Using Pelvic Bone Marrow Hounsfield Units to Predict Cytopenia During Anal Cancer Chemoradiation

E214

International Journal of Radiation Oncology  Biology  Physics

dose. Binary logistic regression evaluated the correlation between radiation dose and pathologic outcome with age >65, chemotherapy type and duration as covariates. The logistic regression coefficients were used to create a linear predictor function using PTV dose to predict node-negative resection, with neoadjuvant chemotherapy type and duration as model variables. Results: From 2010-2015, 83 patients treated with neoadjuvant SBRT followed by surgical resection and 78 patients with complete dosimetry data were included. In this patient cohort 67% were younger than 65, 50% received 4 months of neoadjuvant chemotherapy, and 58% of patients received FFX-based chemotherapy. Neoadjuvant pancreas SBRT was prescribed to a dose of 25-33 Gy in 5 fractions. PTV V20 of >99.5% was associated with improved PFS (17.3 mo vs. 11.9 mo, P Z 0.013, HR: 0.328, 95% CI: 0.136-0.801) and OS (23.6 mo vs. 12.6 mo, P Z 0.007, HR: 0.204, 95% CI: 0.064-0.649). PTV V20 Gy was also correlated with increased rates of node-negative resection (P Z 0.018, OR: 8.361, 95% CI: 1.42-49.1), trended with increased rates of margin-negative resection (P Z 0.056, OR 0.238: 95% CI: 0.052-1.076), and near pathologic complete response (P Z 0.068, OR 5.31: 95% CI 0.883-32.0). Dose to 99% of the PTV (D99) was found to correlate with node-negative resection (P Z 0.0029, OR: 1.20, 95% CI: 1.02-1.40). PTV D99 of 20 Gy predicted for a 41% rate of node-negative resection and PTV D99 of 40 Gy predicts for a 97% node-negative resection rate in patients who receive at least 4 mo of FFX-based chemotherapy. Conclusion: Neoadjuvant PDA SBRT patients who received a PTV V20 Gy of >99.5% were found to have improved OS, PFS, node-negative resection, and trended towards improved rates of margin-negative resection and near pathologic complete response. Further dose escalation predicts for improved node-negative resection. PTV V20 coverage and D99 dose escalation are potential dosimetric parameters that may improve patient outcome. Author Disclosure: L. Chen: None. Y. Cao: None. A. Narang: None. Z. Cheng: None. L.M. Rosati: None. O.Y. Mian: None. S.P. Robertson: None. T.R. McNutt: None. A. Hacker-Prietz, PA-C: None. J.M. Herman: Consultant; Merrimack, Oncosil. Research Grant; Nucletron.

proportional hazards modeling. Estimates for DFS and OS were calculated using Kaplan-Meier methods. Results: The pancreatic, anterior and bile duct SRMs were not significant predictors of DFS and OS irrespective of distance. However, increasing posterior margin clearance up to 2 mm was an independent significant predictor of DFS (P Z 0.001) and OS (P Z 0.001). Margin clearance greater than 2 mm did not significantly impact DFS (P Z 0.08) or OS (P Z 0.08). Dichotomizing the posterior SRM at 2 mm revealed it to be an independent predictor of OS (HR 0.31; P Z 0.008) and DFS (HR 0.46; P Z 0.05) with median OS of 23.2 vs. 60 months and median DFS of 13.9 vs. 27.3 months for < 2mm vs. >/Z 2 mm, respectively. After adjusting for covariates, posterior SRM >/Z 2 mm was a significant predictor of OS in patients who received adjuvant chemotherapy (HR 0.31; P Z 0.03), yet this difference was mitigated in patients receiving adjuvant chemoradiation therapy (HR 0.40; P Z 0.19). Conclusion: Among all SRMs in PDA, the posterior margin is the most clinically significant. Achieving SRM of at least 2 mm has a significant impact on clinical outcomes. The addition of radiation therapy to adjuvant chemotherapy mitigates the negative outcome associated with a posterior SRM of less than 2 mm. Comprehensive archival validation in a larger data set may help redefine what constitutes R1 resection in PDA, as well as the role of adjuvant radiation therapy. Author Disclosure: R. Tuli: None. A. Osipov: None. J.K. Rutgers: None. D. Dhall: None. J. Naziri: None. S. Chopra: None. Q. Li: None. A.E. Hendifar: None. N.N. Nissen: None.

