Tool Predicts PsA Progression From Psoriasis, No Fancy Biomarkers Needed

A risk-prediction algorithm was modestly accurate for identifying patients with psoriasis who are likely to progress to psoriatic arthritis (PsA), researchers reported.

Based on patients' demographics and clinical presentations routinely assessed during office visits, the model was about 72% effective in predicting 1-year risk of progression, and 75% effective (using slightly different input data) for 5-year risk, according to Lihi Eder, MD, PhD, of Women's College Hospital in Toronto, and colleagues.

These findings were based on the model's performance with 635 psoriasis patients, of whom 51 actually developed PsA within 1 year of input data collection; 71 did so within 5 years, the researchers reported in Arthritis & Rheumatology.

An area under the receiver operating characteristic curve (AUC) of 0.723 was obtained for 1-year prediction, using the following inputs:

• Age
• Male sex
• Family history of psoriasis
• Back stiffness
• Nail pitting
• Joint stiffness
• Use of biologic medications
• Patient global health
• Pain severity

For 5-year risk, the model's AUC was 0.749 and achieved this accuracy with fewer inputs: presence or absence of morning stiffness, pain, and psoriatic nail lesions; scores for psoriasis severity and fatigue; and use of phototherapy or systemic non-biologic drugs.

Eder and colleagues dubbed the model PRESTO, for Psoriatic Arthritis Risk Estimation Tool, and have made it publicly available on the internet as a risk calculator. To take a hypothetical example, PRESTO estimated that a patient with a 10-year history of psoriasis, using an oral corticosteroid, morning stiffness, some mild pain, and moderate fatigue, but with no nail lesions, has a 5-year risk for PsA development of 9%.

Something like 30% of psoriasis patients eventually develop PsA, earlier studies have found. However, it hasn't been obvious how to tell which patients are most likely to show this type of worsening. A recent "points to consider" document from a European Alliance of Associations for Rheumatology (EULAR) committee recommended that clinicians watch for certain signs, such as persistent pain and joint imaging abnormalities. But the EULAR panel didn't try to quantify the extra risk associated with such signs and symptoms.

Other groups have attempted to find biomarkers that could underpin a risk prediction tool. Last year, for example, researchers from the University of Rochester in New York reported developing a model based on serum levels of leukotriene B4 and glycoursodeoxycholic acid sulfate that yielded an impressive AUC of 0.94. However, no validation in an independent patient sample was conducted, and it could also be sensitive to different assays for these biomarkers.

PRESTO, while perhaps less accurate, has the advantage of using routinely collected data, as well as plausibility insofar as most of the input variables were already considered, as in the EULAR document, to be risk factors for psoriasis-to-PsA progression.

To develop the model, Eder and colleagues examined 29 different variables under multivariate logistic regression and other statistical tools to find the best fit. "A priori, we considered an AUC of greater than 70% acceptable," the group noted.

In addition to PRESTO's potential for informing management of individual patients, the researchers argued that it could also be useful for enrolling patients in PsA prevention trials. Just such a study is now underway, they observed, and others are likely.

Limitations to PRESTO included the relatively small patient cohort that was its basis -- this "prevented inclusion of previously reported risk factors that were present in only a few patients but may have a strong effect size," Eder and colleagues acknowledged. Also, like the serum biomarker-based model, it was not validated in an independent patient group.

Comment