International Journal of Radiation Oncology Biology Physics
expression of SMAD4 characterized patient derived xenografts (PDX), and identify profiles that may predict response to radiation. Materials/Methods: With IACUC approval, 20 PDAC PDXs characterized for KRAS, p53 and SMAD4 were implanted into bilateral flanks of mice and irradiated. Left flank tumors received 5 Gy/day for 2 days using the right flank as a control (n Z 5 per group). Tumor dimensions were obtained twice weekly (volume Z (LxW2)/2); tumor growth inhibition (TGI) was calculated (treated tumor volume/control tumor volume). Gene expression data (Affymetrix GeneChip) was preprocessed with quantile normalization and log2 transformation. Methylation microarrays (Illumina 450k) were processed with quantile normalization, beta-value scale, and M-value. Key pathways of differentially expressed genes were investigated with KEGG pathway enrichment analysis. Results: Comparison of SMAD4 mutant (n Z 10) and intact (n Z 10) tumors revealed 621 and 602 differentially expressed genes following gene expression and methylation array analysis, respectively. Prominent KEGG pathways included endocytosis, melanogenesis, tight junction, ascorbate and aldarate metabolism, steroid hormone biosynthesis, and androgen and estrogen metabolism. Three genes with the strongest inverse correlation between levels of methylation and mRNA expression included ACOX2 (r Z -.56, p Z .01), CTAGE5 (r Z -.7, p Z .0004), and LY75 (r Z .61, p Z .004). Gene expression and methylation analysis as a function of relative TGI revealed 301 and 447 differentially expressed genes, respectively, and these were enriched in the pathways in cancer, melanoma, and aldosterone-regulated sodium reabsorption pathways. Irradiated tumors revealed a varying range of TGI (-58% to +72%). Using univariate linear regression to correlate TGI with gene expression and methylation revealed 5 target genes (C6orf125, CFD, DNAI1, GLI3, and TRIM24) with strong inverse correlation (r Z .53-.61, p<.02). Conclusions: Correlation of gene expression with methylation profiles of SMAD4-characterized PDAC revealed gene signatures suggestive of metabolic and hormonal dysregulation, which may contribute to the aggressive nature of SMAD4-mutant tumors. Further investigation is underway to identify potential molecular drug targets, and a signature to predict response to RT and patterns of failure. Author Disclosure: R. Tuli: None. Z. Liu: None. R. Nv: None. A.J. Surmak: None. C. Iacobuzio-Donahue: None. M. Hidalgo: None. J.M. Herman: None. S.H. Lin: None.
resectable tumors groups. The resectable tumor group showed higher regression rate mean value from pre-therapy to early-therapy and pretherapy to post-therapy (46.718.7%, 67.717%) respectively than the unresectable tumor group (23.233.4, 35.349.1) respectively. The resectable tumor group Amp mean values: 2.70.6, 31.2, 2.70.8 (pretherapy, mid-therapy, and early-therapy) respectively were higher than the unresectable group Amp: 1.80.6, 2.60.4, 2.41.2, The resectable tumor group Kep mean values were 3.51.5, 3.22.1, 2.20.7 min-1(pre-therapy, mid-therapy, and early-therapy); only mid-therapy Kep was higher than the unresectable group: 5.42.5, 2.10.1, 4.42.4 min-1. The resectable tumor group Kel mean values: 0.060.03, 0.040.02, 0.050.03 min-1(pre-therapy, mid-therapy, and early-therapy) were higher than the unresectable group Kel: 0.010.03, 0.020.01, 0.010.02 min-1. Resectable tumor group tumor regression rate from pre-therapy to midtherapy is positively correlated to mid-therapy Amp (r Z 0.76), and negatively correlated to mid-therapy Kel (r Z -0.73). In addition, tumor regression rate from pre-therapy to post-therapy is negatively correlated to post-therapy Amp (r Z -0.79), Kep (r Z -0.87), and kel (r Z -0.61). No correlation was found between tumor regression rate and DCE parameter for the unresectable group. Conclusions: PNT responsiveness and tumor resectability might be associated with lower kep, high Amp and Kel, concurrent with higher tumor regression rate; these parameter correlations are essential at each time point of the treatment to further improve long-term outcome prediction. Acknowledgment: This research is funded by NCI: R21-CA121582 & P30CA16058. Author Disclosure: J.C. Grecula: E. Research Grant; Funded by NCI: R21-CA121582 & P30CA16058. S. Elias: None. K. Thelen: None. M. Knopp: None. G. Otterson: None. P. Ross: None. E. Kassis: None. M. Welliver: None. M. Villalona-Calero: None. K. Shilo: None. S. Lo: None. G. Jia: None. W.C. Yu: None. B. Yuh: None. S. Ghosh: None. E. Bertino: None. N. Mayr: None.
