Objective.To test whether machine learning (ML) models trained on tidal breathing flow time series can discriminate between individuals with and without respiratory disease and predict lung function indices obtained from conventional pulmonary function testing.Background.accurate assessment of respiratory function in infants and young children is challenging because conventional pulmonary function testing requires sophisticated equipment and/or active patient cooperation. Tidal breathing measurements, in contrast, can be obtained non-invasively with little or no patient cooperation and at low cost, yet their clinical utility has been limited. We hypothesized that sufficiently long tidal breathing flow time series contain clinically relevant information that can be extracted using a recurrent neural network known as a long short-term memory (LSTM) network.Approach.We evaluated LSTM models in two scenarios within the Basel-Bern Infant Lung Development cohort. First, we assessed the ability of a model trained on flow and derived volume time series to detect bronchopulmonary dysplasia (BPD) in 329 infants. Second, we examined whether a model trained on tidal breathing flow alone could predict forced expiratory volume in one second (FEV1) in 135 school-age children. Signals were filtered and normalized prior to model training, and performance was evaluated on held-out test datasets.Main results.For BPD detection, the model achieved 97.0% accuracy, 100% specificity, 91.7% sensitivity, 100% precision, and an F1-score of 95.7%. For FEV1prediction, Bland-Altman analysis showed a mean bias of -0.009 l (95% CI -0.091-0.074), with limits of agreement of -0.416 l and 0.399 l. The mean relative prediction error was 13.7%.Significance.These findings demonstrate that temporal patterns in tidal breathing flow signals contain diagnostically and functionally relevant information. ML applied to tidal breathing measurements may provide a low-burden, minimal-cooperation approach for early respiratory disease detection and functional assessment across early life stages.

Withdrawal Statement The authors have withdrawn this manuscript because unresolved issues concerning contribution assessment and authorship attribution remain under discussion among the co-authors. Therefore, the authors do not wish this work to be cited as a reference. If you have any questions, please contact the corresponding author.

Bronchopulmonary dysplasia (BPD) is the most common respiratory complication after preterm birth. Early preventive measures are important to reduce further damage on lung tissue. Thus, early discrimination between infants with and without BPD is of high importance. A low sample entropy (SampEn) of time series of the oxyhemoglobin saturation (SpO2-SampEn for short) is associated with an increased risk of hypoxemic events in neonates. We hypothesized that preterm infants have a lower SpO2-SampEn compared to term infants. Moreover, that infants with BPD have a lower SpO2-SampEn compared to those without BPD. Preterm infants < 32 w gestation and healthy term infants were eligible for study. We recorded SpO2 over 90 min through the clinical monitoring system with a sampling frequency of 0.98 Hz and calculated the SpO2-SampEn at 32 w postmenstrual age (PMA), 36 w PMA, and at discharge. We included 95 term and 180 preterm infants, of whom 44 (24.4%) developed BPD. SpO2-SampEn was lower in preterm infants compared to term infants. SpO2-SampEn was lower in infants with BPD compared to infants without BPD at 32 w PMA. However, gestational age was the only predictor of SampEn at 32 w PMA. This difference between infants with and without BPD was no longer present at 36 w PMA and discharge. SpO2-SampEn can be utilized to discriminate between preterm and term infants and between preterm infants with and without BPD. However, confounding factors such as caffeine therapy, gestational age and the natural boundary of 100% of SpO2 values have to be considered.

Introduction: We assessed whether longitudinal measurements of sample entropy (SampEn) of heart rate time series reflect postnatal maturation in preterm infants and evaluated its predictive value at 32 weeks of postmenstrual age (PMA) for estimating discharge home. We further compared SampEn of preterm infants at discharge with that of term infants. METHODS:We conducted a prospective study at the University Children's Hospital Basel, Switzerland, from 2018 to 2022. We included preterm infants born before 32 weeks of gestation and a control group of term infants. Heart rate was recorded using the clinical monitoring system. We assessed preterm infants at 32 and 36 weeks of PMA, and at discharge, term infants were evaluated between 3 and 28 days of life. SampEn was calculated from 90-min recordings using custom analytical software. RESULTS:We obtained valid data from 183/183 preterm infants (mean [range] 28.4 [23.3-31.7] weeks of gestation) and from 80/104 (76%) term infants. In preterm infants, SampEn increased from 32 to 36 weeks of PMA (0.35 vs. 0.40; p < 0.01) without further increase to discharge. SampEn was positively associated with maturation and growth, and negatively with complications of prematurity, particularly with bronchopulmonary dysplasia. SampEn at 32 weeks of PMA did not improve clinical predictions of PMA at discharge. At discharge, SampEn did not differ significantly between preterm and term infants. CONCLUSION:SampEn of heart rate time series increased with postnatal maturation in preterm infants, reaching values of term infants at discharge. It was negatively associated with complications of prematurity but its prognostic value for discharge timing is limited. .

