FeaturedScience

Bad Science Journalism: “Eat Nuts, Live Longer”

Perhaps you’ve seen the latest in nut-related research tearing up your news feed? Apparently, “scientists say” that eating nuts will make you live longer. As a Licensed Scientician myself, I was doubtful of the over-simplified claims presented by most news articles, so I decided to take a closer look at the primary research.

I’m specifically picking on this article from Time, but most other articles say the same thing (and they’re likely just regurgitations of a press release). According to the article, the study is based on a huge sample of people (around 120,000), but they’re all health professionals (nurses and doctors), so it’s not exactly representative of the normal population. That aside, there are other confounding issues.

Those who reported regularly consuming nuts were less likely to die from a variety of diseases, most significantly cancer, heart disease and respiratory diseases.

Is this article actually saying that people who eat nuts don’t die from the most common causes of death? But they did die, right? Or did they live forever? Clearly, the takeaway message is that nuts prevent death.

People who ate nuts seven or more times a week, in fact, enjoyed a 20% lower death rate after four years than individuals who did not eat nuts. Nut eaters also tended to be leaner, more physically active, and non smokers.

Because they ate nuts, right? Do Peanut M&M’s count? Or do you think that people who like nuts also tend to lead a healthier lifestyle?

“Anybody want a peanut?” (source)

Prior studies found similar connections between nuts and longer life, but the large size of this study gives the association more support. How many nuts does it take to extend lifespan? That’s not clear, and the scientists say that the findings don’t imply any cause and effect relationship between nuts and later death, but the correlation is worth investigating further.

“How many nuts does it take to extend lifespan????” (I thought this paragraph was the funniest.) Ten a day? Twenty? If I eat 100 nuts a day, will I live longer than everyone else? Or will I die from a blocked bowel? I know “scientists say” that there isn’t necessarily a cause and effect relationship, but c’mon, JUST TELL ME HOW MANY NUTS TO EAT TO ATTAIN IMMORTALITY.

The research was funded by the National Institutes of Health and the International Tree Nut Council Nutrition Research & Education Foundation. But the Tree Nut Council had no part in deigning the study or reporting results. [sic]

Ah, so the authors are in the pockets of Big Nut, you say?

Now onto the actual research paper (as much fun as it was to read the bad science journalism). Most articles that mentioned this study neglected to cite the original article, so I had to do some digging, but I found the article (for free!) in the New England Journal of Medicine.

Scientific journals are written for other scientists, so I get that as a journalist (who may not necessarily have taken a science class) it’s hard to verify the evidence in the article and it’s easier to just reword whatever press release you receive. (The current state of journalism is a whole other topic that I’m not really going to discuss beyond that.) However, most people don’t have the resources (or the drive) to look up scientific articles (if they are even available to the public), so they rely on journalists to accurately describe legitimate studies. But instead, we get science journalism that is constantly contradicting itself (e.g. Egg yolks are bad!! Cholesterol! No wait, egg yolks are healthy cholesterol! Scratch that, egg consumption is linked to disease!!!) and giving bad advice. The takeaway message (promoted by the Nut Council) from this study is that eating nuts leads to longer life, but the truth is much more complicated than that.

If you actually look at the data, there are a few surprises. Here is Table 1:

table1

According to the data, there is definitely a trend of people who “eat nuts” also living longer (although it’s not a terribly significant difference, it’s a mean of 57.5 vs 61.8 years and the error bars overlap). Although, the first thing I noticed about this table was that it only lists means–and I would be much more interested to know the median age of death instead of the mean, because the median is less likely to be skewed by outliers. (And this is why bio majors are required to take Statistics in college too.)

If you look at the other stats in Table 1, people who eat nuts also drink twice as much alcohol, are more likely to take multivitamins, exercise more, are less likely to smoke, and eat more fruits & veggies. Wait a second, scroll back to that first item…OMG you guys! NUT CONSUMPTION MAKES YOU DRINK!!! (Is that why they’re always at bars?) Or maybe, in fact, people who eat nuts just tend to do other things associated with healthy lifestyles? (Except that drinking thing. Oh wait, maybe that adds antioxidents, right? Leave my evening wine alone.)

nuttable3snip

Table 3 lists the causes of death. But it lists it in absolute numbers instead of relative percentages. I’ve taken the time to crunch the numbers below to compare the percentages of causes of death within Nut Consumption categories.

