Application of the RTOG recursive partitioning analysis classification to brain metastases from colorectal cancer

Abstract

4104 Background: Colorectal cancer infrequently causes brain metastases (BMs). Recently, the incidence of BMs from colorectal cancer (CRC) has been reported to be increasing as 3 % of all metastatic CRCs, especially in patients (pts) with lung metastases (LMs). The RTOG previously developed three prognostic classes for BMs using a recursive partitioning analysis (RPA) classification, including Karnofsky performance status, controlled primary tumor, extracranial metastases, and an age. However, the relevance of this classification for CRC remains unclear, because only a few CRC pts were included in the RTOG database. The present studies retrospectively evaluate the usefulness of RPA classification for BMs from CRC. Methods: The subjects were consecutive 290 metastatic CRCs without symptomatic BMs who treated with any chemotherapy initiated in our institution between 2002 and 2005. The treatment of BMs, including whole brain radiation therapy (WBRT), surgery, and stereotactic radiosurgery (SRS), were indicated for pts whose life expectancy was considered to exceed 3 months (M). Results: BMs were detected in 20 pts out of 133 (15%) with LMs at base line, with a median follow-up time of 17.7 M. In the remaining 157 pts, BMs were found in 3 pts out of 41(7%) who had developed LMs during follow-up and 2 pts out of 116 (2%) without LMs. In total, overall incidence of BMs was 9% (25/ 290). Median duration from the diagnosis of LMs to BMs was 12 M (range: 4.4–33.6). According to RPA, the 25 pts with BMs were classified into Class I 0 (0%), Class II 12 (48%) and Class III 13 (52%). Treatments for BMs were performed to all pts of Class II, and to 6 pts (46%) of Class III. Median overall survivals from detecting BMs were Class II 9.7 M and Class III 3.0 M (logrank test; p=0.02). In the 18 pts treated for BMs, median BMs progression-free survival of Class II and Class III were 3.0 M and 3.1 M. Conclusions: It is suggested that the incidence of BMs from CRC is more common than previously reported. Overall survivals for BMs reproduce the results from RPA classification. No significant financial relationships to disclose.