Brain metastasis (BM) is a leading cause of death in patients with non-small cell lung cancer (NSCLC). EGFR mutations in primary NSCLC lesions have been associated with sensitivity to EGFR tyrosine kinase inhibitor (TKI). Therefore, it has become important to understand EGFR mutation status in BM lesions of NSCLC, and its clinical implications. BM samples of 136 NSCLC patients from South China, in which 15 had paired primary lung tumors, were retrospectively analyzed for EGFR mutation by amplification mutation refractory system (ARMS). Effect of BM EGFR mutations on progression-free survival (PFS) and overall survival (OS) was evaluated by Kaplan-Meier curves and log-rank test. EGFR mutations were detected in 52.9% (72 of 136) of the BM lesions, with preference in female and never-smokers. A concordance rate of 93.3% (14 of 15) was found between the primary NSCLC and corresponding BM. Positive prediction value of testing primary NSCLCs for BM EGFR mutation is 100.0 %, and negative prediction value is 87.5%. Median PFS of BM surgery was 12 and 10 months (P = 0.594) in the wild-type and mutant group, respectively. Median OS of BM surgery was 24.5 and 15 months (P = 0.248) in the wild-type and mutant group, respectively. In conclusion, EGFR mutation status is highly concordant between the primary NSCLC and corresponding BM. The primary NSCLC could be used as surrogate samples to predict EGFR mutation status in BM lesions or vice versa. Moreover, EGFR mutations showed no significant effect on PFS or OS of NSCLCs with BM.