3545 Dynamic-Contrast Enhanced MR and Volume Regression Rate as a Preoperative Predictive Assay in Patients With Non-Small Cell Lung Cancer J.C. Grecula,1 S. Elias,1 K. Thelen,2 M. Knopp,1 G. Otterson,1 P. Ross,1 E. Kassis,1 M. Welliver,1 M. Villalona-Calero,1 K. Shilo,1 S. Lo,3 G. Jia,4 W.C. Yu,5 B. Yuh,5 S. Ghosh,1 E. Bertino,1 and N. Mayr5; 1Ohio State University, Columbus, OH, 2Saint Mary’s University, Winona, MN, 3UH Seidman Cancer Center, Cleveland, OH, 4Louisiana State University, Louisiana, LA, 5University of Washington, Seattle, WA Purpose/Objective(s): To investigate dynamic contrast enhanced (DCE)MR pharmacokinetic parameters: Amplitude of contrast enhancement (Amp), rate constant (Kep) and elimination rate (Kel), and T2 weighted (T2W)-MR volume regression rate for assessment of preoperative neoadjuvant chemoradiation therapy (PNT) responsiveness in lung cancer patients and resectability successfulness. Materials/Methods: 8 stage IIIA non-small cell lung cancer patients (5 females; 3 males; median age 60 years; range 47-73 years) were treated with 45 Gy conformal radiation therapy in 25 fractions with 2 courses of concurrent cisplatin (50 mg/m2; Days1, 8, 29,36) and etoposide (50 mg/m2 Days 1-5 & 29-33). 3 multi-parametric 3T MRI (pre-, early-and posttherapy) were performed in all patients. DCE-Amp, Kep and Kel and 3dimensional tumor region of interest were obtained using commercially available software. Results: Multi-parametric MRI of lung cancer showed gradual reduction of 3D tumor volume on T2W MRI and appreciated changes in the DCE Amp, Kep and Kel; all appeared to differentiate unresectable tumors from
3546 Circulating Tumor DNA as a Biomarker for Pancreatic Adenocarcinoma E. Osmundson, A.M. Newman, S.V. Bratman, D.M. Klass, L. Zhou, J. Pai, T.A. Longacre, A.A. Alizadeh, A.C. Koong, and M. Diehn; Stanford University, Stanford, CA Purpose/Objective(s): Reliable identification of patients likely to benefit from primary or adjuvant local therapy for pancreatic adenocarcinoma (PAC) remains elusive. This is in part due to the prevalence of radiographically occult micrometastases, and limitations of available biomarkers to detect residual disease. Therefore, novel biomarkers for PAC that reliably assess for the presence of micrometastatic or residual disease are needed. We recently developed Cancer Personalized Profiling by Deep Sequencing (CAPPSeq), a novel next-generation sequencing based technique that allows for the ultrasensitive and specific detection of circulating tumor DNA (ctDNA) in plasma. The objective of this study was to determine the feasibility of applying CAPP-Seq for quantitation of ctDNA in patients undergoing primary and adjuvant therapy for PAC. Materials/Methods: CAPP-Seq employs a hybrid capture technique, using a “selector” that is tailored to the cancer type of interest. To design a PACspecific CAPP-Seq selector we used a custom bioinformatics approach to identify recurrently mutated genomic regions in PAC using publically available whole exome sequencing data. Biotinylated DNA oligonucleotides targeting these mutated regions were synthesized. Next, we tested the feasibility of extracting sufficient genomic DNA for tumor genotyping (criterion: 4ng) from PAC formalin fixed paraffin embedded (FFPE) specimens collected via fine needle aspirates (FNAs), core biopsies, or surgical resection. CAPP-Seq was then used to identify tumor-specific mutations in these PACs. Finally, ctDNA levels were quantitated from preand post-treatment blood samples prospectively collected from patients undergoing therapy for PAC at our institution using a customized bioinformatics pipeline.
