Inhaled corticosteroids for asthma may increase risk for adrenal suppression
Click Here to Manage Email Alerts
Key takeaways:
- Twelve of 810 plasma metabolites were significantly associated with adrenal insufficiency.
- Patients with low cortisol levels had significantly reduced levels of seven steroid metabolites.
SAN DIEGO — Plasma metabolite profiles were significantly associated with cortisol levels among patients with asthma who use inhaled corticosteroids, according to a poster at the American Thoracic Society International Conference.
Cortisol is a biomarker for adrenal suppression, Dung (Ivy) T. Tran, PhD, MS, postdoctoral researcher at Mass General Brigham and Harvard Medical School, and colleagues, wrote. Previous studies have indicated that long-term exposure to glucocorticoids in high doses could lead to immunosuppression, Tran told Healio.
“In this situation, the adrenal gland will not function normally and cannot produce enough cortisol in the body,” Tran said. “It can create some potential effects and serious health issues for asthma patients on long-term [inhaled corticosteroids (ICS)]. We focused on asthma patients that are using ICS to see if they were at increased risk of developing adrenal suppression after long-term use.”
Under the guidance of her mentor, Amber Dahlin, PhD, MMSc, assistant professor of medicine and an associate epidemiologist in the Channing Division of Network Medicine at Harvard Medical School and Mass General Brigham, Tran and her colleagues sought to identify metabolite biomarkers related to adrenal suppression using metabolomics profiling combined with electronic medical records data to build and test predictive models.
The study’s independent discovery and replication cohorts included plasma samples from 735 and 597 adults with asthma, respectively, who used ICS. The researchers selected 810 metabolites to profile for potential association with plasma cortisol levels and other information from patients’ electronic medical records.
After that, Tran said, the researchers performed network analysis to determine which biochemical pathways related to adrenal suppression might be perturbed. The investigators then built and tested machine learning models to uncover potential biomarkers to predict risks for adrenal suppression in the patients.
“We identified metabolites that would increase or decrease along with fluctuations in the cortisol levels in the human body,” Tran said.
Twelve metabolites had strong, significant associations (P < .05) with adrenal suppression in the discovery cohort, the researchers said. Patients with adrenal insufficiency had significant reductions in seven steroid metabolites and five nonsteroid metabolites. The researchers also independently validated five metabolites involved in unique metabolic pathways.
“We used several machine learning models to test the predictive performance of these metabolites,” Tran said.
Among all tested machine learning models, regularized logistic regression performed the best, the researchers said. Using network analysis, the researchers then identified changes in multiple biochemical pathways corresponding to these metabolites. Through validation, the researchers also replicated their results for the five plasma metabolites.
“These five metabolites could serve as predictive biomarkers for developing adrenal suppression,” Tran said.
Based on these findings, the researchers said adults with asthma who used ICS had significant differences in their plasma metabolite profiles when stratified into groups with or without adrenal insufficiency.
“Using this model and these metabolite biomarkers, we can predict which patients are at increased risk for adrenal suppression,” Tran said. “From that, asthma patients at increased risk can potentially switch to alternative medications. They may also require more intensive monitoring of their adrenal function during ICS use.”
The researchers also plan on refining their process.
“In the future, we will seek improvements in both patient data and prediction techniques to obtain models with better accuracy and reliability,” Tran said. “In the next step, we will integrate genomics and metabolomics to see which genes are involved.”