The China Study, Wheat, and Heart Disease; Oh My!

(Not only is this woefully, frustratingly, absurdly belated, but it’s also not yet finished. But I hate being a blog tease, so here’s part one!)

If you’ve been following along with the previous China Study entries (and the wild drama that ensued), you know that I’ve been promising an entry on wheat for a while now, mostly because this little snippet snagged so many eyes:

Correlation between wheat flour and coronary heart disease: 0.67

That’s a value straight from the original China Study data. Could the “Grand Prix of epidemiology” have accidentally uncovered a link between the Western world’s leading cause of death and its favorite glutenous grain? Is the “staff of life” really the staff of death? Bwah ha ha.

Damning as it seems, a single unadjusted correlation isn’t enough to make that leap. Actually, nothing in this post will be enough to make that leap, because A) it’s epidemiological data and not a controlled study, and B) correlation isn’t causation anyhow. You know the drill.

So my goal here isn’t to prove anything about wheat. Mostly, I want to see if I can find a confounder that’s creating a false association between wheat and heart disease in the China Study data. Something wheat-eating regions have in common that makes them more susceptible to ticker troubles. Because really, folks, this is serious business:

And when we pluck out the wheat variable from the 1989 China Study II questionnaire—which has more recorded data—and consider potential nonlinearity, the outcome is even creepier:

Wowza! By the way, wheat flour also correlates significantly with hypertensive heart disease and stroke, but I’m mainly going to look at coronary heart disease in this post. (And although wheat looks like it could have a nonlinear relationship with heart disease, with the highest wheat eaters having disproportionately steeper rates than non-wheat eaters, I’m going to treat it as linear for the sake of this analysis. That way, the worst that’ll happen is we’ll underestimate the potential effect of wheat, which—for now—is better than overestimating it.)

Since I’m not trying to dissect our friend Campbell’s claims anymore, I’ll be using the China Study II data (from 1989) because it recorded more descriptive variables about diet and blood samples.* And because it’s already available online. (Not that I don’t love typing thousands of numbers onto my computer by hand. Three cheers for data-entry-induced carpal tunnel!)

*Quickie note: If you want to play with the China Study numbers yourself, I recommend not just using the “all vascular diseases” variable, because it includes rheumatic heart disease—a condition spawned by rheumatic fever and generally unrelated to diet. Lumping diseases with different etiologies together dilutes the strong correlations you can find by looking at each disease independently. Try checking out stroke (M065 STROKE), ischaemic heart disease (M063 IHD), and/or hypertensive heart disease (M062 HYPTENS)  along with all vascular diseases (M059 ALLVASC).

Here’s the problem with looking at wheat and heart health. Along with correlating pretty darn strongly with heart disease, wheat-eating regions boast  a number of other factors possibly involved as well—some as protective agents and some as causative. For instance, wheat flour correlates significantly and inversely with:

  • Plasma folate concentrations (and consequently, homocysteine status)
  • Fish intake and DHA levels
  • Yearly green vegetable consumption
  • HDL cholesterol
  • Vitamin C intake

And it correlates significantly and positively with:

  • Height, weight, and BMI
  • Blood pressure
  • Latitude (as a possible marker for vitamin D status)
  • Yearly milk intake
  • Polyunsaturated fatty acid intake

Since all of these variables also associate (inversely or positively) with heart disease, it’s possible they could be confusing the “0.67” figure we’ve cited for wheat. Could some other, non-grain component of the wheat eaters’ diets predispose these folks to heart disease?

On the bright side, China’s wheat eaters are less likely to drown than the wheat-shunners (r = -0.68 for the youngsters under 34). Maybe they’re all buoyant from celiac bloat.

And in case you’re wondering, here are some heart disease risk factors (the ones Conventional Nutritional Wisdom likes to toss around) that don’t positively correlate with wheat. That means we probably can’t blame ’em for wheat’s dirty deeds. Out of curiosity, though, I’ll still include them in some of my models just to see how they behave in relation to wheat with heart disease.

  • All meat intake (r = -0.35)
  • Red meat intake (r = -0.30)
  • Animal fat intake (r = -0.35)
  • Saturated fat intake (r = -0.40)
  • Total animal protein intake (r = -0.27)
  • Total fat intake (r = -0.43)
  • Fat as a percentage of total calories ( r = -0.41)
  • Total cholesterol (r = -0.05)
  • Apolipoprotein B (r = 0.02)
  • Daily alcohol intake  (r = -0.37)

Mostly, what I’m looking for is a little somethin’-somethin’ that both wheat flour and heart disease have in common. A shared variable that could be slyly—and wrongfully—framing wheat as our heart-harming villain.

So how do we untangle all these variables? I’m using two methods: multiple regression analysis and stratification. Multiple regression is a handy way of looking at two or more variables and seeing how each one behaves when the others are held constant, and stratifying data can work similarly by divvying up data into groups that share or exclude a certain variable. (For the stats junkies out there, I’m using ordinary least squares for the regressions, and I’m running each model two times: once with the data as-is, and once with any non-normally-distributed dependent variables transformed (via natural log) for more reliable statistical significance testing. I’m also checking for linearity between the variables before creating each model, since a nonlinear relationship will be underestimated with linear regressions.)

