Recent arguments have suggested that tumor DNA in cancer patients could be found in plasma, but different points remain unclear. Using a series of 117 head and neck squamous cell carcinoma tumors, our goals for this study were: (a) to quantify the amount of plasma DNA; (b) to evaluate the presence of plasma tumor DNA; and (c) to analyze the clinical relevance of tests based on plasma DNA analyses. Low levels of plasma DNA were found in most samples, but all were successfully amplified. Two different methods were used to detect tumor-specific genetic alterations: (a) microsatellite instability at UT5085 with an established sensitivity of 1:500; and (b) p53 mutation screening. Of the 117 tumors typed at UT5085, 65 demonstrated bandshifts (55%). Plasma and tumor DNA a showed similar alteration in only one case among these samples, and the prevalence of tumor DNA in plasma was estimated to be <2% using microsatellite analysis. Tumor DNA was detected in plasma at a higher prevalence (2 of 11 cases) when using p53 mutant allele-specific amplification. These results showed that in plasma, tumor DNA is largely diluted by normal DNA. By comparison with previously published studies, the prevalence of microsatellite alterations in plasma in this series of head and neck squamous cell carcinomas is very low, despite the fact that a large series of tumors was analyzed. To explain this discrepancy, we analyzed the possibility of PCR artifacts as suspected by the presence of loss of heterozygosity in two plasma DNA samples without a similar tumor DNA alteration. When DNA concentrations were under the threshold of detection (<100 ng/ml), we demonstrated that PCR artifacts could occur at random, and, if misinterpreted, these false genetic alterations could artificially enhance the frequency of plasma DNA alterations. This may have been suspected in previously published series, but it has never been discussed before. Microsatellite analysis on plasma DNA is difficult to interpret and can frequently be misleading. Plasma DNA should be analyzed with very sensitive and specific methods such as mutant allele-specific amplification, which excludes artifacts but requires specific optimization that is probably not compatible with routine and clinical use.