Vascular endothelial growth factor (VEGF) is secreted by various human cancer cells and plays a key role in cancer angiogenesis and metastasis. Recently, evidence of VEGF storage in blood cells including platelets has been reported. The serum VEGF levels were reported to increase during clotting as a result of its release from platelets, and plasma sample instead of serum was recommended for measuring the circulating VEGF more accurately. However, platelets have been implicated in tumor metastasis since circulating tumor cells forming aggregates with platelets were observed. The purpose of this study was to clarify which is an optimal specimen to measure VEGF in cancer patients, serum or plasma. We measured serum and plasma VEGF levels and platelet counts in 173 cancer patients and 42 healthy people, and found that serum VEGF levels were significantly higher than matched plasma VEGF and the VEGF difference (serum VEGF - plasma VEGF) correlated with platelet counts (r=0.624, p<0.05) in both cancer patients and healthy controls. We selected cancer patients with normal platelet counts (130-400x103/microl, Plt-normal cancer group). Interestingly, serum VEGF levels were higher in Plt-normal cancer group than in healthy controls. The theoretical platelet-derived VEGF in serum, calculated based on actual blood platelet counts (pg per 106 platelets), was also significantly higher in Plt-normal cancer group than in normal controls. It is, therefore, suggested that, although the serum VEGF levels are affected by blood platelets, platelet-derived VEGF also reflect biology of cancer cells, and that serum would be the more useful specimen for measurement of circulating VEGF in cancer patients for prognosis.