Serum albumin is an established predictor of survival in numerous cancers but its prognostic value in central nervous system tumours has not been established. Here we have examined prognostic factors in 685 patients with histologically proven glioblastoma multiforme (GBM), the majority of which (n = 549) had pre-operative serum albumin assayed. Mean serum albumin was 34.7 g/l (SD 5.2). Post-operative survival was significantly less for patients with hypoalbuminaemia (<30 g/l, n = 82) than for patients with normal albumin level (median 2.3 vs. 5.6 months, P < 0.001 Log-rank test). Furthermore, patients with lower normal albumin (30-40 g/l, n = 371) had significantly shorter survival compared against patients with albumin in the upper normal range (40-50 g/l, n = 96; median 5.1 vs. 8.8 months, P < 0.001). Multivariate Cox regression showed the independent predictors of survival were age, debulking surgery, chemoradiotherapy, and serum albumin (Hazard Ratio 0.97 per g/l, P < 0.005). This study suggests pre-operative serum albumin level is a significant predictor of survival in patients with GBM. Further studies are needed to examine the relationship between albumin and other known prognostic factors, and to determine if pre-operative serum albumin is a clinically useful predictor of survival.