Table 3.

Multivariable analysis of potential factors related to DM at diagnosis or during follow-up

Bold numbers represent significant values (p < 0.05). A (multivariable) backwards Cox regression analysis was performed to produce 4 different models to predict distant metastasis, based on statistically significant variables obtained in univariate analysis. Model I included patients with complete information on the variables age, gender, primary tumour site, T- and N-stage (n = 1399). Model II included the patients of model I of whom histological differentiation grade was also known (n = 997). Model III included patients of model II of whom angioinvasion was also known (n = 473). Model IV included patients of model I who also of whom histopathological capsular extension of cervical lymph nodes was also known (n = 186)

*Early stage represents T1 and T2 tumours, advanced stage represents T3 and T4 tumours

§Only surgically treated cases were included (n = 647)

¶Only cases with surgically treated positive cervical nodes were included (n = 221)