[Ed Note: Table replaced on 27Nov2013 updated math–I had incorrectly added “all causes” to the rest of the numbers and didn’t account for the split in cardiovascular disease. The numbers have been updated in the paragraph below too. Thanks to commenters for pointing this out!]

The people who ate nuts less than once per week had a 38.0% chance of dying of cancer. The people who ate nuts more than five times a week had a …drumroll please… 36.2% chance of dying of cancer. WOW, what a difference. I can totally see why the author of that Time article wrote that people who are nuts were “less likely to die from a variety of diseases, most significantly cancer, heart disease and respiratory diseases.” (Although that doesn’t hold up for the cardiovascular claim.) Interestingly, people who ate no nuts were no more or less likely to die of a certain disease than the group who ate ALL THE NUTS. (That probably has something to do with the fact that people who don’t eat nuts may be more likely to have nut allergies–which doesn’t mean that they’re not otherwise leading a healthy lifestyle.) 

My point is: this study seems awfully nut-focused when in fact there are a lot of different reasons that people who are disposed to eat more nuts may “live longer” (although the difference may be statistically significant, the error bars overlap too much in this area for my satisfaction).

The Tree Nut Council may not have had anything to do with this study other than funding it, but they sure as shit have a lot to do with the press releases being thrown around touting the benefits of nuts (instead of, say, just switching to a “healthier lifestyle”). The authors mention following the Mediterranean Diet but then are quick to mention that nuts are a part of that recommended diet (but gloss over the fact that the diet includes overall healthier fats and nutrients).

Aw man, I so wanted the Scientists to tell me how many pecan pies I could eat this Thanksgiving! (Or even, chocolate pecan pie? Because that has the benefit of added antioxidents, right?)

Featured Image

Mary

Mary Brock works as an Immunology scientist by day and takes care of a pink-loving princess child by night. She likes cloudy days, crafting, cooking, and Fall weather in New England.

Related Articles

42 Comments

  1. “The people who ate nuts less than once per week had a 45.1% chance of dying of cancer. The people who ate nuts more than five times a week had a …drumroll please… 44.2% chance of dying of cancer.”

    There’s a glitch in this statement. Those percentages are of deaths from all causes in your chart. Cancer went from 17.1 to 16.0.

    Otherwise, this is a great send-up of “science journalism” with the “science” part removed. Reading journal articles actually isn’t hard and it doesn’t require a science degree, it just requires some basic knowledge and the willingness to practice. Very similar to reading judicial opinions, actually, especially SCOTUS. You’d think that judicial opinions would be full of jargon and unreadable by lay persons, but it isn’t true, and it isn’t true for science journals either. All you need is a willingness to do it, and journalists sadly are often not willing on either count.

    1. Thanks for pointing that out–it’s fixed! I should really fire my proofreader ;)

      The only thing that makes reading a journal article hard is the fact that they’re usually behind a paywall, unless you know someone who works in science and has access. But this article was free to the public, so the journalists really had no excuse. It really irked me that many of them didn’t even cite the article or list the name of it or anything, so I had to do some Google-fu to find it.

  2. What they don’t say is that people who eat lots of nuts are 47 times more likely to get run over by a bus or eaten by a bear* than people who don’t eat nuts.

    Like 86% of all statistics, these numbers are made up on the spot.

    * True, made up fact: Of the 4.7 people eaten by bears in the US in the last decade, only 0.1 of them didn’t ever eat nuts.

  3. Is this all self report data? I can’t imagine it’s not. If so, “Health professionals who say they eat nuts might or might not live slightly longer than those who don’t” would be a more accurate headline.

  4. You misread the table. What is tabulated is age, not age at death. We don’t know if people who eat nuts live longer because virtually all of them are still alive.

    The demographics of the “eats many nuts” and the “eats nuts less than once per week” are different, such that it is only the adjusted death rate that shows a decrease in the “eats many nuts” group.

