Overall and relative Survival for Cancer Patients

Abstract

In this thesis, basic concepts of survival analysis such as censoring, truncation and survival functions are described. Measures of survival ( i.e overall survival, net survival and relative survival ratio) and regression models such as Cox regression for overall hazard ratios and excess hazard regression model for excess hazard ratios are discussed. Cox regression model estimates the overall risk(hazard) whereas excess mortality pro- vided by relative survival estimates the risk due to cancer. Kaplan-Meier curves are used to estimate the survival curve, to estimate regression coefficient, partial likelihood estimate is used. The main focus is to study the comparison between overall survival and relative survival ratio and apply this on non-curable colon and rectum data, derived from a research project on patients who received non-curative treatment due to an incurable disease or other reasons preventing curative surgery. The data is obtained from Cancer Registry of Norway, Norweigan patient registry and population data from Statistics Norway between 2008 and 2015. The results pro- vided by comparison shows how much change the risk of death is if death only because of cancer is considered and when other causes of death are involved. Regression analysis is done in two ways, 1. univariate analysis in which each covariate affect the analysis individually and 2. multivariate analysis in which all covariates together affect the analysis. The software R is used for analysis and to plot survival curves and other graphs used.