In these, multivariate analysis testing for differences in survival among statistically significant subgroups of FIGO stage, tumour size, depth in cervical wall, number of positive nodes, degree of lymph vascular invasion, pathological parametrial involvement, infiltration to vagina, ovarian metastases, and histological grade was carried out.
Student's t-test and the Mann-Whitney U multivariate analysis test were used for continuous variables, χ and Fisher's χ tests were used for proportions, and logistic regression was used for effects of factors on mortality.
Importantly, multivariate analysis tests controlled for diabetic therapy, diabetic polyneuropathy, serum creatinine, and levels of GGT, showed that Ang-2 remained significantly higher among diabetic subjects with macrovascular disease when compared to those without macrovascular complications.
This was shown in a univariate logistic regression model and a multivariate longitudinal analysis, testing for possible confounders.
Pairwise multivariate regression analysis, testing the association between clinical end point and DNA methylation score and/or clinical parameters, was performed by employing linear Cox Proportional Hazard models.
The models of multivariate analysis were tested for their appropriateness before deciding to interpret the results.
Using univariate and multivariate analysis we tested for factors associated with any missing data.
The multivariate analysis was tested with the Cox's regression model of survival time (Cox, 1972).
All independent variables in the multivariate analysis were tested for multicolinearity.
