptoms, laboratory results, and outcomes. The analysis compared fatal and non-fatal cases to identify predictors of mortality.

Procedures:

Key procedures included measuring initial YF viral load (VL) in log10 copies/mL, assessing clinical symptoms at admission (e.g., fever, jaundice), and evaluating laboratory markers such as AST (aspartate aminotransferase), INR (international normalized ratio), lactate, and WBC (white blood cell count). Multivariate logistic regression was used to identify independent predictors of death. Viral load dynamics were tracked over time, and molecular diagnostics were evaluated for their diagnostic utility beyond the World Health Organization’s (WHO) recommended 10-day window for PCR testing. The study also analyzed vaccine efficacy using relative risk (RR) and explored potential associations between VL patterns and antibody-dependent enhancement (ADE) of disease, though this analysis was limited by sample size. Ultrasound findings, including renal cortex echogenicity, were correlated with mortality outcomes.

Primary Findings:

The in-hospital mortality rate was 24%. Male patients (87.6% of the cohort) faced a 4.66-fold higher risk of death (p < 0.01). Fatal cases were associated with older age (median 49.5 vs. 46 years in non-fatal cases, p < 0.01) and higher initial YF viral loads (≥4.45 log10 copies/mL, odds ratio [OR] 12.2). Clinical symptoms linked to mortality included coma, oliguria (reduced urine output), and hematemesis (vomiting blood). Laboratory markers of severe disease included elevated AST (median log-transformed OR 3.65), prolonged INR (OR 7.40), elevated lactate (OR 4.57), and increased WBC (OR 4.33). Multivariate analysis confirmed age (OR 1.06 per year), sex, viral load, AST, INR, lactate, and WBC as independent predictors of death.

Secondary Findings:

High viral loads persisted in fatal cases until Day 9 post-symptom onset (DPS), suggesting VL’s prognostic value even in late-presenting patients. Molecular diagnostics detected YFV RNA in serum up to 26 DPS, exceeding WHO’s 10-day PCR recommendation. A potential rise in VL between DPS 12–15 correlated with antibody appearance, raising questions about ADE, though sample limitations prevented conclusive analysis. Among 96 vaccinated patients, 6 were diagnosed via serology alone (susceptible to false positives due to cross-reactivity), while 90 were PCR-confirmed, hinting at possible waning immunity or reporting errors. The study’s mortality rate (24%) was lower than regional rates (36–44%), attributed to excluding late-stage transfers and milder cases. Ultrasound findings like renal cortex echogenicity mirrored prior studies’ associations with mortality.

Interpretation:

The findings emphasize liver injury (elevated AST), coagulopathy (INR), systemic inflammation (WBC), and metabolic derangement (lactate) as central to YF mortality. The lower viral load threshold for predicting death (≥4.45 log10 copies/mL) compared to Kallas et al.’s 5.1 log10 threshold in an ICU cohort reflects differences in patient populations. The extended PCR utility beyond 10 DPS underscores its critical role in flavivirus-endemic regions, where serological cross-reactivity complicates diagnosis. The protective trend of vaccination (RR 0.545) suggests efficacy but highlights challenges in interpreting breakthrough infections due to incomplete records. Factors like vaccination timing (<1 year or <10 days pre-symptom) may explain some failures, though data gaps limit conclusions.

Strengths:

Strengths include the large sample size (283 patients) and real-world clinical data, providing actionable insights for risk stratification. The integration of viral load dynamics and molecular diagnostics offers a nuanced understanding of YF progression. The study’s focus on non-ICU patients fills gaps in existing literature, which often prioritizes critically ill cohorts.

Limitations:

Limitations stem from the retrospective design, which introduced variability in data collection and potential biases. Non-standardized clinical records limited analysis of comorbidities, which were overrepresented (e.g., alcohol use) compared to state averages. The exclusion of IgG testing for prior flavivirus infections may have obscured interactions like ADE or cross-protection. Small sample sizes for certain analyses (e.g., ADE) and incomplete vaccination histories weakened conclusions about vaccine efficacy.

Design:

An observational, longitudinal study design was employed, analyzing data from the GBD 2021 study for the period 1990–2021. The study population included all residents of China, with analyses stratified by age, gender, and geographic region. Data were aggregated from over 200 sources, including hospital records, mortality registries, and environmental datasets.

