Abstract Background Renal interstitial fibrosis is a well-established prognostic indicator for kidney health, typically assessed via invasive biopsy. Diffusion-weighted MRI provides a non-invasive alternative by measuring the cortico-medullary difference in apparent diffusion coefficient (ΔADC), an inverse correlation with fibrosis has been reported in monocentric studies. This study evaluates the reproducibility of ΔADC across two centres and validates its diagnostic performance in native and allograft kidneys. The performance of T1 and T2 mapping is also assessed. Methods In this prospective cohort study, patients with chronic kidney disease (CKD) and kidney allograft recipients from two hospitals underwent renal biopsy and MRI within one week. Correlations between ΔADC, T1 and T2 mapping and histological fibrosis were assessed in the overall cohort and subgroups based on kidney type and biopsy indication. ROC analysis evaluated the performance of ΔADC for low and high fibrosis. Additional analyses examined correlation between MRI parameters, eGFR, and histological inflammation. A multivariable logistic regression identified predictors of high fibrosis. Results Among 337 patients (121 CKD, 216 allografts), ΔADC correlated negatively with fibrosis (R=-0.42), consistently across centres and kidney type (native R=-0.49; allograft R=-0.39) and regardless of biopsy indication; T1 correlated only in native kidneys with fibrosis, whereas T2 correlated with degree of inflammation in biopsies with low fibrosis. ΔADC showed the best diagnostic accuracy for low (25%) and high (50%) fibrosis (AUCs: 0.71 (CI95% 0.66;0.77) and 0.83 (CI95% 0.76;0.91); cut-offs: 41.5 and -32 × 10⁻⁶ mm²/s, respectively). ΔADC, eGFR, proteinuria, and diabetes independently predicted high fibrosis (model AUC = 0.95 (CI95% 0.90; 0.99). Conclusion ΔADC reliably reflects renal fibrosis across two independent centres and kidney type, supporting diffusion MRI as a non-invasive tool for fibrosis assessment.
Berchtold et al. (2026) studied this question.