AimWe evaluated whether sample entropy of heart rate time series could serve as a biomarker for guiding caffeine cessation in preterm infants treated for apnoea of prematurity (AOP). We also assessed associations of sample entropy with weeks of gestation, clinical morbidity, AOP frequency and caffeine reinitiation.MethodsWe conducted a prospective single-centre study at the University Children's Hospital Basel, Switzerland, from July 2019 to June 2020. We included 61 hospitalised preterm infants born before 32 weeks of gestation. Heart rate was derived from the clinical standard monitoring system at caffeine cessation, 3 days later, and at discharge. Sample entropy was calculated from 90-min recordings using custom-written analytical software.ResultsWe obtained valid data from 44/61 infants (72%) with a mean of 28.4 weeks of gestation (range: 24.0-31.7). Twenty-eight (64%) were male. Sample entropy at caffeine cessation was positively associated with weeks of gestation (R2 = 0.15, p = 0.01) and negatively with bronchopulmonary dysplasia (R2 = 0.18, p = 0.01). It did not predict AOP frequency or caffeine reinitiation.ConclusionSample entropy at caffeine cessation is associated with maturation at birth and bronchopulmonary dysplasia but does not predict AOP within 3 days of cessation. Further studies should assess longitudinal measurements to predict respiratory control in preterm infants. Trial Registration: : NCT04303494ConclusionSample entropy at caffeine cessation is associated with maturation at birth and bronchopulmonary dysplasia but does not predict AOP within 3 days of cessation. Further studies should assess longitudinal measurements to predict respiratory control in preterm infants. Trial Registration: : NCT04303494

BACKGROUND:Host and environment early-life risk factors are associated with progression of wheezing symptoms over time; however, their individual contribution is relatively small. We hypothesised that the dynamic interactions of these factors with an infant's developing respiratory system are the dominant factor for subsequent wheeze and asthma. METHODS:In this dynamic network analysis we used data from term healthy infants from the Basel-Bern Infant Lung Development (BILD) cohort (435 neonates aged 0-4 weeks recruited in Switzerland between Jan 1, 1999, and Dec 31, 2012) and replicated the findings in the Protection Against Allergy Study in Rural Environments (PASTURE) cohort (498 infants aged 0-12 months recruited in Germany, Switzerland, Austria, France, and Finland between Jan 1, 2002, and Oct 31, 2006). BILD exclusion criteria for the current study were prematurity (<37 weeks), major birth defects, perinatal disease of the neonate, and incomplete follow-up period. PASTURE exclusion criteria were women younger than 18 years, a multiple pregnancy, the sibling of a child was already included in the study, the family intended to move away from the area where the study was conducted, and the family had no telephone connection. Outcome groups were subsequent wheeze, asthma, and healthy. The first outcome was defined as ever wheezed between the age of 2 years and 6 years. Week-by-week correlations of the determining factors with cumulative symptom scores (CSS) were calculated from weeks 2 to 52 (BILD) and weeks 8 to 52 (PASTURE). The complex dynamic interaction between the determining factors and the CSS was assessed via dynamic host-environment correlation network, quantified by a simple descriptor: trajectory function G(t). Wheeze outcomes at age 2-6 years were compared in 335 infants from BILD and 437 infants from PASTURE, and asthma outcomes were analysed at age 6 years in a merged cohort of 783 infants. FINDINGS:CSS was significantly different for wheeze and asthma outcomes and became increasingly important during infancy in direct comparison with all determining factors. Weekly symptoms were tracked for groups of infants, showing a non-linear increase with time. Using logistic regression classification, G(t) distinguished between the healthy group and wheeze or asthma groups (area under the curve>0·97, p<0·0001; sensitivity analysis confirmed significant CSS association with wheeze [BILD p=0·0002 and PASTURE p=0·068]) and G(t) was also able to distinguish between the farming and non-farming exposure groups (p<0·0001). INTERPRETATION:Similarly to other risk factors, CSS had weak sensitivity and specificity to identify risks at the individual level. At group level however, the dynamic host-environment correlation network properties (G(t)) showed excellent discriminative ability for identifying groups of infants with subsequent wheeze and asthma. Results from this study are consistent with the 2018 Lancet Commission on asthma, which emphasised the importance of dynamic interactions between risk factors during development and not the risk factors per se. FUNDING:The Swiss National Science Foundation, the Kühne Foundation, the EFRAIM study EU research grant, the FORALLVENT study EU research grant, and the Leibniz Prize.

