"Although age and certain cardiovascular risk factors such as diabetes and smoking were included in their multivariate models, residual confounding may still explain the associations seen." . . . . .