Statistical Modeling of the Incidence of Breast Cancer in NWFP, Pakistan

Salah UDDIN
Arif ULLAH
NAJMA
Muhammad IQBAL

Keywords

Logistic regression, backward elimination procedure, Brown method, Wald statistic

Abstract

Breast cancer is the most common form of cancer that affects women. It is a life threatening disease and the most common malignancy in women through out the world. In this study an effort has been made to determine the most likely risk factors of breast cancer and to select a parsimonious model of the incidence of breast cancer in women patients of the age 50 years and above in the population of North West Frontier Province (NWFP), Pakistan. The data were collected from a total of 331 women patients, arriving at Institute of Radiotherapy and Nuclear Medicine Peshawar, NWFP, Pakistan. Logistic regression model was estimated, for breast cancer patients, through backward elimination procedure. Brown tests were applied to provide an initial model for backward elimination procedure. The logistic regression model, selected through backward elimination procedure contains the factors Menopausal status (M), Reproductive status (R), and the joint effect of Diet and family History (D*H). We conclude that menopausal status; reproductive status and the joint effect of diet and family history were the important risk factors for the breast cancer. Separate models were then fitted for married and unmarried breast cancer patients. The best-selected model for married females is of factors Feeding (F), R, M, (D*H), whereas the best selected model for unmarried females has only one main factor Menopausal status. We conclude that breast feeding, reproductive status, menopausal status and the joint effect of diet and family history were the important risk factors of breast cancer in married women and the menopausal status was the important risk factor of breast cancer in unmarried women.