Procedures:

Key methods included age-standardized incidence (ASIR) and mortality rates (ASMR), Joinpoint regression to identify inflection points in trends, and age-period-cohort models to disentangle demographic and temporal effects. Bayesian projections were used to forecast trends until 2046. Environmental and behavioral risk factors (e.g., air pollution, smoking rates) were analyzed alongside healthcare access metrics. Demographic data from China’s National Bureau of Statistics were integrated to account for population aging and urbanization.

Primary Findings:

Age-standardized asthma incidence declined by 34% (from 524.81 to 364.17 per 100,000) and mortality by 42% (from 5.82 to 1.47 per 100,000) between 1990 and 2021. Despite falling rates, absolute case numbers increased due to population growth and aging. Males consistently had higher incidence (1.5–2x higher than females) and mortality rates. Age-specific trends revealed peak incidence in children under 5 and rising burden in adults over 60. Mortality was concentrated in elderly populations. Urban areas showed greater declines in incidence than rural regions (e.g., smoking rates fell from 27% to 20% among adults from 2010–2020). Projections suggest ASIR will drop to 330 per 100,000 and ASMR to 0.69 per 100,000 by 2046, but total annual cases may reach 4.5 million and deaths 80,000 due to demographic shifts.

Secondary Findings:

Joinpoint analysis identified steady declines in incidence and mortality without significant inflection points. Age-period-cohort models indicated that earlier birth cohorts faced higher risks, while period effects reflected systemic improvements in healthcare and environmental policies. Pandemic lockdowns temporarily reduced asthma cases (likely due to reduced pollution and outdoor exposure), but long-term impacts remain uncertain. Comorbidities like obesity and COPD emerged as growing concerns, particularly in aging populations.

Interpretation:

Declines in asthma rates reflect improved healthcare access, smoking cessation policies, and environmental regulations (e.g., air quality improvements). However, aging populations and rising comorbidities threaten to offset these gains. Urban-rural disparities highlight inequities in healthcare access, with rural areas lagging in asthma management. The pandemic’s transient effect underscores the role of environmental factors but cautions against overinterpreting short-term trends. Genetic and socioeconomic factors, such as shared pathways with COPD and mental health risks in adolescents, warrant further investigation.

Strengths:

The study leverages robust GBD data and advanced statistical models (e.g., age-period-cohort analysis) to disentangle demographic and temporal influences. Its long-term (31-year) scope and Bayesian projections provide actionable insights for policy planning.

Limitations:

Data uncertainties in regional asthma reporting and underdiagnosis in rural areas may skew estimates. Provincial-level disparities were not fully captured, limiting targeted interventions. Pandemic effects were only partially accounted for, and genetic factors were not directly measured.

Design:

The longitudinal cohort study (2000HIV) included 1,478 discovery participants and 320 validation participants. A parallel healthy cohort (200FG) provided additional validation. Participants were categorized into pre-pandemic, post-lockdown (unvaccinated/uninfected), past SARS-CoV-2 infection, and vaccinated groups. Data collection spanned virological, immunological, and epigenetic measurements.

Procedures:

Proteomic profiling identified 2,367 proteins using mass spectrometry. Cytokine production assays measured responses to stimuli like imiquimod and Streptococcus pneumoniae. DNA methylation was analyzed via EWAS (epigenome-wide association studies), with 2,743 CpG probes retained after quality filtering. Statistical methods included rank-based ANCOVA (adjusting for age, sex, seasonality), NMDS clustering, and pathway enrichment (clusterProfiler, STRING). Validation used Wilcoxon signed-rank tests on paired samples from the healthy cohort. Methylation data were processed with the minfi R package, applying BACON normalization and immune cell proportion adjustments.

Primary Findings:

Lockdowns induced significant immune shifts: reduced systemic inflammation (e.g., lower IL-6, TNF-α) but heightened innate immune responsiveness (increased monocyte cytokine production post-stimulation). Vaccination dampened pro-inflammatory cytokines (TNF-α, IL-1β) while boosting anti-inflammatory IL-1Ra. Only 11 proteins reverted to pre-pandemic levels post-vaccination. Proteomic PCA clusters showed lockdowns drove the largest separation, linked to downregulated innate pathways and upregulated platelet aggregation signaling. DNA methylation analysis revealed lockdowns altered 1,000+ CpG sites in immune genes (e.g., TLR4, NFKBIA), while vaccination affected 500+ sites (e.g., IL6, TNF). Methylation changes inversely correlated between lockdowns and vaccination (R²=0.091, p<2.2e-16). Past mild SARS-CoV-2 infection had minimal long-term immune effects.

