Year: 2021
Type: conference-abstract
Abstract: Rationale: Although inhaled corticosteroids (ICS) are the recommended first-line therapy for subjects with persistent asthma, a significant proportion of subjects continue to have exacerbations despite therapy. There is increased evidence indicating that genetic variation substantially influences ICS treatment response. The objective of this study is to develop polygenic risk prediction of ICS response in subjects with asthma. Methods: Study subjects included five independent cohorts inclusive of pediatric (n=440) and adult (n=927) asthma subjects treated with ICS (n=1,371; 831 females, 540 males). Of these, 823 were randomly selected to form a training set for model development, and the remaining 548 subjects formed a test set for model validation. Poor response to ICS was defined as asthma exacerbations (i.e. emergency department visit, or hospitalization due to asthma, or the need for oral corticosteroids) occurring while on ICS. A genome-wide association study was first performed on the training set to identify SNPs associated with ICS response. Linkage-disequilibrium-independent associations were obtained by clumping and thresholding was applied to remove SNPs with p-value 1x10 -5 . Using the selected SNPs (n=630), LASSO regression models were constructed to predict ICS response with varying regularization parameters and evaluated using 10-fold cross validation on the training set. The most optimal model was then selected for validation on the independent test set. We further developed separate models integrating SNPs and phenotype data (including sex, age, and body mass index) to predict ICS response. Results: A total of 323 (39.2%) subjects in the training set and 199 (36.3%) subjects in the test set experienced asthma exacerbation while on ICS. The most optimal model for predicting ICS response derived from SNPs selected through GWAS achieved an AUC of 0.95 (sensitivity: 0.83; specificity: 0.93) in the training set. When tested on the independent test set, the model achieved an AUC of 0.71 (sensitivity: 0.48; specificity 0.82). The addition of phenotype data did not improve the performance of the predictive models. Models integrating both SNPs and phenotype data achieved an AUC of 0.95 (sensitivity: 0.81; specificity: 0.93) in the training set, and an AUC of 0.71 (sensitivity: 0.48; specificity: 0.81) in the test set. Conclusion: Our study demonstrates the feasibility of utilizing polygenic risk models to predict ICS response in subjects with asthma. Such a tool has the potential to enable individualized therapy. Future research should explore integration of richer phenotype data to improve risk discrimination and investigate the generalizability of the approach.
Language: English
Subfield: Physiology
Field: Medicine
Domain: Health Sciences
Open Access status: bronze
OpenAlex ID: W4251251186