Διαχρονική μεταβολή επιβίωσης παρουσία συγχυτικών παραγόντων : στατιστικές μέθοδοι και εφαρμογή τους στον προχωρημένο καρκίνο του μαστού
Ξυραφάς, Αλέξανδρος Α.
The purpose of the present study was to investigate and present the statistical methodology for exploring possible improvement in survival through time for diseases which might be treated with different therapies in different time periods. Methods of Meta-analysis, Survival Analysis and Marginal Structural Models, a method of causal inference, are presented. These statistical methods were applied on real data from patients with Advanced Breast Cancer, participating in the Hellenic Cooperative Oncology Group Phase II and Phase III Clinical Trials, conducted in Greece from 1990 to today. A short description on Cancer is included here for better understanding of the disease. The main question concerned the possible alteration of survival time -as well as of time to progression (ttp)- in the presence of other confounding factors of patients with Advanced Breast Cancer, during the last fifteen years. The effect of Randomization was also explored (randomized vs non-randomized studies), while the Heterogeneity between studies was also tested using the statistical package RevMan. The analysis points out an increase to survival time -as well as to time to progression (ttp)- through calendar time for patients with Advanced Breast Cancer during the last fifteen years. In the analysis surrogate variables like “year of chemotherapy commencement” and “semester of chemotherapy commencement” were used (not taking in mind the clinical trial that they took part in). The Randomization effect was found significant in the presence of other prognostic variables in the Cox model, while lack of Heterogeneity concerning the important variables in the Cox models allowed us to use methods of Survival Analysis in order to estimate the prognostic factors.