Radiation and Cancer Physics

SS 27 - Physics 7 - Special Session: Outcome Analysis and Modeling

195 - The Role of Heart-Related Dose-Volume Metrics on Overall Survival in the RTOG 0617 Clinical Trial

Tuesday, October 23
3:35 PM - 3:45 PM
Location: Room 214 C/D

The Role of Heart-Related Dose-Volume Metrics on Overall Survival in the RTOG 0617 Clinical Trial
M. Thor1, J. O. Deasy1, C. Hu2, H. Choy3, R. U. Komaki4, G. Masters5, G. R. Blumenschein6, K. M. Forster7, J. H. Oh1, V. S. Kavadi8, S. Narayan9, R. D. Timmerman10, C. G. Robinson11, J. S. Greenberger12, D. Biggs5, M. Augspurger13, J. Meng14, and J. D. Bradley15; 1Memorial Sloan Kettering Cancer Center, New York, NY, 2Sidney Kimmel Comprehensive Cancer Center at Johns Hopkins, Baltimore, MD, 3University of Texas – Southwestern Medical Center, Department of Radiation Oncology, Dallas, TX, 4Department of Radiation Oncology, The University of Texas MD Anderson Cancer Center, Houston, TX, 5Christiana Care/Helen F. Graham Cancer Center, Newark, DE, 6UT Southwestern/Simmons Cancer Center-Dallas, Dallas, TX, 7Community Medical, Clovis, CA, 8Texas Oncology, Fort Worth, TX, 9Saint Joseph Mercy Hospital, Ann Arbor, MI, 10Department of Radiation Oncology, University of Texas Southwestern Medical Center, Dallas, TX, 11Washington University School of Medicine, Department of Radiation Oncology, St. Louis, MO, 12UPMC-Shadyside Hospital, Pittsburgh, PA, 13Baptist Cancer Center, Jacksonville, FL, 14The Ottawa Hospital Cancer Centre, Ottawa, ON, Canada, 15Washington University School of Medicine, St. Louis, MO

Purpose/Objective(s): To investigate the potential role of heart substructures in a dose-volume histogram (DVH) model for overall survival (OS) after radiotherapy (RT) for stage III Non-Small Cell Lung Cancer.

Materials/Methods: The study included the 437 patients with complete dose and OS data. For modeling, the cohort was randomly split into a training and a set-aside validation dataset (N=306, 131) maintaining the rate of prescription dose level, i.e., 60Gy/74Gy in 2Gy fractions (Training/Validation: 58/59% and 42%/41%). Within training, model building was performed using Cox regression modeling with 190 covariates related to disease, fractionation-corrected dose assuming α/β=3Gy (atria, pericardium, ventricles, and Lung; A, P, V, L), patient, or treatment. A candidate predictor was suggested by a p-value≤0.0003 (Bonferroni corrected) with one best DVH variable considered for each structure. Mutlivariable regression analysis (MVA) was conducted with forward variable selection and a retention criterion of p≤0.05 from a likelihood ratio test. Univariate and MVA were conducted on 1000 bootstrapped replicates. The most frequently selected MVA models (≥10% of bootstrap replicates) were subject to validation. In training and validation, model discrimination was assessed using the C-index, and the distribution of prescription dose levels was compared between the high (>84th percentile) and the low (<16th percentile) risk groups in validation.

Results: The median follow-up time was 22 and 27 months in training and validation (overall mortality rates: 73% and 73%), respectively. Candidate predictors involved only DVH variables, and the best predictors were: A: minimum dose to the hottest 45% (AD45), P: mean dose to the hottest 55% (PMOH55), V: mean of the hottest 5% (VMOH5), and L: mean dose (LDmean). Four candidate MVA models were selected in ≥10% of replicates: M1: AD45+PMOH55, M2: AD45+PMOH55 +VMOH5, M3: AD45+PMOH55+LDmean, and M4: PMOH55 +VMOH5+ LDmean. Discrimination of all four models decreased slightly in validation vs. training data (C-index: 0.80-0.82, vs. 0.84-0.87). The high risk group included predominantly patients from the 74Gy arm (M1-4: 55-60%), and the low risk group predominantly came from the 60Gy arm (M1-4: 53-63%).

Conclusion: Due to correlations among DVH variables, this type of analysis can only suggest an association, but not isolate a causal role in mortality. Nonetheless, heart substructure dose-volume metrics were robustly selected in classifying the likelihood of death after RT, more so than lung dose-volume metrics. In particular, the most frequently selected models on bootstrap included DVH variables of the atria and pericardium (AD45, PMOH55), and this was confirmed on set-aside validation data. These results support efforts to avoid heart irradiation when feasible preferably using state-of-the-art planning and delivery techniques such as intensity-modulated image-guided RT.

Author Disclosure: M. Thor: None. J.O. Deasy: Chair, Research Committee; AAPM. C. Hu: None. H. Choy: Research Grant; Celegene. Advisory Board; Bayer, EMD. R.U. Komaki: None. G. Masters: None. G.R. Blumenschein: None. K.M. Forster: None. V.S. Kavadi: Partner; Texas Oncology. Medical Director; US Oncology. R.D. Timmerman: Research Grant; Varian Medical Systems, Accuray, Inc, Elekta Oncology. C.G. Robinson: Research Grant; Varian Medical Systems, Elekta. Speaker's Bureau; Varian Medical Systems, DFINE. Advisory Board; Radialogica. Stock Options; Radialogica. D. Biggs: None. M. Augspurger: None.

Maria Thor, PhD, MS

Memorial Sloan Kettering Cancer Center: Research associate: Employee


Send Email for Maria Thor


195 - The Role of Heart-Related Dose-Volume Metrics on Overall Survival in the RTOG 0617 Clinical Trial

Attendees who have favorited this

Please enter your access key

The asset you are trying to access is locked. Please enter your access key to unlock.

Send Email for The Role of Heart-Related Dose-Volume Metrics on Overall Survival in the RTOG 0617 Clinical Trial