Daniil Lisik

Background: Sleep disturbances are highly prevalent in chronic obstructive pulmonary disease (COPD) but are not incorporated into current phenotypic classifications. We examined whether multidimensional sleep data identify clinically relevant patterns in COPD and whether these patterns are associated with all-cause mortality. Methods: In a clinical cohort of 205 patients with COPD, home-based sleep assessment was performed using the WatchPAT™ device. Fifteen complementary variables reflecting sleep duration and architecture, nocturnal oxygenation, sleep-disordered breathing, and daytime sleepiness were analysed using Partitioning Around Medoids clustering. The resulting phenotypes were characterised by pulmonary function, gas exchange, functional capacity, systemic biomarkers, and comorbidity burden. Longitudinal follow-up was available for 199 participants; survival was assessed using Kaplan-Meier analysis and Cox regression adjusted for age, sex, post-bronchodilator FEV₁ and positive airway pressure (PAP) treatment. Results: Five distinct sleep phenotypes were identified: preserved sleep physiology (good sleepers, 20.5%), curtailed sleep duration (short sleepers, 33%), altered sleep architecture with REM suppression (light sleepers, 18.5%), predominant nocturnal hypoxaemia (hypoxic sleepers, 20%), and severe sleep-disordered breathing (apnoeic sleepers, 8%). Despite similar airflow limitation and COPD symptom burden, phenotypes differed in gas exchange, functional capacity, systemic inflammation, and cardiometabolic comorbidities. During a median follow-up of 4.2 years, 56 participants (28.1%) died. Compared with good sleepers, hypoxic sleepers had a higher hazard of death after adjustment for age, sex, and FEV₁ (HR 2.80, 95% CI 1.05– 7.44). Following adjustment for PAP treatment, a higher mortality hazard was also observed among untreated apnoeic sleepers (HR 6.16, 95% CI 1.56– 24.38). Conclusion: Multidimensional sleep phenotyping identified clinically distinct clusters of patients with COPD beyond conventional spirometric severity and apnoea–hypopnoea index (AHI)-defined sleep-disordered breathing. These findings support sleep as a clinically relevant dimension of COPD heterogeneity, and warrant prospective validation of sleep-related traits and their clinical outcomes.

Background: Asthma is a heterogeneous disease with wide variation in clinical presentation and prognosis. Allergic and non-allergic asthma are two common phenotypes with important pathophysiological differences. Whether they differ in mortality, however, is not known. Aims and objectives: To determine the association of allergic and non-allergic asthma with all-cause mortality in the general population. Methods: Survey data from two population-based cohorts (N=55,841; baseline age 16-75 years) in Sweden were used. Asthma was based on self-reported disease with current medication use and/or wheeze/attacks of shortness of breath in the past year. Allergic asthma was defined as asthma with ever allergic rhinitis/conjunctivitis. National mortality register data (5-21 years) were linked to the cohort data. Cox proportional hazards models were used, adjusting for age, sex, education, smoking status, and cohort. Results: Across 582,175 person-years, 3,030 deaths occured. The hazard ratio (HR) of all-cause mortality was increased in both non-allergic (HR 1.32, 95%CI 1.09-1.6) and allergic asthma (HR 1.57, 95%CI 1.33-1.85) compared to no asthma (Figure 1). Conclusions: Both allergic and non-allergic asthma carry a higher mortality risk than no asthma. Further research is needed to elucidate mechanisms. erj;68/suppl_70/PA2704/F1 F1 F1 Figure 1. Overview of study design and mortality analyses.

