Margriet Kwint

Chapter 7 134 Dosimetric analysis The physical RT-dose was converted to Normalized Total Dose (NTD) for 2 Gy per fraction with an α/β-ratio of 10 Gy for AET. With the NTD corrected dose, esophageal dose-volume-histograms (DVH) were computed and dose-volume-parameters V50 and V60 were derived. The grade ≥2 AET probability was calculated as (16, 17); V50-model: V60-model: The grade ≥3 AET probability was calculated as (16, 17); V50-model: V60-model: Validity of the model was assessed as the ability to predict the number of grade ≥2 and grade ≥3 AET events (calibration). A receiver operating characteristic curve (ROC-curve) was used to analyze the predictive ability of the V50 and V60 prediction models. The area under the curve (AUC) was calculated to distinguish between those who develop grade ≥2 AET. For clinical use, the ability to identify the true positive (sensitivity) patients is more important than the false negative patients (specificity). Therefore, the optimal cut-off point of the model (probability to predict AET) was based on the highest sensitivity. The following sensitivity analyses for quality assurance were performed. A Mann Whitney test was used to compare the median V50 and V60 before and after 1 st of June 2015 (Since 1 st of June 2015 the dose on the irradiated mediastinal lymph nodes was de-escalated to 58.08 Gy in 24 fractions). Thereafter a ROC-curve was used to analyze the predictive ability between the 2 periods (before and after dose de-escalation) for grade ≥2 AET for both V50- and V60-models. The Delong-test was used to analyze the significance differences in AUCs (24, 25). All statistical analyses were performed using IBM SPSS (version 22).