    The raw death rate is (5,203+10,287)/(521,763+1,202,005) = 0.90% vs. (1746)/(164,042) = 1.06%

    There may be (and likely are) unknown differences between the different groups that account for the differential death rates. The assumption of the paper is that all differences are either known (and can be corrected for by using the “adjusted death rates), or are equal in the different groups so they exactly balance out, except for the difference in nut consumption.

    Cause of death and age of death and death rate are independent variables. That there is no great change in the ratios of individual causes of death does not mean that there is not a great change in the age of death or the death rate.

    The “eats many nuts” group is a lot smaller (~1/10 the size) than the “eats nuts less than once per week” group. So the “eats few nuts” group likely has more rare causes of death.

  5. OK, maybe I’m being thick, but how do you get 44.2% “cause of death percentage” (all causes/never) from 5203 deaths with N=521,763? Percent of what?

    1. The N is total number of human years (the lifespans of each goup added up). It says it at the bottom of the table, I believe.

      1. You misreported what the table said, and your percentages are way off. N= 521,763 is the number of people in that group, not the number of years.

        I think you are misreading what the age in the table means. I presume that means the age of the person at the study end or study beginning, or average age during the study. I am pretty sure it does not mean the age at which the people died. Given that most all of the study population is still alive, age at death cannot be determined.

        “The people who ate nuts less than once per week had a 17.1% chance of dying of cancer. The people who ate nuts more than five times a week had a …drumroll please… 16.0% chance of dying of cancer.”

        That is not what the table says.

        There were (521,763+1,202,005) people who ate nuts less than once per week. (1,883 + 3,904) died of cancer. That is (5,787)/(1,723,768) = 0.336%. People who ate nuts 7 times or more were 632/164,402 = 0.384%.

        What the table says, is that if you are going to die, then if you eat a lot of nuts the chances you will die of cancer are 632/1746 = 36.20% and not 5,787/(5,203+10,287) = 37.36% if you eat few nuts.

        1. No, you misread the table. Read the description of Table 3 in the actual study, all the way through to the end. “N denotes number of person-years.” NOT the number of people in the group. Person-Years.

          There are only ~120k people in the study period, so how could 1.2 million be in ONE GROUP? The answer is: it’s not people. It’s person-years. Let me google that definition for you: “The product of the number of years times the number of members of a population who have been affected by a certain condition (years of treatment with a given drug).”

      2. Ah, person-years, gotcha, now that I read the original table – conveniently located right at the end of the fine print at the bottom! (Kicks self, ALWAYS read the primary source..)
        The other thing that bamboozled me was the term “all causes” which is clearly not what it seems (a grand total) but rather another category of “other causes”.

        1. I agree…the authors could’ve written this paper a bit more clearly. I guess it didn’t matter though since journalists were all “NUTS EQUALS LIFE”.

        2. Weirdly, if you add up the number of deaths in a column except for “all causes” and “other causes”, you get close to the amount for “all causes” but not quite. I really don’t like this table, period. (I was going to say “it doesn’t add up” but that’s a really obvious joke ;)

          1. LOL likewise I was going to say something about Big Nuts in pockets but I don’t want to touch that

  6. This article was a breath of fresh air. I don’t know why journalists haven’t figured out that associational studies should be greeted with skepticism!

  7. Dear Mary,

    Thank you for your article. I try to make sense of all the conflicting information on nutrition and this helps. It was also a funny read. There are still things I don’t understand and it’s a lack of education on my part, but how do you get to 44.2 percent for all causes. I’ve tried to divide and multiply various variables, but I’m stuck. Could you please point me to the right direction. I would mean a lot to me.

    Thanks, Janet

    1. In table 3, divide the number of deaths for one parameter by the sum of the total number of deaths in the corresponding column (and multiply by 100 for percent). Don’t be confused by the N at the top if the column–that is not the number of people in the column but rather the person-years.

      1. Oh thank you and with your help I could reproduce the percentages. Feeling real smart now ;-). Can I bother you once more? You mentioned ‘error bars overlap’ and I went looking for them, because I want to learn. I guess you need a better understanding of statistics to get that comment. But if it’s something relatively simple, could you elaborate on that?

