People whose first thousand days fell under sugar rationing had lower adult cancer rates

The analysis uses 64,761 UK Biobank participants born between 1951 and 1956, comparing those whose first 1,000 days fell under rationing with those born after it ended, and following them for cancer diagnoses and markers of biological ageing.

Hazard ratios reported for the rationed group were 0.31 for liver and intrahepatic bile duct cancer, 0.60 for rectal, 0.59 for lung, 0.48 for prostate and 0.64 for breast cancer. The abstract does not carry confidence intervals for these, so they are reported here as stated and not treated as settled magnitudes.

Leukocyte telomere length was 0.05 standard deviations longer in the rationed group, which the authors convert to roughly 2.2 fewer years of biological ageing. That conversion is their own arithmetic rather than a measured outcome.

Circulating Granzyme B was lower in the rationed group. The abstract does not give an effect size for it.

The numbers

Participants	64,761 born 1951 to 1956
Leukocyte telomere length, rationed vs not	+0.05 SD
Authors' conversion of that difference	about 2.2 fewer years of biological ageing
Liver and bile duct cancer hazard ratio	0.31, no interval given in the abstract
Prostate cancer hazard ratio	0.48, no interval given in the abstract

Why this might happen

Proposed by the authors This is the explanation the authors offer in their discussion. This study did not test it.

A note on where this comes from. The published paper is behind a publisher block, so the reasoning below is taken from the authors' own discussion in the preprint version of the same analysis, and may not match the final text.

They propose two complementary pathways rather than one. The behavioural pathway is that early restriction produced a lasting shift in what tastes normal, so the rationed cohort was still eating less sugar and eating better five decades later.

The biological pathway rests on the two markers they measured, longer leukocyte telomeres and lower circulating Granzyme B, which they read as reduced chronic immune activation and slower cellular ageing.

Neither pathway was tested against the other. The study cannot say how much of the cancer difference runs through fifty years of eating habits and how much through anything set down in infancy.

Drawn from Discussion of the earlier medRxiv version of this analysis, since the published paper is not openly reachable.

What this does not show

It does not show that anyone's childhood sugar intake was measured. Nobody in this study was observed eating sugar as a child. The exposure is the birth cohort, and everyone born under rationing is treated as restricted whether or not they were.
It does not isolate sugar. The end of rationing in 1953 sat inside a broader postwar shift in food supply. Whatever else changed on that timeline is bundled into the estimate.
The cancer numbers arrive without intervals. Several site specific hazard ratios are reported with no confidence intervals and no stated multiplicity correction in the abstract. A hazard ratio of 0.31 with an unknown interval is a headline, not a measurement.
UK Biobank is not the population. It is a healthy volunteer cohort, recruited in middle age, which skews towards the well and the willing.
The telomere finding is small. Five hundredths of a standard deviation, translated into years by the authors. The translation is a convention, not an observation.

Where this leaves us

The idea that very early sugar exposure leaves a durable mark now has a large natural experiment behind it rather than only animal work and small cohorts, though the effect may be running through a lifetime of eating habits rather than anything fixed in infancy.

British people born between 1951 and 1956 who later volunteered for UK Biobank, compared by which side of September 1953 their first 1,000 days fell on.

A cohort with measured childhood intake, rather than a policy date standing in for it, would separate the infancy exposure from the fifty years of eating that followed.

Caveats worth holding

Site specific hazard ratios are reported in the abstract without confidence intervals, so they are unconfirmed as magnitudes and are treated that way here.
Dietary data were collected once, when participants were in their late fifties, so no lifetime intake trajectory exists.
Exposure is assigned by birth date, not by measured intake, which makes this an intention to treat comparison between birth cohorts.
An earlier version of this analysis was posted as a medRxiv preprint in November 2025 and circulated then. Coverage dated 2025 refers to the version that had not been peer reviewed.