Volume 90 Number 1S Supplement 2014 Results: The PAC-specific CAPP-Seq selector covered w135 kb and targeted 979 genomic regions from 925 recurrently mutated genes. Although comprising <0.005% of the human genome, the selector identified a median of 11 mutations per tumor and was able to identify multiple mutations in >96% of PACs. Using an optimized protocol, we extracted adequate tumor genomic DNA for CAPP-Seq genotyping from 95% of FFPE samples, including 13 of 13 FNA/core samples (mean Z 178 ng; range Z 4-1075 ng) and 9 of 10 surgical samples (mean Z 680 ng; range Z 3-2351 ng). Finally, we used CAPP-Seq to identify tumor-specific mutations and to quantitate ctDNA in pre- and post-treatment plasma samples. Conclusions: CAPP-Seq is a promising method for the ultrasensitive and specific quantification of ctDNA in patients with PAC. Isolation of tumor DNA from PAC FFPE FNA/core biopsy specimens using our optimized protocol provides sufficient tumor DNA for genotyping using CAPP-Seq. Ongoing analyses are exploring the prognostic and predictive utility of ctDNA analysis in PAC. Author Disclosure: E. Osmundson: None. A.M. Newman: None. S.V. Bratman: None. D.M. Klass: None. L. Zhou: None. J. Pai: None. T.A. Longacre: None. A.A. Alizadeh: None. A.C. Koong: None. M. Diehn: None.
3547 Blinded Evaluation of Sinogram Affirmed Iterative Reconstruction in Radiation Therapy Planning Images N. Nguyen,1 G. Charron,2 and D. Roberge2; 1Juravinski Cancer CentreMcMaster University, Hamilton, ON, Canada, 2Universite´ de Montre´alCentre Hospitalier de l’Universite´ de Montre´al, Montreal, QC, Canada Purpose/Objective(s): The current standard reconstruction algorithm for CT scans is filtered back projection (FBP). Recently, an alternative algorithm of iterative reconstruction (IR) has been increasingly implemented in the realm of diagnostic CT imaging, allowing decoupling of spatial resolution and image noise while reducing radiation doses. We studied a variant of IR, known as “Sinogram affirmed iterative reconstruction” (SAFIRE) and assessed its potential to improve the quality of radiation therapy (RT) planning images. Materials/Methods: Raw CT data sets of patients planned for brain, spine, head and neck, lung, breast, gastrointestinal, liver, gynecological, genitourinary and limb tumors were included. A total of 50 scans were acquired using our standard imaging protocols and reconstructed using FBP and SAFIRE levels 1, 3 and 5 (higher levels referring to more noise correction). For each disease site, two to seven scans were selected. For each site, 2 to 3 specialized radiation oncologists evaluated the 3D image sets in a blind fashion. Using a visual analogue scale, they assessed the image sharpness, noise, perceived ease in delineating gross tumor/clinical target volume (GTV/CTV) and organs at risk (OAR) and overall appreciation of the planning images. Inter-observer correlation was calculated with the Spearman correlation coefficient (r). Generalized estimating equations (GEA) assessed the differences in the mean score for each criteria, between reconstructions, adjusted for observer status. When there were significant differences, pairwise comparisons (PC) were done to compare the least-squares means. The preference for each scan was rank ordered for each observer. The mean rank across all observers and the ranks occurrences were computed. GEA was calculated again for the mean ranks; PC were done when differences existed. Results: The sharpness of borders had r Z -0.22-0.53, the ease of GTV/ CTV delineation r Z -0.28-0.53, the ease of OAR delineation r Z -0.470.42, the image noise r Z -0.34-0.38 and the overall appreciation r Z -0.17-0.38. Although there were discrepancies between physicians, SAFIRE levels 3 and 5 had consistently higher scores than FBP scans and were the highest rated scans for all criteria and for all disease sites (p Z 0.02 and p Z 0.015, respectively). Paradoxically, although SAFIRE level 5 scored well on average, it was ranked as worst the most often. SAFIRE level 3 was consistently well ranked for all criteria. Conclusions: This report is the first to report the potential benefit of IR in RT planning scans. Although highly processed images polarized observers, the use of IR is globally preferred over standard FBP for planning CT
Poster Viewing Abstracts S817 scans. The preference for IR is seen for all disease sites, for all facets of images. This work will lead to clinical implementation of intermediate IR processing and further study investigating quantitatively IR’s impact on contouring. Author Disclosure: N. Nguyen: None. G. Charron: None. D. Roberge: None.
