*Originally posted to LinkedIn May 2026*
In a recent JAMA Network article reassessing evidence of alcohol harm over time, the authors concluded that there was no significant benefit to moderate drinking; they effectively eliminated the J-curve. One reason the J-curve disappears is that the researchers treated self-reported abstainers as unreliable and removed them from the dataset. They used those who reported very low levels of consumption as a proxy for abstainers.
That seems like a reasonable adjustment. But if the issue is the health impact of abstinence versus moderate drinking, removing abstainers from the analysis effectively undermines the purpose of the study.
It does, however, support the narrative.
The study was published in a highly respected journal and widely circulated (at times triumphally) as evidence that alcohol has no positive health value.
This illustrates the power of narrative in steering research. When certain elements are accepted–for example, that alcohol is harmful–the analysis can begin to move in that direction. Adjusting the analysis to address results that do not fit expectations (often referred to as “torturing the data”) begin to seem reasonable.
The result is often a set of findings that appear more consistent and more certain than the underlying data would suggest. It creates the impression of consensus in a field where there are, in fact, many complicating variables.
But it tells a clear story.
At that point, it becomes harder to tell where the evidence ends and the interpretation begins. The story doesn’t just emerge from the data; it starts to organize it.
The challenge is not just evaluating the evidence, but understanding how it is being shaped.
Leave a Reply