CT Blood Perfusion Imaging Parameters in Thoracic Malignancy Patients

S628

International Journal of Radiation Oncology  Biology  Physics

of radiation pneumonitis (RP); 2) Comparison of dosimetric parameters (V5- 30, and fV5-30: the percentage of total and functional lung volume irradiated with >5- 30 Gy or Mean lung dose (MLD), fMLD: mean dose administered to total or functional lung and the incidence of RP. RP was evaluated using the CTCAE ver 4.0. Statistical significance was defined as p < 0.05. Results: Grade 1, 2, and  3 RP was observed in 13, 32, and 6 patents, respectively during a median 12 month follow-up. The mean percentages of LAA-860 and -910 in patients with grade  3 RP were significantly lower than those of grade 1or 2 RP (LAA-860, 48.0% vs 66.7%, p Z 0.024, LAA-910, 75.2% vs 91.4%, p Z 0.0032). There were no significant differences in F10-70. There were significant differences of univariate analysis assessing the incidence of grade  2 RP and V5- 30 (fV5- 30) or MLD (fMLD) in total lung and each functional lung as follows; in total lung, V20 > 30% (p Z 0.039), V30 > 20% (p Z 0.0111), and MLD > 15 Gy (p Z 0.0141), and in F10, fV30 > 20% (p Z 0.0422), in F30, fV30 > 20% (p Z 0.0305), fMLD > 15 Gy (p Z 0.0498), in LAA-860, fV30 > 20% (p Z 0.0266), fMLD > 15 Gy (p Z 0.0223), in LAA-910, fV30 > 20% (p Z 0.0107), fMLD > 15 Gy (p Z 0.0183), respectively. Conclusions: Functional image-guided radiation therapy planning appears to predict risk factors for RP and may be effective in preserving functional lungs to reduce the incidence of severe RP in future. Author Disclosure: T. Kimura: None. T. Nakashima: None. Y. Doi: None. N. Imano: None. T. Katsuta: None. T. Okabe: None. M. Kenjo: None. S. Ozawa: None. Y. Murakami: None. Y. Nagata: None.

toxicity and excellent rates of local, regional, and distant control. Robotic SBRT using real-time fiducial tracking appears to be a safe and effective treatment for patients who are not optimal candidates for surgery due to age or comorbidities such as severe COPD. Author Disclosure: J.M. Samuelian: None.

3059 Curative Treatment of Stage I Non-Small Cell Lung Cancer Using Robotic Stereotactic Radiation in Patients With COPD: A VA Population Study J.M. Samuelian; Baylor College of Medicine, Houston, TX Purpose/Objective(s): Patients with chronic obstructive pulmonary disease (COPD) have a high risk of lung cancer and postsurgical complications. Stereotactic body radiation therapy (SBRT) has been utilized in patients with early non-small-cell lung cancer (NSCLC) deemed medically inoperable with good results. In this study, we report our local control, survival and toxicity results after robotic SBRT in patients with stage I NSCLC. Materials/Methods: A retrospective analysis of the first 50 patients with stage I NSCLC treated between January 2011 and September 2012 was evaluated. Median age was 70 (range, 54-90). 88% of the cohort had known COPD prior to treatment, with 34% GOLD stage III; 68% admitted to continued tobacco use at time of consultation, and 30% required continuous home oxygen supplementation. Workup included CTscan, PET/CT, tissue biopsy confirming NSCLC, and fiducial placement in all patients. 62% of patients also underwent endobronchial ultrasound (EBUS) with lymph node sampling to confirm the absence nodal metastasis. Median gross tumor volume (GTV) >was 5.9 cc (range, 0.5-98 cc), and a uniform 5 mm expansion was used to create a planning target volume (PTV). SBRT was delivered using a robotic delivery system with a respiratory tracking system for real time in-treatment fiducial tracking. Median prescribed dose was 56 Gy in 4 fractions (range, 42-70 Gray in 4-5 fractions) to the PTV. Kaplan Meier and Cox proportional hazard ratio models were used for locoregional control and survival analysis. Tumor response was based on post-treatment serial PET/CT scans. Toxicity was graded according to CTCAE v4.0. Results: Median follow-up was 23 months. The actuarial 2-year local control (LC), regional failure free survival, distant metastasis free survival (DMFS), and overall survival (OS) were respectively 95%, 96%, 92% and 81%. Local relapse was observed in 2 (4%) patients, regional relapse in 3 (6%) patients, and distant relapse in 6 (12%) patients. Treatment was very well tolerated with no grade  4 toxicities, and only one grade 3 toxicity (pneumonitis). Toxicities otherwise included 2 patients (4%) with grade 2 pneumonitis, 2 patients (4%) with grade 2 costal tenderness, and one patient (2%) with rib fracture. Conclusions: Robotic SBRT in a veteran population with stage I NSCLC proved tolerable despite significant patient comorbidities, with very little