3548 Evaluation of an Integrated Minimally Interactive Tool for the Segmentation of Relevant Oar in Lung Cancer T. Fechter,1 J. Dolz,2 H. Kirisli,2 A. Chirindel,1 S. Adebahr,1 T. SchimekJasch,1 M. Vermandel,3 L. Massoptier,2 and U. Nestle1; 1 Universitaetsklinik Freiburg, Freiburg im Breisgau, Germany, 2 AQUILAB, Lille, France, 3Centre Hospitalier Universitaire, Lille, France Purpose/Objective(s): Radiation therapy aims at delivering the highest possible dose to the GTV while minimizing the irradiation of surrounding healthy tissue. Therefore an accurate delineation of organs at risk (OAR) is an absolute prerequirement for radiation treatment planning (RTP). However, it is maybe a very time consuming and tedious task. In this work we present and evaluate a software prototype which integrates fully automatic and minimally interactive algorithms to perform segmentation for relevant OAR in lung cancer. Materials/Methods: The software prototype was implemented within MITK and the body contour, lung, heart, spinal cord, trachea and central bronchi were delineated. Lung, heart and spinal cord were segmented interactively using seed points provided by the user. Between five to seven seed points were needed to obtain satisfactory results. The remainder of the OAR was segmented fully automatically. Quantitative evaluation was performed for lung, heart, spinal cord and body contour. The Dice’s coefficient (DC) were computed between the contours generated with our software and the reference standard contoured by a clinical expert for a real RTP. For trachea and central bronchi a qualitative visual evaluation was performed by a clinical expert as no reference standard was available for these OAR. The contour analysis was carried out on planning CT scans for 12 patients treated with SBRT for locally advanced lung tumors. Results: Mean DC obtained was higher than 90% for lungs, heart and body, with a maximum standard deviation (SD) of 2.2%. For spinal cord, the mean DC decreased to 79.7%, with a SD of 10.7%. The results demonstrate strong agreement between segmentation provided by the proposed software prototype and reference standard segmentations used for RTP. Qualitative visual analysis suggests that the proposed method for trachea and central bronchi segmentation could be a surrogate for the manual contours used in a treatment planning. Conclusions: The proposed minimally interactive software prototype offers a fast way to contour and demonstrates good agreement to standard expert delineated contours in the majority of analyzed OAR. For the spinal cord a relatively decreased concordance was noted, which was not surprising, as the clinical expert used different anatomic landmarks (vertebral canal) in order to overcome the limited visualization of the spinal cord on CT. Changing the prototype in a way to follow the bony limits of the spinal canal will increase concordance. An extended user testing will explore the possible use of this tool in clinical practice. Author Disclosure: T. Fechter: None. J. Dolz: None. H. Kirisli: None. A. Chirindel: None. S. Adebahr: None. T. Schimek-Jasch: None. M. Vermandel: None. L. Massoptier: None. U. Nestle: None.
3549 An Electronic Medical Record Search Engine for Clinical Outcome Research A. Gopal, J.J. Gordon, J. Kim, and I.J. Chetty; Henry Ford Health System, Detroit, MI Purpose/Objective(s): Among common first steps in a study of treatment response is a search of the radiation therapy EMR for patient records meeting study criteria, e.g. “Lung tumors treated with SBRT over the last 5