The glass of red that was never medicine
For thirty years the data seemed to show that light drinkers outlived teetotallers. Then somebody asked who the teetotallers actually were.
This is a story about a flaw in a comparison group, and how long a comfortable finding can survive on it.
The J-curve is one of the most repeated ideas in popular health. Plot alcohol against mortality and the line dips before it climbs: light drinkers appeared to do better than people who drank nothing at all. A glass of red, went the conclusion, is good for the heart.
The problem was never the arithmetic. It was the question of who ends up in the "does not drink" column.
Abstainer bias
People stop drinking for reasons. Illness is a common one. So the group of non-drinkers in a cohort study quietly fills up with former heavy drinkers, people on medication, and people whose health already told them to stop.
Compare a light drinker with that group and light drinking looks protective. It is not doing anything. The reference group is sick.
A 2023 systematic review and meta-analysis in JAMA Network Open tested this directly. It pooled 107 cohort studies, nearly 4.8 million participants and over 425,000 deaths, then applied adjustments for abstainer bias and for how the reference group was defined.
Once the comparison was made fairly, the benefit of light drinking disappeared.
Occasional drinking and low-volume drinking showed no significant mortality benefit. The increased risk appeared at roughly more than three drinks a day for men and two for women. The J had been an artefact of who was standing at the bottom of it.
Being fair to the other side
This is not unanimous, and we will not pretend it is. Some researchers argue the corrections themselves introduce bias, and the disagreement in the literature is genuine and ongoing. What has changed is the default: the protective effect is no longer the safe assumption, it is the claim that has to be defended.
Why this is on a site like ours
Because it is the cleanest example we know of a pattern this whole journal exists to point at.
A finding that flattered a habit people already had, that suited an industry with money to spend, survived for decades on a methodological flaw that was not hidden — it was published, discussed, and simply less convenient than the conclusion.
That is the same shape as "fragrance is a trade secret", and as a bag of bath salts sold on a mineral claim it cannot support. Not lies. Comfortable readings, left uncorrected because nobody with a budget wanted to correct them.
What we do
- We will not call wine healthy. Enjoyable is a different word, and an honest one.
- Never in a sauna, and never near cold water. Those combinations are how people actually get hurt.
- And we notice the pattern. When a finding is unusually convenient for everybody selling something, that is the moment to read the methods section.
