ESTRO 37 Abstract book

ESTRO 37

S556

independent cohort, and to identify models for future patient-specific dose escalation. Material and Methods Eight models relevant to dose escalation in patients with stage III NSCLC treated with chemo-radiotherapy (chemo- RT) were identified: four ≥Grade 2 acute esophagitis (AE) models (Huang et al 2012, Huang et al 2017, Oh et al 2017, Wijsman et al 2015), and four ≥Grade 2 radiation pneumonitis (RP) models (Appelt et al 2014, Huang et al 2011, Oh et al 2017, QUANTEC (Marks et al) 2010). Model- specific parameters were applied to the corresponding variables in a stage III NSCLC cohort (N=193) treated with chemo-RT to 50-80Gy@1.8-2.0Gy/fraction in 2004-2014. The endpoints were ≥Grade 2 AE and ≥Grade 2 RP (rates: 48% and 16%). Model performance was assessed using area under the receiver-operating curve (AUC) and p- values, and model calibration between predicted and observed rates in quintiles using the Hosmer-Lemeshow test (p HL ). Results Predictability of the four AE models was overall higher than that of the four RP models (AUC: 0.66-0.71 vs. 0.51- 0.61; p<0.0001 vs. 0.057-0.55; Table; Figure). Except for two AE models, and one RP model (p HL : 0.10, 0.91, 0.99), calibration was reasonable (p HL : 0.32-0.71). The best- performing AE model (Huang et al 2012) included concurrent chemotherapy and mean esophagus dose (D mean ), while the best-performing RP model (Appelt et al 2014) included age>63years, current smoker, former smoker, inferior/middle primary tumors, obstructive lung disease, and sequential chemotherapy) in addition to lung D mean .

Conclusion Among the investigated AE and RP models, those including both treatment and patient characteristics in addition to dose overall performed better than ‘dose only models’. Apart from esophageal and lung D mean we will consider age, obstructive lung disease, primary tumor location, smoking status (RP), and chemotherapy timing (AE, RP) for patient-specific dose-escalation in stage III NSCLC. Since predictive power of the investigated RP models proved limited, we are currently investigating additional variables in modeling this complication. PO-0998 Feasibility study for high energy photon small beam dosimetry audits using OSL and RPL dosimeters T. Bokulic 1 , P. Kazantsev 1 , P. Wesolowska 1 , W. Lechner 2 , J. Izewska 1 1 International Atomic Energy Agency, Dosimetry and Medical Radiation Physics Section, Vienna, Austria 2 Medical University of Vienna / AKH Wien, Department of Radiation Oncology, Vienna, Austria Purpose or Objective A feasibility quality audit study was designed with the objective to: (a) measure high energy photon small beam output factors (OF) with radiophotoluminescent dosimeters (RPLDs), optically stimulated luminescence dosimeters (OSLDs) and a pinpoint ionization chamber (IC) and verify their agreement with the treatment planning system (TPS) calculated OFs and the reference dataset based on measurements in a number of radiotherapy centres, and (b) check the agreement of beam profiles obtained from the (TPS) modelled dose profiles and from the 2D dose distributions measured with gafchromic films. Material and Methods RPL glass rod GD-302M (Asahi Glass Co.) and NanoDot OSLDs from Landauer dosimeters were used in small beam OF measurements. The response of solid state dosimeters was validated through measurements with the pinpoint IC (PTW31014). Small field output factors were measured in a scanning water phantom (10 cm depth,100