RESEARCHERS have built and tested prediction models that use symptoms and blood markers present at psoriasis onset to identify which patients are most likely to develop psoriatic arthritis, potentially guiding earlier rheumatology referral and closer monitoring.
Undiagnosed Psoriatic Arthritis Carries Long-Term Consequences
Psoriatic arthritis is known to be underdiagnosed in people with psoriasis, and delayed diagnosis is linked to worse long-term joint outcomes, while early treatment improves prognosis. Researchers therefore set out to build prediction models to identify people with new psoriasis who warranted rheumatologist referral, as well as those with subclinical joint disease at highest risk of progressing to clinical psoriatic arthritis, and to pinpoint the strongest predictors.
Following a Swedish Cohort From First Skin Lesion
The team analysed data from the Stockholm Psoriasis Cohort, an inception cohort that enrolled participants within a year of their first psoriasis lesion on non-hairy skin between January 2001 and December 2005. Predictors were recorded at enrolment. Two referral models were developed, one incorporating laboratory biomarkers and one without, alongside two models for subclinical disease, one with a three-year horizon and one with a 15-year horizon. Recursive partitioning and penalised regression were used for modelling, with performance assessed through discrimination, calibration and net benefit across risk thresholds informed by a clinician survey. Of 628 participants without joint disease at enrolment, 347 (55%) were female and 281 (45%) were male, with a median age of 40.9 years.
Discrimination Was Good but Model Stability Varied
Among the 628 participants, 83 (13%) had concomitant psoriatic arthritis at enrolment. Recursive partitioning using laboratory biomarkers stratified participants into four groups based on arthralgia, fatigue, psoriasis phenotype, high-sensitivity C-reactive protein and psoriasis disease activity, with risks of concomitant disease ranging from 1% to 62%. The penalised regression model with a 15-year horizon included nine variables, with the three-year model using a subset of these. All models showed good discrimination, with an optimism-adjusted area under the curve of 0.76 to 0.84, and reasonable calibration. Pain, HLA-B27 and systemic inflammation were the strongest predictors of future disease.
Models Could Guide Referral but Need Validation
The authors concluded that predictors of psoriatic arthritis were already present at psoriasis onset and that the models could support referral and monitoring decisions, though external validation was needed before clinical use.
Reference
Svedbom A et al. Predicting psoriatic arthritis in new-onset psoriasis: development of multivariable prediction models from an inception cohort study. Lancet Rheumatol. 2026;DOI:10.1016/S2665-9913(26)00115-3.
Featured image: Maryna on Adobe Stock
- Author:
