Nuts are a big part of my diet.  It’s my habit to eat handfuls of nuts through the day, and a few times a week to incorporate almonds or cashews or peanuts into a main course.  Perhaps I should be cheering that a headline in ScienceDaily last week told us, “Nuts and peanuts (but not peanut butter) linked to lower mortality rates, study finds”.  So, is the study good news for me?  I can answer with assurance: “Probably.”  

The headline referred to this study, just published in the Journal of Epidemiology.  The researchers in Maastricht looked back at data from a Dutch survey on diet and mortality conducted from 1986-1997.  They found mortality rates of nut-eaters were 23% lower than people who reported eating no nuts.  23% lower mortality corresponds to 2½  years of life extension [how to calculate].  Threshold for the benefit was quite low at a few ounces per week, and more was not better.  (Past studies indicated that perhaps more is better.  I eat about 2 pounds of nuts in a week, perhaps over the top, because I like them and because I’m on a low-carb vegetarian diet.  There is no data in that range.)  Peanuts were found to be just as good as “tree nuts” (almonds, cashews, walnuts, Brazils, etc.) but peanut butter had no benefit whatever.

By itself, a finding like this is hard to translate into a dietary recommendation.  There are qualitative problems with methodology.  People are different, and a diet that is right for one person may be all wrong for another.  And if we eat more nuts, are we adding more calories?  Or are we eating less of something else?

There is also the quantitative problem of cross-correlations–correlation does not necessarily imply causation.  People who eat nuts are likely to be richer and better educated and more careful about their diets, likely to be eating less unhealthy snacks, less meat, less carbohydrates.  Any of these things could produce an incidental statistical association between nut consumption and longevity, with no indication that eating nuts confers a benefit.

Nut consumers were on average somewhat younger, leaner (in women), drank more alcohol, ate more vegetables and fruits, were less often hypertensive or never smokers (women), but were higher educated and more often used supplements, or postmenopausal hormone replacement therapy (HRT). Women with the highest nut consumption less often reported diabetes. [ref]

 

Cross-checking and corroboration

It’s common to correct with multivariate analysis, but multivariate analysis doesn’t work very well if there are more than a few variables, and it’s hard to know in advance which are the relevant ones.  Statistics ends up being an art as much as a science.

So the study gains credibility when previous studies, with different methodologies in different populations, come to the same conclusion.  There are several, the biggest and best of which are this one and this one.  With write-up of the new study, the authors include a “meta-analysis” of these past studies.  This is another layer of statistics which combines previous results to come up with a conclusion stronger than any one study could draw.  Meta-analysis is a pursuit that can keep a data geek happy and productive for weeks on end.  Happiness and productivity are both positively correlated with life expectancy ☺.

Though a self-identified stat geek, I confess to being unfamiliar with Cochran, Begg and Orsini; nevertheless a paragraph like this adds to the credibility of a study in my eyes:

In these analyses, the HR estimate for each study was weighted by the inverse of the variance of the log HR to calculate the summary HR and its 95% confidence interval (CI). Heterogeneity between studies was estimated using the Cochran’s Q test and I2 (the proportion of variation in HRs attributable to heterogeneity).  Publication bias was assessed by the Begg test.  In addition, we performed dose-response meta-analyses using generalized least squares regression described by Orsini et al. with restricted cubic splines (four knots, at 5th, 35th, 65th and 95th percentiles) to investigate potential nonlinearity in the dose-response relationship.

The multi-study meta-analysis results closely paralleled the present study.  All the diseases of old age were lower in people who ate nuts; cancer was marginally lower, and cardiovascular disease much lower.  Some of the older studies found less benefit for peanuts than for tree nuts, and beyond a few ounces a week, there was ambiguity about whether more nuts offered more benefit.

 

Chocolate

There is comparable statistical evidence for the benefit of eating chocolate!  For years, I have refused to take these studies seriously, figuring that they are funded by a consumer industry that is eager to rehabilitate its junk-food image.

