Work

Clinical and functional airway profiles in children with asthma with and without sickle cell disease: a comparative study and explainable machine-learning analysis

Year: 2026
Type: article
Abstract: Background Asthma in children with sickle cell disease (SCD) may show functional characteristics that differ from asthma without SCD, but comparisons are vulnerable to differences in source-cohort ascertainment. We compared clinically available characteristics in children with asthma with and without SCD and explored the degree to which a restricted set of harmonized variables discriminated the two groups. Methods This secondary comparative analysis used two July 2026 analytic datasets from French Guiana. The non-SCD cohort included 180 children aged 5-17 years with physician-diagnosed asthma. The SCD source dataset contained 161 children who underwent respiratory assessment; 95 with an archived confirmed-asthma classification formed the SCD-asthma group. Primary analyses were restricted to variables measured comparably and completely in both groups: age, sex, health-insurance status, place of residence, and bronchodilator responsiveness (BDR; archived historical definition >12% FEV1 increase). Logistic regression adjusted for these covariates. Penalized logistic regression, random forest, and gradient boosting were evaluated by repeated stratified five-fold cross-validation (20 repeats). Results The final comparative cohort comprised 275 children (180 without SCD; 95 with SCD). Children with SCD and asthma were older (10.4 ± 3.3 vs. 9.2 ± 3.3 years; p = 0.003), more often had health-insurance coverage (88.4% vs. 66.1%; p < 0.001), and more often had recorded BDR (91.6% vs. 59.4%; p < 0.001). In adjusted analysis, BDR remained strongly associated with SCD-asthma group membership (aOR 7.58, 95% CI 3.34-17.22), as did age (aOR 1.14 per year, 95% CI 1.05-1.24) and insurance coverage (aOR 3.67, 95% CI 1.72-7.82). Model discrimination was moderate (AUC 0.762-0.768). Conclusion Children with asthma and SCD showed an overlapping rather than clearly separate profile, with greater recorded bronchodilator responsiveness and modest differences in age and healthcare-context variables. Because the source cohorts used different diagnostic pathways and the archived files lack raw pre/post-bronchodilator values and GLI z-scores, these findings should be interpreted as associations requiring prospective confirmation using harmonized contemporary respiratory criteria.
Language: English
Field: Medicine
Open Access status: gold
Estimated APC paid: $3,801
OpenAlex ID: W7236700763
FWCI:
Cites: 18
Cited by:
Related to: 10
Parts: 1