Development and internal–external cross-validation of a patient-reported definition for acute calcium pyrophosphate crystal arthritis

Objective To develop and validate a patient-reported definition of acute calcium pyrophosphate (CPP) crystal arthritis in people with crystal-proven CPP deposition (CPPD) disease.Methods Consecutive patients with crystal-proven CPPD disease from seven centres across four countries were enrolled in a cross-sectional study. In each centre, patient-reported outcomes on the features of acute CPP crystal arthritis were collected. The expert opinion of an independent rheumatologist was the reference standard. We developed definitions based on multivariable logistic regression model with backward selection of predictors and classification and regression tree (CART) approaches.Results Two hundred and forty-six patients [mean age 73.2 years (s.d. 10.7), 65.9% female] were enrolled. At the time of the assessment, acute CPP crystal arthritis was diagnosed in 96/246 (39.0%) participants. Patient-reported joint warmth, patient-reported joint swelling, time from pain onset to peak, and self-reported acute CPP crystal inflammatory arthritis were included in the multivariable logistic model. This model had good discrimination (optimism-adjusted c-index: 0.92; 95% CI: 0.89, 0.95) and calibration (optimism-adjusted calibration slope: 0.95; 95% CI: 0.71, 1.19; optimism-adjusted calibration-in-the-large: 0.005; 95% CI: -0.37, 0.37) in the internal validation. Probability threshold >= 0.53 had sensitivity of 0.83 (95% CI: 0.74, 0.90) and specificity of 0.86 (95% CI: 0.79, 0.91). Performances were similar in the internal-external cross-validation. The CART identified patient-reported acute CPP crystal inflammatory arthritis, followed by joint swelling and joint warmth as the most informative variables for ascertaining acute CPP crystal arthritis [sensitivity 0.83 (95% CI: 0.72, 0.91) and specificity 0.83 (95% CI: 0.74, 0.90)].Results Two hundred and forty-six patients [mean age 73.2 years (s.d. 10.7), 65.9% female] were enrolled. At the time of the assessment, acute CPP crystal arthritis was diagnosed in 96/246 (39.0%) participants. Patient-reported joint warmth, patient-reported joint swelling, time from pain onset to peak, and self-reported acute CPP crystal inflammatory arthritis were included in the multivariable logistic model. This model had good discrimination (optimism-adjusted c-index: 0.92; 95% CI: 0.89, 0.95) and calibration (optimism-adjusted calibration slope: 0.95; 95% CI: 0.71, 1.19; optimism-adjusted calibration-in-the-large: 0.005; 95% CI: -0.37, 0.37) in the internal validation. Probability threshold >= 0.53 had sensitivity of 0.83 (95% CI: 0.74, 0.90) and specificity of 0.86 (95% CI: 0.79, 0.91). Performances were similar in the internal-external cross-validation. The CART identified patient-reported acute CPP crystal inflammatory arthritis, followed by joint swelling and joint warmth as the most informative variables for ascertaining acute CPP crystal arthritis [sensitivity 0.83 (95% CI: 0.72, 0.91) and specificity 0.83 (95% CI: 0.74, 0.90)].Conclusion We developed and initially validated easy-to-use patient-reported definitions for acute CPP crystal arthritis for use in clinical trials and observational research in CPPD.