External Validity of a Nomogram Predicting Risk of 6-Month and 9-Month Distant Brain Failure After Initial Stereotactic Radiosurgery Alone for Brain Metastases in an Independent Multi-institutional Patient Cohort

Abstract

Patients (pts) treated with stereotactic radiosurgery (SRS) alone for brain metastases (BM) are at increased risk of distant brain failure (DBF). A nomogram for predicting individualized 6 and 9-month freedom from DBF after SRS was recently published based on a single institution database (Ayala-Peacock et al., Neuro Oncol 2014). This nomogram has not undergone external validation in an independent patient population. The goal of this study was to assess the external validity of this nomogram in a multi-institutional independent patient cohort. The records of consecutive pts with BM treated with SRS alone at 2 tertiary care cancer institutes between 2005-2013 were reviewed. Inclusion criteria was the same as the Ayala-Peacock study and included patients with BM treated with initial SRS alone, no previous cranial radiotherapy, up to 13 synchronous BM, and either non-small cell lung (NSCLC), breast, renal cell (RCC), or melanoma primary cancer. Additionally, our validation cohort excluded pts with less than 3 months MRI follow-up. The nomogram variables included age, sex, race, primary histology, systemic disease status, and number of BM to calculate 6 and 9-month freedom from DBF using the Kaplan-Meier method. Discrimination was assessed using the Harrell’s c-index and calibration was evaluated using calibration plots and correlation between expected and observed freedom from DBF. After exclusions, 281 patients with 416 BM were eligible and made up the validation cohort. The median patient age was 61 years old, 48% were male, and 78% were white. Primary site was NSCLC for 49%, melanoma - 25%, breast - 19%, RCC - 8%, and 63% had active systemic disease at the time of SRS. The median imaging follow-up period was 13.4 months (range 3 – 124.7 months). Freedom from DBF at 6-months and 1-year in the Ayala-Peacock study was 59% and 30% compared with 64% and 50% in the validation cohort. DBF occurred within 6 months from SRS in 99 pts (35%) and within 9 months in 119 pts (42%) in the validation cohort. The Harrell’s c-index for both the 6-month and 9-month DBF nomogram using the validation cohort was 0.55, indicating poor discrimination ability. Calibration plots for the 6-month DBF nomogram demonstrated underestimation of DBF risk at the lower end of predicted risk and overestimation at the upper end with moderate correlation (r2=0.72). Calibration plots of the 9-month nomogram demonstrated systematic overestimation of DBF risk with moderate correlation (r2=0.68). The Ayala-Peacock et al. nomogram for predicting 6 and 9-month freedom from DBF after SRS alone for BM was not found to be externally valid in an independent multi-institutional cohort and cannot be recommended for clinical use at this time. These results reinforce the importance of validating predictive models in independent cohorts.