Development and external validation of a machine learning model to predict central sensitization symptom burden in psoriatic arthritis

Objectives Central sensitization (CS) is a general mechanism of pain amplification. In psoriatic arthritis (PsA), it contributes to pain that is disproportionate to inflammation. We aimed to develop and externally validate machine learning (ML) models to identify clinically significant central sensitization symptom burden in PsA from routinely available clinical and patient-reported variables. Methods Models were developed in a cohort of 223 patients with PsA (CASPAR criteria) and externally validated in 50 patients from two referral centers. The outcome was a Central Sensitization Inventory (CSI) total score ≥40. Logistic Regression (LR) and Random Forest (RF) were developed using nested 5-fold stratified cross-validation with model-specific regularisation. The final models were then refitted on the entire development cohort, recalibrated using logistic recalibration, and independently evaluated in the external validation cohort. Results Elevated CS symptom burden (CSI ≥40) was present in 63.7% of patients internally and 70% externally. Both models discriminated well internally (AUROC 0.83 LR, 0.81 RF) and externally (AUROC 0.78, 95% CI 0.65–0.90, LR; 0.84, 95% CI 0.72–0.93, RF), with comparable performance between models, and were adequately calibrated after recalibration, with wider uncertainty in the external cohort. The main predictors, consistent across both models, were VAS pain, pain frequency, morning-stiffness duration, DAPSA and sex. Conclusions ML models built on routine variables can identify patients with PsA who have clinically significant CS symptom burden, with good internal and acceptable performance on preliminary external validation. LR and RF performed comparably, with LR offering a transparent, interpretable model and RF greater modelling flexibility. Such models may become an easily implemented assessment of elevated CS symptom burden in clinical practice for PsA patients’ management, pending further validation in larger, more diverse cohorts.

Authors

Institutions

Publication Details

Journal
Frontiers in Immunology
Published
2026-11-19
DOI
https://doi.org/10.3389/fimmu.2026.2013699
Primary Topic
Spondyloarthritis Studies and Treatments
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Development and external validation of a machine learning model to predict central sensitization symptom burden in psoriatic arthritis

Paolo Soda, Francesca Tecce, Riccardo Terribili, Rosa Sicilia et al.
Frontiers in Immunology
Spondyloarthritis Studies and Treatments
article

Development and external validation of a machine learning model to predict central sensitization symptom burden in psoriatic arthritis

Paolo Soda, Francesca Tecce, Riccardo Terribili, Rosa Sicilia, Paola Cipriani, Guglielmo Giannini, Piero Ruscitti, Vanda Franco, Luca Navarini, Marta Vomero, Annalisa Marino, Damiano Currado, Francesco De Vincenzo, Francesca Saracino, Roberto Giacomelli, Stefano Gentileschi, Erika Corberi, Francesca Trunfio, Valerio Guarrasi, Onorina Berardicurti, Leonardo Frascà, Ludovica Lamberti, Marco Salmè
article en

Abstract

Objectives Central sensitization (CS) is a general mechanism of pain amplification. In psoriatic arthritis (PsA), it contributes to pain that is disproportionate to inflammation. We aimed to develop and externally validate machine learning (ML) models to identify clinically significant central sensitization symptom burden in PsA from routinely available clinical and patient-reported variables. Methods Models were developed in a cohort of 223 patients with PsA (CASPAR criteria) and externally validated in 50 patients from two referral centers. The outcome was a Central Sensitization Inventory (CSI) total score ≥40. Logistic Regression (LR) and Random Forest (RF) were developed using nested 5-fold stratified cross-validation with model-specific regularisation. The final models were then refitted on the entire development cohort, recalibrated using logistic recalibration, and independently evaluated in the external validation cohort. Results Elevated CS symptom burden (CSI ≥40) was present in 63.7% of patients internally and 70% externally. Both models discriminated well internally (AUROC 0.83 LR, 0.81 RF) and externally (AUROC 0.78, 95% CI 0.65–0.90, LR; 0.84, 95% CI 0.72–0.93, RF), with comparable performance between models, and were adequately calibrated after recalibration, with wider uncertainty in the external cohort. The main predictors, consistent across both models, were VAS pain, pain frequency, morning-stiffness duration, DAPSA and sex. Conclusions ML models built on routine variables can identify patients with PsA who have clinically significant CS symptom burden, with good internal and acceptable performance on preliminary external validation. LR and RF performed comparably, with LR offering a transparent, interpretable model and RF greater modelling flexibility. Such models may become an easily implemented assessment of elevated CS symptom burden in clinical practice for PsA patients’ management, pending further validation in larger, more diverse cohorts.

Frontiers in ImmunologyVol. 17
University of Siena (IT), Università Campus Bio-Medico (IT), University of L'Aquila (IT), Saint Camillus International University of Health and Medical Sciences (IT), European University of Rome (IT), Campus Bio Medico University Hospital (IT), Link Campus University (IT), Umeå University (SE)
Spondyloarthritis Studies and Treatments
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon