Predicting Psoriatic Arthritis in Psoriasis Patients - A Swiss Registry Study.
Details
Serval ID
serval:BIB_C055CD3EF40D
Type
Article: article from journal or magazin.
Collection
Publications
Institution
Title
Predicting Psoriatic Arthritis in Psoriasis Patients - A Swiss Registry Study.
Journal
Journal of psoriasis and psoriatic arthritis
ISSN
2475-5311 (Electronic)
ISSN-L
2475-5303
Publication state
Published
Issued date
04/2024
Peer-reviewed
Oui
Volume
9
Number
2
Pages
41-50
Language
english
Notes
Publication types: Journal Article
Publication Status: ppublish
Publication Status: ppublish
Abstract
Psoriatic arthritis (PsA) is a prevalent comorbidity among patients with psoriasis, heavily contributing to their burden of disease, usually diagnosed several years after the diagnosis of psoriasis.
To investigate the predictability of psoriatic arthritis in patients with psoriasis and to identify important predictors.
Data from the Swiss Dermatology Network on Targeted Therapies (SDNTT) involving patients treated for psoriasis were utilized. A combination of gradient-boosted decision trees and mixed models was used to classify patients based on their diagnosis of PsA or its absence. The variables with the highest predictive power were identified. Time to PsA diagnosis was visualized with the Kaplan-Meier method and the relationship between severity of psoriasis and PsA was explored through quantile regression.
A diagnosis of psoriatic arthritis was registered at baseline of 407 (29.5%) treatment series. 516 patients had no registration of PsA, 257 patients had PsA at inclusion, and 91 patients were diagnosed with PsA after inclusion. The model's AUROCs was up to 73.7%, and variables with the highest discriminatory power were age, PASI, physical well-being, and severity of nail psoriasis. Among patients who developed PsA after inclusion, significantly more first treatment series were classified in the PsA-group, compared to those with no PsA registration. PASI was significantly correlated with the median burden/severity of PsA (P = .01).
Distinguishing between patients with and without PsA based on clinical characteristics is feasible and even predicting future diagnoses of PsA is possible. Patients at higher risk can be identified using important predictors of PsA.
To investigate the predictability of psoriatic arthritis in patients with psoriasis and to identify important predictors.
Data from the Swiss Dermatology Network on Targeted Therapies (SDNTT) involving patients treated for psoriasis were utilized. A combination of gradient-boosted decision trees and mixed models was used to classify patients based on their diagnosis of PsA or its absence. The variables with the highest predictive power were identified. Time to PsA diagnosis was visualized with the Kaplan-Meier method and the relationship between severity of psoriasis and PsA was explored through quantile regression.
A diagnosis of psoriatic arthritis was registered at baseline of 407 (29.5%) treatment series. 516 patients had no registration of PsA, 257 patients had PsA at inclusion, and 91 patients were diagnosed with PsA after inclusion. The model's AUROCs was up to 73.7%, and variables with the highest discriminatory power were age, PASI, physical well-being, and severity of nail psoriasis. Among patients who developed PsA after inclusion, significantly more first treatment series were classified in the PsA-group, compared to those with no PsA registration. PASI was significantly correlated with the median burden/severity of PsA (P = .01).
Distinguishing between patients with and without PsA based on clinical characteristics is feasible and even predicting future diagnoses of PsA is possible. Patients at higher risk can be identified using important predictors of PsA.
Keywords
classification, machine learning, predictive models, psoriasis, psoriatic arthritis, real word, registry, statistics
Pubmed
Create date
25/09/2024 11:33
Last modification date
26/09/2024 7:19