Secondary Findings:

Validation in the 200FG cohort confirmed lockdown-associated hyper-responsiveness of immune cells and vaccination-induced cytokine modulation. EWAS identified top CpG sites: cg08255374 (UBE2F/RAMP1) for lockdowns and cg10675725 (GSDMD/MROH6) for vaccination. Pathway analysis highlighted lockdown effects on immune regulation and ubiquitination, while vaccination targeted lymphocyte and HLA-related pathways. DNA methylation levels were globally reduced during lockdowns (p<2.2e-16) and increased post-vaccination (p=1.12e-10), suggesting partial reversal of lockdown effects but persistent pandemic-driven changes.

Interpretation:

Lockdowns likely reduced pathogen exposure, lowering systemic inflammation but priming innate immunity. Vaccination’s broader immune modulation—reducing pro-inflammatory cytokines and enhancing anti-inflammatory responses—may influence susceptibility to non-COVID conditions. The inverse methylation patterns between lockdowns and vaccines suggest opposing mechanisms: lockdowns may suppress immune activation via epigenetic silencing, while vaccines induce protective training through methylation changes. These findings underscore the pandemic’s role in reshaping immune resilience, with implications for chronic disease management in vulnerable populations.

Strengths:

The longitudinal design, multi-omics approach, and validation in a healthy cohort strengthened conclusions. Robust statistical methods (e.g., EWAS, pathway enrichment) provided mechanistic insights. Adjustments for ancestry, batch effects, and immune cell proportions minimized confounding.

Limitations:

Cross-sectional comparisons for vaccination effects limited causal inferences. Self-reported SARS-CoV-2 infection data may have introduced bias. The study focused on asymptomatic PLHIV, limiting generalizability to other populations.

Design:

The study employs a comparative analysis of ML models using observational data from three Swiss stations. The primary dataset (Lugano-Università) spans eight years, supplemented by validation data from two additional urban/suburban stations. Key predictors include O₃, NO₂, NOₓ, sulfur dioxide (SO₂), non-methane volatile organic compounds (NMVOC), temperature, radiation, and particulate matter (PM₂.₅/PM₁₀). Missing data (<1% for most parameters) are excluded rather than imputed to avoid bias. The analysis uses 10-fold cross-validation and statistical metrics (MAE, MSE) to evaluate model performance.

Procedures:

Models tested include linear methods (best subset selection, ridge, lasso, PCR, PLS), non-linear regression with interaction terms, and ANNs (multilayer perceptrons [MLP], recurrent neural networks [RNN], long short-term memory [LSTM], gated recurrent units [GRU], and convolutional neural networks [CNN]). Best subset selection uses criteria like adjusted R² and AIC/BIC to identify optimal predictors. Shrinkage methods (ridge/lasso) address multicollinearity, while ANNs are tuned for layer configurations, activation functions (ReLU, swish), and regularization (dropout). All models are implemented in R (version 4.3.2) and Python (TensorFlow 2.13.1/Keras), with normalization applied to input variables.

Primary Findings:

The non-linear regression model with 12 predictors (radiation, temperature, NO₂, NOₓ, SO₂, NMVOC, and interaction terms) achieved the lowest MAE (9 µg/m³) and MSE (144 µg²/m⁶) in Lugano, outperforming linear models and most neural networks. MLPs with 2–3 hidden layers (e.g., 128–64 units) yielded MAEs of 10–14 µg/m³, comparable to non-linear models, while CNNs underperformed due to the lack of spatial/temporal structure in daily averages. LSTM/GRU models showed minimal improvement over MLPs. Regional validation revealed Lugano's lower MAE (9 µg/m³) versus Zürich/Dübendorf (11–13 µg/m³), attributed to localized traffic and meteorology. Models trained on Lugano data generalized poorly to northern stations.

Secondary Findings:

Best subset selection identified NO₂, NOₓ, SO₂, NMVOC, temperature, and radiation as critical predictors, aligning with ozone chemistry. Shrinkage methods highlighted differences: ridge prioritized NMVOC/CO, while lasso excluded PM₂.₅. PLS explained ~81% variance with fewer components than PCR. Neural networks required extensive hyperparameter tuning (e.g., swish activation, 32–64 units) but exhibited parameter instability (up to 0.8 standard deviation in bootstrap analysis). Residual comparisons showed systemic biases in RNNs versus non-linear models.

