The BRFSS is a standardized telephone survey that annually assesses key behavioral risk factors and chronic diseases among adults aged 18 years or older in all U.S. states, District of Columbia, and territories. The median cooperation rate among states was 75.0% in 2009 (5
). BRFSS data have consistently provided valid and reliable estimates compared with national household surveys (6
Diabetes status was ascertained by asking participants, “Have you ever been told by a doctor that you have diabetes?” Responses were coded as “yes,” “yes, but female told only during pregnancy,” or “no.” Gestational diabetes was coded as “no” diabetes.
Cancer status was ascertained by asking participants, “Have you ever been told by a doctor, nurse, or other health professional that you have cancer?” Responses were coded as “yes,” or “no.” For those who answered “yes” to this question, the following questions were asked, “With your most recent diagnoses of cancer, what type of cancer was it?” The survey included 10 major cancer sites/tracts and 29 cancer types: breast cancer, male reproductive tract (prostate cancer, testicular cancer), head/neck (head and neck cancer, oral cancer, pharyngeal cancer, thyroid cancer), gastrointestinal tract (colon cancer, esophageal cancer, liver cancer, pancreatic cancer, rectal cancer, stomach cancer), leukemia/lymphoma (Hodgkin’s lymphoma, leukemia, non–Hodgkin’s lymphoma), skin (melanoma, other skin cancer), lung cancer, urinary tract (urinary bladder cancer, kidney cancer), and other sites (heart cancer, bone cancer, brain cancer, neuroblastoma, other).
Demographic characteristics included sex, age (year), and race/ethnicity (non–Hispanic white, non–Hispanic black, Hispanic, and other). Health-related risk factors included health insurance coverage (none vs. any type of health insurance), smoking status (current smoker, former smoker, and never smoked), heavy drinking (consuming more than two drinks per day among men and more than one drink per day among women), BMI (kg/m2), and physical inactivity.
We estimated the crude prevalence for cancer of all sites and specific types of cancer according to diabetes status. We estimated the prevalence ratios and their 95% CIs using log-linear models with a robust error variance estimator (8
). The equality in the prevalence estimates was tested with a two-sample t
test. We considered results with a two-tailed P
value <0.05 or the 95% CI of a prevalence ratio estimate that did not include 1 to be statistically significant.
We conducted all analyses using SUDAAN 9.0 software (Research Triangle Institute, Research Triangle Park, NC) to account for the complex sampling design.