Alright folks, I’ll be honest. I was not expecting my China Study critique, which started as a nerdy personal project pursued in the wee hours of the morn, to generate much interest. Like most of my weird projects, I figured it would be briefly perused by a few number-lovers before fading quietly into the abyss of cyberspace.
Instead, it went viral and racked up 20,000 page views within 24 hours.
I’m surprised, but equally thrilled. My self-marketing skills are pretty dismal, and it was only by the grace of all the bloggers who featured my critique that this page-view boom occurred. Thank you to everyone who helped spread the word. I owe y’all!
This post is going to be quite long (no shocker there) and, in places, a bit more technical than the last. I know not everyone digs science mumbo-jumbo, so I’ll try to keep that to a minimum and explain things like journal quotes in simpler terms.
First, I’d like to address a couple points I’ve seen crop up in reader comments and emails I’ve received.
One: My graphs and simple statistical explanations. The graphs I posted were not intended to stand as new hypotheses or conclusions about the data. I apologize if I didn’t make this abundantly clear. Their sole purpose was to demonstrate, to the general layperson, how raw correlations (in the instances Campbell used them) can be misleading—as well as show how dramatically a single confounder can affect a correlation and make a positive trend appear where there may not be one at all. The graphs and explanations were meant to be illustrative, not exhaustive.
Two: Bias in Campbell’s representation of the data. This is a point I feel has been overlooked by some critics who’ve myopically targeted my use of statistics.
My biggest concern is with the way data appears to be cherry-picked to create a “plant foods are good” and “animal foods are bad” dichotomy when the actual data from the China Study (as well as from Campbell’s own research) does not reflect this.
For instance, when citing the anti-disease effects of plant foods, Campbell points to inverse correlations with biomarkers for plant food consumption as well as plant food intake itself. One example is in Claim #5 when he notes stomach cancer is inversely associated with plasma concentrations of beta-carotene and vitamin C (biomarkers) as well as with green vegetable intake (a plant food). (Both of these claims are based on uncorrected correlations, by the way.)
Yet when citing the purportedly harmful effects of animal foods, Campbell relies on blood markers (usually total cholesterol or apo-B) but fails to find direct relationships between disease and the animal foods themselves. He still indicts animal foods as harmful, but comes to this conclusion by enlisting the help of intermediary variables. And as I explained in the last post and will continue explaining in this one, the link between cholesterol levels and the diseases Campbell links them to are not even as straightforward as he suggests.
To those who approach this discussion already believing animal foods are generally unhealthy, this bias is subtle and might not be obvious. But to those who approach this discussion from a place of neutrality, the bias is unmistakable.
Response to Campbell
Now, onto business.
In case you haven’t heard yet, the much-discussed T. Colin Campbell wrote a response to my critique of his book. If you haven’t already done so, hop on over and read it on Tynan.net.
Let me preface this with something important. When it comes to science, my motto is an old line from Dragnet (which, having no TV, I’ve never actually watched): “Just the facts, Ma’am.” Or sir. Science itself should be cool, neutral, and somewhat soulless. As far as I’m concerned, personal conflicts, drama, mudslinging, grudges, and other flurries of emotion should be locked out of science’s doors and banned for life.
For this reason, I want to make it clear that even though I disagree with Campbell’s interpretation of the China Study data, I have no interest in launching any personal vendetta against him. I know some readers are none too pleased with the man, but I do believe he’s trying to promote a message he deeply believes will help others. I won’t be participating in any character attacks, regardless of how I feel about his interpretations.
That said, I’m a bit disappointed Campbell didn’t offer a more revealing glimpse into his own methods of analysis. Here’s a secret: If he wants to silence his critics, all he has to do is publish the details of his process—which, apparently, he has already written up:
A more appropriate method is to search for aggregate groups of data, as in the ‘affluent’ vs. ‘poverty’ disease groups … I actually had written material for our book, elaborating some of these issues but was told that I had already exceeded what is a resonable [sic] number of pages. There simply were not enough pages to go into the lengthy discussions that would have been required–and I had to drop what I had already written.
I’ve emailed Mr. Campbell and asked him to consider publishing this material somewhere as a downloadable PDF or in another accessible form. Then the rest of us can study his methodology, look for oversights, and hopefully replicate his findings. I’ll update this when or if he responds.
UPDATE: Campbell has informed me via email:
To go back and fetch the material that I had previously written would take a lot of time that I don’t have. Also, much of it is in my peer-reviewed 300+ scientific papers.
Well, shucky darns. Although he doesn’t have time to fetch already-written material, he does have time to craft a more thorough response to my critique than the one published on Tynan.net, which he’ll be posting on his website sometime soon. Hopefully he’ll provide more details about his methods there.
Some responses to specific parts of Campbell’s letter
To clarify who’s saying what, quotes will always be italicized and indented. All bold parts of the quotations are my own emphasis and not the original author’s.
Campbell: She claims to have no biases–either for or against–but nonetheless liberally uses adjectives and cutesy expressions that leaves me wondering.
News flash: I was an English major with a creative writing emphasis. Cutesy is my thang. When the occasion calls for it, I can become Formal Monotone Academic Denise—but seeing as this is a blog and I needed to keep readers hooked for 9,000 words, I figured a more colloquial tone would be best.
Also, I wasn’t aware adjectives indicated bias, and if that’s the case, boy am I ever in trouble. You know what else? I sometimes use adverbs. That’s right. Evil adverbs. I learned them from Stalin when we worked together in the ’40s (oops, did I say that out loud?).
Campbell: As far as her substantive comments are concerned, almost all are based on her citing univariate correlations in the China project.
Actually, they’re based on the univariate correlations that Campbell cited first.
If you read my critique, you’ll see that Campbell’s claims align with the raw and uncorrected data, which—as I tried to illustrate—can be misleading due to the influence of other variables implicated with disease.
But it seems my critique wasn’t enough to convince some Campbell supporters that he did not use exhaustive analytical methods under some important circumstances, so I’ll present examples straight from his peer-reviewed papers.
First, let’s look at “Diet, Lifestyle, and the Etiology of Coronary Artery Disease: The Cornell China Study” published in the November 1998 issue of the American Journal of Cardiology. One statement from the paper, from the section discussing “Diet-Coronary Artery Disease Relations,” notes the following:
The combined coronary artery disease mortality rates for both genders in rural China were inversely associated with the frequency of intake of green vegetables (r = -0.43, p<0.01)…
Remember the “Green Veggie Paradox” from my last post, which pointed out that frequency of green vegetable consumption may be a geographical marker for southern regions where heart disease rates are lower, instead of an actual protective agent against heart disease? Well, here’s that paradox again. In a peer-reviewed article. Co-authored by Campbell. Using the raw data (-0.43). And neither him nor the rest of his team made adjustments, ran more sophisticated analyses to account for confounding variables, or even mentioned other factors that could explain the correlation between frequent green-vegetable consumption and healthier hearts.
Had Campbell tried to understand the apparent discrepancy between frequency of green vegetable consumption (which had a strong inverse association with coronary heart disease) and the amount of green vegetables consumed (which had a weak positive association with coronary heart disease), he may have realized there was more to our fibrous friends than meets the eye. For instance, geography is closely tied to heart disease in the China Study data, with lower latitudes exhibiting lower rates. And if frequency of green vegetable consumption strongly reflects geography, it seems any researcher committed to accuracy would want to tease apart these variables before citing them in a scientific paper.
In this article, Campbell also employs several other unadjusted correlations straight from the monograph:
The combined coronary artery disease mortality rates for both genders in rural China were inversely associated with … plasma erythrocyte monounsaturated fatty acids (r = 0.64, p<0.01), but positively associated with a combined index of salt intake plus urinary sodium (r = 0.42, p<0.01) and plasma apolipoprotein B (r = 0.37, p<0.01).
These numbers are all raw correlations. Campbell didn’t conduct a deeper statistical analysis on any of it to account for potential confounders, such as lifestyle habits or other dietary factors that might accompany specific biomarkers.
These apolipoproteins, in turn, are positively associated with animal protein intake (r = 0.26, p <0.05) and the frequency of meat intake (r = 0.32, p<0.01) and inversely associated with plant protein (r = 0.37, p <0.01), legume (r = 0.26, p<0.05), and light colored vegetable intake (r = 0.25, p <0.05).
Again, we have a match with the uncorrected data. And again, Campbell and his team didn’t appear to run multiple variable regressions or any other analyses to see if the raw data was accurate. (And notice how Campbell can’t say animal protein itself associates with heart disease, but has to pull a connecting variable into the picture to make his theory fit.)
Why didn’t Campbell pay more attention to the role of confounders? Why did he accept the raw data, which showed plant foods as protective and an animal-food biomarker as harmful, without conducting deeper analyses?
