Gradient-Boosted Decision Trees Predict Treatment Success for Psoriasis
Gradient-boosted decision trees perform better than logistic regression for predicting biologic therapy leading to treatment duration of at least three years
By Dermsquared Editorial Team | August 17, 2022
For adults with moderate-to-severe psoriasis, gradient-boosted decision trees perform better than logistic regression for predicting specific biologic therapy leading to a treatment duration of at least three years without discontinuation, according to a study published online Aug. 17 in JAMA Dermatology .
Mia-Louise Nielsen, from the University of Copenhagen in Denmark, and colleagues conducted a population-based study involving adult patients treated for moderate-to-severe psoriasis with biologics to identify the optimal biologic therapy. A total of 2,034 patients with a total of 3,452 treatment series were included in the study, assuming a success criterion of three years of sustained treatment.
The researchers found that gradient-boosted decision trees and logistic regression could predict a specific cytokine target associated with successful treatment (accuracies, 63.6 and 59.2 percent, respectively) and had top 2 accuracies of 95.9 and 93.9 percent, respectively. Gradient boost and logistic regression had accuracies of 48.5 and 44.4 percent when predicting specific drugs resulting in success, top 2 accuracies of 77.6 and 75.9 percent, and top 3 accuracies of 89.9 and 89.0 percent, respectively.
"These findings could potentially be useful for clinicians in future treatment decision processes for choosing of appropriate biologic therapy for patients with psoriasis," the authors write.
Several authors disclosed financial ties to the pharmaceutical industry.