There is no consensus on the optimal protocol for isolation of circulating tumor DNA. We report our comparison of several extraction methods and variables that may affect yield, quality, and contamination of tumor DNA. DNA was extracted from the plasma and serum of five healthy volunteers by means of four different commercially available kits and DNA yield was quantified by real-time PCR. DNA was extracted using the optimum kit from the plasma and serum of an additional 10 healthy volunteers and 10 patients with small cell lung cancer (SCLC) to compare yield and DNA fragment size in plasma versus serum and in those with SCLC versus controls. Time to sample processing was also examined. We found that DNA yield was greatest using the QIAamp Viral Spin Kit. Delayed time to processing led to increased DNA concentrations in serum, but not plasma. The plasma DNA concentration in SCLC patients was significantly higher than in healthy volunteers (24.5 ng/mL versus 5.1 ng/mL, P= 0.002). In contrast, there was no significant difference in serum DNA concentrations between controls and patients that may be explained by the wide variability and range of DNA concentrations in serum. A significantly higher proportion of longer fragments (272 bp/60 bp) was observed in the plasma DNA extracted from patients with SCLC than in healthy controls (13% versus 8%, P= 0.04). There was absence of DNA fragments of 512 bp in healthy control plasma, but faint bands were observed in serum, which is thought to be due to cellular contamination. We conclude that plasma is a more reliable source of tumor DNA then serum and have optimized a robust procedure for plasma tumor DNA isolation that can be applied to translational research studies.