ESTRO 38 Abstract book

S207 ESTRO 38

OC-0405 Registry-based modelling of early mortality following radiotherapy of lung cancer L. Stervik 1 , N. Pettersson 2 , J. Scherman 3 , C.F. Behrens 4 , C. Ceberg 5 , S. Engelholm 6 , G.F. Persson 7 , M. Pøhl 8 , A. Hallqvist 9 , J. Nyman 9 , I.R. Vogelius 10 , A. Bäck 2 1 University of Gothenburg, Department of Radiation Physics, Gothenburg, Sweden ; 2 Sahlgrenska University Hospital- University of Gothenburg, Department of Therapeutic Radiation Physics- Department of Radiation Physics, Gothenburg, Sweden ; 3 Skåne University Hospital, Department of Radiation Physics, Lund, Sweden ; 4 Herlev and Gentofte Hospital- University of Copenhagen, Department of Oncology, Copenhagen, Denmark ; 5 Lund University, Department of Medical Radiation Physics, Lund, Sweden ; 6 Skåne University Hospital, Department of Oncology, Lund, Sweden ; 7 Herlev and Gentofte Hospital- University of Copenhagen, Department of Oncology- Department of Clinical Medicine, Copenhagen, Denmark ; 8 Rigshospitalet- University of Copenhagen, Department of Oncology, Copenhagen, Denmark ; 9 Sahlgrenska University Hospital, Department of Oncology, Gothenburg, Sweden ; 10 Rigshospitalet- University of Copenhagen, Department of Oncology- Department of Clinical Medicine, Copenhagen, Denmark Purpose or Objective To demonstrate registry based outcome modeling using multivariable models of 90- and 180-day mortality rates following conventionally fractionated, curative radiotherapy treatments (RTs) for non-small cell lung cancer (NSCLC) as exemplar. Material and Methods The scripting capabilities of a modern oncology information system was used to automatically identify patients and extract dose volume information. We considered all patients available in the oncology information systems diagnosed with NSCLC treated with conventionally fractionated, curative RT between 2002 and 2017 at two Danish and two Swedish hospitals. Date of death was available through national registries or hospital records. Exclusion criteria were previous dose exposure in the thoracic region, another RT concurrently or during the observation period for early mortality (90- or 180-days), and/or dose or treatment time outside of normal range (see Figure 1). Maximum likelihood estimation and univariable logistic regressions (LRs) for the binary endpoints death within 90- and 180-days from the treatment start were used. Predictor variables with a p-value for the regression coefficient less than 0.1 in likelihood ratio tests were considered eligible for multivariable LRs to model the same endpoints. The statistical significance of each multivariable model as a whole was evaluated with likelihood ratio tests. Results Data were automatically extracted for 2018 patients. Applying exclusion criteria, 1721 patients and 1621 patients remained for modelling of the 90- and 180-day mortality rate, respectively. Figure 1 shows a CONSORT diagram of the patient cohort. The 90- and 180-day mortality rates were 2.9% (50/1721) and 10% (170/1621), respectively. MLD and patient age were significantly associated with mortality after both 90 and 180 days (all p<0.01). Predicted 90-day mortality rate (Figure 2, upper panel) of 5% for patients aged 50, 60, 70 and 80 years were reached at MLD of 34, 28, 22 and 16 Gy, respectively. Calibration plots showed a good agreement between observed and modelled 90- (Figure 2, lower panel) and 180-day mortality rates.

Using median C20 to stratify the patient data resulted in two significantly different survival distributions when all calcifications were used (p=0.003, log-rank test) but border-line for calcifications only in the heart. Figure 2 shows the corresponding Kaplan-Meier survival estimates.

Conclusion Our results suggest that the risk of death increases by 27% for every extra cm 3 of calcifications receiving at least 20 Gy. Moreover, the loss of significance when analysing data to only the calcifications within the heart suggest an important dose/effect relation for calcifications outside the heart and these should not be ignored. We are planning to validate of our findings in an external cohort.