ESTRO 38 Abstract book

S401 ESTRO 38

Results 188 patients were included. Mean age was 65 years, 50% was male, 69% had adenocarcinoma. Seventy percent received whole-brain RT (95% 20Gy/5#, 2% 25Gy/10# and 3% 39Gy/13#) and 30% underwent stereotactic RT (84% a single fraction of 15, 18, 20 or 21Gy). Ninety-one percent had a KPS ≥70, 70% had extracranial metastases at the time of brain metastases diagnosis. Forty-one percent had 1-2 BM, 46% 3-10 and 13% had >10 BM. Median OS was 4.9 months. The RPA and DS-GPA models did not show significant differences in OS between the risk groups (fig 1). Independent significant risk factors for mortality were non-adenocarcinoma, KPS <70, brain metastases (1-2 vs . 3-10 or >10) and age ≥65 years. Treatment type (whole- brain or stereotactic RT), fractionation schemes, extracranial metastases or mutation status (EGFR, KRAS, ALK/Ros) were not prognostic. Next, the LUMC-BM score was composed using these risk factors and classified patients into three risk groups (tab 1). The median OS for groups 1 (n=94), 2 (n=52) and 3 (n=26) were 6.9, 3.9 and 1.9 months (log rank p<0.0001), respectively. The hazard ratio’s for mortality in group 2 and 3, relative to group 1, were 2.1 (95% CI 1.4-3.1) and 5.9 (95% CI 3.6-9.8) respectively.

Results A total of 62 patients with evaluable % change NLR were available for analysis. The median time interval between blood draw and SBRT was 33 and 55 days for pre- and post- SBRT NLR, respectively. The overall median follow-up was 25.4 months. Of the 62 patients, 38 (61.3%) demonstrated an increase in NLR following SBRT. Key patient characteristics (age, gender, race, KPS, smoking, comorbidities, steroid use, peripheral/central location, stage, mutation status) were well balanced, except for higher rates of hyperlipidemia and less involvement of right upper lobe in the NLR % change >0 cohort. Treatment characteristics (number of fractions, dose/fraction, total dose, BED) did not differ by cohort. A statistically significant OS advantage was seen in patients when stratified by NLR % change: 12-, 24- and 36- month OS rates were 100% vs. 78.1%, 95.7% vs. 50.0%, and 87.7% vs. 45.0%, amongst patients with NLR % change ≤0 vs. >0, respectively. On univariate analysis, NLR % change (p=0.001), female gender (p=0.047), KPS (p=0.047), COPD (p=0.031), and total dose (p=0.013) were significant factors for OS. NLR % change >0 remained significant for predicting inferior OS (Hazard Ratio: 6.16 [95% CI: 1.70- 22.28], p=0.006) on multivariable analysis confirmed; all other factors were not significant. Other disease outcomes were not significantly associated with % change NLR: 2-year lobar recurrence (0% vs. 7.0%), nodal recurrence (4.4% vs. 18.2%), second primary (0% vs. 7.1%), or distant recurrence (13.3% vs. 22.7%) [all NLR % change ≤0 vs. >0]. Conclusion The percent change in NLR after SBRT, a potential marker for radiation-induced inflammatory response, is inversely related to OS in patients with early-stage lung cancer. If prospectively validated, NLR is a simple, systemic marker that can be easily used to guide subsequent management. PO-0777 Predicting overall survival after radiotherapy for brain metastases in patients with NSCLC N. Knotter 1 , N. Horeweg 1 , I. Coremans 1 , R. Wiggenraad 2 , Y. Van der Linden 1 1 Leiden University Medical Center LUMC, Department of Radiotherapy, Leiden, The Netherlands ; 2 Haaglanden Medical Center, Department of Radiotherapy, Leidschendam, The Netherlands Purpose or Objective Brain metastases (BM) are common in non-small cell lung cancer (NSCLC) patients. The current standard treatment is palliative radiotherapy (RT), although its effect on symptoms and survival is debatable. The QUARTZ study suggested that only patients with a good performance status and expected prolonged survival benefit [1]. To aid in predicting overall survival (OS), recursive partitioning analysis (RPA) and Diagnosis specific graded prognostic score (DS-GPA), based on USA trial patients, were developed. Their use is currently widespread [2, 3, 4]. A recent Dutch study reported that OS in patients outside of trials is worse than these models suggest [5]. We therefore investigated whether these models correctly predict OS, and searched for additional prediction factors. Subsequently, a prognostic model was constructed to fit Retrospective cohort study of all patients with NSCLC treated with RT for BM at Leiden University Medical Center (2011-2018). Patient characteristics, data concerning fractionation schemes and follow-up were extracted from electronic patient charts. Kaplan-Meier’s methodology was used to estimate OS. Uni- and multivariable Cox regression analyses were performed to identify significant prognostic factors and to define our prediction model. Performance of the three models was compared using the log-rank test. NSCLC patients with BM. Material and Methods

Conclusion Our new prognostic model more accurately predicts OS in NSCLC patients with brain metastases outside of trials than current widespread used RPA and DS-GPA prediction models. The validation study of our model is currently being performed.