And for anyone not familiar with statistics terminology, here’s a really quick rundown of what you need to know to understand the numbers in this post:

  • r = the Pearson product-moment correlation coefficient  between two variables. It can range from -1 to 1. When it’s zero or close to zero, there’s pretty much no relationship between the variables. When it’s a negative number (like r = -0.50), there’s an inverse relationship between the variables, meaning one increases as the other decreases. When it’s a positive number (like r = 0.50), there’s a positive relationship between the variables, meaning they increase and decrease hand-in-hand. The closer to -1 or 1 r is, the stronger the association. R can never prove cause and effect, though—it only indicates an relationship of some sort.
  • beta = the standardized coefficient for each variable in the multiple regressions I’ll  be running. This is a lot like r, in the sense that it shows how well a specific variable is predicting the outcome (eg, heart disease) and also ranges from -1 to 1. But in the case of beta, we’re also controlling for the effects of other variables, so this number tends to be more accurate than r.
  • p = the probability that our results are just a fluke. P indicates how likely it is that we’d get a value of a test statistic that’s as extreme (or more extreme) as the one we have based on chance alone. Having a p-value of less than 0.05 indicates a high level of significance and means that our results are pretty sound. The lower the number, the more confident we can be that we’ve got something legit.
  • r-squared = percent of variance explained. This number shows what proportion of the outcome (eg, heart disease) can be explained by the variables in a particular model (eg, wheat and HDL cholesterol). The higher the number, the more successfully the variables are predicting the outcome. (“Predicting” is a misleading way of putting it, though, since we still aren’t looking at proof of cause-and-effect—only a relationship.)

Preliminary theories

It’s no secret that I’m less-than-enamored with wheat. We parted ways long ago (he got me allergic and then ran off with some floozy—classy, eh?). Nonetheless, I don’t like pointing fingers where they shouldn’t be pointed, so I’ll entertain some alternative theories that could explain wheat’s apparent association with heart disease.

1. Folate deficiency. In northern China, about 40% of the population qualifies as folate deficient (compared to only 6% in the south)—a geographical trend that corresponds nicely with wheat consumption. Being low in folate tends to elevate homocysteine, which—you guessed it—is an independent risk factor for heart disease. So maybe it’s not the wheat itself causing mischief, but the fact that low-vegetable, wheat-centered diets in China tend to breed folate deficiency and hike up homocysteine.

On top of that, in the China Study II data, wheat flour positively correlates (r = 0.30, p<0.05) with childhood death from neural tube defects—a category of birth defects often related to folate deficiency. Although the China Study data didn’t document homocysteine levels (darnit), the 1989 data did measure plasma folate. That means we’ll be able to test whether folate levels could be obscuring the true relationship between wheat and heart disease.

2. Vitamin D deficiency. For the most part, wheat-eating regions in China are in the northern half of the country—a hotspot for vitamin D deficiency, which is strongly linked to heart disease. Given the pretty convincing correlation between latitude and heart disease mortality, it’s possible that vitamin D is playing a role in this mess. Are the wheat-eaters merely suffering from low levels of the ol’ Sunshine Vitamin due to their unfortunate geographical placement, and getting more heart disease as a result? Sure seems possible.

3. Low intake of DHA. In an earlier publication, Campbell and his crew already determined that fish and DHA intake appears protective against heart disease in the China Study data. Not too surprising, since DHA reduces blood viscosity and can lower other factors associated with heart disease (like triglyceride levels). And considering wheat-eating regions don’t consume much seafood (r = -0.43 for daily fish intake), perhaps DHA deficiency—rather than wheat consumption itself—is to blame for higher rates of heart disease.

4. Combo-abombo. Maybe a mix of low folate, vitamin D deficiency, and DHA deficiency are swirling together into a doomful vortex—some horrible, Bermuda-Triangle-esque zone of heart disease. A zone that just happens to overlay areas of wheat consumption.

5. Unexpected mystery variable. If none of the above can explain the wheat-heart disease link, we’ve still got a verdant jungle of China Study variables to plow through. So plow we shall. I’ll try running a number of common-sense models to see if I can find something that explains heart disease better than wheat alone.

Multiple regression results

Folate. Ah, theory numero uno! Like wheat, folate has a strong, statistically significant correlation with heart disease (r = -0.40, p<0.001), so what happens when we run a model using both folate and wheat as exposures? Initially, it looks like wheat clobbers folate as a predictor (beta = 0.59, p<0.001 versus beta = -0.06, p = 0.39)—which would suggest that, although China’s wheat-eaters tend to have lower folate levels, folate deficiency itself isn’t enough to explain the link with heart disease.

But I’m not ready to dismiss this one just yet. As often happens with plasma measurements and health conditions, folate may have a nonlinear relationship with heart disease—which means multiple regressions (of the linear variety) won’t show the full picture. Indeed, when I make a scatter plot for folate levels and coronary heart disease, it looks like a bit of a curve emerges, with folate being most strongly associated with heart disease when the county average dips below 10 micrograms per liter (or thereabouts). Above that, the correlation is far less dramatic.

So how do we deal with this statistical monkey wrench? For starters, I tried transforming the folate data to make it more suitable for linear regressions, but that didn’t do diddly squat to the results: The numbers were beta = 0.58, p<0.001 for wheat and beta = -0.06, p = 0.31 for folate. So then I tried stratifying the data based on “low” and “high” folate levels (10 or less micrograms/liter versus 10.1 or more micrograms/liter), but both subgroups continued showing wheat as strongly and significantly correlated with heart disease while folate was off the hook.

Just to cover my bases (and because I’m a stubborn son-of-a-gun), I kept playing  with the numbers for a while longer to see if I could excavate anything new. Nope. Bottom line: It looks like wheat is predictive of heart disease whether or not folate levels are low, whereas folate is mostly predictive of heart disease only in the presence of high levels of wheat consumption.