Background: Host and environment early-life risk factors are known to be associated with the evolvement of subsequent wheeze in childhood. They include sex, maternal atopy, viral infections, and pollutants, however their individual relative contribution is small. We hypothesized that the dynamic interactions of these determining risk factors with an infant´s developing respiratory system are the dominant factor for the subsequent wheeze. Methods: We used data from the Basel-Bern Infant Lung Development (BILD) and the Protection Against Allergy STUdy in Rural Environments (PASTURE) prospective birth cohorts. We made week-by-week correlations of the determining factors with cumulative symptom scores (CSS), which showed temporal changes during infancy. The complex dynamic interaction between the host and environment factors and the CSS resulted in a dynamic host–environment correlation network. We quantified this in a simple descriptor, trajectory function G(t). Results: Using a standard logistic regression classification G(t) was able to distinguish between the group of infants with subsequent wheezing disorders at 2-6 years when compared to the healthy group (AUC 0.85 in BILD and 0.99 in PASTURE). Furthermore, in PASTURE, G(t) also showed clear differences between the groups of infants from a farming and non-farming environment (AUC of 0.78). Conclusions: In comparison to individual risk factors, our results illustrate that the dynamic host–environment correlation network properties have a better discriminative ability for identifying the group of infants with subsequent wheeze. Results from this study are consistent with the hypothesis stated by the 2018 Lancet Commission on Asthma.

Background: Children with asthma present fluctuations in lung function that reflect the stability of airway tone and the level of disease control. Objective: To examine whether subgrouping of asthmatic children according to patterns of FEV1 fluctuation allows for the identification of distinct clinical phenotypes. Methods: In the context of lung function variability study (NCT04163146), fluctuation-based clustering (Delgado-Eckert E, Thorax 2018) was applied to twice-daily FEV1 measurements recorded over a period of 3 months in 74 children with mild-to-moderate asthma. Results: Three clusters were identified (Figure: 1) Cluster 1 (63.5% of children) with decreased short-term variability and moderate long-term FEV1 fluctuation; it consisted of children with less allergy and excellent asthma control. 2) Cluster 2 (23%) with increased short-term variability and marked long-term FEV1 fluctuation; it comprised mostly allergic children who required more intensified treatment to maintain disease control. 3) Cluster 3 (13.5%) with increased short-term variability but minimal long-term FEV1 fluctuation; it consisted mostly of overweight and allergic children who despite the escalating medication maintained marginal level of disease control. Conclusions: Patient phenotyping according to FEV1 fluctuations may identify asthmatic children who could benefit from closer monitoring and specific treatment strategies.

