FRI0542 OBTAINING HIGH POSITIVE PREDICTIVE VALUES FOR THE DEVELOPMENT OF CLINICALLY APPARENT ARTHRITIS IN PATIENTS PRESENTING WITH CLINICALLY SUSPECT ARTHRALGIA; IS IT FEASIBLE?

Abstract

Background:The hypothesis that initiation of DMARD-treatment before arthritis becomes apparent could permanently modulate the disease process, such that persistent RA is prevented, is being studied in several ongoing trials. Essential for such studies is the ability to accurately predict clinically apparent inflammatory arthritis (IA). However there are two hurdles: first, it is insufficiently known whether it is possible to obtain high positive predictive values (PPV) in patients presenting with clinically suspect arthralgia (CSA). Second, none of current predictive models is validated in independent cohorts. We here aimed to evaluate the first question, incorporating improved markers of MRI-detected subclinical inflammation that were recently identified but have not yet been combined with other known predictors.[1]Objectives:To assess the feasibility of achieving high PPVs in prediction of IA-development in patients with CSA by combining clinical, laboratory and imaging parameters.Methods:580 patients with CSA were consecutively included in the Clinically Suspect Arthralgia (CSA)-cohort and followed on the development of IA, determined by physical examination of joints. Unilateral contrast-enhanced 1.5 Tesla MRIs were made of MCP(2-5), wrist and MTP(1-5)-joints at baseline and scored in line with the RAMRIS. The number of locations with subclinical inflammation (0/1-2/≥3) and the presence of MCP peritendinitis were defined as described previously.[1] Other studied clinical and laboratory variables were based on the literature: initial localisation of complaints (small/large joints), functional disability (health assessment questionnaire (HAQ) ≥1), ACPA-positivity (Anti-CCP2), RF-positivity (IgM-RF) and elevated CRP.[2,3] LASSO Cox regression with a 10-fold cross-validated shrinkage parameter was used for predictor selection. Regression coefficients were rounded to the nearest number ending in .5 or .0 and multiplied by two, resulting in a weighted score. Kaplan Meijer curves were used to obtain PPVs of this weighted score and the area under the curve (AUC) was determined at 2-year follow-up.Results:Mean age was 44, 78% was female, and 18% progressed to IA within 2 years. The following parameters were selected with LASSO: RF-positivity, ACPA-positivity, HAQ≥1, >2 locations of subclinical inflammation and presence of MCP-extensor peritendinitis. Based on the beta of LASSO-regression, patients were assigned 2 points for the risk-factors ACPA-positivity and >2 locations of subclinical inflammation, 1 point for RF-positivity and presence of MCP-extensor peritendinitis and 0 points for HAQ≥1. Kaplan Meijer curves show PPVs of 8%, 9%, 30%, 54%, 73%, 79% and 86% at two years (Figure 1). This model yielded an AUC of 0.79.Figure 1.Kaplan Meijer curves on inflammatory arthritis development stratified for number of points based on LASSO regression. Legend: Points were based on the regression coefficients yielded by Cox LASSO-regression. 2 points were assigned for the risk factors ACPA-positivity and >2 locations of subclinical inflammation and 1 point was assigned for RF-positivity and presence of MCP-extensor peritendinitis.Conclusion:High PPVs for IA-development can be achieved in patients with CSA by weighting a combination of known predictors. Although encouraging, these data are based on one observational cohort study and have not been validated in independent cohorts, limiting the relevance. To support future research in the field of arthralgia, it is needed that different research groups work together to come to risk estimations that are validated and accepted.