So, theory #1 doesn’t pan out. Bugger. But bear in mind, we’re using folate mostly as a marker for elevated homocysteine, so these results don’t mean that homocysteine itself isn’t playing a role. Other causes of high homocysteine, such as B12 deficiency, weren’t documented in the China Study data. So this is an issue that’ll have to remain annoyingly unresolved. Another bugger!

Onto the next theory: latitude. Could the folks living in northern wheat-eating regions have lower vitamin D levels, leading to more heart problems—and creating a false link between wheat and cardiovascular disease? I admit, this was my favored theory after folate, but it ain’t holdin’ water. When I run wheat and latitude together as potential contributors to heart disease, wheat remains strongly predictive (beta = 0.65, p<0.001), while latitude diminishes (beta = -0.01, p=0.96). It’s pretty clear that the raw correlation between heart disease and latitude (which is 0.43, p<0.01) is just an echo of the relationship between heart diseases and wheat-eating regions, which are typically northern.

Okay, so that’s two strikes for Denise’s heart disease theories. What about fish and DHA? Are the wheat eaters suffering due to their fishless (and low-in-DHA) diets rather than from wheat itself? Alas, it doesn’t look likely. When I run these things together as exposures for heart disease, wheat stays strongly predictive (beta = 0.68, p<0.001) while the fishies do not (beta = 0.08, p = 0.47). Likewise, DHA teeters out into statistical insignificance (beta = 0.06, p = 0.30) when used in a model with wheat.

(Wait, I know what you’re thinking! “Why does it look like fish and DHA contribute positively to heart disease?” It’s because many of the fish-eating regions are more industrialized, and—in the absence of wheat—the fish-heart disease relationship is confounded by other factors like more desk work, more smoking (especially manufactured cigarettes), less physical activity, more vegetable oil consumption, and so forth. When we add some more variables to the model that take away the “city effect” associated with fish—such as apo-B, tobacco use, or percentage of the population employed in agriculture—then both fish and DHA turn inverse again. Although wheat, it should be mentioned, stays rock-steady in its high coefficient and statistical significance.)

Other stuff

Milk. Is moo juice a cardiovascular foe obscuring the relationship between wheat and heart disease? Probably not, according to the data—which isn’t surprising, given how few counties even drink the stuff. When running daily milk intake alongside wheat intake, wheat keeps its positive correlation (beta = 0.67, p<0.001) and milk actually turns a bit inverse, though not significantly so (beta = -0.07, p=0.47). No model shows a significant association between milk and cardiovascular disease, so I’m crossing this one off the list of potential confounders.

Blood pressure, BMI, corn, millet, sorghum, rice, added animal fat, added vegetable fat, total fat, total animal food, total carbs, total protein, percent of calories from animal protein, and all the smoking/tobacco variables I tried became statistically nonsignificant (in relation to heart disease) when thrown into a model with wheat.

Income is positively associated with heart disease when wheat is held constant, but it still doesn’t put a ding in wheat’s association with heart disease.

Models with more variables

So apparently comparing wheat + one other independent variable isn’t enough to explain the Wheat Effect. Not even a little bit. But maybe, just maybe, a bigger combination of variables will do the trick. Perhaps wheat-eating regions just host a collection of heart-harming factors (low folate, low vitamin D, low EFAs, and so forth) that, together, are more powerful predictors of disease than the variable wheat.

Here are the variables I’m interested in looking at. Some could be causative and some could be preventative:

  • Wheat consumption
  • Corn consumption
  • Millet consumption
  • Rice consumption
  • Total blood cholesterol
  • LDL cholesterol
  • HDL cholesterol
  • Apolipoprotein-B
  • DHA levels
  • Folate levels
  • Latitude
  • Added vegetable oil
  • Blood pressure
  • Weight
  • BMI
  • Total fat intake
  • Total monounsaturated fat intake
  • Total polyunsaturated fat intake
  • Total saturated fat intake
  • Percent of calories as fat
  • Percent of calories as carbohydrates
  • Total animal protein intake
  • Total plant food intake (by weight)
  • Total animal food intake (by weight)
  • Green vegetables (daily, not yearly)
  • Vitamin C intake
  • Total sodium intake
  • Poultry consumption
  • Egg consumption
  • Red meat consumption
  • All meat consumption
  • Fish consumption
  • Dietary cholesterol intake
  • Percent of the population currently smoking
  • Percent of the population who have ever smoked tobacco
  • Percent of the population smoking manufactured cigarettes
  • Percent of the population pipe smoking
  • Percent of the population smoking cigars
  • Percent of the population working in industry (typically less physical activity)
  • Percent of the population working in agriculture (typically more physical activity)

I won’t bore you with the results of every single combination I tried (over 100), so here’s the gist. No matter what model I use, wheat always adds unique variance. That means wheat (or an undocumented variable associated with wheat) is contributing something to heart disease that these other variables can’t account for. No combination out of the above bumped the association between wheat and heart disease out of the “statistically significant” zone.

Incidentally, one model had the best fit out of all the others for explaining heart disease:

  1. Wheat consumption (beta = 0.62, p<0.001)
  2. Apolipoprotein B (beta  = 0.38, p<0.001)
  3. Total cholesterol (beta = -0.22, p<0.05)

Note that the number for total cholesterol is inverse, meaning higher cholesterol was associated with less heart disease—at least in this specific model. Unless you’re an Ancel Keys groupie, this may actually be quite plausible.

Anyway, here’s the important point. No matter what variables I adjust for, I can’t make the correlation between wheat flour and heart disease go away. Sorry, wheat! Neener neener.

Cardiovascular disease: The only “Western” problem without “Western” risk factors

Here’s a mystery for ya.

