Leptin concentrations are usually determined in serum or EDTA-treated plasma. Our aim was to evaluate the precision and accuracy of leptin measurement from serum and various kinds of plasma. Blood samples were collected from healthy volunteers and divided into aliquots of serum and plasma with administration of K<sub>2</sub>-EDTA, citrate, fluoride and lithium-heparinate, respectively. Leptin concentrations were measured by a specific radioimmunoassay. There were clear differences between the leptin values from the different matrices. The coefficient of variation (CV) was highest in serum samples (CV: 11.9 %) and very low in K<sub>2</sub>-EDTA-treated plasma samples (CV: 5.3%). Leptin measurements from serum yielded significantly lower results than the respective K<sub>2</sub>-EDTA-treated plasma samples (p < 0.05). The measurement in Li-heparinate-treated plasma revealed similar results to that in K<sub>2</sub>-EDTA-treated plasma. The lowest CV was determined in fluoride-treated plasma (5.2%). A higher CV was determined in citrate-treated plasma (5.8%) and the values were about 25% lower than in the respective K<sub>2</sub>-EDTA-treated plasma samples (p < 0.0001). Lower leptin measurements in citrate-treated plasma may partially, but not in total, be a consequence of dilution by anticoagulating additives. Because of the low CV K<sub>2</sub>-EDTA- and Li-heparinate-treated plasma samples appear most reliable for the measurement of leptin. The usage of these sample matrices is therefore recommended.