More evidence on the J-curve

Another study out on the alcohol-health J-curve. This one uses 13 linked waves of the US National Health Interview Survey series, 1997 to 2009, to look at all-cause mortality, cancer, and cardiovascular disease (CVD) and drinking.*

Lifetime abstainers are taken as baseline, so there is no sick-quitter confound. There could be confounding if those with poor health never begin drinking, but the authors run a sensitivity test excluding those with poor medical histories.

Here's the main table. Model 2 has all the covariates; Model 1 just has demographic covariates.

What do we see here?

Former drinkers have worse characteristics than abstainers - so there's something to the sick-quitter hypothesis. But we already knew that. DiCastelnuovo & Donati showed that the J-curve isn't as deep if you exclude former drinkers.

Light (less than 3 drinks per week) and moderate (3-14 drinks per week for men, 3-7 for women) drinkers, in this study, see a reduction in all-source mortality - their relative risk is just under 0.8 where a lifetime abstainer is 1.0. All-source mortality is the only one we should really care about unless you have particular family histories that you want to factor in. But it is interesting to note that they only find increased cancer risk for heavy drinking - cancer is the one that's had most recent coverage. And note too that they find a stronger J-curve for women than for men - again, the opposite of what you might have concluded from all of the shouting about breast cancer risk.

And here's the more granular J-curve. Again, just what we'd expect.

The Herald covered the study, but neither linked to the study nor contrasted it with prior Herald stories about how a drink will make you get cancer and die.

Time covered it as well (linking to the study), and wrote:

In an accompanying editorial, researchers from the Mediterranean Neurological Institute in Italy wrote that the new findings “supported the conclusion that the J-shaped relationship between alcohol consumption and mortality risk cannot be dismissed, and should guide the formulation of public policies.”

The editorial also addresses the fact that women are sometimes advised to limit alcohol to very low levels because it’s been linked to increased breast cancer risk. While younger adults may not see substantial health benefits from moderate drinking, the editorial argues, “for most older persons, the overall benefit of light drinking, especially the reduced [cardiovascular disease] risk, clearly outweigh possible cancer risk.”

Nigel Latta was giving me heck the other day for defending the J-curve against his preferred anti-alcohol advocate, Jennie Connor. He wondered why the cancer risk isn't listed on the bottle. I'd be pretty happy for it to be - if it followed what the quote above. "For most older persons, the overall benefit of light drinking, especially the reduced cardiovascular disease risk, clearly outweigh possible cancer risk." Like he said, let's see them put that on the bottle. I suspect it would be illegal for them to do so as it's a health claim, but it's nice to think about.


* Note that the National Health Interview Survey is open data. They de-identified it, and anybody in the world can download it just by clicking the link. It is here. You cannot do that for basically any New Zealand health data. The Otago longitudinal survey is closely held by the Otago people. The Ministry of Health's Health Survey has some cross-tabs up, but you can't download the underlying data series. There are Confidentialised Unit Record Files available for the NZ Health Survey, but you have to go through a cumbersome application process to get access - and it would be near impossible for someone not based in New Zealand to get it without having a NZ-based coauthor. I love how American practice is just to de-identify things and put 'em up. I hate how New Zealand's default is "Well, maybe somebody might be able to re-identify, so we won't release anything."