ESTRO 2022 - Abstract Book

S83

Abstract book

ESTRO 2022

Conclusion Patients included in CONCORDE trial with major protocol deviations had a reduced survival and increased toxicities. RTQA remains therefore essential in clinical trials. In the subgroup of patients treated per protocol or with minor deviations only, dose escalation did not appear to improve survival in esophageal cancer patients treated with exclusive CRT.

OC-0109 External validation of a prediction model for two-year mortality in esophageal cancer cohorts

M. Berbee 1 , C. Muijs 2 , F. Voncken 3 , L. Wee 1 , E. Oldehinkel 2 , A. Schaaf van der 4 , J. Reitsma 5 , E. Schuit 6

1 Maastro, Radiation Oncology, Maastricht, The Netherlands; 2 University Medical Center Groningen, Radiation Oncology, Groningen, The Netherlands; 3 The Netherland Cancer Institute/Antoni van Leeuwenhoek, Radiation Oncology, Amsterdam, The Netherlands; 4 University Medical Center Groniningen, Radiation Oncology, Groningen, The Netherlands; 5 University Medical Center Utrecht, Julius Center, Utrecht, The Netherlands; 6 Utrecht University Medical Center, Julius Center, Utrecht, The Netherlands Purpose or Objective Although chemo-radiotherapy has been shown to improve the oncological outcome of esophageal cancer (EC) patients, thoracic radiotherapy may also cause long term toxicity and increase the risk of non-cancer related death due to radiation exposure of normal tissues. For lung cancer patients a model to predict 2-year mortality using GTV volume and mean heart dose (MHD) has been developed and validated previously (https://nvro.nl/images/documenten/rapporten/LIPP_longen_final_01122019.pdf). We here aimed to validate this model for EC patients. Materials and Methods Five EC patient cohorts from 3 different Dutch radiotherapy centers were used for model validation. Two cohorts of patients (n=170) were used for external validation in the setting of definitive chemoradiotherapy (dCRT) and 3 cohorts (n=568) for validation in the neo-adjuvant setting (nCRT). External validity was assessed in terms of calibration (i.e., agreement between predicted and observed risk) by calibration plots, and discrimination (i.e., ability of the model to distinguish between those who died within 2 years and those who did not) by assessment of the c-statistic. If indicated, the model was updated by adjustment of the intercept, or intercept and slope (“recalibration”), or regression coefficients (“revision”) of the model. An additional logistic regression analysis assessed the confounder-adjusted association between the MHD and the outcome as well as the incremental value of the MHD on top of other known predictors of 2-year mortality (GTV, age, gender, histology, N stage, WHO performance status and tumor length). The incremental value was assessed using a likelihood ratio test (LRT) and the integrated discrimination index (IDI). Results For the dCRT patients the model showed good calibration after adjustment of the intercept (Figure 1). The c-statistic was 0.67 (95%CI: 0.58 to 0.75). For nCRT group the model needed recalibration (Figure 2). The c-statistic of the recalibrated model equaled 0.62 (95%CI: 0.57 to 0.67). The additional analysis showed that the MHD was independently associated with 2-year mortality (OR = 1.31; 95% CI 1.03-1.61) and had added value on top of other prognostic factors (LRT p-value = 0.018 and IDI 0.0073; 95%CI 0.0012-0.13; p-value = 0.019).