3060 Temporal Variation in Total Lung Volume and Lung V20 in Lung Cancer Patients Treated With Radiation Therapy D. Bergsma, P. Youn, S. Dhakal, and M.T. Milano; University of Rochester, Rochester, NY Purpose/Objective(s): Dose-volume histogram (DVH) parameters, such as lung V20, play an important role in radiation treatment planning for thoracic malignancies including pulmonary toxicity risk prediction. The lung V20 is defined as the percentage of normal lung receiving at least 20 Gy and is dependent on the total lung volume (TLV). A lung V20 of 3035% represents the general threshold of acceptable normal lung dose. TLV may vary based on inspiratory effort which is dependent on a variety of host factors. This retrospective study was designed to assess intrapersonal variation in TLV, as measured by CT, and the effect this might have on the calculated lung V20 in patients undergoing radiation therapy for treatment of lung cancer. Materials/Methods: The analysis included 62 patients from a single institution who were treated with curative intent for locally advanced NSCLC or limited stage SCLC between December 2010 and May 2013. All patients had a staging PET-CT and treatment planning CT (TPCT) which were both obtained during free breathing. Patients were treated with 3D conformal radiation therapy (n Z 55) or IMRT (n Z 7) using standard fractionation. Lung contours were generated in accordance with RTOG structure delineation guidelines. Normal lung volume was defined as TLV minus GTV in the vast majority of cases. The staging PET-CT was registered with the TPCT and contoured applying the aforementioned methodology. TLV and lung V20 were calculated for both the PET-CT and TPCT. Results: The range of absolute difference in TLV (given as volume and percentage) between individual patients’ staging PET-CT and TPCT was 23-1605 cc (0.5-30.0%) with median of 271 cc (7.1%) [p Z 0.091, 2-sided paired t-test]. The 25th, 75th and 90th percentiles of absolute difference in calculated TLV were 3.8%, 13.3% and 18.1% respectively. The median lung V20 was 30.0% for PET-CTs and 30.5% for TPCTs. The range of absolute difference in lung V20 was 0.1%-13.7% with median of 2.0% (p Z 0.258, 2-sided paired t-test). The 25th, 75th and 90th percentiles of absolute difference in lung V20 were 1.0%, 3.9% and 5.1% respectively. Conclusions: There were measurable differences in the TLV and lung V20 between staging and planning CTs for individual patients. Although the majority of patients exhibited minor changes in TLV and lung V20, which were not statistically significant as a group, it is noteworthy that 10% of patients had variation in TLV and lung V20 greater than 18% and 5% respectively. As the study median lung V20 approached the general threshold of acceptability, this variation may be of significance in regards to radiation treatment planning. Further study regarding clinical characteristics associated with greater variation and whether they affect the risk of pulmonary radiation toxicity are necessary. Author Disclosure: D. Bergsma: None. P. Youn: None. S. Dhakal: None. M.T. Milano: None.

3061 Negative Correlation Between Quantitative FDG PET/CT and MAA SPECT/CT Blood Perfusion Imaging Parameters in Thoracic Malignancy Patients S.R. Bowen, M.J. Nyflot, P.E. Kinahan, G.A. Sandison, R. Miyaoka, S.A. Patel, H.J. Vesselle, J. Zeng, and R. Rengan; University of Washington, Seattle, WA Purpose/Objective(s): Increased 18F-FDG PET/CT uptake has been shown to correlate with aggressive tumor phenotype and poor clinical