This might be the answer:
The principal hypothesis of this study was that the greater the dietary proportion of a variety of good-quality plant-based foods, the lower the rate of chronic degenerative diseases.
Essentially, Campbell and his team approached the data set specifically looking for trends showing plant foods to protect against disease (and, perhaps, showing animal foods to be harmful).
If you’ll recall, the China Study has 8,000 statistically significant correlations. That’s a lot. Enough, in fact, to find pretty much anything you want if you look hard enough—especially if you use a bit of sloppy science and cite raw correlations or chains of variables when they suit your needs.
Of course, that’s not the only China-Study-based paper showcasing analytical shortcomings. Let’s look at “Fish consumption, blood docosahexaenoic acid and chronic diseases in Chinese rural populations” published in the September 2003 issue of Comparative Biochemistry and Physiology. This paper examines the role of fish and the essential fatty acid DHA in relation to several diseases.
Campbell and his crew’s methodology for studying the variables:
Pearson’s correlation coefficient was used to explore the relationship between variables. The two-tailed test of significance was used to examine the significant differences within variables.
Alright, this is your standard high school stuff: examining the linear relationship between two variables. No multiple variable regressions. No adjustments for confounding variables. And from these rudimentary correlations, Campbell and his team cite a number of observations about the relationships between fish, other meat, total lipids, blood markers, and disease, ultimately concluding:
[T]he protective nature of DHA or aquatic foods is intrinsic and global, with implications for health world wide. The decline in sea and fresh water food consumption in many regions last century could be an adverse, contributory factor to the increasing risk of chronic diseases and the rise in mental ill health …
Researchers concluded from these raw correlations that the DHA is associated with lower risk of many chronic diseases. But might this effect become even more pronounced through different statistical models—namely ones that account for confounding variables?
It seems likely, and here’s why. In this paper, Campbell and his team noted that diabetes is positively associated with DHA in the China Study data, despite other research showing the opposite:
Diabetes showed a positive but non-significant relation with DHA in Fig. 2, which meant no clear-cut conclusion about the efficacy of lower DHA level in diabetes even though a negative association has been found between DHA and triglycerides in plasma [in previous research]. … [We] have no explanation for the positive correlation with diabetes.
No explanation, eh? I’ve got one. In this paper, Campbell and other researchers determine that fish is the most significant source of DHA in the studied counties. And we know from the China Study monograph that fish-eating regions tended to have high intakes of processed starch and sugar compared to other counties—a correlation of 0.58. Could processed sugar and starch intake be skewing the relationship between DHA and diseases like diabetes? If so, why didn’t Campbell et al run more appropriate analyses to account for this?
Still not convinced Campbell’s methods are less than perfect? Here’s some more. From “Diet and chronic degenerative diseases: perspectives from China,” published in the May of 1994 issue of the American Journal of Clinical Nutrition:
Intakes of 14 complex carbohydrate and fiber fractions were obtained in this study to determine whether particular fiber fractions were associated with particular diseases, especially cancers of the large bowel. … Based on an overview of the univariate correlations, colon and rectal cancer mortality rates were consistently inversely correlated with all fiber and complex carbohydrate fractions except for pectin, which showed no correlation.
So here we have Campbell and his team using univariate correlations to look at the relationship between fiber and colorectal (large bowel) cancer. No adjustments made for potential confounding variables. And from these correlations, he concludes:
[T]here is evidence of a weak inverse relationship between cancer of the large bowel and the intake of multiple complex carbohydrate and dietary fiber fractions.
In other words, the fiber fractions seemed to protect against colorectal cancer across the board. But is this an accurate inference?
Had Campbell looked more closely at the data (instead of assuming the raw figures were accurate, as he seems fond of doing when it supports anti-disease properties of plants), he would’ve noticed something striking. The correlations between those 14 fiber fractions and colorectal cancer seem to mirror the correlations between the fiber fractions and schistosomiasis infection.
Okay, I know what you’re thinking. “What’s Denise blathering on about this time?” Let’s back up for a minute.
Schistosomiasis (also called bilharzia) is a parasitic disease known to raise risk of colorectal cancers. If you get infected with one of these lovely worms, they’ll lay eggs that travel to your liver, intestine, or bladder, where they can cause permanent damage and inflammation. How fun!
