Weily Soong, MD, FAAAAI

In chronic spontaneous urticaria (CSU), the identification of clinical phenotypes enables early prediction of non-response to H1 antihistamines alone or in combination with H2 antihistamines/leukotriene receptor antagonists (LTRAs), according to findings published in The Journal of Allergy and Clinical Immunology.

“A substantial proportion of [patients with] CSU do not achieve control with guideline-recommended first-line therapies,” Weily Soong, MD, FAAAAI, and colleagues wrote. “This study identified predictors associated with a treatment-refractory phenotype.”

Dr. Soong and colleagues conducted a retrospective cohort study of adult patients with CSU based on EMR data from 2007-2020. The study included patients starting therapy with H1 antihistamines alone or in combination with H2AH/LTRA/doxepin/corticosteroids following a diagnosis of CSU. The researchers classified patients as responders and non-responders using treatment continuation and treatment switch, respectively, for 12 months after the start of treatment as proxies of therapeutic efficacy. They also noted that adjunctive corticosteroid prescriptions and heterogeneous switching patterns indicated non-responsiveness. Machine-learning models of non-responsiveness were used to analyze demographic variables, clinical symptoms, and treatment regimens as candidate predictors.

Treatment History Enables Prediction of Non-Response

The study included 17,062 patients (mean age, 47; 77% female),87% of whom were determined to be non-responders. Of 3,804 variables originally used, 52 were determined to be predictive by the final models, with an area under the curve of 0.77 for both models.

The LASSO model determined that baseline CSU treatment characteristics, including the use of LTRA (OR, 1.94) and corticosteroids (OR, 2.21), specialists seen, including gastroenterology (OR, 1.30) and emergency providers (OR, 1.59), and related extracutaneous conditions, including depression (OR, 1.24) and chronic pulmonary disease (OR, 1.26), predicted non-response compared with response. Based on the random forest model, other predictors included baseline prescriptions for narcotics/other analgesics (non-responders: 38%; responders: 24%), inhaled corticosteroids (6% vs 3%), and abdominal pain (15% vs 9%).

“Identifiable clinical phenotypes facilitate early prediction of non-response to H1 [antihistamines] alone/combined with H2 [antihistamines]/LTRA/doxepin/corticosteroids in CSU, highlighting the importance of personalized treatment strategies for control,” Dr. Soong and colleagues wrote. “Patients at risk of non-response may be prioritized for earlier initiation of advanced therapies.”