Volume 90  Number 1S  Supplement 2014 outcome in both esophageal and non-small cell lung cancer patients. Similarly, increased tumor perfusion has been suggested to inversely correlate with aggressive tumor biology. However, it remains unclear if tumor perfusion truly reflects tumor biology or simple hemodynamics. We utilized 99mTc-MAA SPECT/CT scans to calculate tumor perfusion in thoracic malignancy patients and performed a comparative analysis with tumor FDG uptake. Materials/Methods: Nine patients with histopathologically confirmed thoracic neoplasms (7 non-small cell lung cancer, 1 esophageal cancer, 1 oligometastatic ovarian cancer) underwent pre-treatment FDG PET/CT and MAA SPECT/CT perfusion scans. Images were rigidly co-registered to the planning CT, on which lung and gross tumor volumes (GTV) were delineated. Maximum FDG standardized uptake value (SUVmax) in GTV, PERCIST-defined peak FDG SUV (SUVpeak) in GTV, maximum MAA perfusion uptake ratio in GTV normalized to lung (GTV:LUmax), and peak MAA perfusion uptake ratio (GTV:LUpeak) were calculated. Pearson correlation Rp between FDG and MAA perfusion image parameters were estimated and tested for statistical significance of linear relationships. GTV size was included in linear correlation and multivariate linear regression analysis as a potential confounder to account for imaging partial volume effects and heterogeneous tumor cell population. Results: Tumor FDG uptake and MAA perfusion appear to be inversely correlated and independent of tumor volume. FDG SUVmax and FDG SUVpeak achieved high negative linear correlations to MAA GTV:LUmax (Rp Z -0.84, p Z 0.004) and MAA GTV:LUpeak (Rp Z -0.83, p Z 0.005), respectively. GTV size was not significantly correlated to FDG SUVmax (Rp Z 0.56, p Z 0.11), FDGpeak (Rp Z 0.63, p Z 0.07), MAA GTV:LUmax (Rp Z -0.27, p Z 0.47), or MAA GTV:LUpeak (Rp Z -0.47, p Z 0.20). Adding GTV size to the linear regression models of FDG SUV vs MAA uptake ratios did not improve goodness-of-fit or statistical significance (FDG vs MAA + GTV: R2 > 0.68, p < 0.01; FDG vs MAA: R2 > 0.64, p < 0.005). Conclusions: Tumor perfusion, as quantified by MAA SPECT/CT, was inversely correlated with FDG PET/CT uptake, suggesting that this parameter is reflective of tumor biology and not simple payload delivery. Calculating tumor perfusion using this approach with readily available MAA SPECT/CT may serve as an indicator of tumor phenotype. Prospective studies in larger patient cohorts will be required to confirm these data. Author Disclosure: S.R. Bowen: None. M.J. Nyflot: E. Research Grant; GE Healthcare Research Contract. P.E. Kinahan: E. Research Grant; GE Healthcare Research Contract. O. Partnership; Co-founder, PET/X LLC. G.A. Sandison: None. R. Miyaoka: E. Research Grant; Philips Healthcare Research Contract. S.A. Patel: None. H.J. Vesselle: None. J. Zeng: None. R. Rengan: None.

3062 Patterns of Regional Failure and Evaluation of LN Dose in EarlyStage NSCLC Treated With SBRT J. Zhung,1,2 T.A. DiPetrillo,1,2 D. Wazer,1,2 J.T. Hepel,1,2 L. Price,3 and K.L. Leonard1,2; 1Tufts Medical Center, Boston, MA, 2Rhode Island Hospital-Warren Alpert Brown Medical School, Providence, RI, 3Tufts Clinical and Translational Science Institute, Boston, MA Purpose/Objective(s): SBRT is used for curative intent in medically inoperable early stage NSCLC with 4 year regional control rates of w80%. It is hypothesized that incidental low dose to regional lymph nodes (LN) may contribute to regional control. The purpose of this study is to test the hypothesis that incidental radiation dose to regional LNs affects regional control in SBRT for early stage NSCLC. Materials/Methods: Patients with newly diagnosed T1-T2 N0 NSCLC were treated with IMRT-based SBRT at this institution from January 2008 to January 2013. Initial disease was confirmed by biopsy or PET. LN disease was excluded on CT/PET. Mediastinoscopy or EBUS was performed only in equivocal cases. Thoracic LN stations were retrospectively contoured according to the IASLC and Michigan CT-based contouring atlas. Dose to LN was calculated using the original treatment plan. Upon