The link with colorectal cancers isn’t something I’m just pulling out of the air, by the way. It’s pretty well established. Some references:
- “Schistosoma japonicum and colorectal cancer: an epidemiological study in the People’s Republic of China“: Prevalence of infestation with Schistosoma japonicum was highly correlated with mortality from colorectal cancer in 89 communes in four counties of Jiangsu province, China (rank correlation coefficient = 0.68) in 1973-75, and with incidence of colorectal cancer in 24 communes of Haining county, Zhejiang province in 1977-79.
- “Correlations of colon cancer mortality with dietary factors, serum markers, and schistosomiasis in China“: [P]revalence of schistosomiasis was significantly correlated with increased colon cancer mortality.
- “Schistosomiasis and its prognostic significance in patients with colorectal cancer. National Cooperative Group on Pathology and Prognosis of Colorectal Cancer“: This paper analyses 430 cases of colorectal cancer complicated with schistosomiasis. The 5 year survival rate was 45.6%, lower than that without schistosomiasis. … The infection of schistosome should be considered as one of the important factors in prognosis. (In other words: Schistosomiasis infection increases the mortality rate of colorectal cancer sufferers.)
- “A cohort study on the causes of death in an endemic area of schistosomiasis japonica in Japan“: These results suggest that schistosomiasis japonica is one of the important risk factors for cirrhosis of the liver, cancer of the liver and cancer of the colon.
With that in mind, it seems pretty obvious that Campbell would want to look at a schistosomiasis infection in relation to colon cancer occurrence, especially since 1) it’s pretty common in Asia and 2) it could be a confounding variable. In fact, schistosomiasis has a correlation of 0.89 with colorectal cancer mortality in the China Study data. (If you’re having a déjà vu moment, you’re not crazy: I wrote about this in the previous entry as well.)
So what does this have to do with fiber?
The fiber fractions Campbell cites as having a “weak inverse relationship” with “cancer of the large bowel” also have a somewhat stronger inverse relationship with schistosomiasis. In other words, fiber is already likely to be associated with less colorectal cancer simply because those who eat more of it tended to have less of another significant risk factor.
It might help to represent this visually, so here’s a graph plotting each fiber fraction’s correlation with schistosomiasis and colorectal cancer. These are the fiber fractions corresponding to the x-axis numbers:
- Total fiber
- Total neutral detergent fiber
- Hemi-cellulose fiber
- Cellulose fiber
- Lignins remaining after cutin removed
Bottom line: Is the inverse relationship between fiber and colorectal cancer legitimate, or is that correlation influenced by schistosomiasis rates? Given the relationship between these variables, shouldn’t Campbell have run a more thorough analysis on the data?
I sure think so. But he didn’t. Again, he seems to readily accept uncorrected correlations when they prove his theory.
So, what happens when we do adjust for confounding variables? Let’s look at another of Campbell’s peer-reviewed papers: “Erythrocyte fatty acids, plasma lipids, and cardiovascular disease in rural China” published in the December 1990 issue of the American Journal of Clinical Nutrition. Here were their statistical methods:
To adjust for the effect of other factors in the relationship between two variables, ordinary least-squares multiple-regression analysis was used. Natural logarithmic transformations of the mortality rates (the dependent variable in the models) were used to obtain a normal distribution of the outcome variable for reliable statistical significance testing of the regression coefficients.
No uncorrected correlations here. And the results:
Within China neither plasma total cholesterol nor LDL cholesterol was associated with CVD [cardiovascular disease]. The results indicate that geographical differences in CVD mortality within China are caused primarily by factors other than dietary or plasma cholesterol.
Did you catch that? After adjusting for confounding variables, researchers found that cholesterol was not associated with cardiovascular disease in the China Study data. And that includes both blood cholesterol and cholesterol from food.
Let that sink in for a moment.
Nah, this is pretty big: Give it two moments.
Now, let’s look at Campbell’s next point, which flows quite nicely from the last:
Diseases of affluence and diseases of poverty
Campbell: A more appropriate method is to search for aggregate groups of data, as in the ‘affluent’ vs. ‘poverty’ disease groups, then examine whether there is any consistency within groups of biomarkers, as in considering various cholesterol fractions.
If you’re unfamiliar with Campbell’s disease-clustering strategy, you can read “From Diseases of Poverty to Diseases of Affluence” to get a feel for it (although be warned, the formatting is a little wonky). In essence, Campbell examined the China Study data and identified two distinct groups of diseases that were generally associated with each other—with one group representing diseases common to developing nations and the other representing “Western” afflictions.
