Density Overrides that Best Predict Bowel Changes during Pencil Beam Scanning Proton Radiotherapy for Prostate Cancer

Purpose/Objective(s)

Changes in patient anatomy during pencil beam scanning proton radiotherapy (PBSPT) can result in substantial differences in dose distribution. In patients with prostate cancer receiving elective nodal irradiation (ENI), interfraction variability in bowel position or bowel filling can highlight this uncertainty. Such changes are often managed by assessing the plan on CT dataset copies with the bowel contour overridden to other densities. The ideal Hounsfield Unit (HU) overrides remain unclear, but extremes (HU=0 water and HU=-1000 air) are often utilized. The aim of the present study is to determine the bowel density overrides that are most predictive of bowel changes during a course of PBSPT.

Materials/Methods

Consecutive patients, from a single institution, receiving PBSPT with ENI for intact prostate cancer were retrospectively reviewed for the period between 10/2020-12/2021. Proton radiotherapy consisted of 50.4 Gy in 28 fractions to the pelvic lymph node stations with a simultaneous integrated boost to 70 Gy in 28 fractions to the prostate. Individually contoured loops of small bowel, large bowel, and rectum were overridden to HU of 0 to -1000 in 100 HU intervals. A bowel evaluation structure, consisting of a bowel bag cropped 3 cm away from the high-risk volume, was created; its dose volume histogram was reviewed; and values for the maximum point dose (Dmax), V105%, V45Gy, V20Gy were recorded for each plan. The dosimetry of bowel override plans were compared to the patient's Quality Assurance CT (QACT) scans. The bowel override plan which most closely matched the dosimetry of each QACT for each dosimetric parameter was determined, and the prevalence of closest prediction for each override value was tabulated.

Results

A total of 32 QACTs from 12 patients were included. For Dmax, HU=0, HU=-100, and HU=-700 predicted most accurately with 10 (31.3%), 5 (15.6%), and 4 (12.5%) closest predictions, respectively. For V105%, HU=-700 and HU=-1000 performed best with 6 (18.8%) and 7 (21.9%), respectively. For V45, HU=0 and HU=-500 predicted best with 13 (40.6%) and 5 (15.6%), respectively. For V20, HU=0, HU=-200, and HU=-600 predicted best with 18 (56.3%), 4 (12.5%), and 3 (9.4%), respectively. Overall, HU=0, -100, -600, -700, and -1000 predicted best with 42 (32.8%), 11 (8.6%), 10 (7.8%), 12 (9.4%), and 12 (9.4%) out of the 128 total dose metrics, respectively. This created a relative bimodal distribution with peaks in the 0 to -100 and -600 to -700 ranges. The nominal plan predicted best in 13 scenarios.

Conclusion

Bowel density override plans remain an important tool for assessment of PBSPT robustness to bowel filling changes throughout the course of therapy as the nominal CT relatively poorly predicts QACT dosimetry. The relatively bimodal distribution of accurate predictions in this study would suggest that the most predictive overrides would be HU∼0 and HU∼-650. Future investigation into multidataset robust optimization with these overrides to improve plan robustness is planned.