Unsupervised phenotypic clustering for determining clinical status in children with cystic fibrosis

Conventional chest physiotherapy compared to other airway clearance techniques for cystic fibrosis

BACKGROUND

Cystic fibrosis (CF) is an inherited life-limiting disorder. Over time persistent infection and inflammation within the lungs contribute to severe airway damage and loss of respiratory function. Chest physiotherapy, or airway clearance techniques (ACTs), are integral in removing airway secretions and initiated shortly after CF diagnosis. Conventional chest physiotherapy (CCPT) generally requires assistance, while alternative ACTs can be self-administered, facilitating independence and flexibility. This is an updated review.

OBJECTIVES

To evaluate the effectiveness (in terms of respiratory function, respiratory exacerbations, exercise capacity) and acceptability (in terms of individual preference, adherence, quality of life) of CCPT for people with CF compared to alternative ACTs.

SEARCH METHODS

We used standard, extensive Cochrane search methods. The latest search was 26 June 2022.

SELECTION CRITERIA

We included randomised or quasi-randomised controlled trials (including cross-over design) lasting at least seven days and comparing CCPT with alternative ACTs in people with CF.

DATA COLLECTION AND ANALYSIS

We used standard Cochrane methods. Our primary outcomes were 1. pulmonary function tests and 2. number of respiratory exacerbations per year. Our secondary outcomes were 3. quality of life, 4. adherence to therapy, 5. cost-benefit analysis, 6. objective change in exercise capacity, 7. additional lung function tests, 8. ventilation scanning, 9. blood oxygen levels, 10. nutritional status, 11. mortality, 12. mucus transport rate and 13. mucus wet or dry weight. We reported outcomes as short-term (seven to 20 days), medium-term (more than 20 days to up to one year) and long-term (over one year).

MAIN RESULTS

We included 21 studies (778 participants) comprising seven short-term, eight medium-term and six long-term studies. Studies were conducted in the USA (10), Canada (five), Australia (two), the UK (two), Denmark (one) and Italy (one) with a median of 23 participants per study (range 13 to 166). Participant ages ranged from newborns to 45 years; most studies only recruited children and young people. Sixteen studies reported the sex of participants (375 males; 296 females). Most studies compared modifications of CCPT with a single comparator, but two studies compared three interventions and another compared four interventions. The interventions varied in the duration of treatments, times per day and periods of comparison making meta-analysis challenging. All evidence was very low certainty. Nineteen studies reported the primary outcomes forced expiratory volume in one second (FEV1)and forced vital capacity (FVC), and found no difference in change from baseline in FEV1 % predicted or rate of decline between groups for either measure. Most studies suggested equivalence between CCPT and alternative ACTs, including positive expiratory pressure (PEP), extrapulmonary mechanical percussion, active cycle of breathing technique (ACBT), oscillating PEP devices (O-PEP), autogenic drainage (AD) and exercise. Where single studies suggested superiority of one ACT, these findings were not corroborated in similar studies; pooled data generally concluded that effects of CCPT were comparable to those of alternative ACTs. CCPT versus PEP We are uncertain whether CCPT improves lung function or has an impact on the number of respiratory exacerbations per year compared with PEP (both very low-certainty evidence). There were no analysable data for our secondary outcomes, but many studies provided favourable narrative reports on the independence achieved with PEP mask therapy. CCPT versus extrapulmonary mechanical percussion We are uncertain whether CCPT improves lung function compared with extrapulmonary mechanical percussions (very low-certainty evidence). The annual rate of decline in average forced expiratory flow between 25% and 75% of FVC (FEF25-75) was greater with high-frequency chest compression compared to CCPT in medium- to long-term studies, but there was no difference in any other outcome. CCPT versus ACBT We are uncertain whether CCPT improves lung function compared to ACBT (very low-certainty evidence). Annual decline in FEF25-75 was worse in participants using the FET component of ACBT only (mean difference (MD) 6.00, 95% confidence interval (CI) 0.55 to 11.45; 1 study, 63 participants; very low-certainty evidence). One short-term study reported that directed coughing was as effective as CCPT for all lung function outcomes, but with no analysable data. One study found no difference in hospital admissions and days in hospital for exacerbations. CCPT versus O-PEP We are uncertain whether CCPT improves lung function compared to O-PEP devices (Flutter device and intrapulmonary percussive ventilation); however, only one study provided analysable data (very low-certainty evidence). No study reported data for number of exacerbations. There was no difference in results for number of days in hospital for an exacerbation, number of hospital admissions and number of days of intravenous antibiotics; this was also true for other secondary outcomes. CCPT versus AD We are uncertain whether CCPT improves lung function compared to AD (very low-certainty evidence). No studies reported the number of exacerbations per year; however, one study reported more hospital admissions for exacerbations in the CCPT group (MD 0.24, 95% CI 0.06 to 0.42; 33 participants). One study provided a narrative report of a preference for AD. CCPT versus exercise We are uncertain whether CCPT improves lung function compared to exercise (very low-certainty evidence). Analysis of original data from one study demonstrated a higher FEV1 % predicted (MD 7.05, 95% CI 3.15 to 10.95; P = 0.0004), FVC (MD 7.83, 95% CI 2.48 to 13.18; P = 0.004) and FEF25-75 (MD 7.05, 95% CI 3.15 to 10.95; P = 0.0004) in the CCPT group; however, the study reported no difference between groups (likely because the original analysis accounted for baseline differences).

