Encrypted login | home

Program Information

Fitting Three NTCP Models to Treatment Outcome Data of Cardiac Toxicity After Radiation for Stage III Non-Small Cell Lung Cancer

P Mavroidis

P Mavroidis*, K Wang , M Eblan , B Jensen , J Dooley , L Marks , S Das , Univ North Carolina, Chapel Hill, NC


SU-I-GPD-T-95 (Sunday, July 30, 2017) 3:00 PM - 6:00 PM Room: Exhibit Hall

Purpose: To estimate the radiobiological parameters of three popular NTCP models that describe the dose-response relations of cardiac toxicity following radiotherapy for non-small cell lung cancer (NSCLC). To identify the parts of heart that best correlate with the manifestation of these clinical endpoints. Finally, to evaluate the goodness-of-fit of the NTCP models.

Methods: The clinical data of 112 patients with Stage III NSCLC (ECOG PS 0-1), who were treated to 70-90 Gy on six dose-escalation trials, were analyzed. The dose volume histograms of left and right ventricles, left and right atrium, left anterior descending artery, heart and pericardium were calculated. The records of the patients were reviewed for symptomatic cardiac events (symptomatic pericardial effusion, acute coronary syndrome, significant arrhythmia and combination of them). Baseline risk was assessed using the WHO/ISH score. The Lyman-Kutcher-Burman (LKB), Relative Seriality (RS) and Logit NTCP models were used to fit the clinical data. The fitting of the different models was assessed through the area under the receiver operating characteristic curve (AUC) and the Odds Ratio methods.

Results: The AUC values were highest for left ventricle regarding acute coronary syndrome (LKB model: D50 = 78.9Gy, m = 0.51, n = 0.49), right ventricle regarding significant arrhythmia (LKB model: D50 = 95.6Gy, m = 0.56, n = 0.44), left atrium regarding symptomatic pericardial effusion (LKB model: D50 = 93.2Gy, m = 0.42, n = 0.71) and pericardium regarding the combination of all the symptoms (LKB model: D50 = 55.6Gy, m = 0.70, n = 0.45).

Conclusion: It was shown that the examined NTCP models could fit the clinical data well. The model parameter values were derived for four structures that correlated best with the symptoms examined. Those parameter sets can be used in the form of NTCP objective functions in treatment plan optimization.

Contact Email: