Previous: Private Coverage Often Isn't Enough For Children
Next: More People Should Get the HPV Vaccine



View count:81,663
Last sync:2022-11-20 07:15
Nutrition studies are really, really bad a lot of the time. Sometimes researchers don't do careful work, and the systems in place don't always prevent weak research from being published. Add that to the fact that media outlets tend to overblow stories about food and cancer, and it's a recipe for research disaster.

Follow Tamar on Twitter @TamarHaspel

Related HCT episodes:
1. Low Fat, Low Carb, Whatever!

Be sure to check out our podcast!

Other Healthcare Triage Links:
1. Support the channel on Patreon:
2. Check out our Facebook page:
3. We still have merchandise available at
4. Aaron's book "The Bad Food Bible: How and Why to Eat Sinfully" is available wherever books are sold, such as Amazon:

John Green -- Executive Producer
Stan Muller -- Director, Producer
Aaron Carroll -- Writer
Mark Olsen – Art Director
Meredith Danko – Social Media


#healthcare #healthcaretriage #nutrition
How bad is nutrition science? So much worse than you think! That's the topic of this week's Healthcare Triage.


Not long ago, I was made nauseous by some news on a new nutrition study. You might have seen headlines like this one at CNN: "A Small Glass of Juice or Soda a Day is Linked to Increased Risk of Cancer, Study Finds." You might also have seen my tweet on it. I appreciate you keeping me in business, but, you know.

To the research! ...If I can call it that.

This was published in the BMJ. I mean, this is not a predatory journal here; the BMJ. Sugary drink consumption and risk of cancer: results from NutriNet-Santé prospective cohort.

This was a cohort study using data from the French NutriNet-Sante cohort. There were more than 100,000 participants followed up for a median of five years. Members of the cohort measured the consumption of sugary drinks and diet sodas through repeated 24-hour diet diaries.

The main outcome of interest was the risks of breast, prostate, and colorectal cancers, as well as overall cancer rates.

The researchers adjusted for age, sex, energy intake without alcohol, sugar intake from other dietary sources, alcohol, sodium, lipid in fruit and vegetable intakes, body mass index, height, physical activity, smoking status, the number of 24-hour dietary records, family history of cancer, education, and a host of clinical conditions.

And, they found that drinking sugary beverages was associated with an increased risk of overall cancer (hazard ratio 1.18) for each drink per day; breast cancer (ratio 1.22). Oh, and diet beverages were not, but let's just stick a pin in that and move on. 100% juice was associated with cancer with a hazard ratio of 1.12.

But, I'm not going to ask you to overthink any of this, because it's a perfect example of how nutrition research can often fail us. We find small odds ratios or hazard ratios in big cohorts that are statistically significant but not really clinically significant, and then we run with it. We see an association, which is all this is, and assume causality. We cherry-pick the study, don't discuss other findings, or talk about other diseases, and we run with it. But, this is so much worse.

The first clue is something I noticed in the abstract. See the part I highlighted there? The dietary records were designed to register participants usual consumption for 3,300 different food and beverages items. 3,300! Man, it's lucky that their main outcome of interest, beverages, was significant, right? Oh, no.

And, for the next part, I have to acknowledge Tamar Haspel, who's a James Beard-winning Washington Post columnist who writes on food-related stuff. She's amazing! And, all she asked in addition to my crediting her here was for you to follow her on Twitter, @TamarHaspel. So, go do that.

At the bottom of the abstract, you'll see a study registration. These pre-registrations exist so researchers can't change the rules after the trial to try and find significant results. They can't, for instance, make up new outcomes just because they were the ones they found to be significant. I was too lazy to go check the study registration; Tamar wasn't.

Here are the primary outcome measures that they planned to analyze in that pre-registration: All cancers, cardiovascular disease, type-2 diabetes, all other metabolic syndromes, hypertension, depression, migraines, rheumatoid arthritis, cognitive decline, quality of life, weight, and height. Basically, everything under the Sun. All diseases. And, what do they plan to run against these? All the nutrients they're collecting. All the foods they're collecting. All the dietary patterns they're collecting. And then, I guess, they're throwing all these things against the wall and whichever are positive are going to get published.

The purpose of pre-registration is to force scientists to declare their hypotheses a priori. If they register every possible outcome and declare thousands of potential independent variables, and they selectively report only on the positive results, pre-registration is useless. Why do they publish only on these select cancers? Why cancer first? Why sugary beverages? Why do you think. If they'd selected one of the likely gazillion negative findings, do you think they would have been published in the BMJ? 

Everyone's at fault here. The registration, the scientists should know better; the journal, too. But, all of this would be only in the ivory tower if the media and all of us didn't fall for it again and again and again. But, this case is special. If we don't take the tools, like pre-registration, seriously, then there's no hope. We depend on these safeguards to work, and we depend on journals to take them seriously. Do better everyone.


Hey, did you enjoy this episode? You might enjoy this other episode on low-fat and low-carb diets.

We'd like to thank our research associate, Joe Sevits, and, of course, our surgeon admiral, Sam. And, like them, you too can support the show at

And, speaking of nutrition and things that are better, buy my book, The Bad Food Bible