Multivariate analysis for prognostic value was performed using the Cox-regression model.
In multivariate analysis, the prognostic value of cystatin C was compared to the values of pN-stage, UICC pTNM-stage, extranodal tumour spread and the presence of tumour emboli in the lymphatic vessels.
In a multivariate analysis the prognostic value of survivin in terms of disease-free and overall survival was found to be independent of both TNM stage and molecular parameters.
However, in multivariate analysis, its prognostic value remained significant in partial patients, including male patients, patients at younger age (≤54), patients with moderately differentiated tumors, patients with distant metastasis, patients without esophageal tumor resection and patients with palliative treatment.
For both definitions of the Her2+ subtype, multivariate analysis showed prognostic value of the Morisita index independent of and stronger than those of node, size and grade in both cohorts (Her2-amplified discovery: p = 0.027, HR = 0.44, 95%% CI: 0.21 0.91; validation: p = 0.0031, HR = 0.03, 95%% CI = 0 0.3; Table 2).
In order to have greater insight into the stability of the final Cox multivariate models for prognostic value of change in HRQoL from baseline, and thus evaluate the importance of a single variable being included as an independent factor, we conducted a bootstrap resampling technique based on 5,000 bootstrap-generated simulation datasets.
However, when using multivariate Cox analysis, the prognostic value of EpCAM expression was not statistically significant, possibly due to the limited sample size in our study.
In multivariate analysis, no independent prognostic value for p53 expression was found.
Univariate or multivariate analysis for the prognostic value of gene signatures was performed by the Cox proportional hazard-regression model.