AUTHORS' CONCLUSIONS

We are uncertain whether CCPT has a more positive impact on respiratory function, respiratory exacerbations, individual preference, adherence, quality of life, exercise capacity and other outcomes when compared to alternative ACTs as the certainty of the evidence is very low. There was no advantage in respiratory function of CCPT over alternative ACTs, but this may reflect insufficient evidence rather than real equivalence. Narrative reports indicated that participants prefer self-administered ACTs. This review is limited by a paucity of well-designed, adequately powered, long-term studies. This review cannot yet recommend any single ACT above others; physiotherapists and people with CF may wish to try different ACTs until they find an ACT that suits them best.

Cystic Fibrosis Screen Positive, Inconclusive Diagnosis Genotypes in People with Cystic Fibrosis from the U.S. Patient Registry

Rationale: Variants within the cystic fibrosis (CF) transmembrane conductance regulator gene, CFTR, that are of unknown significance or are categorized as non-CF causing may be observed in persons with CF. These variants are frequently detected in children with inconclusive newborn screen results and, in some cases, may be associated with a benign presentation in early childhood that progresses to a CF phenotype later in life. Objectives: To analyze data from individuals enrolled in the U.S. Cystic Fibrosis Foundation Patient Registry who have received a diagnosis of CF and who have variants found in a population of children with a CF screen positive, inconclusive diagnosis (CFSPID). Methods: This retrospective review analyzed registry data from individuals with a diagnosis of CF who also harbor one or more variants of interest because of their frequency within a CFSPID population and/or their interpretation as non-CF causing. Three groups were defined by the number of CF-causing variants identified (CF-Cx2, CF-Cx1, and CF-Cx0), which were reported in addition to the variant(s) of interest. Multivariate quantile regression modeling of the outcome for forced expiratory volume in 1 second (FEV1) generated a disease severity score for each person determined by six selected variables. Median scores were calculated for the three groups. Results: Patients carrying one CF-causing variant and at least one variant of interest (CF-Cx1) had higher median disease severity scores compared with those carrying CF-Cx2, suggesting a milder phenotype (P < 0.05). However, there was no statistically significant difference in scores between CF-Cx2 and the two other groups combined (CF-Cx1 and CF-Cx0; P = 0.33). Analysis revealed that the CF-Cx1 and CF-Cx0 groups, when compared with the CF-Cx2 group, had later median diagnoses (8 years vs. newborn; P < 0.0001), lower median sweat chloride (48 mmol/L vs. 94.5 mmol/L; P < 0.0001), lower prevalence of pancreatic insufficiency (29% vs. 78%; P < 0.0001), and higher median FEV1% predicted (95% vs. 87%; P = 0.0002). Conclusions: Individuals with CF who have specific variants frequently identified in children with CFSPID have a similar range of disease severity scores compared with those who have two CF-causing variants, but a milder phenotype overall. Variants that should be given careful scrutiny because of their high prevalence are G576A+R668C, T854T, R75Q, F1052V, R1070W, R31C, and L967S.

Using machine learning to impact on long-term clinical care: principles, challenges, and practicalities

The rise of machine learning in healthcare has significant implications for paediatrics. Long-term conditions with significant disease heterogeneity comprise large portions of the routine work performed by paediatricians. Improving outcomes through discovery of disease and treatment prediction models, alongside novel subgroup clustering of patients, are some of the areas in which machine learning holds significant promise. While artificial intelligence has percolated into routine use in our day to day lives through advertising algorithms, song or movie selections and sifting of spam emails, the ability of machine learning to utilise highly complex and dimensional data has not yet reached its full potential in healthcare. In this review article, we discuss some of the foundations of machine learning, including some of the basic algorithms. We emphasise the importance of correct utilisation of machine learning, including adequate data preparation and external validation. Using nutrition in preterm infants and paediatric inflammatory bowel disease as examples, we discuss the evidence and potential utility of machine learning in paediatrics. Finally, we review some of the future applications, alongside challenges and ethical considerations related to application of artificial intelligence. IMPACT: Machine learning is a widely used term; however, understanding of the process and application to healthcare is lacking. This article uses clinical examples to explore complex machine learning terms and algorithms. We discuss limitations and potential future applications within paediatrics and neonatal medicine.

