15. Don’t Trust the American Heart Association (Part 2)

15. Don’t Trust the American Heart Association (Part 2)

The UK Biobank is a continuing program to collect comprehensive health data from volunteers, then track what happens to them over the following many years. Later, researchers can “mine” the data for retrospective studies.

In the UK Biobank Circulating Ketones Study, 90,000 people were surveyed. The researchers tabulated Day 1 ketone-body (KB) counts. On the basis of that single Day 1 measurement, the 90,000 were divided into three groups of 30,000 each — lowest third, medium third, and highest third by KB count. Then the researchers tracked how many heart attacks, strokes, etc., each group had in the following years. The study’s unqualified conclusion was that having more ketones in the blood is dangerous, and therefore a keto-type diet is dangerous.

By all means read the paper, but in case you are short on time or intimidated by numbers, here are some points that the American Heart Association paper authors took some pains to disguise or outright hide:

• The study was funded by the Wellcome Foundation, a major charity that was established with pharmaceutical-industry money related to GlaxoSmithKline. We can be reasonably certain that researchers who receive grants from Wellcome are conscious of where the grant money ultimately came from, and what kinds of study “findings” are likely to lubricate more research grants from Big Pharma in the future. In addition, two of the study’s authors acknowledged directly receiving money from pharmaceutical companies. So much for impartiality.
• Ketones are not an “emergency” energy-delivery system when carbohydrates aren’t available, as the paper tries to claim. The authors seem ignorant of the history of life on earth. Agriculture and carb-rich crops only arose 10,000 years ago. For 2 million years of hominid existence, hunter-gatherers would gorge on a food source (typically a prey animal or a fruit bush that they quickly stripped bare), store some body fat, then go hungry for days or weeks, burning that fat as ketones until they found the next food source. So much for the researchers’ understanding of anthropology.
• The study’s division into 3 tertiles (⅓ groups) was based on a single KB reading for each participant, at the beginning of their average 13+ years in the study. That’s like asking assorted teenagers how many calories they ate on their 13th birthday, then correlating that single calorie measurement with their future earnings, or antisocial behavior, or whatever. So much for logic.
• Hazard Ratio is a number that appears in almost every kind of clinical study. It tries to convey how doing or having something (like taking a drug or having a certain blood pressure) changes your risk of something bad happening, like dying. Usually Hazard Ratios are linear (e.g., how the amount of dosage of a new drug relates to efficacy) or bipolar (e.g., how being married, or not being married, relates to longevity). In this case, having started out by classifying the subjects into thirds according to total Ketones (a linear measurement), the researchers shifted gears and eventually reported Hazard Ratio in a scale conjured out of thin air, which appears to be some kind of exponential scale, with deliberately mangled syntax: “Total KB per 10-fold increase” (The only reasonable meaning of this bad wording is: “Per 10-Fold Increase in KB, what was the extra risk of dying?”).
• The departure from linearity is clearly a desperation move by the researchers. Here’s a close analogy. Suppose you were bent on proving pork eaters die younger than others, but between 0 and 300 grams, you found no correlation between daily pork intake and death. Not to worry. You simply change the x-axis scale to exponential: “In order to get the study conclusion we needed to make our research sponsors (chicken farmers) happy, we ten-tupled the pork to 3 kilos daily, at which daily consumption level people fortunately started dying, so now we can say with a clear conscience that this study proves pork is dangerous.” So much for intellectual honesty.
• A study of this kind is supposed to eliminate “confounding factors” as much as possible. This means that the different groups ought to be as similar as possible in terms of all relevant factors. For example, it would obviously be ridiculous to test a lung cancer drug vs. placebo when the genuine-drug group (the one the study hopes will be healthier) consists of health-conscious regular exercisers, while the placebo group consists of heavy smokers. And yet…
• The High-KB tertile had 2000 (about 7% of 30,000) more present or past smokers than the Low-KB group. Having taken pains to describe the rigorous measures the Biobank employs to ensure accurate blood sampling, the study organizers then did not correct for smoking habits by winnowing the sample sizes down from 30,000 as needed to have similar smoking history in all the three groups. So much for conscientiousness.
• The 30,000 in the High-KB group also included 1000 more persons with diagnosed diabetes. See comment above.
• In the manner of most clinical studies, the report punctiliously presents numerous comparative measurements for each of the three tertiles, such as average Blood Pressure, Hba1C, LDL, etc. But it suspiciously does NOT report the actual average ketone levels of the three tertiles. This is a ketone level study, right? In brief, the most important numbers in the entire study were simply not reported. It’s like saying you’ll divide pro basketball players by height (shortest third, middle third, tallest third), then omitting to tell the reader what the average height of each group was. An inconsequential lapse? Surely not. The researchers were avoiding revealing that almost all people eating a normal diet have very close to 0 ketones.
• Buried in Annex 1, it is finally reported that the median KB (for the entire 90,000 people in the study) was 0.063 millimoles per liter — close to 0 as we would expect. Since 8 times that, 0.5, is the minimum ketosis for persons on a keto-type diet, very few or probably none of the study’s 90,000 persons were on any kind of deliberate keto diet. Unfortunately the UK Biobank does not record the health datum “Are you on a keto diet, or not?” needed to study whether keto diets are dangerous. So the researchers were trying to draw a conclusion from the wrong population sample.
• The Higher KB levels recorded in the higher tertile were therefore not related to any conscious dieting plan but must have been involuntary, due to underlying health issues.
• The “health issues” which can plausibly raise ketone levels ever so slightly despite steady high-level carb consumption (230-260 grams for all the tertiles) are: alcoholism, cocaine and marijuana use, drug addiction, early-stage heart failure, chronic high stress levels, and general metabolic dysfunction. Do any of these sound like they might have contributed to heart attacks or mortality?

The structure of this study was not only deeply flawed from the beginning, but the report findings were simple-minded or carelessly spun out of thin air, presuming a reading audience with no capacity for analytical thought. How this paper could have been published at all is a scientific mystery all on its own. And why the Wellcome Foundation decided to waste so much money on it given its obvious design flaws is another mystery (but one you can solve for yourself).

The American Heart Association doesn’t care. It just wants to continue pontificating from its fake-scientific pulpit.

Do Not Trust the American Heart Association.