Background: As a surrogate for eosinophilic activation, eosinophil‐derived neurotoxin (EDN) is a potential clinical biomarker. However, EDN levels and discriminatory ability in different asthma phenotypes are unknown. We quantified serum and nasal lavage fluid (NLF) EDN levels and assessed the potential to differentiate clinically defined asthma phenotypes in an adult‐representative sample. Methods: A total of 1499 serum and 386 NLF samples from individuals with current asthma obtained from the West Sweden Asthma Study were analyzed for EDN. Eosinophilic asthma was defined as blood eosinophil count of ≥ 300 cells/mm ³ , T2‐high asthma was defined as blood eosinophil count of ≥ 300 cells/mm ³ or fractional exhaled nitric oxide (FeNO) levels of ≥ 25 ppb. Other asthma phenotypes were defined based on presence of atopy, chronic rhinosinusitis (CRS), nasal polyposis, and obesity. Results: Subjects with eosinophilic asthma had higher serum and NLF EDN levels than those with non‐eosinophilic asthma. In ROC analyses, serum EDN provided excellent discrimination between eosinophilic and non‐eosinophilic asthma (area under curve [AUC] = 0.84, 95% CI = 0.82–0.86); the corresponding AUC for NLF EDN was 0.67 (95% CI = 0.62–0.73). Serum and NLF EDN were higher in atopic asthma, T2‐high asthma, and asthma with nasal polyposis compared to their counterparts, but not in asthma with CRS. The highest absolute values of serum and NLF EDN were observed in the high eosinophil + FeNO group, followed by the high‐eosinophil‐only and high‐FeNO‐only groups. Conclusion: Both serum and NLF EDN were higher in those with eosinophilic compared to non‐eosinophilic asthma. However, only serum EDN appears to distinguish eosinophilic asthma in ROC analyses.

Background: Machine learning (ML)-based cluster analysis is a common method for subtyping medical conditions and presentations of disease states. Recent advancements in algorithms and increasing access to vast healthcare data have further increased the use of such ML models. However, practical implementation is lacking, largely due to methodological limitations and insufficient reporting and clinical contextualization in the extant literature, with no existing guidelines for this purpose. Objective: To propose a general framework to assess and ensure clinical utility of ML-derived clusters. Discussion: The proposed framework encompasses three domains, focusing on clinical relevance, stability and generalizability, and ease of identification, enabling researchers and readers to ensure clinical utility of ML-generated subgroups.

Rapid advancements in artificial intelligence (AI), in combination with increased availability of rich large-scale clinical data, has paved the way for promising implementations in both diagnosing/subtyping as well as managing of sleep disordered breathing (SDB). A central strength of AI in this regard is how it facilitates analysis of complex multidimensional/modal data, for example pertaining to comorbidities and so-called treatable traits. However, the utility of such applications remains somewhat limited, as most AI models are so-called “black boxes”, for which it is not intuitive to assess how an output was arrived at. This poses challenges for ensuring patient trust and controlling bias. Explainable AI (XAI), which constitutes techniques to either distill “black box” models to simpler ones or to elucidate influential patterns by surrogate “white box” models, may provide such insights. Adoption of XAI within this field, however, is still at a nascent stage, and overall, the optimal role of AI alongside clinicians continues to be unclear. This review provides a clinically oriented overview of state-of-the-art implementations of AI and XAI in SDB. In addition, we discuss limitations and trade-offs with XAI and propose a general framework for personalized management of obstructive sleep apnea using XAI and treatable traits.

Background: Fractional exhaled nitric oxide (FE_NO) is used to differentiate asthma inflammatory phenotypes and guide its management. However, data on FE_NO reference values in a representative adult population is limited. We aim to derive reference values and determinants of FE_NO in a representative adult population.Methods: The West Sweden Asthma Study is a clinical‐epidemiological population‐representative study of randomly selected adults in Western Sweden. From this cohort, 943 subjects participated in comprehensive clinical investigations, including skin prick testing (SPT), specific immunoglobulin E (sIgE) analysis, and FE_NO measurement. Clinical allergy was defined as co‐occurrence of atopy (positivity to SPT or sIgE) and self‐reported allergic symptoms to the same allergen family. FE_NO levels were analysed in relation to the presence or absence of clinical allergy, asthma, and other factors.Results: The 95^th percentile of FE_NO ranged from 34 parts per billion (ppb) in those between 30 and 40 years old to 52 ppb in those ≤ 30 years old in the entire sample (N = 943), and from 26 to 37 ppb in those without clinical allergy, asthma, or chronic obstructive pulmonary disease (COPD) (n = 587), depending on age. Sex, smoking, clinical allergy, atopy, asthma, and hypertension influenced FE_NO levels, meanwhile, age, asthma, clinical allergy, and reversibility‐related variables were significant determinants of FE_NO levels.Conclusion: The 95^th percentile (upper normal limit) for FE_NO ranges from 34 to 52 ppb overall, and from 26 to 37 ppb in those without clinical allergy, asthma, or COPD, depending on age. These findings provide a guide for interpreting FE_NO in the general population.