Lung clearance index to characterize clinical phenotypes of children and adolescents with cystic fibrosis

Background

Lung clearance index (LCI) is accepted as an early marker of lung disease in cystic fibrosis (CF), however the utility of LCI to identify subgroups of CF disease in the paediatric age group has never been explored. The aim of the study was to characterize phenotypes of children with CF using LCI as a marker of ventilation inhomogeneity and to investigate whether these phenotypes distinguished patients based on time to pulmonary exacerbation (PE).

Methods

Data were collected on patients with CF aged < 18 years old, attending the CF Center of Milan during outpatient follow-up visits between October 2014 and September 2019. Cluster analysis using agglomerative nesting hierarchical method was performed to generate distinct phenotypes. Time-to-recurrent event analysis investigated association of phenotypes with PE.

Results

We collected 313 multiple breath washout tests on 125 children aged 5.5–16.8 years. Cluster analysis identified two divergent phenotypes in children and adolescents of same age, presenting with almost normal FEV₁ but with substantial difference in markers of ventilation inhomogeneity (mean LCI difference of 3.4, 95% Confidence Interval [CI] 2.6–4.2). A less severe phenotype was associated with a lower risk of PE relapse (Hazard Ratio 0.45, 95% CI 0.34–0.62).

Conclusions

LCI is useful in clinical practice to characterize distinct phenotypes of children and adolescents with mild/normal FEV₁. A less severe phenotype translates into a lower risk of PE relapse.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12890-022-01903-5.

Implementation of prognostic machine learning algorithms in paediatric chronic respiratory conditions: a scoping review

Machine learning (ML) holds great potential for predicting clinical outcomes in heterogeneous chronic respiratory diseases (CRD) affecting children, where timely individualised treatments offer opportunities for health optimisation. This paper identifies rate-limiting steps in ML prediction model development that impair clinical translation and discusses regulatory, clinical and ethical considerations for ML implementation. A scoping review of ML prediction models in paediatric CRDs was undertaken using the PRISMA extension scoping review guidelines. From 1209 results, 25 articles published between 2013 and 2021 were evaluated for features of a good clinical prediction model using the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) guidelines.

Most of the studies were in asthma (80%), with few in cystic fibrosis (12%), bronchiolitis (4%) and childhood wheeze (4%). There were inconsistencies in model reporting and studies were limited by a lack of validation, and absence of equations or code for replication. Clinician involvement during ML model development is essential and diversity, equity and inclusion should be assessed at each step of the ML pipeline to ensure algorithms do not promote or amplify health disparities among marginalised groups. As ML prediction studies become more frequent, it is important that models are rigorously developed using published guidelines and take account of regulatory frameworks which depend on model complexity, patient safety, accountability and liability.

Physical Activity and Its Association with Traditional Outcome Measures in Pulmonary Arterial Hypertension

Rationale Limitation of physical activity is a common presenting complaint for patients with pulmonary arterial hypertension (PAH). Physical activity is thought to be determined by cardiopulmonary function, yet there are limited data that investigate this relationship. Objective We aimed to study the relationship between right ventricular function and daily activity and its impact on health-related quality of life (HRQoL) in PAH. Methods Baseline data for 55 patients enrolled in PHANTOM, an ongoing multicenter randomized controlled trial of anastrozole in PAH were used. Post-menopausal women and men were eligible and underwent six-minute walk testing, echocardiography and completed HRQoL questionnaires. Each patient wore an accelerometer for 7-days. Multivariable linear regression models were used to study the association between tricuspid annular plane systolic excursion (TAPSE) and vector magnitude counts, and between daily activity and HRQoL. Principal component analysis and K-means clustering were used to identify activity-based phenotypes. K-nearest neighbors' classification was applied to an independent cross-sectional cohort from the University of Pennsylvania. Results The mean age of patients in PHANTOM was 61 years. 67% were women with idiopathic PAH as the most common etiology. A 0.4 cm increase in TAPSE was associated with an increase in daily vector magnitude counts (ß:34000, 95%CI:900-67000, p=0.004) after adjustment for age, sex, body mass index, etiology of PAH and wear time. A 1-standard deviation increase in vector magnitude counts was associated with higher six-minute walk distance (ß: 56.1 meters, 95%CI:28.6-83.7, p<0.001) and lower emPHasis-10 scores (ß:-3.3, 95%CI:0.3-6.4, p=0.03). Three activity phenotypes, low, medium, and high, were identified. The most active phenotype had greater six-minute walk distances (p=0.001) and lower emPHasis-10 scores (p=0.009) after adjustment for age, sex, body mass index, WHO functional class and parenteral prostacyclin use. Phenotypes of physical activity were reproduced in the second cohort and were independently associated with six-minute walk distance. Conclusion Better right ventricular systolic function was associated with increased levels of activity in PAH. Increased daily activity was associated with greater six-minute walk distance and better HRQoL. Distinct activity-based phenotypes may be helpful in risk stratification of PAH patients or provide novel endpoints for clinical trials.