Main Session
Sep
29
PQA 03 - Central Nervous System, Professional Development/Medical Education
2691 - Normal Tissue Complication Probability Model Predicting the Risk of Radiation Necrosis after Stereotactic Radiosurgery for Small Brain Metastases
Presenter(s)
Sreenija Yarlagadda, MD - Miami Cancer Institute, Miami, FL
S. Yarlagadda1, R. P. Tolakanahalli1,2, Y. Zhang3, D. J. Wieczorek1, T. Bejarano1, Y. C. Lee1,2, M. Hall1,2, R. H. Press1,2, M. W. McDermott4, A. Gutierrez1,5, M. P. Mehta1, and R. Kotecha1,2; 1Department of Radiation Oncology, Miami Cancer Institute, Baptist Health South Florida, Miami, FL, 2Herbert Wertheim College of Medicine, Florida International University, Miami, FL, 3TD – Artificial Intelligence and Machine Learning, Baptist Health South Florida, Miami, FL, 4Department of Neurosurgery, Miami Neuroscience Institute, Baptist Health South Florida, Miami, FL, 5Florida International University, Herbert Wertheim College of Medicine, Miami, FL
Purpose/Objective(s):
The traditional methods of evaluating whole-brain cumulative dose to predict an individual lesion-based risk of radiation necrosis (RN) are limited in the modern era with contemporary stereotactic radiosurgery (SRS) platforms; hence, the aim of the present study was to develop a Lyman-Kutcher-Burman (LKB) normal tissue complication probability (NTCP) model to predict (RN) risk for small brain metastases (SBM; = 2cm) using local dose evaluation.Materials/Methods:
The treatment plans and outcomes of consecutive patients with intact SBM treated with single fraction SRS to 20-24 Gy between January 2017 and July 2022 and with at least one follow-up assessment were evaluated. Each lesion was followed from the date of SRS to the date of RN (graded by CTCAE v5) or last follow-up. The locoregional at-risk normal brain was contoured as 1 cm volume around each lesion, excluding bone and ventricles and volumetric doses were recorded. Logistic regression models were used to estimate the relationship between any-grade RN and each volumetric dose (V8, V10, V12, V15), and NTCP modeling was performed using a modified LKB model.Results:
Data from 1497 SBM treated over 358 SRS courses in 235 patients were analyzed. The prescription dose was 20 Gy for 297 lesions (20%), 22 Gy for 439 lesions (29%), and 24 Gy for 761 lesions (51%). With a median follow-up of 12 months, 52 (3.5%) any grade RN events were observed of which 21 were grade 1, 21 were grade 2, 8 were grade 3, 2 were grade 4. The constrained optimization of the NTCP LKB model for any-grade RN estimated model parameters of m, n as 0.09, 0.32 with TD50 and TD5 of 51.5 Gy and 24.2 Gy respectively. Logistic model fitting for Vx50 demonstrated V8Gy of 42.24 cc (?50=0.88), V10Gy of 32.09 cc (?50=0.87), V12Gy of 23.3 cc (?50=0.88), and V15Gy of 16.29 cc (?50=0.88). V12Gy of 5 cc, 10 cc, 15 cc had RN probabilities of 5.9%, 11.9%, 22%, respectively. Similarly, V15Gy of 5 cc, 10 cc, 15 cc had RN probabilities of 8%, 19.8%, 42%, respectively.Conclusion:
In the modern era, for SBM treated with single fraction SRS, the locoregional at-risk normal brain has TD5 of 24 Gy. V12 < 3.8 cc and V15 < 2.6 cc were associated with NTCP risk of 5%.