Genetic abnormalities in SCCHNs are frequent and may be useful for screening, follow-up and prognosis. A biopsy or resection generally is utilized to identify these alterations but analysis of scraped or exfoliated tumor cells has been proposed as simpler and more versatile. It is unknown how well genetic abnormalities in scrapes reflect those in the tumor. Therefore, we compared DNA alterations in tumor scrapes obtained prior to treatment with alterations in microdissected tumor biopsies. Eight primary squamous-cell carcinomas of the head and neck (SCCHNs) were examined at 14 loci to determine loss of heterozygosity (LOH) at sites on 3p, 9p, 11p, 11q and 17p and amplification of cyclin D1 (CCND1). All biopsies contained DNA alterations, but only 3/8 scrapes contained unequivocal abnormalities; 4/8 contained subtle alterations that could not have been definitively identified without comparison to the paired biopsies. Overall, 22 alterations were detected in the biopsies: 8/22 were found unequivocally in the scrapes; 7/22 were identifiable in scrapes only after the biopsy alterations were defined and 7/22 were absent from scrapes. One LOH in scrape, but not biopsy, DNA was found. Discrepancies between scrapes and tumors tended to increase if multiple tumor samples were examined. We conclude that DNA alterations can be detected in scrapes of SCCHNs but may inaccurately reflect the tumor's complex genetic abnormalities. This may be due to contamination of scrapes with normal cells or to genetic heterogeneity within the tumor not represented in the scrape. Although examining scrapes of SCCHNs is an attractive technique, its clinical utility may have limitations.