Pulmonary MRI and Cluster Analysis Help Identify Novel Asthma Phenotypes

Abstract

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 ¹²⁹ Xe MRI, volume-matched CT, spirometry and plethysmography within a 90-minute visit.

Field strength/sequence: Three-dimensional gradient-recalled echo ¹²⁹ Xe 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 ¹²⁹ Xe MRI in combination with CT biomarkers is feasible and may challenge currently used paradigms for asthma phenotyping and treatment decisions.

Evidence level: 3 TECHNICAL EFFICACY: Stage.

Keywords: 129Xe MRI; asthma; asthma phenotypes; cluster analysis.