Ten Studies, One Cohort, an Exposure Nobody Measured

The pattern underneath the finding

Every one of those ten papers draws on the same cohort: roughly 64,000 UK Biobank volunteers born between 1951 and 1956. None of them were ever asked how much sugar they actually ate as children. The exposure the researchers are measuring is not sugar. It is year of birth. Anyone born before September 1953 in Britain also lived through meat rationing, egg rationing, butter rationing, a different childhood infectious disease environment, different smoking prevalence around them, and seventy years of everything else that separates one birth cohort from another. Crediting the entire gap in disease rates to one of those variables, sugar specifically, is the classic fallacy of assuming that because one thing came before another, it caused it, dressed up with a regression table.

The critique that lays this out in detail comes from Christopher Snowdon, writing in The Critic on August 5, 2026. He is worth naming and worth being fair to on both sides. He writes for the Institute of Economic Affairs, a free market think tank that has long opposed sugar taxes, and that is a real interest a reader deserves to know about. It does not make his arithmetic wrong. I checked the npj Aging paper directly rather than taking his word for it, and the numbers he cites from it are accurate. He also notes something the headline percentage hides: in the dementia analysis, the raw rates across the three birth cohorts he examines are 2.6 percent, 2.5 percent and 2.1 percent, and dementia actually fell in the younger, non rationed group before statistical adjustment produced the reported protective effect in the other direction.

The test case that settles it

The strongest part of the critique is also the most checkable. A different research group used the exact same UK Biobank data to conclude that early life exposure to wartime nighttime blackouts, the mandatory lights out periods enforced across Britain during the war, reduced mental disorder risk by 21 percent and diabetes risk by 38 percent, including in people who were still fetuses at the time. Same cohort. Same method. A conclusion nobody, including the sugar rationing researchers, would actually stand behind. That is what a study design produces when it cannot separate "born a few years earlier" from whichever specific childhood exposure a researcher happens to be interested in that year.

None of this means the ten papers are fraudulent or badly intentioned. They are published in real journals, including Science, Nature Communications, Molecular Psychiatry and npj Aging, five of them share a lead author, and natural experiments are a legitimate and sometimes genuinely powerful research tool when the event they lean on isolates one clear variable. The problem here is narrower: this particular natural experiment, a national rationing policy that ended on one date, changed dozens of things about childhood at once, and sugar was never independently measured as one of them.