2523 Margin Distance in Resected Pancreatic Cancer Independently Predicts Survival: Implications for Adjuvant Radiation Therapy R. Tuli, A. Osipov, J.K. Rutgers, D. Dhall, J. Naziri, S. Chopra, Q. Li, A.E. Hendifar, and N.N. Nissen; Cedars-Sinai Medical Center, Los Angeles, CA Purpose/Objective(s): Controversy exists regarding prognostic significance of margin clearance and what constitutes a positive margin (R1) in pancreatic ductal adenocarcinoma (PDA). We performed a comprehensive archival analysis of all surgical resection margins (SRM) to determine the effect on clinical outcomes in PDA. Materials/Methods: With IRB approval, we identified 105 patients with resected stage I-III PDA from 2007-2014. SRMs were determined for pancreatic, anterior, bile duct and posterior margins (posterior surface, uncinate and vascular groove). Three pathologists reviewed all archival surgical specimens and recategorized each margin as: tumor at ink, <0.5 mm, 0.5-1 mm, >1-2 mm or >2 mm from the inked surface. The significance of SRM and other clinical variables was assessed on disease-free survival (DFS) and overall survival (OS) using multivariate Cox

2524 Using Pelvic Bone Marrow Hounsfield Units to Predict Cytopenia During Anal Cancer Chemoradiation A.Y. Lee, S. Thomas, and S. Liauw; University of Chicago, Chicago, IL Purpose/Objective(s): Cytopenias are common during chemoradiation for anal cancer, yet predictors of hematologic toxicity are lacking. Differences in marrow activity might be distinguished by measuring Hounsfield units (HU) on CT. We evaluated whether pre-treatment bone marrow HU correlate with blood count nadirs during chemoradiation. Materials/Methods: Nineteen patients with stage I-III squamous cell carcinoma of the anal canal were treated with definitive chemoradiation at a single academic institution from 2009 to 2013. Chemotherapy consisted of concurrent 5-FU and mitomycin C (days 1, 28). Median RT dose was 54 Gy to the primary tumor, 45 Gy to the true pelvis, and 30.6 Gy to the whole pelvis, by IMRT (n Z 12) or 3D planning (n Z 7). Pelvic bone marrow HU were measured on simulation CT scans by a study-blinded body radiologist using a 1 cm diameter spherical region of interest. Based on prior PET studies of marrow activity, HU were measured at the bilateral superior acetabula, femoral necks, first sacral alae, and fifth lumbar vertebral body. HU were correlated with nadirs for white blood cells (WBC), neutrophils (ANC), hemoglobin, and platelets. Raw HU from significantly associated locations were normalized by standard deviation and averaged to create an “HU score.” Results: Median nadirs in WBC, ANC, and platelet counts occurred in week 3 of treatment, and week 6 for hemoglobin. Grade 3 nadirs were seen in nine patients for WBC, ten patients for ANC, one patient for hemoglobin, and one patient for platelets. Lower WBC and ANC nadirs correlated with higher HU for the superior acetabula and femoral necks, but not the sacral alae or lumbar vertebrae. No significant associations

Abstract 2524; Table 1.

Mean HU (Std dev.) Linear Regression (P-values) WBC ANC Hgb Platelets

L Sup. Acet.

R Sup. Acet.

