Koos Boeve

108 Chapter 6 square test (or Fisher’s exact when appropriate) was used. Binary logistic regression analysis was used to evaluate the value of multiple variables in predicting occult nodal metastases. For interobserver agreement of cyclin D1 expression during the validation phase, the ICC between both observers (S.M.W. and K.B.) was analyzed. If not mentioned otherwise, a 2-sided p-value < 0.05 was considered significant. All statistical analyses were performed using SPSS 21.0 Statistical Software (IBM, New York, USA). Ethical justification Because the remaining tissue after the clinical diagnostic process was used, no ethical approval was required according to Dutch National Ethical Guidelines (www.federa.org) . Anonymous or coded use of leftover tissue for scientific purposes is part of the standard treatment agreement with patients in our center [21]. RESULTS Descriptive Analysis Table 2 shows descriptive IHC and FISH results. For cyclin D1, FADD and cortactin, protein expression could be scored in at least 1 core in 96%, 97%and 96%, respectively, of the tumors. In case of multiple scored cores, mean nuclear staining (%) for cyclin D1 and maximum cytoplasmic intensity for FADD and cortactin was used as overall protein expression score. Examples of IHC staining patterns, including staining pattern of normal tonsil epithelium are illustrated in Figure 1. Normal tissue showed weak/moderate staining for FADD and cortactin and some nuclear stained cells for cyclin D1 near the basal layer. For FADD and cortactin, strong positive staining was considered as overexpression. Tumors with a nuclear cyclin D1 in at least 15% of tumor cells are considered as overexpressed cyclin D1, based on ROC curve analysis. Of the scored tumors, 39% showed overexpression of cyclin D1, 19% of FADD and 15% of cortactin. To address the possibility of tumor heterogeneity, the ICC was determined for tumors with three scored cores. This revealed a very good consistency of expression of these three proteins within the tumor; cyclin D1 (0.89), FADD (0.89) and cortactin (0.90). Examples of FISH images of CCND1 are illustrated in Figure 2. FISH results were available for 88% of the tumors, 19 tumors were excluded because of lack of fluorescence signal or insufficient tumor cells. Correlation CCND1 copy number and cyclin D1 protein expression For 139 tumors, both CCND1 copy number analysis by FISH and cyclin D1 protein expression by IHC were scored. Overall, CCND1 copy number results are significantly correlated with increased nuclear cyclin D1 expression, see Supplementary Figure 1. This correlation was