Brain metastases are frequent in patients with lung cancer and a major cause of morbidity and mortality. Finding a biomarker predicting brain metastases could facilitate early start of treatment and thereby reduce morbidity and possibly improve overall survival. Previous studies suggest S100B as a possible biomarker for this purpose. This prospective study enrolled 185 patients with newly diagnosed stage IV non-small cell lung cancer (NSCLC). A total of 22 patients had brain metastases verified by magnetic resonance imaging or computed tomography at the time of enrollment. Serum S100B levels were measured in blood samples collected prior to any treatment from 22 patients who had brain metastases at enrollment and from 50 patients randomly selected from the remaining 163 patients without brain metastases at enrollment. No statistically significant difference was found in the levels of serum S100B between patients with and without brain metastases [range 0.018-0.209 µg/l, mean 0.049 µg/l, 95% confidence interval (CI), 0.032-0.061 µg/l] and (range 0.016-0.130 µg/i, mean 0.044 µg/l, 95% CI, 0.037-0.051 µg/l), respectively, (P=0.852). Univariate analysis of prognostic factors for S100B indicated a correlation (P<0.2) with sex (P=0.088) and histology (adenocarcinoma vs. squamous cell carcinoma/others) (P=0.028). In the multivariate analysis only histology (P=0.029) remained statistically significant. Conclusion: The present study found no significant correlation between the level of serum S100B and the presence of brain metastases in patients with advanced NSCLC. The clear cut-off of S100B in patients with and without brain metastases reported in other studies could not be verified in this study. Further studies investigating the role of S100B as a biomarker for brain metastases in non-small cell lung cancer are warranted.