Objective: To develop and validate a clinical prediction model for rheumatoid arthritis-associated interstitial lung disease (RA-ILD) based on multicenter cohorts, and to evaluate the predictive value of MUC5B polymorphisms (rs35705950 and rs868903) and the efficacy of a combined model in a prospective multicenter cohort. 

Methods: A multicenter, two-phase study design was employed. In the discovery phase, 446 RA patients from 5 medical centers were retrospectively enrolled, and multivariate Logistic regression was used to identify independent risk factors for RA-ILD and construct a clinical prediction model. In the validation phase, 238 RA patients from 3 medical centers were prospectively enrolled and underwent MUC5B genotyping (rs35705950 and rs868903). External validation of the clinical model was performed first, followed by development of a combined clinical-genetic model. Model discrimination was assessed using receiver operating characteristic (ROC) curve analysis, calibration was evaluated using the Hosmer-Lemeshow test.

Results: Multivariate Logistic regression in the discovery phase identified smoking history [OR=2.00, 95%CI: 1.01-3.96, P=0.045], increased rheumatoid factor (RF) [OR=1.01, 95%CI: 1.01-1.02, P=0.037], elevated DAS28 score [OR=1.20 95%CI: 1.03-1.41, P=0.023], increased serum CA19-9 [OR=3.70, 95%CI: 1.43-10.4, P=0.002], elevated lactate dehydrogenase (LDH) [OR=1.01, 95%CI: 1.01-1.02, P=0.000], and disease duration ≥2 years [OR=1.73, 95%CI: 1.07-2.86, P=0.029] as independent risk factors for RA-ILD. The clinical prediction model demonstrated an area under the ROC curve (AUC) of 0.785 (95%CI: 0.74-0.83). In the prospective validation cohort, MUC5B rs35705950 and rs868903 genotypes showed significant differential distribution between RA-ILD and RA-only patients (P<0.05). External validation yielded an AUC of 0.777 (95%CI: 0.72-0.84) with good calibration. The combined model incorporating MUC5B genotypes achieved an AUC of 0.786 (95%CI: 0.73-0.85), with no statistically significant difference compared to the clinical model alone (P=0.51).

Conclusions: The RA-ILD clinical prediction model developed from multicenter cohorts demonstrates good predictive performance , with stable reliability confirmed by prospective multicenter validation. Although MUC5B polymorphisms are associated with RA-ILD, the combined model did not significantly improve predictive efficacy, suggesting that clinical indicators play a dominant role in multicenter RA-ILD risk stratification. This clinical model holds broad potential for clinical application.