Do plant or animal proteins contribute to risk of dying?
Does eating animal or plant-based proteins affect our risk of dying?
A new study published in JAMA seems to suggest that it does. But does this study add anything to the field of nutritional science? Or is it yet another example of healthy user bias and low hazard ratios, making the study’s true contribution more akin to statistical noise?
The timing is interesting, as we recently wrote about two other nutritional epidemiology studies that are clear examples of healthy user bias rather than meaningful additions to scientific health literature.
Despite the enthusiasm for the new JAMA article, I’m afraid it is more of the same.
The paper starts off acknowledging that higher-protein diets can lead to weight and fat mass loss, and including protein to replace carbs improves BP, lipids, and glucose. But then the authors pose the question of whether we have data to support choosing plant protein sources over animal protein.
To answer this question, the researchers looked at the data from the NIH-AARP Diet and Health Study, which includes over 400,000 people with a mean age of 62 at enrollment. The participants were enrolled in 1995 in California, Florida, Louisiana, New Jersey, North Carolina, and Pennsylvania, as well as Atlanta and Detroit. To assess their dietary habits, participants filled out an initial food frequency questionnaire.
Mortality data was collected for 16 years, but importantly, there was no follow-up assessment of food intake.
That’s right. No follow-up assessment.
This means that the authors are basing their supposedly important conclusions on a one-time food recall questionnaire with no follow up for 16 years! When it comes to data collection, this is about as bad as it gets.
But let’s discuss what they found anyway.
The simple conclusion is that eating more plant-based protein was associated with a lower risk of dying from various causes. That is what the headlines say and that is what many will think this study proves.
But guess what else eating plant-protein correlated with —
- Higher educational level
- Lower total energy intake (difference of over 500 calories per day!)
- Higher fiber fruit and vegetable intake
- More physical activity
- More likely to use vitamins
- Less likely to smoke (29% vs 7%!) or drink alcohol
- Less likely to report their health as poor or fair.
Hopefully you get the picture that this is a perfect example of healthy-user bias. In the 1990s, doctors, dietitians, and public health campaigns were warning us about eating eggs and meat. Those who continued to eat eggs and meat likely cared less about their health and had less healthy lifestyles. These data confirm that hypothesis.
But what about the mortality differences? Were they so robust that they might overcome the question of healthy user bias?
Not even close.
Overall mortality had an HR 0.95 for men and women for every standard deviation of more plant-based protein. The difference was not significant for cancer, and there was an HR of 0.97 for heart disease in men but it wasn’t significant for women. This is a minimal association that further highlights the weak quality of the data.
And here are some of my favorite points.
More plant protein intake was associated with a decrease in death from injury and accident with an HR of 0.90 for men, an even stronger association than heart disease! If someone can explain the physiologic mechanism of how eating animal protein reduces accidents and injuries, I would love to hear it. Because to me, this is one of the best examples of healthy user bias I have seen in a while; this finding forces us to seriously question the rest of the data.
Why don’t the headlines proclaim, “eating more plant protein reduced accidents and injuries”?
Because it doesn’t fit the narrative. It doesn’t make sense. It is a result of statistical noise, not science.
The only problem is that it isn’t any different for heart disease, cancer stroke and overall death.
The authors go further in the assessment, claiming that substituting 3% of dietary intake from plant protein for animal sources will result in significant mortality improvements. Keep in mind, there was no substitution here. This was not a substitution trial. This is a mathematical equation based on poor data with razor-thin hazard ratios.
When it comes to data quality, this is about as weak as it can get.
One last note: white meat protein did not show a significant mortality difference. Why might this be?
In addition to eating less meat and eggs, the other predominate public health message in the 90s was to eat more chicken. This is yet another likely example of healthy user bias.
I have to stop and ask myself, have I been overly critical of this study? Could I be missing something profound that should shape how we choose to eat?
I don’t think so. In fact, I believe it is quite the opposite. Studies like this one can fuel the movement that believes that we all need to eat less meat, that our kids should have vegetarian school lunches and government food programs should emphasize plant proteins.
The bottom line is the quality of this data is not strong enough to support those conclusions. That is the message we need to shout from the roof tops.
It may be an interesting study that can spur further research, including randomized controlled trials, but by itself, it should not inform how we choose to eat. The more we realize that, the better the entire field of nutritional science will be.
Thanks for reading,
Bret Scher, MD FACC
Start your FREE 30-day trial!
Get delicious recipes, amazing meal plans, video courses, health guides, and weight loss advice from doctors, dietitians, and other experts.Join now