        1. The “error bars” are the standard deviations. If you look at table 1, the ages are listed at mean +/- standard deviation (SD). If you subtract the SD from the mean, you get the lower end of the range, and if you add you get the upper end. Comparing the ranges, you’ll find a lot of overlap. If two means are statistically different from each other, the ranges shouldn’t have any overlap.

  8. Or even, chocolate pecan pie? Because that has the benefit of added antioxidents, right?

    Definitely chocolate pecan pie. Chocolate is made from beans, and beans are a vegetable. Therefore, chocolate is a vegetable and must be good for you. That’s science, that is.

  9. I’d like to see what the breakdown of income levels is for the subjects as well as weight and eating habits. If peanuts are excluded from this study then income seems a likely confounder because tree nuts are not an inexpensive snack and higher income usually relates to better health care and disease outcomes.

    1. That’s also a great point. Wait, maybe eating nuts makes you more successful? Don’t mention that to the nut council ;)

  10. I think I’ve figured out the deal with mortality numbers. It’s not exactly clear from Table 3, but they look at cardiovascular diseases overall and also look at two specific cardiovascular diseases, heart disease and stroke. It’s a bit clearer in Table S1 in the Supplementary Appendix (http://www.nejm.org/doi/suppl/10.1056/NEJMoa1307352/suppl_file/nejmoa1307352_appendix.pdf). So if we just add up all the numbers for specific-cause mortality in the column, we end up with a number greater than the all-cause mortality because we double-count heart disease and stroke. Using just the numbers from overall cardiovascular disease gives the expected answer. For example, in the Never nut consumption group, 1883 + 1355 + 376 + 327 + 69 + 80 + 79 + 1034 = 5203, the same number as the all-cause number of deaths.

    Also, with regards to Table 1, all the characteristics listed are potential confounders that the study authors tried to control for in the analysis. If you look at the footnote in Table 3, it says that the analyses were adjusted for all the factors in Table 1, meaning that the authors already took these factors into account. For example, age is not being used as an outcome. The authors are NOT trying to say that because those in higher nut consumption groups are older, that supports the conclusion that eating nuts causes people to live longer. Rather, they are showing how the distribution of age may be different between the groups, so they need to make sure it will not confound their analysis.

    So, I don’t think it is right to just say “Healthier lifestyles of nut consumers can explain the association” and move on, because the researchers already realized that and attempted to control for it. The right questions to ask are if the ways they tried to control for healthy lifestyle are sufficient. If weight is a confounder, does adjusting for BMI take care of that? Is fruit, vegetable, and red meat intake an adequate proxy for overall eating habits/diet? Is smoking poorly controlled for because it was assessed as a binary current smoker/not current smoker variable rather than by smoke exposure levels?

    Finally, the numbers that the researchers are basing their conclusion on are the hazard ratios. Basically, they found that the death rate in the highest group of nut consumers is 0.83 times that in the never group (with non-overlapping error bars), which is about a 17% reduction.

    1. I agree with you, the study authors are more nuanced about its overall findings. Someone else pointed out that socioeconomic status could play a role and I don’t believe that was pointed out in the study. There was also some basic mention of the Mediterranean Diet, but the authors steered the subject back onto nuts (rather than, say, focusing on people who ate a diet that focused on “whole” foods).

  11. I think the discussions above, in addition to the whole accessibility issue, make the point that figuring these things out is actually not incredibly straight-forward. You have to have a reasonably technical background and a lot of patience to really figure out what the study is saying, then a good amount of experience to know whether what they are saying matters.

  12. Do peanuts count!? Per the “you know where,” peanuts are a legume, which is why they were called “goober peas” in the South during the Civil War. So peanuts are beans…and we all know what beans are good for…but I digress: If peanuts do count, then I’m in for eternity because I eat tons of them. As the good book says: Man can not live by bread alone, he must have peanut butter.

    1. Don’t forget about boiled peanuts. Those are pretty awesome too, and in my neck of the woods, there was always a Boiled Peanut Stand in the summer.

  13. Oh, yes, boiled is the tradition. Big, black kettle in the yard full of fresh pulled peanuts boiling away with an roaring fire underneath.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Back to top button