Multi-institutional Validation of Brain Metastasis Velocity, a Recently Defined Predictor of Outcomes Following Stereotactic Radiosurgery

Abstract

In this study we attempt a multi-institutional validation of brain metastasis velocity (BMV), a recently defined predictor of overall survival (OS) and distant brain failure (DBF) following stereotactic radiosurgery (SRS) for brain metastasis. Patients from nine academic centers were treated with upfront SRS alone for brain metastasis. Of these nine centers, data from the eight institutions not previously used to define BMV were used to construct the validation dataset. BMV was defined as the number of brain metastases at time of first DBF following initial SRS divided by the time (in units of years) to first DBF, consistent with the definition as previously described. Time-to-event outcomes were estimated using the Kaplan-Meier method. Cox proportional hazards methods were used to estimate the effect of BMV on OS following first DBF, with BMV defined both as a continuous variable and also as stratified into low (<4 metastases/yr), intermediate (4-13 metastases/yr), and high risk (>13 metastases/yr) groups. Simple and multiple linear regression were used to identify predictors of BMV (as a continuous variable) from baseline / treatment-related patient variables. Of a total 2829 patients, 737 patients were treated at the institution at which BMV was initially defined, while the remaining 2092 patients were treated at one of the eight other institutions comprising the validation dataset. Of these 2092 patients, 921 (44.0%) experienced DBF, with 786 (37.6%) patients also having a quantifiable number of metastases at time of DBF. Median OS from initial SRS was 11.2 mo (CI: 10.4-11.9 mo). Cumulative incidence of DBF at 6, 12, and 24 mo was 26.7%, 36.6%, and 42.8%, respectively. Median OS following DBF was 7.2 mo (CI: 6.4-8.2 mo). BMV was predictive of OS following DBF, with median OS of 12.5 mo (CI: 11.0-14.8 mo) for BMV < 4, 7.0 mo (CI: 5.9-9.4 mo) for BMV 4-13, and 4.6 mo (CI: 3.8-5.3 mo) for BMV > 13 (log-rank p < 0.0001). BMV was also predictive of OS following DBF as a continuous variable (HR: 1.010, CI: 1.008 - 1.012, p<0.0001). Predictors of BMV on simple linear regression included melanoma histology (p<0.0001), number of initial brain metastases (p<0.0001), progression of systemic disease (p=0.01), diffuse systemic disease (p=0.02), and lowest SRS dose (p=0.02). On multiple regression, only melanoma histology (β: 10.10, SE: 1.89, p<0.0001) and number of initial brain metastases (β: 1.52, SE: 0.34, p<0.0001) remained predictive of BMV (adjusted R2 = 0.06). The results of this multi-institution study validate the previous single-institution finding that BMV at time of initial DBF is a dominant predictor of OS following DBF. Melanoma histology and number of initial brain metastases were validated as predictors of BMV, although these account only for a small amount of variation in BMV.