Koos Boeve

132 Chapter 7 Statistics Numbers (n) with corresponding percentages, mean with standard deviation (SD) and median with interquartile range (IQR) are given for respectively categorical data, normally distributed data and skewed distributed data. Associations between categorical data were tested using the Fisher’s exact test or Chi-square test. The Student’s t test (normally distribution) or Mann-Whitney U test (skewed distribution) were used for the associations between groups with continuous data. True N-status positive (N+) was defined as patients with histological proven metastasis in their SLN (pN+) Also the four patients with a negative SLNB (pN0) and a regional recurrence without local disease during their follow-up were considered as true N-status positive. Associations with true N-status were multivariable tested using a stepwise binary logistic regression model in which variables with a p-value ≤ 0.1 from an univariable analysis were included. The homogeneity of the tumor markers was tested by calculating the intraclass correlation coefficient (ICC) using only cases with three scorable TMA cores. An ICC of 0.61 (substantial correlation) or higher was used as cut-off between a homogeneous or heterogeneous expression of the marker [17,24]. In case of homogeneity (ICC > 0.61) also cases with one of two cores were included. In 88% (FADD) or more (other biomarkers) of the tumors on the TMAs were 2 or 3 cores available (supplementary data 1). The ICC analysis using cases with three available cores showed the lowest ICC of 0.66 for RAB25 and the highest of 0.86 for FADD, representing respectively a substantial and almost perfect reproducibility of measurements (Supplementary data 1). Therefore, all molecular tumor markers were considered as homogeneous and cases with one or two cores on the TMA were included for further analyses. The threshold for significant differences was a p-value ≤ 0.05. All statistical analyses were performed using IBM SPSS Statistics 23 (Statistical Package for the Social Sciences, Inc., Chicago, IL, USA). RESULTS Twenty-six patients had a positive SLN. Four patients with a negative SLNB were diagnosed with a regional recurrence during follow-up. These 30 (34%) patients were considered as true N positives. The other 57 (66%) were considered as true N negatives (Table 1). None of themolecular tumor biomarkerswas associatedwith trueN-status in the total cohort of cT1-2N0 OSCC patients with SLNB neck staging. An univariable analysis of associations between true N-status and clinico-pathological and molecular tumor biomarkers showed significant associations for pT status (p = 0.012), tumor infiltration depth (p = 0.007), perineural invasion (p = 0.018) and infiltrative tumor pattern of invasion (p = 0.001) (Table 1). In a multivariable analysis pT status (OR 3.6, 95% CI 1.1-11.3) and infiltrative tumor pattern of invasion (OR 4.4, 95% CI 1.7-11.7) were found to be independent factors for true N-status.