PREDICT-crFMF score: A novel model for predicting colchicine resistance in children with familial Mediterranean fever.

Abstract

To develop a novel scoring system to predict colchicine resistance in Familial Mediterranean fever (FMF) based on the initial features of the patients. The medical records of patients were analyzed prior to the initiation of colchicine. After generating a predictive score in the initial cohort, it was applied to an independent cohort for external validation of effectiveness and reliability. Among 1418 patients with FMF, 56 (3.9%) were colchicine resistant (cr) and 1312 (96.1%) were colchicine responsive. Recurrent arthritis (4 points), protracted febrile myalgia (8 points), erysipelas-like erythema (2 points), exertional leg pain (2 points), and carrying M694V homozygous mutation (4 points) were determined as the parameters for predicting cr-FMF in the logistic regression model. The cut-off value of 9 was 87% sensitive and 82% specific to foresee the risk of cr-FMF in the receiver operating characteristic. Validation of the scoring system with an independent group (cr-FMF = 107, colchicine responsive = 1935) revealed that the cut-off value was 82% sensitive and 79% specific to identify the risk of cr-FMF. By constructing this reliable and predictor tool, we enunciate that predicting cr-FMF at the initiation of the disease and interfering timely before the emergence of complications will be possible.