In the China Study data, most Western diseases (such as breast cancer, colon cancer, lung cancer, and diabetes) are concentrated in areas that share some key characteristics: more industrial employment, less agricultural work, greater population density, and often higher levels of schooling. Folks here eat more processed starch and sugar, use more polyunsaturated vegetable oils, chug down more beer, smoke more manufactured cigarettes, and typically get less physical activity than their neighbors in pastoral communities.

In other words, the Western-disease-prone-regions are like baby Americas—slowly waddling, diapered and naive, towards the motherly lap of disease.

Most likely, these Western ailments aren’t spawned from a single food or activity, but from a tragic mix of diet choices, lifestyle habits, and environmental factors. For problems like breast cancer and colon cancer and lung cancer, it’s pretty easy to see what the matrix of risk-raisers are from looking at the data: It’s the same combination of things spurring disease in Western nations.

But oddly enough, this isn’t the case for heart conditions. The factors shared by other Western illnesses are not, in most cases, associated with heart disease in this data set. If you’ve read some of the earlier China Study posts, you might remember that I took issue with Campbell’s disease-clustering strategy because heart disease doesn’t fit cleanly with the “diseases of affluence” group, despite his insistence on sticking it there anyway. Unlike the other Western problems, heart disease isn’t associated with eating more sugar, working in industry, drinking more alcohol, using vegetable oils, having higher apo-B levels, or any of the other variables uniting the Western diseases and mirroring the traits common to industrialized countries.

What’s the only thing heart-disease-prone regions have in common with Westernized nations? That’s right: consumption of high amounts of wheat flour.

Food for thought. Kinda spooky.

Wheat eaters: fatter with fewer calories

Here’s some more weirdness. In both China Study I and II, wheat is the strongest positive predictor of body weight (r = 0.65, p<0.001) out of any diet variable. And it’s not just because wheat eaters are taller, either, because wheat consumption also strongly correlates with body mass index (r = 0.58, p<0.001):  

How odd! This aligns with a post Stephan Guyenet at Whole Health Source wrote about wheat consumption and obesity in China, speculating that wheat might wreak metabolic havoc wherever it goes—a trend that becomes apparent when comparing similar populations of wheat eaters and non-wheat eaters, such as in China. But perhaps there’s some confounding going on. What about calorie intake? Are the wheat eaters just scarfing down more food in general, leading to higher weight regardless of wheat consumption? Doesn’t look like it. Running wheat and calorie intake together as predictors with BMI as the outcome, wheat takes the weight-gaining gold:

  • Wheat: beta = 0.56, p<0.001
  • Calorie intake: beta = 0.13, p = 0.19

Unfortunately, we have no way of accounting for energy expenditure through physical activity—but considering wheat-eating regions tend to be pastoral and dominated by agricultural work, it seems they’d be burning through a greater wallop of calories than more sedentary regions. Indeed, independent of calorie intake, there’s a clear association between agricultural work and weight (lower) versus industry work and weight (higher), suggesting these things could be approximate measures of calorie expenditure. So once again, we’ve got a paradox: The wheat eaters are consuming lower or average levels of calories, doing more physical labor, and yet… they’re fatter.

Out of curiosity, I ran a stepwise regression on a bunch of relevant variables to see what combination would best predict BMI. (In statistics, stepwise regression is a really cool, but sometimes totally misleading method for building a statistical model. It involves adding (or winnowing away) variables one by one based on how they behave together and contribute to the outcome—BMI, in this case—until you’ve got a model where each variable offers significant variation and the highest possible percent of explanation (represented as r-squared). Unfortunately, since this process is automated and computers usually don’t understand the whole “biological plausibility” thing, you can wind up with weird models that don’t make sense in the real world. Nonetheless, it can be a worthwhile method if used with caution.)

Setting BMI as the outcome, I chose the following variables as potential exposures:

  • Total calories
  • Total fat
  • Total carbohydrates
  • Total plant food
  • Total animal food
  • Total plant protein
  • Total animal protein
  • Total monunsaturated fat intake
  • Total saturated fat intake
  • Total polyunsaturated fat intake
  • Red meat
  • All meat
  • Fish
  • Poultry
  • Eggs
  • Wheat flour
  • Corn
  • Millet
  • Legumes
  • Starchy tubers
  • Green vegetables (daily, not yearly)
  • Agricultural employment
  • Industrial employment

(I left out milk because so few counties consumed it.)

The best-fitting model for predicting BMI (at 95% confidence)? Drum roll please. Three variables made the cut.

  1. Eating more wheat flour (beta = 0.48, p<0.001)
  2. Eating more polyunsaturated fat (beta = 0.44, p<0.001), and
  3. Eating fewer green vegetables (beta = -0.29, p<0.01).

This model has an r-squared value of 0.53, meaning it predicts a little over half of the variation in BMI—at least in theory. That’s actually pretty high, considering we haven’t directly factored things like physical activity into the equation.

Interesting, eh? All animal foods and total dietary fat, by the way, were completely insignificant in terms of BMI.

Of course, there could be other variables involved that the China Study didn’t cover. Were the higher-BMI folks also more heavily muscled (perhaps from more physical labor), increasing their body weight but not body fat? Are the wheat eaters, some of whom are ethnic minorities in China (especially Turkic and Mongolian), genetically “bigger” than the Han Chinese? There are plenty of unknowns, and alas, no way to clarify them based on this data.

I guess we’ll leave it as a question mark for now.

Grain damage: Do other studies back it up?

