[1140] Frozen Section Analysis of Head and Neck Tumor Resection Margins: Methods, Trends, and Error Rates at a Large Academic Institution

SM Olson, JS Lewis, M Hussaini. Washington University in Saint Louis, Saint Louis, MO

Background: Frozen section (FS) analysis is an essential tool for assessing tumor resection margin status intra-operatively. Most institutions evaluate surgical defect edge tissue provided by the surgeon after the main lesion is removed. This will be an ever more utilized method, particularly as transoral laser microsurgery (TLM) is more frequently utilized, for which defect margin specimens are the only choice. There have been very few large studies evaluating error rates for this type of margin assessment or that analyze methodology.
Design: All head and neck cases with margins evaluated at FS were identified by database search. Margins from all of these cases were evaluated in the routine manner by all faculty members who cover the FS service with two H&E levels at FS and one permanent section on the remaining tissue. All cases were reviewed retrospectively for the study. Finally, a prospective one-year evaluation of obtaining an additional third, deeper H&E FS level has been undertaken.
Results: The preliminary data presented here includes all head and neck tumor cases with margins evaluated by FS during 2007 at Barnes-Jewish Hospital. 1107 total margin specimens were evaluated, 117 showing invasive tumor or dysplasia. There were 32 errors in total (2.9% of all margins). Of these, 16 were sampling errors (FS slides negative, permanent slide positive 1.5% of all margins; 14% of all margins with neoplasia; sensitivity of 86%) and 16 interpretation errors (diagnostic tissue present on frozens which was mis-called 1.5% of all margins). There were 17 bone margins (marrow scrapings), 4 of which had neoplasia. None of these had errors. Review of the main specimens showed that all corresponding bone margins correlated with the frozen results.
Conclusions: The baseline error rate is low (2.9%) which is comparable to published data, but could be improved, particularly when considering that our current method of 2 FS levels had an 86% sensitivity for the lesion (when just considering specimens with neoplasia in them). Sampling errors constitute a significant portion of all errors (50%). This may be decreased by examination of an additional deeper level on FS. Once the prospective data is gathered, the utility of this method change can be evaluated relative to retrospective data. Bone margins evaluated by scraping have strong predictive value as well.
Category: Head & Neck

Monday, March 9, 2009 1:00 PM

Poster Session II # 174, Monday Afternoon