The predictors of psoriatic arthritis are already there on the day psoriasis is diagnosed

Design

prediction model development and internal validation in a psoriasis inception cohort

Population

628 participants enrolled within one year of first psoriasis lesion, Stockholm, 2001-2005; median age 40.9 years

Primary outcome

concomitant psoriatic arthritis, and psoriatic arthritis at 3 and 15 years

Effect

13% had concomitant psoriatic arthritis; risk strata from 1% to 62%; optimism-adjusted AUC 0.76-0.84; strongest predictors pain, HLA-B27 and systemic inflammation

The Stockholm Psoriasis Cohort enrolled people within a year of their first psoriasis lesion, between 2001 and 2005, and followed them. This analysis took the 628 who reported no psoriatic arthritis at enrolment - and found that 83 of them, 13%, already had it.

Recursive partitioning using five factors available at that first visit - arthralgia, fatigue, psoriasis phenotype, high-sensitivity CRP and psoriasis disease activity - split those 628 people into four groups whose risk of concomitant psoriatic arthritis ranged from 1% to 62%. Models for future disease at three and fifteen years were also built, with optimism-adjusted areas under the curve of 0.76 to 0.84 across all models. Pain, HLA-B27 and systemic inflammation were the strongest predictors of psoriatic arthritis developing later. Net benefit analysis, with thresholds set by a clinician survey, supported using these for referral and monitoring decisions but not for deciding on preventive treatment.

Two cautions come from the authors themselves: the models showed instability, particularly the prognostic ones, and none has been externally validated. So this is not a calculator to start using. What is usable today is simpler and does not need a model: a person with new psoriasis who reports joint pain and fatigue, has a raised CRP and has active skin disease sits in a group where nearly two in three already have arthritis. That is a question to ask at the first dermatology visit, not at the third.

• Ask about joint pain, fatigue and inflammatory back symptoms at the first visit for new psoriasis, not later
• Check a CRP at diagnosis of psoriasis - it carried independent predictive weight here
• Treat coexisting arthralgia plus fatigue plus raised CRP as a reason for rheumatology referral rather than watchful waiting
• Do not use these models to select anyone for preventive treatment - net benefit did not support that
• Remember 13% already had psoriatic arthritis at psoriasis diagnosis, so the question is often about detection rather than prediction

Why it matters

It says the information needed to find psoriatic arthritis early is present at the psoriasis diagnosis, where nobody currently looks for it.

Don't overread it

Internally validated models only, showing instability and no external validation - these are not ready to be used as a calculator.

The statistics, in plain English

An area under the curve of 0.76 to 0.84 is respectable discrimination for a clinical prediction model, but 'optimism-adjusted' means it has been corrected for overfitting to its own data and still needs testing on a different population - models routinely lose several points in external validation. The instability the authors report means small changes in the data would produce different models, which is what happens with 83 events and nine variables. The 1% to 62% risk spread is the finding that is robust to all this, because it reflects a real difference between people rather than a modelled coefficient.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.