A.B. Chen

Use of frailty to predict survival in elderly patients with early stage non-small-cell lung cancer treated with stereotactic body radiation therapy

OBJECTIVES Frailty has been shown to increase morbidity and mortality independent of age, but studies are lacking in radiation oncology. This study evaluates a modified frailty index (mFI) in predicting overall survival (OS) and non-cancer death for Stage I/II [N0M0] Non-Small-Cell Lung Cancer (NSCLC) patients treated with Stereotactic Body Radiation Therapy (SBRT). MATERIALS AND METHODS Medical records for all patients with Stage I/II NSCLC treated at our institution with SBRT from 2009 to 2014 were reviewed. A validated mFI score, consisting of 11 variables was calculated, classifying patients as non-frail (0-1) or frail (≥2). Primary endpoint (OS) was analyzed using Kaplan-Meier method and log-rank. Secondary endpoint, non-cancer death, was analyzed using Fine-Grays method, with death from lung cancer as a competing risk. RESULTS Patient cohort consisted of 38 (27.3%) non-frail and 101 (72.7%) frail [median total mFI score 3.0 (range 0-7)]. Median age and pack-year history was 74 and 46years, respectively. Median follow-up among survivors was 38.5months (range 4.0-74.1months). Frailty was associated with a lower 3-year OS (37.3% vs. 74.7%; p=0.003) and 3-year cumulative incidence of non-cancer death (36.7% vs. 12.5%; p=0.02). Frailty remained significant in the multivariate model [OS HR for mFI ≥2: 2.25 (1.14-4.44); p=0.02]. CONCLUSION Frailty is associated with lower OS in older patients with early stage NSCLC treated with SBRT, yet frail patients survived a median 2.5years, and were more likely to die of causes unrelated to the primary lung cancer, suggesting SBRT should be considered even in older patients deemed unfit for surgery.

TU‐E‐BRB‐02: A Decision Support Tool for SBRT Planning Using a Searchable DVH Database

PURPOSE Develop a decision support tool that aids dosimetrists, physicians, and physicists in assessing and improving plan quality through comparison to plans previously used in similar clinical situations. METHODS Software was developed to capture and store DVHs and other clinically relevant treatment plan characteristics in a database. In addition to the plan DVH, the database contains a total of 24 plan characteristics including fractionation, prescribed dose, treatment volume, prior surgery, tumor position, and smoking history. DVH and other plan data was captured from the treatment planning system via exported dicom RT files. Structures in the plan were automatically matched by name to a list of standard structures using a system of regular expressions. Additional fields were entered manually using a simple java interface. As a support tool, a plan under development can be quickly compared to similar plans in the database based on selected plan characteristics. A plot displaying the current and historical DVHs provides an easy visual comparison. Our interface also provides statistics for comparison for each dose/volume level such as average, minimum, maximum and standard deviation. RESULTS DVHs from 111 lung SBRT plans treated from 2009-2011 were imported in accordance with an approved IRB protocol. As an example of data comparisons that can be easily performed to guide plan evaluation, we examined plans prescribing 5400cGy in 3 fractions and found that tumors >7.5cc (n=34) had an average PTV coverage of 94.2% (range: 73.5-95.0%), and tumors =7.5cc (n=35) had an average PTV coverage of 94.9% (range: 81.6-99.6%). CONCLUSION A searchable DVH database was constructed to provide planners, physicists, and physicians with a straightforward means of comparing plans against historic distributions of DVHs. In the future, outcome data will be included in the database to strengthen its functionality as a decision support and research tool.

MO-G-BRF-01: BEST IN PHYSICS (JOINT IMAGING-THERAPY) - Sensitivity of PET-Based Texture Features to Respiratory Motion in Non-Small Cell Lung Cancer (NSCLC).