In the article linked above, Campbell et al describe the first group:
As expected, diseases of poverty are associated more with agricultural than with industrial activity. Areas where these diseases are common are located further inland where mean elevation is higher and overall economic activity, literacy and population density are lower.
And the second group:
In contrast, diseases of affluence are found in the more densely populated rural areas nearer the seacoast where industrial activity and literacy rates are higher and more fish, eggs, soy sauce, beer and processed starch and sugar products are consumed.
More specifically, Campbell defines the “diseases of poverty” as:
- Intestinal obstructions
- Peptic ulcer
- Other digestive disorders
- Pulmonary tuberculosis
- Infectious diseases (other than schistosomiasis)
- Rheumatic heart disease
- Metabolic and endocrine disease (other than diabetes)
- Diseases of pregnancy and birth (other than eclampsia)
And “diseases of affluence” include:
- Stomach cancer
- Liver cancer
- Colon cancer
- Lung cancer
- Breast cancer
- Coronary heart disease
- Brain cancer (ages 0-14 years)
Again, the diseases in each cluster tend to associate positively with each other but inversely with the diseases in the opposite group.
It’s not a bad strategy, really. Campbell uses this two-group method to identify general factors (such as nutritional patterns) related to each disease cluster, taking a holistic view of disease rather than examining ailments through reductionism. This approach aligns with something I very much agree with: that diseases don’t happen in isolation, but that multiple forms of chronic disease can spring from the same cause (poor nutrition, processed foods, unhealthful living, and so forth).
But while I agree with this general method, it’s not without flaws—and the way Campbell employs it to study nutrition and disease requires a few leaps of faith.
First, some problems with the groups Campbell created:
- Not all of the “diseases of affluence” are actually common in affluent countries, raising questions about whether these disease clusters apply outside of China. For instance, the two most prevalent diseases of affluence in the China Study data are liver cancer and stomach cancer—but in the US, a decidedly affluent nation, these diseases make up less than 5% of all cancer deaths.
- Where’s “stroke” on either list? Nowhere to be found. Campbell had to create a third group called “Other” for a few diseases that didn’t fit cleanly into the other two clusters. According to the American Heart Association, stroke is currently the third leading cause of death in America. So what explains its lack of correlation with other diseases of affluence? Campbell offers no insights.
Perhaps more importantly, Campbell makes some excellent observations about the nutritional variables correlating with diseases of affluence, but then dismisses them without any satisfying or even logical explanation. He lists the following correlations between several foods and his affluent disease cluster:
- Processed starch and sugar: 0.51
- Fish (g/day): 0.56
- Beer: 0.59
- Eggs (times per year): 0.31
Since the industrialized areas with diseases of affluence tended to be near the coast, it’s not surprising fish consumption was high. But that’s a pretty hefty correlation with processed starch and sugar, too. Could those refined carbs contribute to diseases of affluence? Eh? Eh?
Apparently not. Campbell doesn’t consider them significant in the China Study data. He states that “beer and processed starch and sugar products are also consumed in much lower quantities [than in the US],” and therefore “consumption of these foods is probably more indicative of general economic conditions and other local circumstances than of biological relationships to disease.” And that’s the last we hear about ‘em.
That’s right, folks.
Here we have evidence that areas in China with the highest rates of Western-type diseases also eat the most processed starch and sugar. Maybe not in the grotesque amounts that Americans eat them, but then again, China’s “affluent disease” rates were also lower than America’s.
But instead of examining the relationship between processed carbohydrates and poor health, Campbell zeros in on another variable associated with industrialized nations and diseases of affluence. And if you’ve been paying attention to this post and the last, that variable won’t surprise you: It’s cholesterol.
By the way, the correlation between Campbell’s affluent diseases (in the aggregate) and cholesterol is 0.48, slightly less than the correlation with processed starch and sugar. And if you’ll recall, Campbell’s own analysis showed that cholesterol levels in the China Study data didn’t associate with cardiovascular disease, a major cause of “affluent” mortality. But I guess that doesn’t matter, because Campbell says so and Campbell has lots of credentials.
But back to Campbell’s response. His statement that a more appropriate method of analysis is to “search for aggregate groups of data, as in the ‘affluent’ vs. ‘poverty’ disease groups, then examine whether there is any consistency within groups of biomarkers” is something I can at least partially agree with. Yet in examining Campbell’s own use of these disease groups, I smell another whiff of bias: He immediately implicates cholesterol (and, as a consequence, animal products) as causative of disease, when at least four other diet variables (most notably processed starch and sugar) are also heavily implicated with diseases of affluence.
