Four-factor risk score for the prediction of interstitial lung disease in rheumatoid arthritis

Abstract

Objective: Interstitial lung disease (ILD) is one of the commonest systemic complications in patients with rheumatoid arthritis (RA) and carries a significant morbidity and mortality burden. We aimed to identify key variables to risk-stratify RA patients in order to identify those at increased risk of developing ILD. We propose a probability score based on the identification of these variables.

Methods: A retrospective, multicentre study using clinical data collected between 2010 and 2020, across 20 centres.

Results: A total of 430 RA (210 with ILD confirmed on high-resolution computed tomography (HRCT)) patients were evaluated. We explored several independent variables for the risk of developing ILD in RA and found that the key significant variables were smoking (past or present), older age and positive rheumatoid factor/anti-cyclic citrullinated peptide. Multivariate logistic regression models were used to form a scoring system for categorising patients into high and low risk on a scale of 0-9 points and a cut-off score of 5, based on the area under the receiver operating characteristic curve of 0.76 (CI 95% 0.71-0.82). This yielded a sensitivity of 86% and a specificity of 58%. High-risk patients should be considered for investigation with HRCT and monitored closely.

Conclusion: We have proposed a new model for identifying RA patients at risk of developing ILD. This approach identified four simple clinical variables: age, anti-cyclic citrullinated peptide antibodies, Rheumatoid factor and smoking, which allowed development of a predictive scoring system for the presence of ILD in patients with RA.

Keywords: Extra-articular manifestation; Interstitial lung disease; Probability score; Rheumatoid arthritis; Risk prediction.

Fig. 1

Receiver operating characteristic curve of the risk scores (the ROC area (95% CI) 0.76 (0.71–0.82). A value of 5 points (with the sign of the cross) showed a sensitivity of 86% and a specificity of 58%. The point was chosen based on the optimal sensitivity and specificity