Retrospective multivariate analysis of risk factors of breast cancer in a high-risk population using a logistic regression model.

Breast cancer (BC) is the most common malignancy among women in western society, and its prevalence varies widely in different Countries. Several risk factors (RF) have been reported, but none was able to give reliable epidemiological information on clinical outcome for each patient population. The aim of this study was to provide information about relationship between BC and different RF to be considered at the time of diagnosis in women residing in a high-risk region of Italy. Patients and Methods: The study included 375 BC cases (median age 60 years, range 26-89) and 1,261 population-based age-matched controls. All patients lived in the same urban area and underwent clinical breast examination spontaneously. Patients with other or previous cancer were excluded. The following parameters were considered: age, family history of BC, menstrual and reproductive factors (age at menarche, menstrual pattern, number of births and abortions, age at first birth, lactation), use of oral contraceptives and hormonal replacement therapy, smoking, alcohol consumption, occupational and sedentary activity, body mass index (BMI). Results: Univariate analysis showed significant differences (p<0.01, Student’s t-test and chi-squared test) between cases and controls in: (1) age at menarche (12.3±1.6 vs. 12.7±1.5 years) and menopause (49.5±4.1 vs. 47.3±5.3 years), (2) number of births (1.4±1.1 vs. 1.8±1.3) and age of first birth (25.2±4.3 vs. 24.4±3.6 years), (4) estrogen replacement therapy (43.9±30.3 vs. 33.7±28.1 months), (5) smoking (5.9% vs. 12.5%), and (6) alcohol abuse (5.7% vs. 2.3%). Multivariate analysis using a logistic regression model showed that only four independent parameters correlated with BC: age at menarche (years), number of births, lactation, and estrogen replacement therapy. Odds ratio for BC calculated from the observed vs. predicted values obtained using the logistic regression function was 5.05 (95%CI 3.6-7.1), while the OR of single variables was <3 (95%CI 1.51-4.32). Conclusions: The results display that the prolonged (>3 years) use of estrogen replacement therapy, is strongly related to the onset of BC. However, in this study, many classical parameters did not result useful as RF, suggesting that to correctly select the high risk population both different primary RF and other environmental and external factors should be considered for each patient.