G. L. Andrade de Sousa1, C. Nguyen1, S. DSilva2, K. Walker2, E. M. M. Janowski2, and K. Wijesooriya3; 1Department of Physics, University of Virginia, Charlottesville, VA, 2University of Virginia, Department of Radiation Oncology, Charlottesville, VA, 3University of Virginia, Department of Radiation Oncology, Department of Physics, Charlottesville, VA
Purpose/Objective(s): Treatments for locally advanced pancreatic cancer often produce severe lymphopenia, which has been correlated with worsened overall survival. Maximum lymphocyte reduction from radiation treatment (RT) to pancreas could be as much as 78% from the pre treatment ALC value, including 81% with grade 3 lymphopenia and a nadir at day 35 following RT initiation. Currently there is no computational model to predict radiation induced immune suppression (RIIS). We hypothesize that we can identify metrics that predict RIIS, thereby identifying actionable modifications in RT plans that maximize immune sparing. Materials/
Methods: We have built and validated a python-based predictive model that uses RT treatment plans, CT datasets, dose maps in DICOM format as input and simulates the dynamics of lymphocytes in the blood flow and lymphatic flow, and the osmosis between organs to predict the doses absorbed by circulating lymphocytes and the time-dependent absolute lymphocyte count (ALC) after RT. All organs in the abdomen encompassing a total dose volume of 40 cGy/fraction, including primary/secondary lymphoid organs, and blood rich organs were manually contoured and evaluated by physicians following the RTOG contouring atlas (heart, inferior vena cava, aorta, celiac axis, superior mesenteric artery, liver, kidneys, portal vein, spleen, stomach, vertebral bodies, duodenum (when present), bowel and the lymph nodes (around the aorta, superior mesenteric artery, celiac trunk, and portal vein)). Model was developed and evaluated with 5-fold cross validation on a cohort of 56 pancreatic cancer patients having varying PTV volumes (6.2 cc – 969.9 cc) and time between beginning of RT and the post-RT blood measurement (3 - 239 days). Predicted values were compared to measured lymphocyte values, which were obtained prior to treatment and at 3 month intervals after treatment. Results: Measurements showed varying amounts of immune suppression due to RT to pancreas (11%, 98%). The ALC difference between prediction and measurement is mean (SD): 0.29 (0.23) x109 cells/L. The accuracy was consistent across a wide range of clinical and treatment variables such as pretreatment ALC, PTV volume, post treatment measurement day of ALC. 82.7% of the patients had an ALC difference between prediction and measurement smaller than 0.5x109 cells/L and 43.6% had a difference smaller than 0.2x109 cells/L. Conclusion: Our model performed well in predicting RIIS for pancreatic cancer patients. It has the capability to be interfaced with treatment planning systems to prospectively reduce immune cell toxicity while maintaining conformity and plan quality criteria from national protocols.
