Volumetric Assessment Of Tumor Vessel Interaction Is A Predictor Of Surgical Candidacy In Patients With Borderline Resectable And Locally Advanced Pancreatic Cancer Treated With Stereotactic Body Radiation Therapy

Surgical resection offers the only chance of cure for non-metastatic pancreatic cancer; however, numerous factors must be taken into account such as tumor size, location, and proximity of the tumor to critical vascular structures. Stereotactic body radiation therapy (SBRT) may improve the likelihood of an unresectable patient to undergo surgery. We developed a method to improve the characterization of tumor location relative to critical vascular structure(s) that affects surgical resectability. Oncospace is a local learning health system that systematically captures clinical outcomes and radiotherapy treatment parameters and can generate overlap volume histograms (OVH)–a measure of spatial relationships between two structures. Minimum distances between the target volume and organs at risk (OAR) based on OVH, GTV, PTV, anatomic location by ICD-9 code, and resected status were queried. Normalized distance from PTV to both kidneys was calculated to determine tumor location within the pancreas. The shape relationship of the GTV to superior mesenteric (SMA), celiac, and common hepatic (CHA) arteries was characterized by distance of involvement (cm), target volume overlap into critical arteries (cc), and radiographic degree of involvement (no involvement, abutment, encasement, and occlusion). Each tumor was further characterized as SMA-dominant or celiac-dominant. Spatial relationships were validated against borderline resectable (BR) and locally advanced (LA) determination, and further compared by resected status by t-test and chi-square test. From 2009-2015, 205 patients were treated with pancreas SBRT to a median total dose of 33 Gy in 5 fractions. Of 163 patients (LA: n = 118, BR: n = 45), 111 were included for analysis of arterial involvement. Patients who underwent surgery (n = 66) were found to have smaller tumor size (P = 0.002) and a further distance from the SMA (P = 0.0005), celiac (P = 0.006), and dominant artery (P = 0.0005). For target volume overlap into critical arteries, resected patients appeared to have less involvement of SMA (P = 0.0006), celiac (P = 0.003), dominant artery (P = 0.0006), and CHA (P = 0.0102). Similar results were found using radiographic degree of involvement of the SMA (P = 0.0008), celiac (P = 0.0272), dominant artery (P = 0.0041), and CHA (P = 0.0032). Our platform enabled systematic analysis of target shape/size, location, and relationship to adjacent arteries. These data suggest that an OVH-based analysis of the distance between PTV, GTV, surrounding OARs, and major arteries may improve our prediction of surgical resectability and decision-making capacity for treatment intent and aggressiveness.