Performance and calibration of the algorithm ASSIGN in predicting cardiovascular disease in Italian patients with psoriatic arthritis

Abstract

The increased cardiovascular (CV) risk is one of the major challenges in the management of patients with psoriatic arthritis (PsA). Recently, EULAR suggested to adapt the already available CV risk algorithms with a 1.5 multiplication factor in all the patients with rheumatoid arthritis (RA), but it is still uncertain if this adaptation could also be applied to patients with PsA. This study aims to evaluate the performance and calibration of the CV risk algorithm ASSIGN and its adaptations for RA (ASSIGN-RA) and according to EULAR recommendations in a cohort of patients with PsA (ASSIGN*1.5). Prospectively, collected data from two Italian cohorts has been analyzed. The discriminatory ability for CV risk prediction was assessed using the areas under the ROC curves. Calibration between predicted and observed events was assessed by Hosmer-Lemeshow (HL) test and calibration plots. For each algorithm, sensitivity and specificity were calculated for low- to high-risk cut-off (20%). One hundred fifty-five patients were enrolled with an observation of 1550 patient/years. Area under the ROC were 0.8179 (95% CI 0.72014 to 0.91558) for ASSIGN, 0.8160 (95% CI 0.71661 to 0.91529) for ASSIGN-RA, and 0.8179 (95% CI 0.72014 to 0.91558) for ASSIGN*1.5. HL tests did not demonstrate poor model fit for none of the algorithms. Discriminative ability and calibration were not improved by adaptation of the algorithms according to EULAR recommendations. Up to 20% of CV events occurred in patients at “low risk”. No difference in performance has been observed between ASSIGN, Progetto CUORE, and QRISK2. ASSIGN could represent a useful tool in predicting CV risk in patients with PsA. Adaptation for RA or according to EULAR recommendations did not show any further improvement in performance and calibration.