Abstract
PURPOSE
We compared multiple response assessment and visualization techniques for characterization of post-SRS treatment outcomes in a cohort of breast cancer patients with long median follow-up.
METHODS
A retrospective IRB-approved review was completed of 75 breast cancer patients treated with SRS for 271 brain metastases with median follow up of 40 months. Tumor dimensions, brain failure events, corticosteroid use, and clinical status were analyzed utilizing RANO-BM, bidimensional product (BDP), and bidimensional sum (BDS) techniques. 46/75 patients were eligible for RANO-BM assessment. Response at each post-treatment assessment were scored as PD, SD, PR, or CR, and the concordance between techniques was determined. A scoring system-based outcome metric labelled ‘average state’ was derived to estimate fractional time/response state by each assessment method. Interactive timeline displays of outcome states were generated.
RESULTS
The concordance of patient response states was determined using either RANO-BM, BDP or BDS among the 46 eligible patients. The overall mean and median concordance between techniques were 0.82 and 0.83, (range 0.52 – 1). The average state for the patient population post-treatment was 1.98 by RANO-BM, 2.29 by BDP, and 2.19 by BDS. For patients excluded from RANO-BM secondary to lack of measurable disease, the average state was determined to be 2.44 by BDP and 2.35 by BDS. The average state for HER2+ vs. HER2- patients was 2.21 vs. 1.75 by RANO-BM, 2.58 vs. 2 by BDP, and 2.39 vs. 1.99 by BDS. An interactive timeline view was generated to display outcome states utilizing the 3 response assessment techniques, and the impact of inclusion of non-target lesions and variable response parameters was assessed graphically.
CONCLUSIONS
These results characterize the concordance and the limitations of multiple outcome assessment methodologies in a post-SRS cohort with long median follow-up. The utility of a novel ‘average state’ outcome metric is demonstrated in this cohort.