Related Experiment Video
Updated: Jan 9, 2026
Predictive Immune Modeling of Solid Tumors
Published on: February 25, 2020
Efficacy of Risankizumab across distinct PsA phenotypes identified with machine learning analytics using data from
Laure Gossec1,2,3, Andra Balanescu4, Maria Antonietta D'Agostino5,6
1Sorbonne Université, INSERM, Institut Pierre Louis d'Epidémiologie et de Santé Publique, Team Pepites, Paris, France. laure.gossec@aphp.fr.
Background:
The development of personalized approaches in psoriatic arthritis (PsA) is challenging due to unclear patient phenotypes and trajectories. Machine learning (ML) could help to identify homogeneous patient groups. The objective of this analysis was to classify patients with PsA into distinct phenotypes using ML.
Methods:
A post hoc analysis of PsA patients treated with risankizumab for ≤ 4 years (196 weeks) from KEEPsAKE 1 and KEEPsAKE 2. The phenotypes were based on baseline demographics and clinical characteristics using unsupervised ML (a finite mixture model). Response to risankizumab at 4 years was defined as minimal disease activity (MDA) and Disease Activity in PsA (DAPSA) low disease activity (LDA).
Results:
A total of 1119 patients were classified into 5 distinct PsA phenotypes: Moderate to High Disease Activity (40.3% of patients) - characterized by lower tender joint count (TJC) and swollen joint count (SJC), dactylitis, and enthesitis; Enthesitis and Large Joints Dominant (20.8% of patients) - characterized by enthesitis and mainly active large joints; Very High Disease Activity (14.0% of patients) - characterized by high TJC/SJC, dactylitis, and enthesitis; Hand Dominant (13.8% of patients) - characterized by active joints primarily in the hands; Dactylitis and Feet Dominant (11.1% of patients) - characterized by dactylitis and active joints primarily in the feet. At 4 years, risankizumab demonstrated efficacy across all phenotypes (MDA and DAPSA LDA; range: 42.9% to 58.8% of patients and 66.0% to 82.0% of patients, respectively), with highest responses observed in the Moderate to High Disease Activity and Dactylitis and Feet Dominant phenotypes.
Conclusions:
Five distinct PsA phenotypes were identified in patients starting risankizumab. Moderate to High Disease Activity, the most frequent phenotype, showed the highest response, though risankizumab demonstrated efficacy across all phenotypes. These results are a first step toward more personalized medicine for patients with PsA.
Trial Registration:
ClinicalTrials.gov: KEEPsAKE 1, NCT03675308; KEEPsAKE 2, NCT03671148.
