ESTRO 2021 Abstract Book

S712

ESTRO 2021

Figure 1.

Figure 2.

Conclusion We have developed a novel classification of parenchymal changes associated with RILD that correlate with late impairment of lung function. The tissue classes respond to radiotherapy in a dose-dependent manner, and have a distinct evolution over time. A DLA has been developed and trained to automate the parenchymal classification, facilitating future studies on larger patient groups. We have demonstrated the potential of using our approach to analyse and understand the morphological and functional evolution of RILD in greater detail than previously possible. PD-0876 First-failure prediction model for locally advanced non-small cell lung cancer - External validation C.M. Lutz 1 , M.M. Knap 1 , D.S. Møller 1 , L. Hoffmann 1 , A.A. Khalil 1 , K.E. Håkansson 2 , G. Persson 3 , S. Bentzen 4 , L. Nygård 2 , I. Vogelius 2 1 Aarhus University Hospital, Department of Oncology, Aarhus N, Denmark; 2 Rigshospitalet, Department of Oncology, Copenhagen, Denmark; 3 Herlev Hospital, Department of Oncology, Copenhagen, Denmark; 4 University of Maryland School of Medicine, Department of Epidemiology and Public Health, Baltimore, USA Purpose or Objective Patients diagnosed with locally advanced non-small cell lung cancer (LA-NSCLC) face high risk of recurrence or death. Relating the risk of failure to pre-treatment parameters may allow more risk-adapted treatment strategies that may lead to improved outcome. Based on retrospective data from a single institution, a competing risk model for first site of failure after definitive chemoradiotherapy for inoperable LA-NSCLC was published by Nygård et al. [J Thorac Oncol. 2018 13:559-567]. The model uses pre-treatment performance status (PS), stage, histology, concomitant chemotherapy, gender, age, tumor size, SUVpeak of the primary tumor and the lymph nodes with the highest uptake to predict the type of failure or death. In this study, we test the published model in an independent external validation cohort. Materials and Methods An external validation cohort (V) consisting of 258 consecutive LA-NSCLC patients treated with definitive chemoradiotherapy (66 Gy/33 fractions, 2012-2018) was collected retrospectively. Cohort V was compared to the model cohort (M, treated 2009-2015) with respect to the pre-treatment parameters in univariate analysis (c 2 or Mann Whitney U test). The observed incidences of loco-regional and distant failures in the validation cohort were compared to the model predictions. The first site of failure in patients with squamous cell carcinoma (SCC) and adenocarcinoma (AC) were examined with stacked incidence plots and Fine and Gray test, to investigate whether the difference in pattern of failure in the model cohort could be confirmed. For this, distant failures were split into intra- and extracranial distant metastasis. Results The validation cohort presented worse PS (PS≥1, V: 61 %, M: 39 %, p<0.001), lower tumor stage (stage I-II, V: 17 %, M: 9 %, p=0.013) and smaller gross tumor volumes (GTV median volume [range], V: 65.4 cm 3 [2.2;440.8], M: 85.3 cm 3 [4.4;802.6], p<0.001). Concomitant chemotherapy was more frequently used (V: 80 %, M: 67 %, p<0.001). No significant differences were observed for gender, age, histology and SUVpeak of primary tumor or lymph nodes. The predicted and observed cumulative incidence rates are in agreement within the uncertainty bounds for both loco-regional failure and distant metastasis (Figure 1). The observed difference in