L Fem. Neck

R Fem. Neck

L Sacral Ala

R Sacral Ala

195.7 (72.5)

184.9 (85.5)

76.6 (74.8)

81.1 (71.8)

26.2 (59.1)

30.0 (64.3)

0.046 0.050 0.258 0.272

0.045 0.059 0.312 0.197

0.084 0.072 0.411 0.120

0.059 0.057 0.434 0.264

0.966 0.954 0.725 0.864

0.769 0.788 0.539 0.625

L5

Normalized HU Score

151.8 (52.3)

N/A

0.324 0.322 0.956 0.434

0.023 0.025 N/A N/A

Volume 96  Number 2S  Supplement 2016 were observed between HU and hemoglobin or platelet nadirs. The HU score demonstrated strong correlation with WBC nadir (P Z 0.023) and ANC nadir (0.025). Patients with a high HU score (> median) had significantly higher rates of grade 3 WBC nadirs (77.8% vs 20.0%, P Z 0.019) and grade 3 ANC nadirs (88.9% vs 20%, P Z 0.005). On multivariate analysis, HU score had a stronger association with ANC nadir (P Z 0.084) and WBC nadir (P Z 0.157) than age, body mass index, or gender (all P > 0.25). Conclusion: Higher marrow HU in the superior acetabula and femoral necks are associated with more severe WBC and ANC nadirs during treatment. We developed a “HU score” that correlates with grade 3 WBC and ANC nadirs. Further validation studies are warranted to determine if and why HU correspond to differential sensitivities to chemoradiation, or if they merely reflect baseline differences such as osteopenia. Author Disclosure: A.Y. Lee: None. S. Thomas: None. S. Liauw: None.

2525 Contouring of Pancreatic Tumor Volume Is Highly Variable on Interobserver Analysis in the Planning of Stereotactic Body Radiation Therapy P.K. Stumpf,1 B. Jones,2 A. Amini,1 S. Chang,3 B. Edil,4 C. Gajdos,4 K.A. Goodman,2 M. McCarter,4 K. McKinney,3 J. Meier,3 S. Pokharel,3 R. Schulick,4 M. Wagh,5 S. Wani,5 and T. Schefter2; 1Department of Radiation Oncology, University of Colorado Denver, Aurora, CO, 2 Department of Radiation Oncology, University of Colorado School of Medicine, Aurora, CO, 3Department of Radiology, University of Colorado, Aurora, CO, 4Department of Surgical Oncology, University of Colorado, Aurora, CO, 5Department of Gastroenterology, University of Colorado, Aurora, CO Purpose/Objective(s): In pancreatic stereotactic body radiation therapy (SBRT), contouring the target volume (TV) can be challenging. Accurate target delineation is necessary for precise treatment delivery, as margins are typically small to allow for dose-escalation inherent to SBRT. It has been shown that size of TV is correlated with increased toxicity (Hoyer, Radiother Oncol 2005, 76: 48-53). Participating clinicians at our institution were asked to contour pancreatic tumor volume to assess observer variability. Materials/Methods: From the group of patients with adenocarcinoma of the pancreas treated with SBRT at our institution in 2014 and 2015, a total of 10 patients were selected who met the following criteria: contrasted CT scan acquired at simulation, presence of fiducials markers, and pathology available from post-SBRT resection. Patients were anonymized, and 11 physician observers from multiple subspecialties including radiation oncology, radiology, interventional gastroenterology, and surgical oncology, all with expertise in pancreatic cancer, were instructed to delineate the pancreatic gross tumor volume (GTV). A consensus GTV was defined for each patient as the union of all voxels that were included by at least 50% of the observers. The sensitivity and positive predictive value (PPV) of each observation was calculated by comparing against this gold standard. Results: The median consensus volume for all ten tumors was 19 cm3  16 cm3. There were significant differences in contours between observers, resulting in large inter-observer variability. Across all observers, the median volume difference as compared to the consensus volume was 15 cm3. The average sensitivity was 72%  20% with a range of 27-97%. The average PPV was 60%  20% with a range of 18-84%. Conclusion: Pancreatic SBRT is a complex, technical procedure in which precise, accurate GTV definition is imperative for safe and effective treatment delivery. The results of this study reveal wide variation in interobserver contours. This verification of the necessity for consensus contouring guidelines is the initial step in our endeavor to assist in standardization of pancreatic tumor delineation for SBRT. Author Disclosure: P.K. Stumpf: None. B. Jones: Research Grant; Varian Medical Systems. A. Amini: None. S. Chang: None. B. Edil: None. C. Gajdos: None. K.A. Goodman: None. M. McCarter: None. K.