But don’t those peer-reviewed, scientific studies tell us wheat is healthy? Alas, the vast majority of studies on grains—especially wheat—showcase at least one of the following problems:

  • They look at the effects of whole grains versus refined grains—not whole grains versus the same diet with no grains at all.
  • Study subjects increase their consumption of whole grains, and this displaces some portion of yuckfoods (processed junk, white-flour products, sugary things, and so forth). As a result, it’s hard to tell whether any health perks are due to the addition of whole grains, or from the reduction of truly-awful-for-you foods. This is particularly true in studies that scout out disease patterns in populations rather than controlled studies that measure specific changes that occur with the addition of whole grains.
  • They don’t adequately account for other factors that often accompany whole-grain consumption, like a greater level of health consciousness, more exercise, other positive diet choices, and so forth.

However, a few gems are lurking in the massive slush-pile of irrelevant studies. This one’s pretty doggone interesting, and it’s from all the way back in 1959: “Comparisons of atherogenesis in rabbits fed liquid oil, hydrogenated oil, wheat germ and sucrose.” You can click on that for the full-text PDF.

As you might guess from the title, this study examines the effects of diet on the development of atherosclerosis—AKA hardening of the arteries. The researchers took cholesterol-infused rabbit food and supplemented it with liquid corn oil (yuck), hydrogenated corn oil (double yuck), wheat germ (mystery murderer?), and sucrose (sweet poison!). Sorry, I dig hyperbole. Anyway, part of the goal was to create an experiment testing the hypothesis that “the geographic differences in the incidence of coronary disease might be related to selective hydrogenation of polyunsaturated fatty acids or to degermination of cereals.”

So now, the moment of truth: Which group had the most severe atherogenesis? Perhaps the one fed the nasty hydrogenated oil, as hypothesized? Ladies and gentlemen, place your bets. From the article:

The most severe atherogenesis occurred in the animals on the wheat germ diet.

Was it a fluke? Probably not:

In an earlier study, we maintained 5 groups of 5 rabbits each for three months on 500 mg of cholesterol daily and rabbit chow supplemented with different fats or with wheat germ. Here also, the animals on the wheat germ diet showed a significantly greater degree of atheromatous lesions than the animals on rabbit chow plus 20% corn oil, cottonseed oil or hydrogenated cottonseed oil, whereas no significant difference was found between the various fats.

So what made the wheat germ contribute to atherogenesis? The researchers state that it’s “difficult to speculate” about the mechanism, which is a scientific way of saying “We dunno.” They suggest the extra dietary protein from wheat germ could be the cause, but from the literature I’ve skimmed so far, it looks like plant proteins don’t have much effect on bunnies (although animal protein does).

Of course, rabbits are truly terrible models for anything that happens in the human body. They’re hardcore herbivores. A mere billowing of the wind is practically enough to spike their cholesterol. But what explains the specific effect of wheat germ on their poor arteries? Could this have implications for humans?

My answer: It’s “difficult to speculate.”

Other studies

Prefer human studies? Me too. Here’s one that initially looks totally irrelevant but is actually pretty interesting: Flaxseed and cardiovascular risk factors: Results from a double blind, randomized, controlled clinical trial. (This also a stellar example of why it’s important to read full-text articles instead of just abstracts, which often don’t tell you diddly about the stuff you want to know.)

This particular study charted the effects of flaxseed on adults with high cholesterol. One group got food with ground flaxseed; the other group got food with added wheat bran. Other dietary elements were the same. (Low fat, low cholesterol. Fun times!)

The results? Ye Olde Flaxseed Group did pretty well: Compared to their baseline measurements, these folks had lower insulin, lower blood glucose, lower C-reactive protein (a marker for inflammation), and better insulin sensitivity (as calculated by HOMA-IR).

But poor Wheat Group was less fortunate. Since the study was about flaxseed, the results of wheat aren’t specifically discussed, but check out “Table 4” in the link above to see the numbers for yourself. The wheat-bran eaters had a 14.9% increase in insulin resistance (calculated by HOMA-IR) and a 9.3% increase in C-reactive protein. In other words, they lost some insulin sensitivity and gained some inflammation—two risk factors for heart disease. Hmm. Was the wheat bran to blame? Some other element of the control diet? It’s impossible to say for sure based on this study, but considering the wheat group’s adverse effects were more dramatic than the flaxseed group’s benefits, it seems a little suspect.

(A rather abrupt end of part one! The next post will have some more studies and speculations on potential mechanisms for wheat as causative of heart disease.)



  1. Hi Denise, while my wife and I were traveling across country yesterday, listening to the XM Doctors show, a doctor called in and reported the following. He had worked in Hawaii a number of years and never had a patient with a low vitamin B-12 value. He moved to Montreal and has found over 500 patients with low B-12 values in just a few years. Some of these had extremely low numbers (200). One of the causes of low B-12 is that B-12 is blocked in the intestines by some factor that prevents it from moving into the intestines. I was thinking that perhaps Hawaiians eat very little wheat while those in Montreal could be expected to consume a lot. Potential for another study? I admire you so much; keep up the good work. There are other statistical treasures to mine.

    1. I’m so glad I found you!
      I’ve read the China Study and found it very interesting, but I never thought of looking at the “wheat” connection.
      In my own little study (need for loss of weight 😉 the only way I can lose is to leave the wheat behind while eating the normal amount of calories. And I thought it was just me!

  2. hmm.. wheat substitute. There are rice flour, tapioca flour, potato flour, quinoa flour – whether these will give the same results as wheat flour you’ll have to test. I have made pancakes with the above mentioned flour types. Making bread is a whole different matter (maybe add yeast or baking soda?)

