Logistic Regression Models: An Empirical Investigation in Risk Factors of Breast Cancer Among Sudanese Females

Abstract

Over the last decade the logistic regression model has become, in many fields the standard method of analysis of data concerned with describing the relationship between a binary variable and one or more explanatory variables. The goal of an analysis using logistic regression is the same as the usual linear regression model where the dependent variable is assumed to be continuous or discrete. The aim of the project described in this thesis is to investigate whether logistic regression methods are able to give improved prediction in risk factor of breast cancer. Attempts are carried to minimize the incidence of diseases, to enhance the awareness of the public regarding the factors causing these diseases. Logistic regression will be used to modeling the probability that a woman developed breast cancer as a function of age, educational level, marital status, family history of breast cancer, age at menarche, parity, contraceptive use, previous benign biopsy (PBB), age at first full term pregnancy, menopausal status, HRT, occupation, obesity (BMI), breast feeding and residence. We introduce a brief description for logistic concepts in statistics with a brief account for logistic distribution function, simple logistic regression analysis and, fitting the simple and multiple logistic regression models, testing for the significance of the coefficients and the confidence interval estimation. Also we introduce the interpretation of the fitted logistic regression model. The goal of any method is to select those variables that result in a “best” model within the scientific context of the problem, seeking for the best model that seems good in explaining the data and assessing to fit an estimated logistic regression model containing those variables. In many statistical inference procedures, we have used chi square distribution based on the likelihood ratio, Score, or Wald test statistics, goodness-of-fit statistics. This study of breast cancer among Sudanese women shows that the women with family history of breast cancer, low education, and early age at menarche are at significantly increased risk of breast cancer. However Parity, marital status and age at 1st full term pregnancy, age at menopause, contraceptive use, HRT, PBB and BMI are insignificantly protected of breast cancer. Summarizing the above results, we conclude that our study showed that the women with family history of breast cancer, low education, and early age at menarche are at significantly increased risk of breast cancer for this population. Our study failed to support the hypothesis that breast feeding, parity, age at 1st full term pregnancy, menopausal status, reduces the risk of breast cancer. The public awareness of this fatal disease must be developed, it may help in early detection of breast cancer, decreasing the mortality and ultimately increasing the probability of survival. Studies of occupational, nutritional and environmental exposures are also needed. There must be a centers of cancer all around at any center there must be division of cancer prevention and population sciences, these divisions configured into a more broadly based scientific program, with leadership and members focusing on the primary goal of identifying means to lessen the risk of cancer through the development of better links between epidemiology, nutrition, nutritional biochemistry, chemoprevention, and molecular biology.