Purpose: This investigation defined patient populations at high-, intermediate-, and low-risk of regional failure (RF) after stereotactic radiosurgery (SRS) lesion treatment using clinical nomograms and recursive partitioning analysis (RPA).
Methods and materials: We created a retrospective database compiling 361 oligometastatic brain metastases patients treated with single-modality Linac-based SRS. Logistic analysis was performed to identify factors to be included in a RPA to predict for cumulative RF at 1-year. A 1-year cumulative RF clinical nomogram was constructed and validated (c-index statistic).
Results: Age, number of brain metastases, World Health Organization (WHO) performance status (PS), and maximum gross tumor volume (GTV) size were found to be statistically significant predictors of the primary outcome. RPA classifications were defined as follows: low-risk (<25% 1-year RF): solitary lesion AND age >55Y; intermediate-risk (25-40% 1-year RF): age ⩽55Y AND solitary lesion OR WHO⩾1 AND 2-3 lesions; and high-risk (>40% 1-year RF): WHO PS=0 AND 2-3 lesions. These classifications were highly statistically significant (p<0.01) for RF. A clinical nomogram (containing patient age, lesion number, largest GTV volume, and WHO PS) for the prediction of 1-year cumulative RF was created (c-index 0.69).
Conclusion: A risk-adapted treatment approach can be applied for BM radiosurgery either using RPA categories and/or nomogram-based risk estimates.
Keywords: Brain metastases; Predictive modeling; Recursive partitioning analysis; Regional failure/control; Stereotactic radiosurgery.
Copyright © 2014 Elsevier Ireland Ltd. All rights reserved.