Aquaporin 1 (AQP1) is one of thirteen known mammalian aquaporins. Its main function is the transport of water across cell membranes. Lately, a role of AQP has been attributed to other physiological and pathological functions including cell migration and peripheral pain perception. AQP1 has been found in several parts of the enteric nervous system, e.g., in the rat ileum and in the ovine duodenum. Its function in the intestine appears to be multifaceted and is still not completely understood. The aim of the study was to analyze the distribution and localization of AQP1 in the entire intestinal tract of mice. AQP1 expression was correlated with the hypoxic expression profile of the various intestinal segments, intestinal wall thickness and edema, as well as other aspects of colon function including the ability of mice to concentrate stools and their microbiome composition. AQP1 was found in a specific pattern in the serosa, the mucosa, and the enteric nervous system throughout the gastrointestinal tract. The highest amount of AQP1 in the gastrointestinal tract was found in the small intestine. AQP1 expression correlated with the expression profiles of hypoxia-dependent proteins such as HIF-1α and PGK1. Loss of AQP1 through knockout of AQP1 in these mice led to a reduced amount of bacteroidetes and firmicutes but an increased amount of the rest of the phyla, especially deferribacteres, proteobacteria, and verrucomicrobia. Although AQP-KO mice retained gastrointestinal function, distinct changes regarding the anatomy of the intestinal wall including intestinal wall thickness and edema were observed. Loss of AQP1 might interfere with the ability of the mice to concentrate their stool and it is associated with a significantly different composition of the of the bacterial stool microbiome.

Background: Estimating the probability of an asthma patient’s response to long-acting beta-2 adrenergic receptor agonists (LABA) from only a few lung function measurements would help clinicians avoid unnecessary drug prescription, contributing to personalized treatment and reduced costs. Aims and objectives: To develop a computational methodology capable of generating estimates of the probability of LABA-responsiveness and assess its accuracy using cohort data. Methods: Using fluctuation-based clustering (Delgado-Eckert E. et al. Thorax 2018; 73.2:107-115) we retrospectively analysed time series of PEF measurements recorded by 79 mild-to-moderate asthmatic adults during the placebo phase of a study aimed at characterizing the effects of salmeterol and salbutamol treatment (described in Thamrin C. et al. Eur Respir J 2009; 33:486–493). This yielded three clusters of patients: 1, composed mainly of salmeterol responders; 2, mainly non-responders; 3, a mixed group. We randomly removed 99% of the data from each time series, using these stripped-down versions as proxies for new patients’ data. Using Softmax Regression, we calculated the probability that a patient would be assigned to one of the three clusters. Based on the proportion of responders in each cluster, we calculated the salmeterol-responsiveness probability of a given “new” patient. We used a Receiver Operating Characteristic (ROC) curve to assess the probability’s classification performance. Results: The LABA-responsiveness probability ROC curve had a mean area under the curve of 0.7. Conclusion: Our method yields LABA-responsiveness probability estimates with good discriminatory power.

Background: In all chronic airway diseases, the dynamics of airway function are influenced by underlying airway inflammation and bronchial hyperresponsiveness along with limitations in reversibility owing to airway and lung remodeling as well as mucous plugging. The relative contribution of each component translates into specific clinical patterns of symptoms, quality of life, exacerbation risk, and treatment success. Objective: We aimed to evaluate whether subgrouping of patients with obstructive airway diseases according to patterns of fluctuation in lung function allows identification of specific phenotypes with distinct clinical characteristics. Methods: We applied the novel method of fluctuation-based clustering (FBC) to twice-daily FEV1 measurements recorded over a 1-year period in a mixed group of 134 adults with mild to-moderate asthma, severe asthma, or chronic obstructive pulmonary disease from the European BIOAIR cohort. Results: Independently of clinical diagnosis, FBC divided patients into 4 fluctuation-based clusters with progressively increasing alterations in lung function that corresponded to patterns of increasing clinical severity, risk of exacerbation, and lower quality of life. Clusters of patients with airway disease with significantly elevated levels of biomarkers relating to remodeling (osteonectin) and cellular senescence (plasminogen activator inhibitor-1), accompanied by a loss of airway reversibility, pulmonary hyperinflation, and loss of diffusion capacity, were identified. The 4 clusters generated were stable over time and revealed no differences in levels of markers of type 2 inflammation (blood eosinophils and periostin). Conclusion: FBC-based phenotyping provides another level of information that is complementary to clinical diagnosis and unrelated to eosinophilic inflammation, which could identify patients who may benefit from specific treatment strategies or closer monitoring.