Poster Viewing Abstracts S629 follow-up, failure location was confirmed on serial CT, PET, or biopsy. Kaplan-Meier plots for failure and univariate (UV) analysis were used to evaluate associations between incidental LN dose and regional failure. Results: 50 patients met criteria for evaluation, with a median follow-up of 16 months. 84% were T1 and 16% were T2. The median treatment delivered was 50Gy using 12 treatment beams in 5 fractions. Mean D95 to LNs are reported in the Table. The 4-year rates of LC, RC and DM were: 88.5%, 81% and 58%. There were 7 regional failures; the most common site being hilar and paratracheal LNs. 71% of regional failures were hilar (one was contralateral) and 86% were paratracheal. Subcarinal and AP window failures were present in 57% and 29% of regional failures. On UV analysis, D95 to ipsilateral hilum was associated with risk for regional failure (p Z 0.026). However, those with regional failure had higher incidental hilar dose (mean D95 Z 6 Gy) than those without regional failure (mean D95 Z 2.2 Gy). Given the low number of failures, UV analysis was limited and revealed that regional failure was not associated with target size (p Z 0.54). 57% of regional failures also failed distantly. Conclusions: SBRT is highly conformal therapy with a high rate of locoregional control in T1-2 N0 NSCLC, despite low incidental dose to regional LNs. The majority of LN failures occurred in the ipsilateral hilar and paratracheal LN. Incidental dose to LNs does not appear to be associated with decreased risk of regional failure. Thus, selection of patients with low risk of microscopic LN disease is crucial and EBUS or mediastinoscopy should be considered for patients at high risk for LN involvement, such as those with central or large T2N0 disease.

Scientific Abstract 3062; Table mary site

LUL RUL RML LLL RLL

Mean D95 to Regional LN according to pri-

2R (Gy)

2L (Gy)

4R (Gy)

4L (Gy)

5 (Gy)

6 (Gy)

7 (Gy)

8 (Gy)

Ipsilateral hilar (Gy)

1.34 3.67 0.76 0.23 0.16

2.52 1.95 0.63 0.30 0.13

0.84 3.36 0.80 0.56 0.24

1.12 1.37 0.72 0.36 0.24

3.04 0.75 2.59 1.69 0.26

0.69 0.87 0.63 0.53 0.19

1.04 1.71 6.74 1.53 1.78

0.07 0.30 1.35 0.32 1.35

1.69 0.84 9.21 3.12 6.27

Author Disclosure: J. Zhung: None. T.A. DiPetrillo: None. D. Wazer: None. J.T. Hepel: None. L. Price: E. Research Grant; NIH CTSA UL1 TR000073. K.L. Leonard: None.

3063 Predictive Models for Assuring Quality and Consistency of Lung IMRT Plans Across Institutions in Cooperative Group Trials Q. Wu,1 Y. Ge,2 L. Yuan,1 C.R. Kelsey,3 F. Yin,1 X. Zhang,4 R. Mohan,5 J.B. Bluett,6 M.T. Gillin,4 Z. Liao,7 C.B. Simone,8 L. Lin,8 K.L. Moore,9 J.D. Bradley,10 J. Galvin,11 and Y. Xiao11; 1Duke University Medical Center, Durham, NC, 2University of North Carolina, Charlotte, NC, 3Duke University, Durham, NC, 4University of Texas MD Anderson Cancer Center, Houston, TX, 5University of Texas MD Anderson Cancer Center, Houston, TX, 6University of Texas MD Anderson Cancer Center, Houston, TX, 7University of Texas MD Anderson Cancer Center, Houston, TX, 8 Hospital of the University of Pennsylvania, Philadelphia, PA, 9UCSD, San Diego, MO, 10Washington University School of Medicine, St Louis, MO, 11 Bodine Cancer Center, Philadelphia, PA Purpose/Objective(s): To build DVH-based prediction models that capture anatomical variations and other institutional variations from multiple lung IMRT plan datasets, and to examine the feasibility of such models for assessing compliance with constraints in multi-institutional clinical trials. Materials/Methods: 69 lung IMRT plans from three institutions were collected (43, 10, and 16 from Institutions 1, 2, and 3, respectively). A PTV anatomical feature distribution analysis was performed to examine the patient anatomical variations among institutions. Based on the analysis, the entire dataset was partitioned into training and validation datasets, ensuring that the anatomy spectrum was represented in both datasets. Hence, the training dataset consisted of 33 (of 43) and 8 (of 16) plans from