Chances are, your favorite news feeds/sources are overflowing with headlines of the following variety: “Protein ‘almost as big a cancer risk as smoking“, “High-Protein Diets In Midlife Linked To High Risk Of Premature Death, Study Says“, or the ever amusing “Now we can’t eat protein. What can we eat?“
Studies like the one referenced – which was carried out by the Longevity Institute of the University of Southern California – make for juicy headlines and a great filler story for slow days in the newspaper business. Thus, it is easy to over-dramatize the findings stemming from them or to draw sweeping conclusions. These kinds of broad-stroke interpretations of nutrition studies are a favorite pet peeve of mine (more evidence in this post), so when I saw the headlines, beyond being generally interested in the topic, I wanted to look further into the study. What did they actually find? Will the results warrant an overnight switch to a vegan diet to boost our life expectancy?
What was their hypothesis?
The researchers point out that there have been a number of studies (my personal addition: including the China Study) linking (particularly animal) protein consumption to occurrences of cancer and other age-related diseases. One hypothesis is that lower protein intake also lowers the activity of growth hormone receptors, which in turn allows you to live longer. There were first studies on yeast and mice that found significant evidence for this hypothesis, and thus the researchers wanted to look into practical evidence in humans.
How did they test it?
They used an epidemiological study of 6,381 US men and women over 50 called NHANES III. These men and women were asked to describe their diet by providing information of what they had eaten the previous day in a method called “24-hour dietary recall”. Then, this data was recorded, and the people were followed over the next 18 years, in which their health status, particular diseases and incidences and causes of death were also specified.
Now, using a statistical method called Cox Proportional Hazard Models, they basically divided their observations into groups – high protein, moderate and low protein intake – and checked whether individuals in the different groups were equally likely to incur particular diseases, or to die within the 18 year follow-up window. They were also able to ‘control’ for some factors, such as percent calories from fat, from carbohydrates, or from animal protein, by trying to hold these constant between individuals and check for differences in outcomes.
What were their findings exactly?
Well, the ones widely quoted in the press are that (according to Forbes) “People from ages 50-65 who ate high-protein diets were four times more likely to die of cancer – this is in the ballpark of smoking risk – compared to people who ate low-protein diets“. This is true, though note that their 95% confidence interval – the range that, if this study were repeated, would in 95% of the cases include the value in question – reached all the way from 1.96–9.56. Thus, this finding could be anything from a twofold to tenfold increase in cancer risk – in other words, this might be an indication that their data does not allow for a very precise estimation of their result of interest. According to the study, subjects in the high-protein group also had “a 74% increase in their relative risk of all-cause mortality (HR: 1.74; 95% CI: 1.02–2.97)” compared to those in the low-protein group.
One other interesting result was that, once they controlled for the percent calories from animal protein (again – using statistical methods to only compare the individuals with similar percent calories from animal protein but different protein intake levels overall), the effect vanished – from which they conclude that “animal proteins are responsible for a significant portion of these relationships“.
What are potential limitations?
The paper is very honest in its methods and points out one major limitation from the get-go: “First, the use of a single 24 hr dietary recall followed by up to 18 years of mortality assessment has the potential of misclassifying dietary practice if the 24 hr period was not representative of a participant’s normal day. ” Basically, you observed people on one day and try to extrapolate relationships that ultimately try to predict causes of mortality. That is rather risky – what if individuals changed their diets significantly over time?
Furthermore, as this article points out, dietary recall studies are often criticized for their accuracy – would you be able to remember every morsel of food you ate yesterday? The article quotes a study that looks specifically at NHANES, saying that ““Across the 39-year history of the NHANES, EI [Employment Insurance] data on the majority of respondents (67.3% of women and 58.7% of men) was not physiologically plausible.” In plain English, if they really ate what they said they’d eaten, they would no longer be alive. In two-thirds of cases.“
I have two additional points to make. First, this study looked at a subset of the population (in particular, women and men over 50 years of age). Only in the population between 50 and 65 did they find a positive association of mortality and protein, whereas the relationship flips over the age of 65. Independently of the (rather complex) explanations they give for this reversal, I would be careful to extrapolate these results onto the entire population – it is not necessarily given that the effect would be similar for a 20-year old, for example.
Finally, what none of the articles mentioned is that these are individuals with lifestyles that cannot be controlled for. If, say, the low-protein group also eats more fruits and vegetables, never eats junk food, smokes less, exercises more, meditates more often, and [insert a multitude of other factors], how do we know that it’s the protein per se that is the cause of the cancer? Capturing all these effects is close to impossible, but without doing so the definiteness of your results can be called into question. The study gave no indication of potential associated lifestyle factors, but they are likely to exist when you study individuals.
So what should we take from the study?
Nevertheless, the results are an interesting stepping stone in figuring out potential relationships between diets, lifestyles and longevity. What I have not talked about is that they also ran more experiments on rodents to support their original hypothesis, which is rather fascinating – but still not necessarily related to their human sample. What I take from it is that the people in the group with low intake of animal protein seemed to have less health problems and live longer, and that I would thus try to emulate them – but not only by eschewing protein in isolation, but as a part of a healthy and active lifestyle in general. And that, I think, is a conclusion worth the research that went into it. No need for flashy headlines.
Did this help you figure out the debate a bit better? Do you have more questions?