Serial peak expiratory flow (PEF) measurements can identify phenotypes in severe adult asthma, enabling more targeted treatment. The feasibility of this approach in children has not been investigated. Overall, 105 children (67% male, median age 12.4 years) with a range of asthma severities were recruited and followed up over a median of 92 days. PEF was measured twice daily. Fluctuation-based clustering (FBC) was used to identify clusters based on PEF fluctuations. The patients’ clinical characteristics were compared between clusters. Three PEF clusters were identified in 44 children with sufficient measurements. Cluster 1 (27% of patients: n=12) had impaired spirometry (mean forced expiratory volume in 1 s (FEV 1 ) 71% predicted), significantly higher exhaled nitric oxide (≥35 ppb) and uncontrolled asthma (asthma control test (ACT) score <20 of 25). Cluster 2 (45%: n=20) had normal spirometry, the highest proportion of difficult asthma and significantly more patients on a high dose of inhaled corticosteroids (≥800 µg budesonide). Cluster 3 (27%: n=12) had mean FEV 1 92% predicted, the highest proportion of patients with no bronchodilator reversibility, a low ICS dose (≤400 µg budesonide), and controlled asthma (ACT scores ≥20 of 25). Three clinically relevant paediatric asthma clusters were identified using FBC analysis on PEF measurements, which could improve telemonitoring diagnostics. The method remains robust even when 80% of measurements were removed. Further research will determine clinical applicability.

Background: The Respiratory system is commonly assessed through different functional and inflammatory parameters, such as FeNO. However, such parameters may display large fluctuations, thus hampering their potential use as respiratory disease biomarkers. Here we systematically compare the ability of the average FeNO and of the relative change in FeNO to discriminate between uninfected individuals and individuals suffering from a RV infection. Methods: FeNO was measured thrice weekly in a prospective cohort consisting of 12 mildly asthmatic and 12 healthy volunteers during 2 months before and 1 month after a RV-16 inoculation. Mild asthmatics were included to obtain a heterogeneous sample. The effectiveness of inoculation and the development of a response was carefully assessed for each cohort participant [Sinha et al., eLife 2019]. Accordingly, each participant was labelled as responder (R) or non-responder (NR). Each participant’s average FeNO over the first 10 days after inoculation (MeanPostFeNO) and the relative difference between each participant’s MeanPostFeNO and their average FeNO over the 2 months prior to inoculation were then used to discriminate between Rs and NRs. The classification performance of these two magnitudes was assessed using Receiver Operating Characteristic (ROC) curves. Results: The MeanPostFeNO ROC curve had an area under the curve (AUC) of 0.585, whereas the AUC of the ROC curve constructed using the relative change in FeNO was 0.852 Conclusion: The relative change in FeNO could serve as a good discriminator of viral infections. On the contrary, the MeanPostFeNO performs nearly like a random classifier, which may affect clinical decision making.

Accurate detection of human respiratory viral infections is highly topical. We investigated how strongly inflammatory biomarkers (FeNO, eosinophils, neutrophils, and cytokines in nasal lavage fluid) and lung function parameters change upon rhinovirus 16 infection, in order to explore their potential use for infection detection. To this end, within a longitudinal cohort study, healthy and mildly asthmatic volunteers were experimentally inoculated with rhinovirus 16, and time series of these parameters/biomarkers were systematically recorded and compared between the pre- and post-infection phases of the study, which lasted two months and one month, respectively. We found that the parameters’/biomarkers’ ability to discriminate between the infected and the uninfected state varied over the observation time period. Consistently over time, the concentration of cytokines, in nasal lavage fluid, showed moderate to very good discrimination performance, thereby qualifying for disease progression monitoring, whereas lung function and FeNO, while quickly and non-invasively measurable using cheap portable devices (e.g., at airports), performed poorly.

Asthma is a dynamic disease, in which lung mechanical and inflammatory processes interact in a complex manner, often resulting in exaggerated physiological, in particular, inflammatory responses to exogenous triggers. We hypothesize that this may be explained by respiratory disease-related systems instability and loss of adaptability to changing environmental conditions, manifested in highly fluctuating biomarkers and symptoms. Using time series of inflammatory (eosinophils, neutrophils, FeNO), clinical and lung function biomarkers (PEF, FVC,FEV1), we estimated this loss of adaptive capacity (AC) during an experimental rhinovirus infection in 24 healthy and asthmatic human volunteers. Loss of AC was estimated by comparing similarities between pre- and post-challenge time series. Unlike healthy participants, the asthmatic’s post-viral-challenge state resembled more other rhinovirus-infected asthmatics than their own pre-viral-challenge state (hypergeometric-test: p=0.029). This reveals loss of AC and supports the concept that in asthma, biological processes underlying inflammatory and physiological responses are unstable, contributing to loss of control.

