Background and purpose: Internal organ motion over a course of radiotherapy (RT) leads to uncertainties in the actual delivered dose distributions. In studies predicting RT morbidity, the single estimate of the delivered dose provided by the treatment planning computed tomography (pCT) is typically assumed to be representative of the dose distribution throughout the course of RT. In this paper, a simple model for describing organ motion is introduced, and is associated to late rectal morbidity data, with the aim of improving morbidity prediction.
Material and methods: Organ motion was described by normally distributed translational motion, with its magnitude characterised by the standard deviation (SD) of this distribution. Simulations of both isotropic and anisotropic (anterior-posterior only) motion patterns were performed, as were random, systematic or combined random and systematic motion. The associations between late rectal morbidity and motion-inclusive delivered dose-volume histograms (dDVHs) were quantified using Spearman's rank correlation coefficient (Rs) in a series of 232 prostate cancer patients, and were compared to the associations obtained with the static/planned DVH (pDVH).
Results: For both isotropic and anisotropic motion, different associations with rectal morbidity were seen with the dDVHs relative to the pDVHs. The differences were most pronounced in the mid-dose region (40-60 Gy). The associations were dependent on the applied motion patterns, with the strongest association with morbidity obtained by applying random motion with an SD in the range 0.2-0.8 cm.
Conclusion: In this study we have introduced a simple model for describing organ motion occurring during RT. Differing and, for some cases, stronger dose-volume dependencies were found between the motion-inclusive dose distributions and rectal morbidity as compared to the associations with the planned dose distributions. This indicates that rectal organ motion during RT influences the efforts to model the risk of morbidity using planning distributions alone.