PURPOSE PET-based texture features are used to quantify tumor heterogeneity due to their predictive power in treatment outcome. We investigated the sensitivity of texture features to tumor motion by comparing whole body (3D) and respiratory-gated (4D) PET imaging. METHODS Twenty-six patients (34 lesions) received 3D and 4D [F-18]FDG-PET scans before chemo-radiotherapy. The acquired 4D data were retrospectively binned into five breathing phases to create the 4D image sequence. Four texture features (Coarseness, Contrast, Busyness, and Complexity) were computed within the the physician-defined tumor volume. The relative difference (δ) in each measure between the 3D- and 4D-PET imaging was calculated. Wilcoxon signed-rank test (p<0.01) was used to determine if δ was significantly different from zero. Coefficient of variation (CV) was used to determine the variability in the texture features between all 4D-PET phases. Pearson correlation coefficient was used to investigate the impact of tumor size and motion amplitude on δ. RESULTS Significant differences (p<<0.01) between 3D and 4D imaging were found for Coarseness, Busyness, and Complexity. The difference for Contrast was not significant (p>0.24). 4D-PET increased Busyness (∼20%) and Complexity (∼20%), and decreased Coarseness (∼10%) and Contrast (∼5%) compared to 3D-PET. Nearly negligible variability (CV=3.9%) was found between the 4D phase bins for Coarseness and Complexity. Moderate variability was found for Contrast and Busyness (CV∼10%). Poor correlation was found between the tumor volume and δ for the texture features (R=-0.34-0.34). Motion amplitude had moderate impact on δ for Contrast and Busyness (R=-0.64- 0.54) and no impact for Coarseness and Complexity (R=-0.29-0.17). CONCLUSION Substantial differences in textures were found between 3D and 4D-PET imaging. Moreover, the variability between phase bins for Coarseness and Complexity was negligible, suggesting that similar quantification can be obtained from all phases. Texture features, blurred out by respiratory motion during 3D-PET acquisition, can be better resolved by 4D-PET imaging with any phase.

Comparative Effectiveness of Intensity-Modulated Versus 3D Conformal Radiation for Stage III Non-small Cell Lung Cancer in the Medicare Population

Purpose/Objective(s): The clinical benefit of intensity-modulated (IMRT) compared to 3D conformal radiation (3D-RT) has not been wellestablished for locally advanced non-small cell lung cancer (NSCLC). We evaluated trends in use of IMRT for stage III lung NSCLC and compared survival and hospitalization outcomes. Materials/Methods: Using SEER-Medicare data, we identified use of IMRT or 3D-RT among 7061 Medicare beneficiaries diagnosed with stage III NSCLC from 2002-2009. Factors associated with use of IMRT versus 3D-RT were identified using multivariable logistic regression. Overall survival and number of hospital days within 90 days of radiation were analyzed using Cox proportional hazard and negative binomial regression models, respectively. Propensity score adjustment was used to control for clinical and demographic variables. Results: IMRT comprised an increasing proportion of conformal treatments for NSCLC, rising from 3.0% in 2002 to 26.8% in 2009. Patients treated at freestanding versus hospital-based facilities were twice as likely to receive IMRT (17.3% vs 9.5%, adj ORZ 2.0, p < 0.01). IMRT use varied by region, with higher rates in the South (12.8%, adj OR Z 1.13) and West (14.2%, adj ORZ 1.25), compared to theNortheast (9.9%, ref) andMidwest (9.1%, adjOR Z 0.88) (overall pZ 0.03) and in urban versus rural areas (12.5% vs 9.9%, adj OR Z 1.52, p < 0.01). Patients with more comorbidities were more likely to receive IMRT, 11.3% (ref) vs 11.7% (adj OR Z 1.03) vs 14.7% (adj OR Z 1.35) for modified Charlson score 0, 1, and 2+, respectively (overall pZ 0.03). We did not find a difference in IMRT use between stage IIIA and stage IIIB patients (11.8% vs 12.4%, adj OR Z 1.11, p Z 0.19). Patients receiving chemotherapy were more likely to receive IMRT (12.8% vs 10.4%, adj ORZ 1.24, p Z 0.02), though there was no difference in IMRT use among patients having surgery (11.4% vs 12.2%, adj ORZ 0.95, pZ 0.68). With propensity score adjustment, IMRTwas associatedwith greater overall survival (adj HRZ 0.91, pZ 0.03) compared to 3D-RT, though there was no difference in survival among patients receiving at least 25 fractions of radiation (adj HRZ 0.99, pZ 0.83). There was no significant difference in number of hospital days in the 90 days following radiation start (mean 5 days, adj HR Z 1.01, p Z 0.89). Conclusions: When radiation is used to treat locally advanced NSCLC, IMRT is increasingly preferred to 3D-RT. However, among patients receiving potentially curative radiation ( 25 fractions) there was no significant difference in overall survival or time spent hospitalized following treatment compared to 3D-RT. Author Disclosure: A.B. Chen: None. L. Li: None. A. Cronin: None. D. Schrag: None.

