Monday, August 27, 2012
SparkPeople trivia question: "True or False: People who eat chocolate live longer than people who abstain from it." SparkPeople wants us to click "true" to this, explaining, "Researchers at Harvard University believe chocolate may help people live longer! A study tracking older men found that those who ate chocolate lived almost a year longer than those who didn't."
"Wrong," says the actuary.
First, one single study from one research group, no matter how prestigious, doesn't make it so. Second, there's a fairly large leap from researchers who "believe" that something "may" be a "help" to the absolute statement that people who do it live longer than people who abstain. Let's jump to an absurd extreme: if your plane goes into a tailspin, should you quick eat some chocolate to increase your chances to survive the crash?
These are not trivial distinctions, nor are they quite as absurd as I myself pretend to make them.
Let's start with a simple extreme - not saying this is how the real world actually is, of course, but merely hypothesizing an extreme for the sake of illustrating a very key point about such research. Let's say that all economically disadvantaged adults spent their food stamps on alcohol, cigarettes, and the occasional bit of food, but in our extreme hypothetical not wasting their food stamps on chocolate. Conversely, in our extreme hypothetical, the affluent would be indulging in chocolate-covered everything. Now the Ivy League researcher comes along and tallies who tends to live longer, the person who has been eating chocolate or the one who has not. Guess what the researcher would report.
Let's do an even clearer one, one we don't even need to stretch into any hypotheticals to make the point. On average, who tends to eat more chocolate, men or women? OK, and do we know who on average tends to live longer, men or women? Guess what a research study that fails to separate out non-chocolate gender differentials is going to indicate? (Or else do the SparkPeople Trivia jump-to-conclusions leap and ask - "True or False: Women live longer than men because they eat more chocolate." You'd be as near truth as the existing SparkPeople trivia question on chocolate is.) I know, yeah, the Harvard study tracked older men, so supposedly avoided this distinction? Except I'm only using this particular well-known distinction to make a point; and unless their research fully eliminated similar distinctions that exist in subgroups of older men, they would make the same blunder.
But this is Harvard we're talking about. Surely they would take such factors into account? Not necessarily in sufficient detail to account for the very minor effects that their research is showing, such as less than a 5mg easing of blood pressure.
On NPR radio, I once gave the American Academy of Actuaries' response to a supposedly prestigious research study that claimed that left handers have an intrinsic higher mortality rate than right handers. The research and analysis was very seriously flawed. Just one of too many missteps to count: their analysis relied on statistical methods that have not been used for mortality studies since the times of the ancient Romans, since it's been known since then that those methods produce unreliable results. Knowing how to perform regression on economic data or on lab rat behavior or on factory widgets does not give a researcher the tools to study mortality, which works with a completely different mathematical model. So sure, left handers who use right-handed equipment may encounter a higher accident rate than their right-handed counterparts, just the opposite can also emerge in practice: knowing the risks from having to do it the hard way for so long, left-handers using right-handed equipment might frequently be more conscientiously cautious than their right-handed counterparts, hence actually LOWER their accident rate. And bluntly, aside from the accident rate, there is NO reliable research that suggests any measurable differential between the underlying mortality rates of left handers versus right handers. No matter how prestigious the researcher.
"Whether chocolate is good or bad for you depends largely on the type of chocolate and the amount you consume." So says a Harvard Medical School article. Eat dark chocolate in moderate amounts, and yes, the flavonoids might - "might," the researchers say, nowhere near as absolute a conclusion as SparkPeople trivia pursuers jump to - improve your health. Even then, researchers consider the effects of chocolate only in combination with other healthy habits - eating a block of dark chocolate daily won't act like an exercise pill replacing getting up and working that heart out.
All things considered, a piece or two of DARK chocolate eaten regularly in balance with veggies and the rest of a healthy diet combined with a lifestyle that includes exercise and avoids unreasonable risks can maybe - still only maybe - give you a few days - but more than likely no more than days, nowhere near the year SparkPeople suggests - of extra life over the hypothetical twin who is EXACTLY like you in every other respect other than that bit of chocolate. But even that stretch is NOT the same conclusion that SparkPeople's trivia question jumps to. Not even if we beg the issue about them needing to shorten it to make the point. Their's remains a statement for which the CORRECT response is "maybe a little, depending on the type and amount of chocolate," but since that was not an option, the only true response to the trivia question is a resounding "false."
Eh, except I went ahead and clicked "true" to collect my SparkPoint. Already, most of the times I get one wrong on SparkPeople trivia is because they trick us with misleading language when we actually do know the true answer or when they misread facts such as they've done here, versus getting one wrong because I actually don't know the answer. But it truly is trivial. Enough more of this, and I'll be getting bored with picking up those SparkPoints, I think. The point of it gets lost when they get silly as they've done here with chocolate.