Estimation of pneumonitis risk in three-dimensional treatment planning using dose-volume histogram analysis

Abstract

Purpose : Investigations to study correlations between the estimations of biophysical models in three dimensional (3D) treatment planning and clinical observations are scarce. The development of clinically symptomatic pneumonitis in the radiotherapy of thoracic malignomas was chosen to test the predictive power of Lyman's normal tissue complication probability (NTCP) model for the assessment of side effects for nonuniform irradiation. Methods and Materials : In a retrospective analysis individual computed-tomography-based 3D dose distributions of a random sample of sol46 20 patients with lung/esophageal cancer were reconstructed. All patients received tumor doses between 50 and 60 Gy in a conventional treatment schedule. Biological isoeffective dose-volume histograms (DVHs) were used for the calculation of complication probabilities after applying Lyman's and Kutcher's DVH-reduction algorithm. Lung dose statistics were performed for single lung (involved ipsilateral and contralateral) and for the lung as a paired organ. Results : In the lung cancer group, about 20% of the patients (9 our of 46) developed pneumonitis 3–12 (median 7.5) weeks after completion of radiotherapy. For the majority of these lung cancer patients, the involved ipsilateral lung received a much higher dose than the contralateral lung, and the pneumonitis patients had on average a higher lung exposure with a doubling of the predicted complication risk (38% vs. 20%). The lower lung exposure for the esophagus patients resulted in a mean lung dose of 13.2 Gy (lung cancer: 20.5 Gy) averaged over all patients in correlation with an almost zero complication risk and only one observed case of pneumonitis (1 out of 20). To compare the pneumonitis risk estimations with observed complication rates, the patients were ranked into bins of mean ipsilateral lung dose. Particularly, in the bins with the highest patient numbers, a good correlation was achieved. Agreement was not reached for the lung functioning as a paired organ. Conclusions : Realistic assessments for the prediction of radiation-induced pneumonitis seem to be possible. In this respect, the implementation of DVH-analysis in 3D planning could be a helpful tool for the evaluation of treatment plans.