Part I | Chapter 7 136 Statistical analysis Descriptive statistics were reported for baseline patient, tumour, and treatment characteristics. Categorical variables were summarised using absolute numbers and percentages. For continuous variables mean and standard deviation or median and interquartile range (IQR) were used, for normally distributed and skewed data, respectively. Significant differences in baseline characteristics between those with and without toxicity were reported. Differences in IPSS between baseline and 1 month and 3 months were assessed using the Wilcoxon signed-rank test. P-values < 0.05 were considered statistically significant. To show the crude effect, correlations between dosimetry parameters and the combined outcome were assessed using univariable logistic regression analysis. Additionally, area under the receiver operating characteristic (ROC) curve (AUC) was calculated for each dose parameter and Pearson’s correlation coefficients were calculated between dose parameters. Multivariable logistic regression analysis was performed, providing odds ratios (OR) corrected for the following available baseline characteristics: age (continuous), diabetes, cardiovascular disease, baseline IPSS (continuous), and alpha-blocker usage at baseline. Dose-effect curves were plotted for a selection of dose parameters that showed the highest correlation with the outcome (bladder Dmean and bladder wall Dmean and V25Gy in cm3) using the corrected OR. Finally, preliminary dose cut-off values for these dose parameters were determined using the ROC-curve, with the optimal constraint – that best discriminates between those with and without the outcome – determined by the Youden index.25 All statistical analyses were performed using SPSS version 26 (IBM® SPSS Statistics, Armonk, New York, United States of America) and R Studio (version 4.1.2, R Foundation for Statistical Computing, Vienna, Austria, https://rstudio.com).
RkJQdWJsaXNoZXIy MTk4NDMw