SU‐E‐J‐139: Real‐Time Motion Management Will Increase the Patient Population Eligible for Lung SBRT

Purpose: Lung toxicities encountered during the early lung SBRT clinical trials have been largely abated by adhering to strict planning limits on lung dose and PTV size. This restricts the maximum safely treatable GTV size in the presence of respiratory motion. We hypothesize that real‐time motion management (e.g. tracking) will enable the maximum GTV size to increase, thereby making more patients eligible for SBRT. Methods: Consensus maximum PTV size (50 cc) was determined from recent clinical trials and our own clinical experience. PTV as a function of GTV was found for four treatment scenarios: 1 cm motion, no tracking; 0.5 cm motion, no tracking; no motion, no tracking; and, arbitrary motion, with tracking. We included a 0.5 cm/1.0 cm asymmetric CTV to PTV margin for no tracking (RTOG recommendation) and a 0.3 cm margin for tracking, as suggested by the results of Rottmann (2010). No GTV to CTV margin is included. A large population study is used to determine the increase in numbers of patients eligible for SBRT. Results: PTV 250%, going from a limit of 13 cc, for tumors moving more than 5 mm, to 33 cc for all motions. This translates to a ∼50% increase in the numbers of patients eligible for lung SBRT. Conclusions: Restrictions on PTV size limit patient eligibility for SBRT, excluding patients with smaller GTVs and respiratory motion. Real‐time motion management will enable a substantial increase in the numbers of patients eligible for SBRT. The project described was supported, in part, by Award Number R21CA156068 from the National Cancer Institute. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Cancer Institute or the National Institutes of Health.

SU-E-T-694: Correlation Between Geometric Information and Dose Fractionation in Lung SBRT Treatment Plans

PURPOSE Previously, a database-driven decision support tool to assist clinical staff in assessing and improving lung SBRT treatment plans was developed. Using planning data collected by this software, we determined a simple metric that can be used for early identification of cases for which it will be particularly difficult to achieve desired dosimetric goals. Early detection of these cases will be of clinical value. METHODS This retrospective IRB-approved study utilizes an existing database of treatment planning data for 121 lung SBRT patients. Patients who had multiple SBRT treatments or more than one lung tumor were excluded. In our department, SBRT patients receive treatment in either 3 fractions of 18Gy or 5 fractions of 10-12Gy. Five fractions are generally used for cases in which planning goals are difficult to achieve in terms of target coverage vs. dose to OAR. Using the specific fractionation as a surrogate for plan difficulty, we examined possible correlations between fractionation and parameters including target size, total prescribed dose, number of fields and distance to OAR. RESULTS 84 patients received 3 fractions of treatment and 37 received 5 fractions. Patients with PTVs greater than 70 cm3 in size and ITVs greater than 40 cm3 were always treated with 5 fractions. These patients constituted approximately 20% of the 5 fraction patients. Tumors smaller than this size were sometimes (e.g. when situated close to an OAR) treated with 5 fractions. No correlation was observed between the fractionation and the number of fields. CONCLUSION With large volume targets we can predict a treatment of 5 fractions. However, this represents a minority of patients. More powerful predictions could be possible by combining target size information with other metrics such as distance to nearby OARs. The ability to predict fractionation early in planning would allow for more efficient scheduling of LINAC time. Kaye Scholar Grant.