Subclassification of Recursive Partitioning Analysis Class II Patients with Brain Metastases Treated Radiosurgically

Purpose: Although the recursive partitioning analysis (RPA) class is generally used for predicting survival periods of patients with brain metastases (METs), the majority of such patients are Class II and clinical factors vary quite widely within this category. This prompted us to divide RPA Class II patients into three subclasses.

Methods And Materials: This was a two-institution, institutional review board-approved, retrospective cohort study using two databases: the Mito series (2,000 consecutive patients, comprising 787 women and 1,213 men; mean age, 65 years [range, 19-96 years]) and the Chiba series (1,753 patients, comprising 673 female and 1,080 male patients; mean age, 65 years [range, 7-94 years]). Both patient series underwent Gamma Knife radiosurgery alone, without whole-brain radiotherapy, for brain METs during the same 10-year period, July 1998 through June 2008. The Cox proportional hazard model with a step-wise selection procedure was used for multivariate analysis.

Results: In the Mito series, four factors were identified as favoring longer survival: Karnofsky Performance Status (90% to 100% vs. 70% to 80%), tumor numbers (solitary vs. multiple), primary tumor status (controlled vs. not controlled), and non-brain METs (no vs. yes). This new index is the sum of scores (0 and 1) of these four factors: RPA Class II-a, score of 0 or 1; RPA Class II-b, score of 2; and RPA Class II-c, score of 3 or 4. Next, using the Chiba series, we tested whether our index is valid for a different patient group. This new system showed highly statistically significant differences among subclasses in both the Mito series and the Chiba series (p < 0.001 for all subclasses). In addition, this new index was confirmed to be applicable to Class II patients with four major primary tumor sites, that is, lung, breast, alimentary tract, and urogenital organs.

Conclusions: Our new grading system should be considered when designing future clinical trials involving brain MET patients.