Reminder: no Class until after Labor Day.
Headlines blared that children growing up when the UK rationed sugar, a severe act resulting from one of their many wars, caused these sugar-deprived children to lose their minds to dementia at lower rates than those children whose sweet-tooths (sweet teeth?) were indulged.
There are at least two peer-reviewed papers making this claim. The first is “Association of Sugar Restriction in Utero Through Age 2 Years on Dementia Risk Later in Life” by Jiazhen Zheng and others in Neurology. And the second is “Early-life sugar rationing, brain aging, and long-term neurodegenerative and psychiatric health outcomes: A population-based natural experiment study” by Xingji Lian and others in npj (sic) Aging.
From the Zheng Abstract:
We analyzed UK Biobank participants born around the end of UK sugar rationing. Exposure was classified as rationing in utero only, in utero plus the first year of life, in utero plus 1–2 years of life, or no exposure. Incident dementia was identified from linked ICD-10 records…
Among 64,737 participants included in the analysis, the mean age at recruitment was 54.6 years, and 56.4% were women; 40,963 were exposed to sugar rationing during fetal and/or early-childhood life, and 23,774 were unexposed. Compared with unexposed individuals, sugar rationing in utero plus the first year of life was associated with lower hazards of all-cause dementia (HR, 0.79; 95% CI 0.66–0.94) and AD (HR, 0.77; 95% CI 0.59–1.00)…
I’ll explain that “hazards” business below. We can forgive the tortured English, which is not the fault of the authors being Chinese, but happens because that is how epidemiologists speak. They do say things like “exposed to not eating sugar.” Let it pass.
The Lian paper had much the same set up.
Now you may be curious, as I was, how they found their comparison group. It couldn’t have been the same kids at the same time sugar rationing was in place. And indeed, they were born later. The unsweetened kids were born late 1951 to 1954, while the sugar-available kids were born after rationing ending in July 1954 to early 1956.
How much sugar did mom, and then child, eat during sugar rationing? Presumably very little. But how much sugar did mom, and then child, eat after rationing ended?
In neither case was “exposure to” sugar, nor, even stronger, how much sugar was eaten, measured, not for moms and not for kids. Regular readers will recognize we have entered the realm of the Epidemiologist Fallacy. This is when an Epidemiologist announces “X causes Y”, but where X is never measured, and the cause is “confirmed” with wee P-values.
Did somebody say wee Ps? Sing It!
Did you sing along? I know you did.
Here’s the Lian paper results, which are pretty much the same as the Zheng paper. These are “hazard” rates for Dementia, Alzheimer’s, and so forth, for the different groups, rationed and unrationed:
You can go to the site linked above for clearer pictures.
“Hazards” are not simple counts or fractions. They do not provide any table or data on something as simple as counts of the number of ex-kids coming down with dementia or not, and their sugar group. Which would be useful for the causal reader to see if, maybe, something interesting was going on. That would have been too easy and not sciencey enough. Instead, they did models. We see the results of models, not data.
They have several models and lots of different measures, besides just sugar groups, all of which are fed into models, and out pops those “hazard” rates, which are defined to be “instantaneous” probabilities of coming down with the named maladies.
Notice first that the difference (see the Y-axis) between groups is trivial, minuscule, even. And with no indication of the uncertainty due to the model predictions. The come the P-values. And two of them aren’t even wee: they are greater than the magic number.
Notice second the models themselves. Feel free to only scan this.
In other words, lots of stuff was tossed into regressions and we get the picture of the Epidemiologist Fallacy in action.
“Wait a minute, Briggs. This is all plausible. Sugar is well known to be bad for you. There could be something to this.”
Whether sugar is bad or good is not something these researchers studied. They have no idea how much sugar any mom or kid ate. Neither do I know.
“Yeah, but you have to admit the correlation is curious. The kids are almost the same age, and the ones in the rationing group did better.”
One, we don’t know that they did better, We only know their models showed, in some cases, and only in some cases, very slight improvements. And two—well, two is the big one. Can you guess what it was? It’s nothing weird or statistical. It’s not some math trick. Think for just a moment and see if you can come up with it.
Think! Don’t peek.
Don’t peek. Think!
The wee Ps, when they got them from their models, were confirming correlations, yes?
“That’s what the theory says.”
The theory is wrong, but let that pass. What else happened in July 1954 besides the end of sugar rationing?
“Uh oh.”
Right, uh oh. All rationing ended, not just sugar. Sugar was not the only thing rationed in Merrie Olde in those years. So were butter and meats, which included bacon and ham, those once British staples. They, too, were restored in 1954.
Which means an alternate explanation is that eating butter, bacon, and ham ward off dementia! And that sugar does nothing one way or the other.
The P-values would be the same. The correlation is the same. The cohorts are the same. The conclusions must therefore be the same.
“But you said we can’t draw causal conclusions from correlations.”
Right, we can’t. The real reason is we have two slightly different samples, and that the causes and conditions of things like dementia are many and various, and we would expect to see differences in rates because of this. (Even though here we haven’t even seen the rates.)
“I guess the real lesson is not to pay so much attention to epidemiology.”
You got it.
Video
All my work is free and user-supported. Here’s how to help:
- Subscribe at Substack (paid or free!)
- Cash App: $WilliamMBriggs
- Zelle: use email: matt@wmbriggs.com
- Buy me a coffee
- Paypal
- Other credit card subscription or single donations
- Hire me
- Subscribe at YouTube
- PASS POSTS ON TO OTHERS
Discover more from William M. Briggs
Subscribe to get the latest posts sent to your email.
It’s also not coincidental that the control group is, by the standards of the study, entirely younger than the test group. Dementia gets worse and more noticeable as you age.