Mucoepidermoid carcinoma - A clinicopathologic study of 80 patients with special reference to histological grading

We sought to review our experience with salivary mucoepidermoid carcinoma ( MEC) over two decades to confirm the validity and reproducibility of histol ogic grading and to investigate MIB-1 index as a prognosticator. Diagnosis was confirmed on 80 cases, and chart review or patient contact was achieved for 48 patients, with follow-up from 5 to 240 months (median 36 months). I mmunohistochemistry with citrate antigen retrieval for MIB-1 was performed on a subset of cases. Kaplan-Meier survival curves were generated for each stage, site, and grade according to our proposed grading system. To address the issue of grading reproducibility, 20 slides were circulated among dye observers, without prior discussion; slides were categorized as low-, inter mediate-, or high-grade according to one's "own" criteria, and then accordi ng to the AFIP criteria proposed by Goode et al.(10) Weighted kappa (K) est imates were obtained to describe the extent of agreement between pairs of r ating. The Wilcoxon signed rank test or the Friedman test as appropriate te sted variation across ratings. There was no gender predominance and a wide age range (15-86 years, median 49 years). The two most common sites were pa rotid and palate. All grade 1 MECs presented as Stage I tumors, and no fail ures were seen for this category. The local disease failure rates at 75 mon ths for grades 2 and 3 MEC were 30% and 70%, respectively. Tumor grade, sta ge, and negative margin status all correlated with disease-free survival (D FS) (p = 0.0091, 0.0002, and 0.048, respectively). The MIB index was not fo und to be predictive of grade. Regarding the reproducibility of grading, th e interobserver variation for pathologists using their "own" grading, as ex pressed by the K value, ranged from good agreement (K = 0.79) to poor (K = 0.27) (average K = 0.49). A somewhat better interobserver reproducibility w as achieved when the pathologists utilized the standardized AFIP criteria ( average K = 0.61, range 0.38-0.77). This greater agreement was also reflect ed in the Friedman test (statistical testing of intraobserver equality), wh ich indicated significant differences in using one's own grading systems (p = 0.0001) but not in applying the AFIP "standardized" grading (p = 0.33). When one's own grading was compared with the AFIP grading, there were 100 p airs of grading "events," with 46 disagreements/100 pairs. For 98% of disag reements, the AFIP grading "down-graded" tumors. This led us to reanalyze a subset of 31 patients for DFS versus grade, for our grading schema compare d with the AFIP grading. Although statistical significance was not achieved for this subset, the log rank value revealed a trend for our grading (p = 0.0993) compared with the Goode schema (p = 0.2493). This clinicopathologic analysis confirms the predictive value of tumor staging and three-tiered h istologic grading. Our grading exercise confirms that there is significant grading disparity for MEG, even among experienced ENT/oral pathologists. Th e improved reproducibility obtained when the weighted AFIP criteria were us ed speaks to the need for an accepted and easily reproducible system. Howev er, these proposed criteria have a tendency to downgrade MEG. Therefore, th e addition of other criteria (such as vascular invasion, pattern of tumor i nfiltration [i.e., small islands and individual cells vs cohesive islands]) is necessary. We propose a modified grading schema, which enhances predict ability and provides much needed reproducibility.