Caren van Roekel

189 Dose-effect relationships of holmium-166 radioembolization in colorectal cancer was done to determine the discriminatory power of tumor dose in response estimation (18). The 95% confidence interval of the area under the curve shows the boundaries of the likely discriminative ability of tumor dose for response in this cohort. Using a threshold of 100% sensitivity for response (CR/PR), the threshold for a minimal mean tumor-absorbed dose was determined and used in the survival analyses. The same threshold of 100% sensitivity for response was used to determine the threshold for a minimal tumor-absorbed dose (lesion-level). The agreement between response according to PERCIST and according to RECIST was analyzed using Cohen’s kappa, with disagreements weighted according to their squared distance from perfect agreement. Overall survival was defined as the interval between treatment and death from any cause. Cox regression models were made using Firth’s correction for small sample bias (17). Analyses were adjusted for the following possible confounders: tumor load, parenchymal dose and the presence of extrahepatic disease at baseline. Inspection of Schoenfeld residuals showed that the proportionality of the hazard assumption was not violated. Analyses were performed using R statistical software, version 3.6.2 for Windows. The following R libraries were used: readxl version 1.3.1, dplyr version 0.8.3, data.table version 1.12.8, lme4 version 1.1-21, nlme version 3.1-143, ggplot2 version 3.2.1, gdata version 2.18.0, gmodels version 2.18.1, ggpubr version 0.2.4, Hmisc version 4.3-0, lmerTest version 3.1.0, foreign version 0.8-72, ggfortify version 0.4.8, logistf version 1.23, grid version 3.6.2, car version 3.0-5, pROC version 1.15.3, ggeffects version 0.14.0, splines version 3.6.2, sjmisc version 2.8.3, rel version 1.4.1 and rcompanion version 2.3.21. We report effect estimates with associated 95%CIs and corresponding two-sided p-values. RESULTS Forty patients were included, with a total of 133 hepatic lesions. Three patients did not have follow-up imaging for tumor-response assessment and were only included in the survival- and toxicity analyses. Patient- and treatment characteristics are summarized in Table 1. 7