Sunday, April 2, 2017

The J-curve is dead. Long live the J-curve!

 There is a resurgence of debate about the validity of the J-curve, especially as it relates to alcohol and cancer. A 2014 report determined that “alcohol use was positively associated with overall mortality, alcohol-related cancers, and violent death and injuries, but marginally to CVD/CHD” (cardiovascular disease). In other words, there was little benefit if any in terms of heart disease but a big upside risk for cancer and accidental or violent demise. Gone was the French Paradox! The J curve is dead! Or not.
Though that statement may be technically true, I looked at look at the data myself and found something different: a strong confirmation of the J-curve for overall mortality, overall cancer deaths, cardiovascular disease, and all “other causes.” This held for both men and women:

    Used under creative commons license from Ferrari P, Licaj I,Muller DC, et al. Lifetime alcohol use and overall and cause-specific mortality in the European Prospective Investigation into Cancer and nutrition (EPIC) study. BMJ Open 2014;4:e005245.

How the J-curve was hidden in plain sight

The J-curve may not be immediately obvious, however, due to some interesting choices made with how the data were presented. For one, rather than taking “nondrinking” as the reference, the authors chose "moderate drinking." When we rotate the chart and draw a reference line for nondrinkers (in purple), the pattern is easier to discern. In fact, looking at the chart for women, the J-curve is strikingly apparent for both “alcohol-related” cancers, most of which would be breast, and “other cancers”:

Another noteworthy feature is the CVD/CHD relationship, which drops significantly from nondrinkers and remains low – even at levels considered heavy drinking. The uptick of the right side of the J never appears. Considering that cardiovascular disease outweighs breast cancer as a cause of death in women by very large margin, the conclusion that alcohol had only a marginal benefit seems an almost bizarre interpretation.
Consider also that this is aggregate data for all types of drinking. Beer drinkers tended to fare worse in this population than wine drinkers, but those who drank primarily spirits were not clearly separated out. For wine drinkers, the bottom line is still a J-shaped curve.