Abstract

  Background and Aim: In biomedical research, we are concerned with exploring the risk factors of disease and classifying the patients based on similarity of their responses. However, traditional methods need to consider the related assumptions that are difficult to establish in biomedical studies. In this study, an alternative analytical method was used for determining the risk factors of breast cancer and classifying patients into groups based on similarity of their features. Advances in the practical and theoretical aspects of tree-based methods were developed by Breiman et al. (1984) in their monograph on classification and regression trees. Tree-based methods have become one of the most flexible, intuitive, and powerful data analytic tools for exploring complex data structures.

  Materials and Methods: In this article, we used the data from a case-control study. The two groups included 628 women who were under 40 years old. The case group included women with positive breast cancer diagnosis, and the control group included the patients who referred to Emam Khomeiny hospital with non-hormone and non-neoplastic disease.

  Results: Of covariate selected to build the classification tree methods, the following variable determined to risk factors for breast cancer: Family history breast cancer, and ovarian cancer, irregular menstruation, none of physical activity, and low age of menarche.

  Conclusions: The simplicity of result interaction in terms of clinical or other relevant patient characteristics made trees an appealing approach in clinical and epidemiologic investigation.