Margriet Kwint

Real world evidence to audit NTCP-models for acute esophagus toxicity 135 7 Results Two hundred and seventeen consecutive patients were included in this study ( Table 1 ). Mean age was 62 years (SD 8.9 year). One hundred and one patients were treated before 1 st of June 2015 when the prescription dose was 66 Gy in 24 fractions on the primary tumor and pathological lymph nodes and 116 patients were treated after dose de-escalation with a prescription dose of 66 Gy in 24 fractions on the primary tumor and 58.08 Gy in 24 fractions on the pathological lymph nodes. Ninety-five patients developed grade 2 AET (44%) and AET grade 3 was seen in 12 patients (6%) according to the electronic toxicity registration. No AET grade 4 and 5 were registered. Due to the low incidence of grade 3 AET (6%) and corresponding low power, further analyses on grade 3 AET were not possible. Therefore, validation of the V50- and V60 NTCP-models for prediction was only performed for grade ≥2 AET. Validation of electronic toxicity registration The period from December 2012 until December 2013 was considered a learning period for the physicians to get used to the toxicity registration in the electronic medical record. Compliance of toxicity registration in de the electronic medical record increased over this year from 48% to 60% of all appointments. After additional training, compliance increased further to an average of 73%, 80% and 80% for the years 2014, 2015 and 2016 respectively. This means that in 73% - 80% of the patients, the absence/presence of toxicities were registered. The results of the sample test to assess the validity of the electronic toxicity registration are shown in Table 2 . For this sample test, the sensitivity of the electronic toxicity registration for grade ≥2 AET (when the independent validation is the gold standard) was 83%, with a specificity of 86%.