A free-breathing CT scan with 4-D respiratory correlation was also obtained to characterize target motion during quiet respiration. If target motion was >5 mm, respiratory gating using the Varian Respiratory Management System™ (Stanford), Cyberknife™ respiratory tracking (Stanford), or the Elekta Active Breathing Coordinator System™ (Hopkins) was utilized during treatment delivery. When available (12 of 18 patients), FDG-PET/CT

scans were fused with simulation scans. SBRT treatment plans were developed using Eclipse™ (Varian, Palo Alto, CA), Multi-Plan™ (Accuray, Sunnyvale, CA), or Pinnacle™ (Philips, Amsterdam, Netherlands). The gross tumor volume (GTV) was contoured Inhibitors,research,lifescience,medical by a radiation oncologist using the simulation scan. An internal target Inhibitors,research,lifescience,medical volume (ITV) was then defined after review of diagnostic imaging, respiratory-correlated

4D-CT, pancreas-protocol CT, and FDG-PET/CT scans. Final planning target volume (PTV) was obtained by an additional 1-3 mm uniform margin expansion of the ITV. The dose was prescribed to the isodose line that completely surrounded the PTV and 6-12 5-FU mouse co-planar fields were used to generate the plan Inhibitors,research,lifescience,medical for non-Cyberknife™ treatments. Dose constraints for organs at risk were employed as follows: duodenum—V15Gy<9 cc, V20Gy<3 cc, V33Gy<1 cc; liver—D50%<12 Gy; stomach—D50%<12 Gy, V33Gy<1 cc; spinal cord—V8Gy<1 cc. Institutional standards for patient-specific dosimetric quality assurance were applied. SBRT delivery For non-Cyberknife™-based treatment (N=11), initial patient position was based on cone-beam CT with alignment to spine. Volumetric kV-imaging was then used to align biliary Inhibitors,research,lifescience,medical stent

and/or fiducials to the digitally-reconstructed radiograph. All fiducials were placed specifically for SBRT image guidance using an endoscopic approach (N=11 patients); complications of fiducial placement were observed in only one patient who experienced Inhibitors,research,lifescience,medical laryngospasm and had to return for repeat endoscopy the following day. Common bile duct stents through were placed endoscopically for relief of symptomatic biliary obstruction and not for purposes of SBRT image guidance, but if a stent was present, then fiducial placement was deemed unnecessary (N=4 patients). If a stent or fiducials were not present, patients were aligned to spine only (N=3). In patients who had previously undergone intra-tumoral fiducial placement, orthogonal kV/MV or kV/kV projection imaging was used to verify fiducial location before first treatment beam delivery and, if indicated, a secondary shift was performed. Active monitoring of treatment delivery accuracy was accomplished using kV and MV projection imaging. For CyberKnife™-based treatment (N=7; fiducials required), initial orthogonal kV/MV or kV/kV projection images were obtained to confirm fiducial location.