We performed regression analyses with polynomial models (quadratic and cubic) using alcohol intake as a continuous variable to examine for possible non-linear relations between alcohol volume and the occurrence of hard coronary events. The parsimonious linear specification was adopted because the second and third order polynomial terms were not consistently significant, whatever the model. For an increase of 10 g in the amount of alcohol consumed a day, the risk of developing hard coronary events in regular drinkers was 1.04 (95% CI 0.97 to 1.12; P=0.31) in Belfast, 0.98 (95% CI 0.93 to 1.04; P=0.59) in the French centres, and 1.01 (95% CI 0.96 to 1.05; P=0.73) in the whole cohort (P=0.11 for interaction centre × alcohol volume).