  3. Regarding this section of the report::
    Cardiovascular disease: The only “Western” problem without “Western” risk factors ….

    I would like to suggest that – as you suggested earlier – Vitamin D3 deficiency is a factor.

    Researchers know that the further we move away from the equator the greater the incidence of chronic disease.

  4. the number one brand in poker, World Series of Poker: Tournament of Champions takes a story-based approach. Take a seat next to Chris Jesus Ferguson as

  5. Love the post! Astonishing amount of work –agree it should be published. Wheat hates me and I hate wheat, believe it’s truly the bane of the human existence! On they say that CT scans of Egyptian mummies reveals rampant heart disease. They ate copious amounts of wheat and it wasn’t even the kinds we have today that have increased protein molecules.

  6. Nice post!

    One potential reason why wheat might help cause heart disease is that it appears to temporarily open up the tight junctures between the gut cells (even in bunnies!). This might cause over-absorption of iron from the environment … normally the gut prevents over-absorption. Iron is known to damage the heart, and there can be a lot of it in well water. Was well water mentioned in the China Study? People who drink from rural wells have a higher chance of getting Parkinson’s, which is another potentially iron-related disorder. The body produces insulin when iron is absorbed in a meal, to sequester it. See:

    The question for me has been: if iron is so regulated in the body, how come much of the US has too much of it? I think wheat might be the culprit in that. The fact wheat bread also has a fair bit of added iron makes it doubly problematic.

  7. So I’m curious. What’s the theory/conclusion (if any) on fermented wheat (bread) compared to plain ol’ wheat germ? I’m also wondering WHY there would be a difference between the two. Did I not read carefully enough?


    In your first two graphs, what are the units on the x-axis? Is it grams/day? If so, it looks like there is no difference in risk of coronary mortality eating 0-300g/day of wheat. However if you eat 300-600g, which, frankly, is a shedload, you double your risk, which is not a huge increase given the magnitude of the difference in behaviour. Though I might be mistaken about this; what are the units on the y-axis?

    I applaud your efforts but, if I’m right or close to right in my reading of the data you’ve presented, a doubling of risk due to eating a whole loaf of bread a day plus some noodles is a smaller biological effect than “Wheat is Murder” would imply.

    What do you reckon?

  9. Great analysis, thanks!

    “Maybe they’re all buoyant from celiac bloat.”, as to why inner “wheat eaters” had less mortality from downing than their coastal counterparts had me ROTFL!

    One needs to take care when someone starts throwing stats at you!

  10. p oor nutrition caused by the actual fatty hard working liver ailment is situated a lot of people enjoy your body, they frequently desire might inside the shortest possible the perfect time to view the influence in which vehicle,p oor nutrition caused by the actual fatty hard working liver ailment is situated a lot of people enjoy your body, they frequently desire might inside the shortest possible the perfect time to view the influence in which vehicle,

  11. What’s the difference between r-squared and beta coefficients? They seem to measure the same thing (how well a variable explains the data).

  12. Use great caution in basing any health decision on correlation studies. In statistics 101 my prof emphasized, correlation is not causation, correlation is not causation! correlation is not causation!! and if you torture the data long enough you can prove anything. I use correlation analysis as well as other mathematical systems in decision making but mostly in the hard science areas where results are more reliable. Medical research is notoriously flawed which leading experts and polymaths agree on, here is a link from one of the worlds best known authorities on the subject. Shockingly 90% of medical research is estimated to be flawed.

    Why are correlations in medical research less reliable indicators than in other sciences? Because the basic units are not consistent, An atom is same today as it was 100 years ago, or the same on Earth as it is on Jupiter, but human biology is very complex and definitely varies from person to person. One of the key questions that should be asked of any study is “to which population do these results apply?” The answer is in the case of medical research, in the strictest sense, only to the population studied. The mathematics of statistical analysis is relatively straight forward and is canned in software programs, but the interpretation of results from any study is a different skill altogether and may even be considered an art.

    I am not taking any stand on the healthfulness of wheat as a food, just frustrated on the unreliability of medical research results. We know drug companies influence medical studies, but I suspect that pop culture ideas and attitudes also pervade in selecting and analyzing the results in some cases. I think that correlation studies are useful as an heuristical tool to promote further analysis when a strong association is indicated.

  13. I agree with Roger. Many bold statements made here throughout the comments, and premature assumptions without hard evidence – which the China Study does not show despite Denise’s correlations. I understand the reason people are so strongly hating on wheat is that they’ve stopped eating it for a few months, or years, and they “feel great”. What could be the cause of that? What are they replacing it with in their diet? Vegetables, fruits? Enjoying the high of hypoxanthine in meat? Or not much, which means a reduction in overall intake of calories that might have been in excess before? Is the weight loss fat or body mass in general – could the loss possibly not actually be a ‘good thing’? (stunted growth, deficiencies) It’s something to think about, and if you want your opinion respected it would be better if it wasn’t only opinion – you know what I mean. Everybody has a friend who is always raving about the latest thing who jumps ship at the next latest thing and blows with the proverbial and hypothetical wind.
    I work out at CrossFit, and I listened to the paleo pitch. It has a lot of holes. I’m going to play along…If that’s the way everyone ate for so long (which is a bold claim in itself), why did they stop? And what happened in the world after they did? Why is a poster above saying we have a population explosion in an agricultural society wherein we should all have short, painful lives as described by another poster? Why are the people who eat strict paleo (a diet ascribed to a people that the “ascribers” invented) so aggressive and short-tempered, and constantly obsessed with what they eat and how it will balance them out and reduce their inflammation? Ok, that was a stereotype, but really – I’ve observed a personality change in a few individuals after eating this way for a time. This is also a diet which gets abused, as it differs in no way from the SAD in that people focus on the meat part and overlook the fruit and veggie part, and still struggle with dessert and recreational substance vices.
    I have a hard time believing that a plant which looks and acts like a lot of other plants is as poisonous as some claim. What hasn’t been mentioned here is that there is a population of people who enjoy a healthy existence and *gasp* eat wheat. No bloating and indigestion. No weight problems. Strong, athletic, brilliant, talented, happy, well-balanced people who manage life just fine. Maybe we should look at them.
    Something’s just not right here. Wheat is valuable. It has a purpose. I believe it’s as food. You may not believe that, but you are still not convincing enough. I would think more should be done to look at
    a) corrupt supply
    b) gross misuse
    This would be more helpful.
    D&C 89

