New Classification Scheme to Predict Survival in Patients with Brain Metastases from Non-small Cell Lung Cancer Treated with Upfront Stereotactic Radiosurgery

Purpose/Objective(s) 

To validate three existing classifications and propose a new system to predict prognoses of patients undergoing upfront stereotactic radiosurgery (SRS) for non-small-cell lung cancer (NSCLC) brain metastases (BMet).

Materials/Methods 

Patients (pts) treated with upfront SRS for NSCLC BMet were retrospectively identified from an IRB-approved database. Variables assessed included gender, age, interval from NSCLC diagnosis (dx) to BMet dx, KPS, control of primary, extracranial metastasis (met), hepatic met, bone met, maximum size of BMet, number of lesions, and whole-brain radiation upfront. Patients were also scored using the RTOG RPA Class, Score Index for Radiosurgery (SIR) and Basic Score for Brain Metastases (BS-BM). Data were analyzed using the log-rank test and proportional hazards model.

Results 

One hundred fifty-seven pts were identified from database. Patients were treated with SRS from 6/91 to 5/04. In univariate analysis, pts with interval from NSCLC dx to BMet dx of >12 months (mos) had longer median overall survival (OS) of 10.6 mos vs. 6.5 mos for ≤12 mos (p = 0.02). Those with KPS of 90-100 had median OS of 9.3 mos, 6.5 mos for 70-80, and 3.1 mos for <70 (p = 0.02). Patients with no hepatic met had median OS of 7.2 mos vs. 5.2 mos for those with (p = 0.02). Those without bone met had median OS of 7.5 mos vs. 6.2 mos for pts with (p = 0.01). Females and those with primary controlled also tended to do better, (p = 0.06 for both). On multivariate analysis, higher KPS (p = 0.001), female gender (p = 0.007), no liver met (p = 0.004), no bone met (p = 0.02), and longer interval from NSCLC dx to BMet dx (p = 0.06) were identified as independent predictors for improved OS. The RPA Class and BS-BM score were associated with OS. The pts with RPA Class I had median OS of 11.0 mos, Class II had 7.2 mos, and Class III had 3.1 mos (p = 0.001). The pts with BS-BM score of 3 had median OS of 9.3 mos, BS-BM of 2 had 8.8 mos, BS-BM of 1 had 6.5 mos, and BS-BM of 0 had 3.7 mos (p = 0.005). The SIR did not predict for OS (p = 0.62).

Based on our identified prognostic factors, a new model was generated. One point each was given for male gender, KPS of 70-80, interval from NSCLC dx to BMet dx ≤12 mos, or presence of bone met. Two points each were given if liver met was present or KPS was ≤60. The pts with a score 0-1 were classified as favorable risk (FR); a score of 2 was considered intermediate risk (IR) 1; a score of 3, IR 2; and a score of >3, unfavorable (UF). Using this scheme, FR pts had a median OS of 11 mos, IR 1 10.4 mos, IR 2 6.1 mos, and UF 3.6 mos (p < 0.001).

Conclusions 

Only the RPA Class and BS-BM could be validated in our group of pts. Given the complexity of determining who may benefit from upfront SRS in NSCLC BMet pts, we have developed our own easy prognostic system tailored to this patient population. This schema will need to be validated in a prospective setting.