Now, for something completely different:
The “Mysterious Tuoli” not so mysterious?
Campbell: [W]e discovered after the project was completed that meat consumption for one of the counties, Tuoli, was clearly not accurate on the 3 days that the data were being collected. On those days, they were essentially eating as if it were a feast to impress the survey team but on the question of frequency of consumption over the course of a year, it was very different.
I’m glad Campbell pointed this out (and I’ll be updating the Tuoli page to reflect it), but meat was not the component I found notable with the Tuoli diet: dairy was. Assuming the frequency questionnaire was more reliable than the three-day diet survey, the Tuoli still consumed dairy most days of the year and still consumed nearly no vegetables (twice per year), nearly no fruit (once per year), and ate wheat as their primary plant food. Not exactly a balanced diet—yet, compared to the rest of China, they remained in good health.
(By the way, a number of you have asked for help finding more information about the Tuoli. A Google search for “Tuoli” doesn’t reel in a whole lot of relevant hits, so you can try the alternative English spelling of “Toli,” or a search for a related group of people called “Uyghur” or “Uygur” in the Xinjiang Autonomous Region of China.)
However, Campbell’s statement about the unreliability of the diet survey for the Tuoli also calls into question the validity of the three-day diet survey as a whole—as well as the significant observations Campbell gleaned from it. For instance, on page 99 of “The China Study,” Campbell notes:
Average calorie intake, per kilogram of body weight, was 30% higher among the least active Chinese than among average Americans. Yet, body weight was 20% lower. How can it be that even the least active Chinese consume more calories yet have no overweight problems? What is their secret?
If Tuoli is any indication, there may not be a secret at all. Since Campbell drew his calorie data from the three-day diet survey, suppose multiple counties tried to impress researchers by “feasting” or otherwise altering their eating habits to reflect greater wealth, prosperity, or food abundance than they actually had. The result? Calorie intake during those three days would be higher than for the rest of the year, leading to an overestimated average calorie intake for the 65 counties studied.
Did Campbell consider this, especially given his awareness about the unreliable records for the Tuoli? Apparently not. On page 101, he states:
Chinese consume more calories both because they are more physically active and because their adoption of low-fat, low-protein diets shifts conversion of these calories away from body fat to body heat. This is true even for the least physically active Chinese.
Physical activity certain plays a role in higher calorie requirements, but eating a low fat, low protein diet may not increase thermogenesis as Campbell suggests—at least not based on the China Study data. Some counties may have simply been showing off by stuffing themselves silly, leading to high average calorie intakes. We’ve got Campbell’s assertion that at least one place did this: How do we know others didn’t as well?
Again, let me highlight what appears to be another link in a chain of bias: Campbell dismisses the low disease rates and high animal protein intake of the Tuoli because the three-day diet survey was inaccurate, yet doesn’t account for potential shortcomings in that diet survey when it helps score brownie points for plant foods.
Campbell: One final note: she repeatedly uses the ‘V’ words (vegan, vegetarian) in a way that disingenuously suggests that this was my main motive.
I understand—and respect—that Campbell was trying to avoid the ethical implications of the word “vegan,” since the term often conveys a complete lifestyle choice rather than just a diet. However, my intent was definitely not disingenuous, nor was I trying to peg a motive on Campbell. My own use of the term “vegan” was simply to describe a completely animal-product-free diet. I apologize if this wasn’t clear from my post.
Campbell: One further flaw, just like the Weston Price enthusiasts, is her assumption that it was the China project itself, almost standing alone, that determined my conclusions for the book (it was only one chapter!).
I guess Campbell missed the 2,135 words I dedicated to his research on casein, including the problems with extrapolating its effects to all animal protein. And the citation of his own research showing it’s a full spectrum of amino acids, not just animal protein, that apparently spurs cancer in aflatoxin-exposed rats. And the insight that a vegan diet provides all amino acids (and thus complete protein) if you eat a variety of plant foods, thereby posing similar purported risks as omnivory in terms of cancer growth. And the question about the apparent unhealthfulness of breastfeeding and exposing young, delicate-bodied children to casein. And the glaring example of bias in Campbell’s treatment of animal versus plant protein in relation to body size and disease.