    Also, on a side note, to the woman bemoaning vaccination. I read the research, the blogs, listened to both sides. What I found was that no proof has been provided that immunization causes the host of ills it’s fingered for (other than anaphylaxis), but it’s clearly proven that immunization does prevent a host of ills. My friend just endured watching her perfectly healthy 6-month old niece die of whooping cough in an outbreak in an area with a 48% immunization rate. She is not the first and it will get worse. Smart? I would say that until it happens to you, you cannot imagine the hurt immunizations can keep us (and others) from.

  14. First – I love Demise’s page. It is my favourite even amoung the other sites that use scientific data to give nutritional advices. But I’m not used to analysing data myself. So I really tried to figure the variables of the 1989 China Study II questionnaire graph using the site Demise referred to ( but I still don’t get them… is that daily wheat in grams? mortality in a mil? I’d be very happy if anyone enlightened me ^_^°

  15. I couldn’t possibly read all of these comments so I read none. You will excuse me then if I am repeating someone’s observation. In the rabbit study mentioned at the end it may have been WGA, wheat germ agglutinin, that caused the atheromatous lesions. It’s known for being quite nasty to endothelium. It’s a lectin like ricin eeeks.

  16. WOW! serious amount fo comments…..I think support for your osition has come in William Davi’s “Wheat Belly”, a cardiologist who was recently on CBS talking about his book & how wheat is poison for us…….I have the book making it’s rounds with my patients.

    1. Hi Carla, Just be sure not to get sucked into the nonsense that wheat is the worst or only grain to be cautious of. True WGA is toxic, and many, many people are being harmed by the gliadins and exorphins. But to suggest that a very “chubbers” patient can drop 50 pounds in a couple months by simply not eating wheat borders on fraud. All grains are “new foods” and as such promote ill health when eaten in large quantities. In my upcoming book (go to for excerpts from various chapters), I list over a dozen serious problems that lead to obesity and some will shock you! The point is that although having a cute name like Grain Gut or whatever is well cute, it oversimplifies the problem of obesity. just like Dr. Lustig saying HFCS is the only cause of obesity.

      Sorry but to get lean there is no magic-bullet-single-food-theory that works. I have been zero grains for 6 months now and my pants size keeps shrinking but my weight has incredibly stayed almost exactly at 165 # but I’m now over 3 inches leaner from a 32 down to a 28/29, oh did I mention that I feel great?

      A simple look at the glycemic indices of a few other grains/starches reveals corn flakes to be way worse in elevating blood sugar clocking in at if I remember correctly 88?, as well as the baked russet potato with a GI over 100.

      Wheat’s paltry contribution is a GI of about 71, durum semolina pasta is about 41 which makes some forms of wheat simply ducky . I am using the Harvard GI’s of 100 common foods as a reference since it is the most accurate and the one professor Willett uses.

      Davis’s book is really just a repackaged low carb diet book. The contents of which have been better touted by many before him such as Barry Sears, Atkins, and Dr Eades. As such he even says as much (in the fine print only of course) when commenting on losing body fat that simply stopping wheat and eating other grains is not likely to cause weight loss (but you just said that if I stop wheat…..duh).

      Which is why I call it a “grain belly” because all grains make you fat, while some grain like corn in a flaked form is a far greater poison as far as glycation and elevating serum insulin and blood sugar.

      There is nothing new in that book. For a better treatise on low carb read the Anti-inflammation Zone/the Omega Zone/Toxic Fat all by Sears, Protein Power, The Paleo Prescription, the Paleo Solution-here he includes useful exercises for the novice which is very nice, and one more: the groundbreaking Paleolithic Prescription first published way back in the 80’s.

  17. I’m not sure why but this web site is loading incredibly slow for me. Is anyone else having this issue or is it a issue on my end? I’ll check back later on
    and see if the problem still exists.

  18. Dear Denis,
    thank you for being so transparent with your analysis !
    I personally find it difficult to believe any study which makes it impossible to reproduce at least the more basic graphs.
    In this case, I followed the link to the “already available online data” from Oxford and downloaded the 1989 data:

    I was amazed at the percentage of missing values and wanted to confirm these with you or anyone else following this blog:
    – 415 out of 621 records have no entry for wheat flour intake
    – From these 206 records 139 do not have any corresponding values for any of the suggested mortality variables M059, M065, M063, M062.

    So when I join the mortality data with the wheat intake column I am left with a pitiful 67 rows, that cannot possibly be true ?

    Thanks a lot!

        1. You’re welcome, Markus. My current plan is to attempt reproduction of all important figures and statistics from these articles. If you’d like to help with that (it’s pretty easy so far) or add some original research, contributions are most welcome!

