For the past three decades or so, the conventional wisdom has been that drinking alcohol at moderate levels is good for us.

The evidence for this has come from many studies that have suggested the death rate for moderate drinkers is lower than that for non-drinkers. In other words, we thought moderate drinkers lived longer than those who didn’t drink at all.

This phenomenon has been communicated with great impact by the J-shaped curve that shows death rates fall as you move from non-drinking to moderate drinking, before rising again as drinking levels increase.

Most of us embraced these studies with enthusiasm. But the findings were probably too good to be true. The problem has always been the potential mixing of many other variables — called confounding factors — with drinking.

The concern was that non-drinkers as a group in many of these previous studies were different to moderate drinkers in many ways in addition to their drinking. Non-drinkers may have been unhealthier to begin with (hence not taking up drinking in the first place), or they may have included recovering alcoholics with poor health.

These confounding factors may have made moderate drinkers look healthier than they actually were (relative to non-drinkers) and thus have led us to associate moderate drinking with better health.