*Statistical Analysis*

Descriptive statistics were used for baseline patient and tumor characteristics. The change in renal function (eGFR) from baseline versus post-treatment at the latest available time point in follow-up was evaluated using the paired sampled *t*-test. Local, regional, distant disease control and overall survival (OS) were estimated using the Kaplan-Meier method. OS was calculated as the time between the first fraction of MRgRT and the date of the last follow-up. LC was calculated as the time between the first fraction of MRgRT and the date of last imaging. Statistical analysis used for plan comparisons was performed using the Wilcoxon Signed-Rank test. A *p*-value of < 0.05 was considered to be statistically significant. Decision tree analysis (CHAID, Chi-square automatic interaction detection) was used to explore predictive pretreatment characteristics and most significant cut-off values to identify patients for

whom daily re-optimization was needed. Baseline volumetric, geometric, and dosimetric parameters, i.e., GTV size (cc), laterality (left, right), location (interpolar, upper or lower pole), V33Gy, V30Gy, V25Gy, and V20Gy for each OAR structure separately or combined in one structure were used as input variables. The qualitative re-optimization benefit variable ("redundant" or "needed") was selected as the target variable for decision tree analysis. The significance level for node splitting was set at *p* < 0.05. Stopping parameters to prevent over-fitting were applied by setting the minimum number of records in a leaf to be at least 10% of the data set. The Statistical Package for the Social Sciences (SPSS) version 26 (IBM® SPSS Statistics, Armonk, NY, USA) was used to perform all statistical analyses.
