Margriet Kwint

Real world evidence to audit NTCP-models for acute esophagus toxicity 131 7 Introduction Randomized clinical trials (RCTs) provide the highest level of evidence to prove treatment efficacy and safety. When evidence from RCTs is lacking, clinical decision- making is typically supported by data from observational studies and clinical registries. Collectively, data obtained from sources outside RCTs are often referred to as ‘real world data’ (RWD), and the evidence derived from aggregation and analysis of such data as real world evidence (RWE) (1-5). RWE is often criticized due to its presumed lack of validity because of the broad patient heterogeneity and the vast amount of missing data compared to data derived from RCTs (1). With the introduction of electronic medical records, a more systematic and accurate recording of treatment related toxicity was implemented in clinical practice (6, 7). Such a toxicity registration, when consistently used for all patients, provides a large amount of data, possibly useful for the evaluation of treatment quality since this data can easily be associated with patient characteristics and/or treatment parameters. Nonetheless, the completeness and validity of this ‘big data’ remains uncertain. Therefore, validation is needed to define the accuracy of such a registration to provide RWE (1, 8, 9). For patients with locally advanced non-small cell lung cancer (NSCLC), the treatment of choice is concurrent chemoradiation (CCRT)(10-12). The addition of chemotherapy provokes a radiosensitizing effect leading to an improved local tumor control and overall survival, compared to radiotherapy only or sequential chemoradiation. But this comes at the cost of an increase of acute esophageal toxicity (AET)(13). Several studies have reported (14-20) on different dosimetric and clinical parameters to predict the risk of AET with the use of Intensity Modulated Radiotherapy (IMRT). We previously investigated the dose-effect relation between dose volume parameters and AET (16) and identified the V50 (volume of the esophagus receiving ≥ 50 Gy) as the most accurate predictor of AET. Palma et al. (17) performed an individual-patient- data meta-analysis and concluded that the V60 (volume of the esophagus receiving ≥60 Gy) was the most accurate predictor for AET. The aim of this work is to assess the validity of a RWD toxicity registration and to show the feasibility of such an infrastructure to audit toxicity prediction models and dose constraints in daily clinical practice. Since June 2015 the dose to the irradiated mediastinal lymph nodes was de-escalated from 66 Gy to 58.08 Gy (60 Gy (EQD2))