Development and external validation of a clinical-genetic risk identification model for rheumatoid arthritis-associated interstitial lung disease

Vollständiger Abstract

Worum geht es in dieser Arbeit?

Objective This study aimed to develop and externally validate a clinical risk identification model for prevalent rheumatoid arthritis-associated interstitial lung disease (RA-ILD) using multicenter Chinese patient cohorts. We also evaluated the discriminative value of MUC5B polymorphisms (rs35705950 and rs868903) in a combined clinical-genetic model for this purpose. Methods A multicenter, two-phase design was adopted. A total of 446 patients with rheumatoid arthritis (RA) from five medical centers were retrospectively enrolled in the discovery phase. Multivariate logistic regression was used to identify independent risk factors associated with RA-ILD and to construct a clinical risk identification model. A total of 238 patients with RA from three medical centers were then prospectively recruited for the validation phase and genotyped for MUC5B (rs35705950 and rs868903). External validation of the clinical model was performed first, followed by the construction of a combined clinical-genetic model. The model’s discrimination was assessed using receiver operating characteristic (ROC) curve analysis, and calibration was evaluated using the Hosmer–Lemeshow test. Results In the discovery phase, smoking history (odds ratio [OR] = 2.00, P = 0.045), elevated rheumatoid factor (RF) (OR = 1.10 per 10 IU/mL, P = 0.037), a higher 28-joint disease activity score (DAS28) (OR = 1.20, P = 0.023), elevated serum carbohydrate antigen 19-9 (CA19-9) (OR = 3.70, P = 0.002), elevated lactate dehydrogenase (LDH) (OR = 1.13 per 10 U L, P < 0.001), and disease duration of ≥ 24 months (OR = 1.73, P = 0.029) were identified as independent risk factors for RA-ILD. Our clinical risk identification model yielded an area under the ROC curve (AUC) of 0.785 (95% confidence interval [CI]: 0.74–0.83). In the prospective validation cohort, we identified a novel association between MUC5B rs868903 and RA-ILD (OR = 2.58, P = 0.001), with both rs868903 and rs35705950 distinguishing patients with RA-ILD from those without ILD (P < 0.05).The combined model incorporating MUC5B genotypes achieved an AUC of 0.786, which did not differ significantly from that of the clinical model alone (P = 0.507). Although the genetic variants significantly improved continuous risk stratification (integrated discrimination improvement [IDI] = 0.020, P = 0.040; continuous net reclassification improvement [NRI] = 0.466, P < 0.001), categorical NRI at clinically relevant thresholds was non-significant. Conclusions This multicenter study validated a clinical risk identification model for RA-ILD. We identified a novel association between MUC5B rs868903 and RA-ILD in a Chinese cohort. Although the incorporation of genetic information improved continuous risk estimation, it provided limited incremental value at standard clinical decision-making thresholds, indicating that clinical indicators play a dominant role in RA-ILD risk identification. The resultant clinical model may have broad potential for routine clinical use.