Background: Epidemiological evidence on the influence of long-term exposure to traffic-related particulate matter (TPM10) on heart rate variability (HRV) is weak. Objective: To evaluate the association of long-term exposure (10 years) with TPM10 on the regulation of the autonomic cardiovascular system and heart rate dynamics (HRD) in an aging general population, as well as potential modifying effects by the a priori selected factors sex, smoking status, obesity, and gene variation in selected glutathione S-transferases (GSTs). Methods: We analyzed data from 1593 SAPALDIA cohort participants aged >= 50 years. For each participant, various HRV and HRD parameters were derived from 24-hour electrocardiogram recordings. Each parameter obtained was then used as the outcome variable in multivariable mixed linear regression models in order to evaluate the association with TPM10. Potential modifying effects were assessed using interaction terms. Results: No association between long-term exposure to TPM10 and HRV/HRD was observed in the entire study population. However, HRD changes were found in subjects without cardiovascular morbidity and both HRD and HRV changes in non-obese subjects without cardiovascular morbidity. Subjects without cardiovascular morbidity with homozygous GSTM1 gene deletion appeared to be more susceptible to the effects of TPM10. Conclusion: This study suggests that long-term exposure to TPM10 triggers adverse changes in the regulation of the cardiovascular system. These adverse effects were more visible in the subjects without cardiovascular disease, in whom the overall relationship between TPM10 and HRV/HRD could not be masked by underlying morbidities and the potential counteracting effects of related drug treatments.

Children with frequent respiratory symptoms in infancy have an increased risk for later wheezing, but the association with symptom dynamics is unknown. We developed an observer-independent method to characterise symptom dynamics and tested their association with subsequent respiratory morbidity. In this birth-cohort of healthy neonates, we prospectively assessed weekly respiratory symptoms during infancy, resulting in a time series of 52 symptom scores. For each infant, we calculated the transition probability between two consecutive symptom scores. We used these transition probabilities to construct a Markov matrix, which characterised symptom dynamics quantitatively using an entropy parameter. Using this parameter, we determined phenotypes by hierarchical clustering. We then studied the association between phenotypes and wheezing at 6 years. In 322 children with complete data for symptom scores during infancy (16 864 observations), we identified three dynamic phenotypes. Compared to the low-risk phenotype, the high-risk phenotype, defined by the highest entropy parameter, was associated with an increased risk of wheezing (odds ratio (OR) 3.01, 95% CI 1.15-7.88) at 6 years. In this phenotype, infants were more often male (64%) and had been exposed to environmental tobacco smoke (31%). In addition, more infants had siblings (67%) and attended childcare (38%). We describe a novel method to objectively characterise dynamics of respiratory symptoms in infancy, which helps identify abnormal clinical susceptibility and recovery patterns of infant airways associated with persistent wheezing.

BackgroundPoor control of body temperature is associated with mortality and major morbidity in preterm infants. We aimed to quantify its dynamics and complexity to evaluate whether indices from fluctuation analyses of temperature time series obtained within the first five days of life are associated with gestational age (GA) and body size at birth, and presence and severity of typical comorbidities of preterm birth.MethodsWe recorded 3h-time series of body temperature using a skin electrode in incubator-nursed preterm infants. We calculated mean and coefficient of variation of body temperature, scaling exponent alpha (T-alpha) derived from detrended fluctuation analysis, and sample entropy (T-SampEn) of temperature fluctuations. Data were analysed by multilevel multivariable linear regression.ResultsData of satisfactory technical quality were obtained from 285/357 measurements (80%) in 73/90 infants (81%) with a mean (range) GA of 30.1 (24.0-34.0) weeks. We found a positive association of Talpha with increasing levels of respiratory support after adjusting for GA and birth weight z-score (p<0.001; R-2 = 0.38).ConclusionDynamics and complexity of body temperature in incubator-nursed preterm infants show considerable associations with GA and respiratory morbidity. Talpha may be a useful marker of autonomic maturity and severity of disease in preterm infants.