Interpretation:

The non-linear model's success reflects its ability to capture non-linear interactions (e.g., radiation³, temperature³) critical to photochemical ozone formation. Radiation and NO₂/NOₓ dominate due to their roles in ozone production and destruction cycles. NMVOCs' inclusion underscores their role as precursors under high NOₓ conditions. Particulate matter's exclusion suggests indirect effects (e.g., aerosol interactions) are negligible at daily scales. Neural networks' comparable performance but higher computational cost highlights a trade-off between interpretability and predictive power.

Strengths:

The study rigorously compares diverse ML approaches, validates across multiple stations, and links findings to atmospheric chemistry. Cross-validation and bootstrap analysis strengthen reliability, while explicit predictor selection criteria enhance transparency.

Limitations:

Missing data exclusion may bias results, and daily averages obscure hourly ozone dynamics. Regional model performance differences suggest spatial variability requires localized calibration. Static input variables ignore ozone's autocorrelation, limiting time-series modeling potential.

Design:

A multicenter observational study using data from the PROVENT-C19 registry, enrolling 1,816 mechanically ventilated patients with COVID-19-induced ARDS from 53 international centers. Participants were followed prospectively or retrospectively, with ethical approval obtained from the Saint Bortolo Hospital Ethics Committee. The study design allowed for analysis of real-world practices while addressing variability in clinical protocols.

Procedures:

Key physiological metrics included changes in oxygenation (PaO₂/FiO₂ ratio), ventilatory ratio (a proxy for dead space ventilation), and respiratory system compliance (Crs) during prone positioning (Delta-PP) and after resupination (Delta-PostPP). Prone positioning duration and ICU mortality were primary outcomes. Statistical analyses included multivariable logistic regression to identify independent predictors of mortality, adjusting for confounders like age, comorbidities, and ventilation parameters.

Primary Findings:

Improvements in oxygenation during prone positioning (Delta-PP PaO₂/FiO₂) strongly predicted reduced ICU mortality, with a 52% lower odds of death per 1 mmHg increase (OR 0.48, 95% CI 0.38–0.59). Post-resupination oxygenation gains (Delta-PostPP PaO₂/FiO₂) also correlated with survival (OR 0.60, 95% CI 0.52–0.68). Non-linear relationships emerged for ventilatory ratio: values exceeding 0.04 during prone positioning and 0.09 post-resupination were linked to increased mortality risk. Respiratory compliance improvements during prone positioning (Delta-PP Crs) independently reduced mortality (OR 0.80, 95% CI 0.65–0.98), while post-resupination changes were non-significant. Prolonged prone duration (≥16 hours) and longer first-cycle durations (each additional hour reduced mortality by 18%, OR 0.82, 95% CI 0.73–0.91) further predicted better outcomes.

Secondary Findings:

Subgroup analyses confirmed consistent effects across prospective and retrospective cohorts. Ventilatory ratio thresholds identified 'negative responders' with elevated mortality risk, suggesting this metric could guide clinical decisions. Prolonged prone sessions (≥24 hours) showed dose-dependent mortality reductions without increased complications like pressure ulcers. Static compliance and oxygenation improvements during the first prone cycle were critical for survival, emphasizing early intervention.

Interpretation:

Enhanced oxygenation and compliance during prone positioning reflect improved lung recruitment and reduced strain, aligning with prior ARDS studies. The ventilatory ratio’s non-linear relationship highlights that worsening dead space may signal unfavorable lung mechanics, warranting re-evaluation of prone strategies. Longer prone durations likely optimize lung recruitment and sustained benefits. These findings underscore the need for personalized protocols tailored to individual physiological responses.

Strengths:

The study’s large international cohort and detailed physiological measurements provide robust evidence for prone positioning’s efficacy in severe ARDS. The inclusion of both prospective and retrospective data enhances generalizability. Standardized metrics like PaO₂/FiO₂ and ventilatory ratio offer actionable tools for clinical decision-making.

Limitations:

Observational design limits causal inference, and unmeasured confounders (e.g., comorbidities, ECMO use) may influence results. The focus on a homogeneous cohort (severe ARDS from COVID-19) restricts extrapolation to non-COVID populations. Variability in prone positioning practices across centers introduces potential heterogeneity.