Everything We Know About Nutrition May Be Wrong

A new study says 40 years of CDC data connecting food and obesity are fundamentally flawed.

National Journal
Brian Resnick
See more stories about...
Brian Resnick
Oct. 10, 2013, 7:52 a.m.

For 40 years, the Cen­ters for Dis­ease Con­trol and Pre­ven­tion has been ask­ing people what they eat in an at­tempt to un­der­stand the con­nec­tions between what we con­sume and how our bod­ies feel. And for 40 years, they may have been do­ing it wrong.

The lim­it­a­tions of the CDC data “make it ex­ceed­ingly dif­fi­cult to dis­cern tem­por­al pat­terns in cal­or­ic in­take that can be re­lated to changes in pop­u­la­tion rates of obesity.”

That’s the claim in a new study pub­lished in the on­line journ­al PLO­Sone. The re­search­ers probe CDC’s Na­tion­al Health and Nu­tri­tion Ex­am­in­a­tion Sur­vey, which has in­ter­viewed Amer­ic­ans about the foods they eat and their life­styles since 1971. From the sur­vey, we learned things about nu­tri­tion that now seem so fun­da­ment­al — that diet and ex­er­cise choices are linked to body weight, that cho­les­ter­ol is linked to heart dis­ease, and so on.

But here’s the prob­lem, ac­cord­ing to the au­thors: All of that data was com­piled by ask­ing people to re­call what they ate.

“Nu­tri­tion sur­veys fre­quently re­port a range of en­ergy in­takes that are not rep­res­ent­at­ive of the re­spond­ents’ ha­bitu­al in­takes,” the au­thors write. “And es­tim­ates of EI [en­ergy in­take] that are physiolo­gic­ally im­plaus­ible (i.e., in­com­pat­ible with sur­viv­al) have been demon­strated to be wide­spread.” Men and wo­men have been found to un­der­re­port cal­or­ies by between 12 per­cent and 20 per­cent, and are more likely to se­lect­ively un­der­re­port eat­ing the bad stuff, such as fat and sug­ar.

Trans­la­tion: We can’t trust hu­man memory as the source of our nu­tri­tion data, be­cause people can un­der­re­port what they eat to an ab­surd de­gree. Their self-re­ports doc­u­mented amounts of food that could not pos­sibly sup­port their sur­viv­al. “In no sur­vey did at least 50 per­cent of the re­spond­ents re­port plaus­ible EI [en­ergy in­take] val­ues,” the au­thors re­port.

The NHANES sur­vey does con­tain many, many ob­ject­ive meas­ures such as phys­ic­al ex­am­in­a­tions and blood work (for in­stance, it found el­ev­ated levels of lead in Amer­ic­ans’ blood work, which lead to the de­creased use of the met­al in gas­ol­ine and soda cans). But it’s not like the CDC can mon­it­or all a per­son eats. Nor is it really feas­able to do large-scale ex­per­i­ments on nu­tri­tion — that is, sep­ar­ate people in­to con­trol and ex­per­i­ment­al groups, have every­one eat the same ex­act things ex­cept for one vari­able, and then com­pile this data over dec­ades.

Gran­ted, in re­cent years, CDC has re­vised i’ts meth­od­o­logy. Since 2001, it has fol­ded NHANES in­to the “What We Eat in Amer­ica Pro­gram, which re­cords food in­take in a more con­trolled man­ner.

But there’s even reas­on to be­lieve that as time went on, and as the sur­vey raised aware­ness of obesity, people’s an­swers be­came even more skewed. The au­thors ex­plain:

“There is strong evid­ence that the re­port­ing of ‘so­cially un­desir­able’ (e.g., high fat and/or high sug­ar) foods has changed as the pre­val­ence of obesity has in­creased. Ad­di­tion­ally, re­search has demon­strated that in­ter­ven­tions em­phas­iz­ing the im­port­ance of ‘healthy’ be­ha­vi­ors may lead to in­creased mis­re­port­ing as par­ti­cipants al­ter their re­ports to re­flect the ad­op­tion of the ‘health­i­er’ be­ha­vi­ors in­de­pend­ent of ac­tu­al be­ha­vi­or change.”

The sur­vey in­dic­ates that health be­ha­vi­or in­flu­ences re­sponses to fu­ture sur­veys. And that’s bad sci­ence. All in all, the au­thors con­clude that the lim­it­a­tions of the NHANES data “make it ex­ceed­ingly dif­fi­cult to dis­cern tem­por­al pat­terns in cal­or­ic in­take that can be re­lated to changes in pop­u­la­tion rates of obesity.”

It doesn’t mean our nu­tri­tion as­sump­tions are wrong. It just means we haven’t proven them, be­cause our meth­ods have been flawed.

What We're Following See More »
Trump Leads Tightly Packed Group Vying for Second
2 hours ago

In one of the last surveys before New Hampshirites actually vote, a Monmouth poll has Donald Trump with a big edge on the Republican field. His 30% leads a cluster of rivals in the low-to-mid teens, including John Kasich (14%), Jeb Bush and Marco Rubio (13% each) and Ted Cruz (12%). On the Democratic side, Bernie Sanders leads Hillary Clinton 52%-42%.