And in fact, just last month there was a spoof done by a science journalist, intending to remind us how easy it is to lie with statistics.  The headline was “Slim by Chocolate”.  Here is  John Bohannon’s account of what he did, and what morals we should draw.  His main point is that we all should use our common sense and be skeptical of sensational health claims.  Who can argue with that?

But the topic he chose incidentally illustrated other points as well.  There are legitimate claims for health benefits from chocolate.  People are complicated, and no two bodies are alike.  Not only do foods affect our metabolisms differently, but even more various are the psychological effects of foods.  There are people who find a little bit of chocolate uniquely satisfying, and it helps them to eat a leaner, healthier diet in many other ways.  There are other people who find chocolate addictive, and the more chocolate they eat, the more they want.

I find it completely plausible that some people are better able to lose weight with chocolate than without.

 

The psychology of eating is the most individual thing about diets, and it plays an essential role.

Time after time, human psychological studies have demonstrated that for the great majority of people, will power is worse than useless in trying to control weight.  (Present company, of course, is excluded.  You and I both have perfect control over what we eat, regardless of what the statistics may say.)  People who set out to lose weight by adhering to a set of rules generally succeed for awhile, then well over 90% bounce back to a weight higher than they started.  Habits in themselves are hard enough to change; but in addition we have powerful and persistent homeostatic impulses whispering in our ears.  The body gravitates to a “set point” in weight and percentage of fat.  The most successful diets all manipulate those signals of craving and satiety with alterations to the body’s biochemistry.

We joke about pregnant women having aversions to some foods and cravings for others.

Methionine is an essential amino acid, and it is normally part of all protein that we eat (though some sources have more than others.)  There have been a lot of studies of methionine restriction, in which mice are fed an artificial diet of re-constituted proteins in which this one amino acid is missing.  Despite the fact that methionine has no distinctive flavor or smell, mice know at some level that they are missing methionine.  Some researchers report that the mice refuse to eat unless there is methionine in their food.

The moral of the story is that our bodies know what they want, and will nag at us until they get it.

For some people, phytochemicals in chocolate can play a positive role in regulating gut biota and controlling anxiety that can lead to nervous eating and other destructive behaviors. This new study comes from University of Aberdeen in Scotland, and was published last week in the British Medical Journal.  All-cause mortality was not compiled, but the study claims that eating more chocolate is associated with less cardiovascular disease, and the group with highest chocolate consumption enjoyed 23% less heart disease.  Here is a meta-analysis of studies in the past that have been less clear and consistent than for nuts.  The average is that people who ate the most chocolate had 25% fewer cardiovascular events compared to people who ate the least.  No studies have been done about cancer, or all-cause mortality.

Interestingly, the non–chocolate-eating group had the highest mean body-mass index, the highest percentage of participants with diabetes, and the highest levels of inactivity. On the other hand, “higher chocolate intake was associated with a higher energy intake, with lower contributions from protein and alcohol sources and higher contributions from fat and carbohydrates.” [ref]

Translation: chocolate eaters weigh less despite eating more.  Did Mr Bohannon stumble onto something that none of us expected?  Probably not.  Here are two studies [one, two] that find just what we would expect, that eating chocolate is associated with weight gain.

 

The bottom line

People who eat nuts and chocolate have lower rates of cardiovascular disease and live longer than people with comparable amounts of body fat who don’t.  If you can adjust your diet to add chocolate and nuts without gaining weight, you will probably benefit.  Remember always that diets are individual and the response of your own body is not the average response.

This is the era of big data.  We are awash in data.  What fun for people like me, who love to extract meaning from numbers!  Still, answers to basic questions remain elusive.  Finding correlations between single foods and particular diseases is a start.  But researchers might remember that our goal is to design diets and life styles that are healthy and adapted for each individual.  We have a long way to go.  More creative and ambitious study designs for the future might help.  I’ll have two examples for you next week.