Association of Race and Breast Cancer Stage. METHODS

Posted by James

We conducted a retrospective cohort study. Inclusion required a diagnosis of breast cancer, and it reported to the TriHealth tumor registry from 1991-2003. There are three TriHealth hospitals in Cincinnati, OH. This is the largest tumor registry in this geographical area. This region is approximately 89% Caucasian and 11% African-American. TriHealth dedicates a full-time nurse whose exclusive responsibility is the oversight of the tumor registry data. This nurse collects data from the patient, their chart and the treating physician. Patients are followed over time for the purposes of prognosis. For each patient, we collected data on race; AJCC stage at diagnosis; and 12 potential confounding variables, including topography, morphology, laterality, age, menopausal age, smoking status, estrogen and progesterone receptor status, marital status, menopausal status, family history of breast cancer in a first-degree relative and insurance status. Due to small number of patients in some stages, AJCC breast cancer stages 0 and 1 were analyzed as separate and combined stages, and stages 2A and 2B, and stages ЗА, 3B and 3C were collapsed in the analysis into stages 2 and 3, respectively. Use of the AJCC stage for research purposes has been utilized by previous authors. Races other than Caucasian and African-American were excluded due to small numbers (N=60).

Univariate analysis using Chi-squared and t tests were performed to evaluate the prevalence of confounding variables by race. Confounding variables found to exhibit statistically significant variation between the two races were included in the regression model. Then, nominal regression analysis for dichotomous variables investigated the adjusted risk between race and each breast cancer stage while controlling for any significant potential confounding variables. Data analysis was conducted using SPSS statistical software (SPSS Inc., Chicago, IL). Utilizing a two-tailed alpha of 0.05 and a beta of 0.10, it was estimated that approximately 5,500 patients would be required for this study to have 90% power to find a significant difference between the two races for any stage if they differ by 20%.