Pulmonary MRI and Cluster Analysis Help Identify Novel Asthma Phenotypes

• Published in: Journal of magnetic resonance imaging 2022-11, Vol.56 (5), p.1475-1486
• Main Authors: Eddy, Rachel L., McIntosh, Marrissa J., Matheson, Alexander M., McCormack, David G., Licskai, Christopher, Parraga, Grace
• Format: Article
• Language: English
• Subjects: 129Xe MRI; Area; Asthma; Asthma - diagnostic imaging; asthma phenotypes; Biomarkers; Bronchodilator Agents - therapeutic use; Bronchodilators; Cluster Analysis; Computed tomography; Diameters; Eosinophilia; Female; Females; Field strength; Functional magnetic resonance imaging; Health services; Humans; Kruskal-Wallis test; Lung - pathology; Magnetic resonance imaging; Magnetic Resonance Imaging - methods; Medical imaging; Pathogenesis; Patients; Phenotype; Phenotypes; Phenotyping; Plethysmography; Population studies; Respiratory tract; Retrospective Studies; Spirometry; Statistical analysis; Statistical tests; Structure-function relationships; Subgroups; Variance analysis; Ventilation; Xenon 129
• ISSN: 1053-1807; 1522-2586
• DOI: 10.1002/jmri.28152

Background Outside eosinophilia, current clinical asthma phenotypes do not show strong relationships with disease pathogenesis or treatment responses. While chest x‐ray computed tomography (CT) phenotypes have previously been explored, functional MRI measurements provide complementary phenotypic information. Purpose To derive novel data‐driven asthma phenotypic clusters using functional MRI airway biomarkers that better describe airway pathologies in patients. Study Type Retrospective. Population A total of 45 patients with asthma who underwent post‐bronchodilator 129Xe MRI, volume‐matched CT, spirometry and plethysmography within a 90‐minute visit. Field Strength/Sequence Three‐dimensional gradient‐recalled echo 129Xe ventilation sequence at 3 T. Assessment We measured MRI ventilation defect percent (VDP), CT airway wall‐area percent (WA%), wall‐thickness (WT, WT* [*normalized for age/sex/height]), lumen‐area (LA), lumen‐diameter (D, D*) and total airway count (TAC). Univariate relationships were utilized to select variables for k‐means cluster analysis and phenotypic subgroup generation. Spirometry and plethysmography measurements were compared across imaging‐based clusters. Statistical Tests Spearman correlation (ρ), one‐way analysis of variance (ANOVA) or Kruskal–Wallis tests with post hoc Bonferroni correction for multiple comparisons, significance level 0.05. Results Based on limited common variance (Kaiser–Meyer–Olkin‐measure = 0.44), four unique clusters were generated using MRI VDP, TAC, WT* and D* (52 ± 14 years, 27 female). Imaging measurements were significantly different across clusters as was the forced expiratory volume in 1‐second (FEV1%pred), residual volume/total lung capacity and airways resistance. Asthma‐control (P = 0.9), quality‐of‐life scores (P = 0.7) and the proportions of severe‐asthma (P = 0.4) were not significantly different. Cluster1 (n = 15/8 female) reflected mildly abnormal CT airway measurements and FEV1 with moderately abnormal VDP. Cluster2 (n = 12/12 female) reflected moderately abnormal TAC, WT and FEV1. In Cluster3 and Cluster4 (n = 14/6 female, n = 4/1 female, respectively), there was severely reduced TAC, D and FEV1, but Cluster4 also had significantly worse, severely abnormal VDP (7 ± 5% vs. 41 ± 12%). Data Conclusion We generated four proof‐of‐concept MRI‐derived clusters of asthma with distinct structure–function pathologies. Cluster analysis of asthma using 129Xe MRI in combination with CT biomarkers is feasibl