Poster Viewing E215 McKinney: None. J. Meier: None. S. Pokharel: Scientific Advisory Board Member; 3D Biopsy, LLC. R. Schulick: None. M. Wagh: None. S. Wani: None. T. Schefter: None.

2526 Radiation Dose Effect in Adjuvant Chemoradiation for Locally Advanced Gallbladder Cancer and Extrahepatic Cholangiocarcinomas R. Beck, M.P. Deek, and S.K. Jabbour; Rutgers Cancer Institute of New Jersey, Department of Radiation Oncology, New Brunswick, NJ Purpose/Objective(s): Gallbladder cancer (GBC) and cholangiocarcinomas are rare cancers with low incidence and short survival. The NCCN recommends surgery as primary treatment when possible. A recent phase II trial of radical surgery followed by chemoradiation shows promising outcomes for patients with locally advanced GBC and extrahepatic cholangiocarcinomas (EHCC). Our study aimed to evaluate whether the dose of radiation in adjuvant chemoradiation was an independent predictor of survival in this patient population. Materials/Methods: We used the National Cancer Data Base (NCDB) data for patients diagnosed between 2004 and 2013 with T2-4 or node positive GBC or EHCC. We included only patients who received a total surgical removal of the primary site followed by adjuvant chemotherapy and external beam radiation. Patients were excluded if they had metastatic disease or received neoadjuvant treatment. Dose of radiation was divided into three groups consisting of 4500 cGy or less, 4501 to 5399 cGy, and 5400 cGy or greater. Survival was also compared using multiagent or single-agent chemotherapy. Kaplan-Meier curves were generated to compare survival among the different groups using the log-rank test. Multivariate Cox proportional hazard models were performed including variables significant on univariate analysis. Results: Six hundred four patients met our inclusion criteria, 347 (57.5%) with GBC and 257 (42.5%) with EHCC. There was no significant difference in demographic or tumor characteristics between the groups on chi squared analysis. Overall, radiation dose had no significant impact on median survival in the whole population (33.2 vs. 33.5 months; P Z 0.78). Dose of greater than 4500 cGy demonstrated improved survival (39.5 vs. 32.6 months; P Z 0.034) among patients with EHCC. No benefit was seen for doses 5400 cGy compared to 4501-5399 cGy (P Z 0.51). Conversely, there was decreased survival for doses greater than 4500 cGy among patients with GBC (29.5 vs. 34.6 months; P Z 0.02). Single-agent chemotherapy demonstrated trend toward improved survival compared to multiagent chemotherapy (35.7 vs. 31.4 months, P Z 0.08). Conclusion: Our analysis indicates that for patients in the United States with clinical T2-4 or node positive gallbladder cancer or EHCC who receive adjuvant chemoradiation, there is no dose-response relationship between radiation dose and survival. There was improved survival among patients with EHCC who received doses greater than 4500 cGy compared to those that received 4500 cGy or less, and worse survival among gallbladder cancer patients receiving doses over 4500 cGy. Single-agent chemotherapy demonstrated improved survival compared to multiagent chemotherapy. This data suggests that doses of greater than 4500 cGy should be used for adjuvant chemoradiation in EHCC, though doses of greater than 4500 cGy may not be beneficial in gallbladder cancer. Prospective studies are needed to validate these findings. Author Disclosure: R. Beck: None. M.P. Deek: None. S.K. Jabbour: None.

2527 Dosimetric Predictors of Local Control After Stereotactic Body Radiation Therapy of Locally Advanced Pancreatic Ductal Adenocarcinoma C. Chin,1 S. Perni,2 T.K. Yanagihara,3 E.P. Xanthopoulos,1 P. Yan,1 and D.P. Horowitz1; 1Department of Radiation Oncology, Columbia