A clinical-radiomics integrated nomogram for predicting 5-year survival in RA-associated interstitial lung disease

OBJECTIVES: RA-associated interstitial lung disease (RA-ILD) shortens survival and impairs quality of life, with prognosis assessment challenging. This study aimed to predict RA-ILD patients' survival via clinical features and radiomics. METHODS: Patients from two centres (n = 230) were divided into training (n = 144), internal validation (n = 37) and external validation (n = 49) sets. Univariate/multivariate Cox regression identified independent clinical predictors (age, lymphocyte count). Radiomics features (n = 1688) were selected via variance thresholding, univariate selection and LASSO-Cox regression (24 optimal features retained). Clinical, radiomics and integrated models (with nomogram) were built; performance was assessed by concordance index (C-index), calibration curves and decision curve analysis (DCA). Patients were stratified by Rad-score (threshold = 0.15) for survival analysis. RESULTS: Age (HR = 1.046, 95%CI = 1.013-1.080, P = 0.006) and lymphocyte count (HR = 1.385, 95%CI = 1.059-1.812, P = 0.018) were independent overall survival (OS) predictors. The integrated model outperformed others (C-index: training = 0.832, internal = 0.816, external = 0.812), with excellent calibration and higher DCA net benefit. High-risk patients (Rad-score ≥ 0.15) had significantly shorter OS than low-risk patients (all P < 0.01). CONCLUSION: The integrated nomogram (age, lymphocyte count, Rad-score) enables precise, user-friendly prediction of RA-ILD 5-year survival. It addresses unmet clinical needs by complementing current guidelines and supporting personalized risk-stratified management.