There's a paragraph of Methods on how they chose to deal with those: Confounding factors Age in years and smoking in three categories (never, past and current smokers) were assessed. Total cholesterol, triglyceride and high-density lipoprotein cholesterol (HDL) were measured by the automated enzymatic technique. We assessed dyslipidemia by total/HDL cholesterol ratio which was proven the best predictor of future atherosclerosis.30 Diabetes was ascertained by medical record review. Subjects were considered to have diabetes if documented diagnoses were in the medical records or if they were being treated for diabetes. To avoid confounding by affluence and high socioeconomic status, we adjusted educational levels, income and private insurance status.