  19. Hi! I know this is kind of off topic but I was wondering which blog platform are you using
    for this site? I’m getting fed up of WordPress because I’ve had problems with hackers
    and I’m looking at options for another platform. I would be fantastic if you could point me in the direction of a good platform.

    1. Update: those links are now dead and replaced by the full reproduction attempt, which lives at

      I get a different correlation values for “Correlation between wheat flour and coronary heart disease” at the top of the article, as well as the following correlations with wheat intake:
      – “Total animal protein intake”
      – “Fat as a percentage of total calories”
      – “Total cholesterol”
      – “Apolipoprotein B”
      – “Daily alcohol intake”

  20. I recently read an article (supposed to published in Lancet) about mummies that were scanned (CT?) for heart disease. Something like 43% had some form of heart disease. The mummies where from various places in the world. Their conclusion- maybe humans have a genetic predisposition for heart

    I began to wonder what all the societies these mummies had in common. Could it be agriculture and SEEDS?

  21. Can someone explain (or point me to a source that can) where correlation figures come from….for example, how do we know that wheat correlates with CAD @ .67)?

  22. The problems with this blog is that most of normal readers don;t know statistics, correlation’s etc. It looks very impressive your blog, so many numbers, long writing, scientifically sounding conclusions. However someone soon might be analysing your entry with similar manner and proving that you are not doing it correctly. You might be famous by that time though!

  23. Great post, I spent some time looking purely at the correlations, and there seems to be a flip side to heart disease promoting or reducing factors that I haven’t quite been able to flush out.
    You may be able to statistically work out what the hell is going on. Have a look at the following variables in CSII dataset:

    * Wheat flour
    * Rice
    * Proline
    * 15:1
    * 18:1
    * Vascular disease mortality
    * Infectious disease and parasitic mortality.

    Wheat flour, Proline and 15:1 on one side and Rice and 18:1 on the other side seem to for the 3 headed and two headed cluster peaks between two scarringly inversely correlated mortality causes. Have a look at these:

    I would really be interested in any tips you would have at looking into this deeper and if you think my conclusion (females should swap pasta for rice) is a sane one. Can we flush out the real factors? Is it wheat versus rice, 15:1 versus 18:1? Can we eliminate anything? Is it anyway pro/anti-inflamation related? It’s fun working with this data set, but the conflated Xiang level data points and their implicit clustering and attenuative nature are so frustrating. I really wish those questionnaires results, etc, would be available in a less clustered way in whatever digital form.

  24. With regard to the question, why does wheat increase HD?, one hypothsis is that wheat promotes the production of small dense LDL particles which are in turn more atherogenic (hope I have spelt that right) or to put it another way small means more prone to oxidisation and attachment within lesions of the artery wall.

  25. Denise if you are looking for a way to fund all this nutritional research I have a shed load of UK horse racing data that would keep you up all night 🙂

  26. I have been taking a look at the raw China Study data and in particular with regard to wheat. I loaded the data up into Excel and tidied up the data to do a simple check on wheat and IHD

    The average incidence of heart disease across all regions is 20.38 (HEART DISEASE AGE 35-69 (stand. rate/100,000)
    The average intake of wheat is 130.08 grams per day

    Now if we look at the numbers for heart disease when the wheat consumption is below average we get heart disease at 15.32
    However for heart disease when wheat is above the average of 130.08 we get heart disease at 28.56

    As we increase wheat to >140.08 we get HD at 29.49
    Wheat > 150.08 HD 29.68
    Wheat > 170.08 HD 30.6

    The other interesting feature is that if you lower wheat consumption below average HD does not drop it remains fairly constant suggesting that there is a level in which when exceeded HD starts to take off. That levels appears to be 210 grams per day, above that and HD starts to take off. In summary the numbers tend to suggest that eating up to 210 grams per day of wheat is not a big deal but after that HD kicks in. Note I am simply relaying the figures here, sure maybe low wheat eaters are non smokers or whatever

  27. In his rebuke of Denise Minger Campbell makes a valid point that univariate analysis is fraught with danger.
    What he means is analysing one variable to an effect can be misleading as other variables may be playing a part.

    For example he cites that with wheat there is a predisposition for higher wheat consuming regions to be lower green plant eating regions. This would tilt the results somewhat if we accept that green plant is protective.

    Indeed there are 23 regions where wheat exceeds grenn plant consumption and 41 where green plant exceeds wheat.

    Looking at wheat alone ands comparing it with average consumption of wheat across all regions, when it is above average heart disease stands at a high 23.24

    But what would we expect to happen if wheat was greater than average wheat consumption but so too was green plant greater than average green plant consumption. Would we expect HD to be below 23.24 ?. Well it comes in at 26.33 from 7 regions

    Wheat greater than average but plant less than average HD = 21.8 from 15 regions

    Wheat less than average but plant greater than average HD = 8.91 from 19 regions

  28. Very interesting indeed, but feeding mice wheat germ doesn’t equal feeding them wheat, anymore than feeding them casein equals eating meat. What kind of wheat? How was the wheat processed? How was it consumed? These are important questions that need answers before I give it up, over my dead (non-heart diseased) body! Those who are genuinely allergic to wheat, yeah, give it up. But I’m not allergic to it, so if I gotta go, I’d rather it be death by Ruben sandwich, than some random accident.

    1. How do you know you are not allergic, perhaps wheat allergy is not a binary condition but a question of degree. Maybe you are on a scale of 1 to 10 you have a 2/10 allergy which is not enough to cause day to day problems but over 50 years it is. One simple experiment you could do is give up wheat for a month or two and see how you feel. Check out whether you lose weight, sleep better, have more energy or perhaps no effect. Be your own experiment, find out what suits you

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s