Easy oversight, I guess. It was a pretty formidable post. As is this one, apparently.
By the way, if anyone had trouble following my train of thought in the casein/wheat/lysine/complete protein section of the critique, Chris Masterjohn has written a more “digestible” article (pun definitely intended) expanding on this subject and probably explaining it better than I did. Yep, that’s the same Chris who wrote a well-known critique of “The China Study” five years ago.
Next up, a very serious and momentous subject:
Does Denise work for the meat and dairy industry/is Denise a cyborg/is Denise a figment of your imagination/is Denise actually Campbell’s employee, son, dog, long-lost daughter, or alter-ego?
Campbell: I find it very puzzling that someone with virtually no training in this science can do such a lengthy and detailed analysis in their supposedly spare time.
Campbell: I have no proof, of course, whether this young girl is anything other than who she says she is, but I find it very difficult to accept her statement that this was her innocent and objective reasoning, and hers alone. If she did this alone, based on her personal experiences from age 7 (as she describes it), I am more than impressed.
Then thank you for the compliment, Mr. Campbell! I’m definitely a singular person, so I’m glad to more-than-impress you.
Initially, I didn’t want to muddy this post with retorts to statements like this, but really. What’s so hard to believe about a 23-year-old Super Nerd deciding to tackle a project out of personal interest? What do I need to show to prove I’ve got a brain in this noggin? College transcripts? 4.0, three scholarships, dean’s list, top 1% of the class? I can say the alphabet backwards, too. That has to count for something.
In all seriousness, I can understand why Campbell would express skepticism that a young person would have the resources or repertoire of knowledge necessary to tackle this sort of project. And I think it may largely be a generational issue. When Campbell was a young’un, he didn’t have access to the internet or online books or PubMed or Google Scholar or any of the other self-educational tools most of us now take for granted. For him, education did necessitate sitting in a room with a teacher, pouring over textbooks, showing up to a physical classroom, and accumulating credentials to prove you’d survived the journey. These days, education can manifest in numerous other forms.
In other words, I’m more flattered than offended.
Other odds and ends
Several readers have raised an issue that probably deserves more attention than I’ve given it so far: the limitations of the China Study itself. Although I’ve focused on examining the errors and biases in Campbell’s conclusions, the fact of the matter is, this study itself is just a big ol’ epidemiological survey—and any analyses it produces, no matter how thorough, are inherently limited due to the nature of the data.
In fact, before Campbell’s “The China Study” was even released, Thomas Billings of Beyondveg.com wrote an excellent overview of the shortcomings of the study itself. I recommend reading this if you want a better understanding of what a study like the China Project can and cannot do.
A note on wheat
I know many of you are particularly interested in the correlation between wheat and heart disease. In my critique’s gargantuan cascade of words, the two little paragraphs about wheat pinged on many readers’ radar (or, perhaps, grain-dar). I’ve already seen the “correlation of 67″ statistic thrown around the ‘net as if it’s solid evidence. Holiest of molies, that spread fast!
Indeed, I feel the China Study may hold important clues—ones that research thus far has simply not explored—about the role of wheat or wheat flour in human health. However, we can’t jump the gun yet. I will be doing some more analysis of the China Study data regarding wheat and other grains, but even if this manages to paint our glutenous friends as the most malicious of dietary villains, it doesn’t prove a darn thing.
As someone who’s massively allergic to wheat, I’d love nothing more than to shove this grain in the corner with a dunce cap and revel in my victory. Karma’s a… female dog in heat. But I can’t do that. Not yet, anyway. Bottom line, this is epidemiological data we’re working with, and it can only show correlations—not causation. Not proof. Not irrefutable evidence.
What I do hope occurs—and feel free to cross your fingers with me—is that this information snags the eye of other nutritional researchers and leads to controlled experiments about the health effects of wheat.
At any rate, I have a couple more China-Study-related posts coming up (including one with the results of multiple variable regressions), and wheat will probably be the subject of the next one. Keep your eyes peeled if this is a subject that interests you.
Summary of this post
For those of you who skipped over everything above and scrolled directly to this part, well… I don’t blame you. However, there’s really only one thing you need to know about this whole ordeal, and this is it:
- Data sets are like people. If you torture them long enough, even when they’re innocent, you’ll eventually squeeze out a false confession.
Some final thoughts, for those who haven’t clicked the “back” button on their browser yet
Although the vast majority of the feedback I’ve received (both positive and negative) has been intelligent, respectful, and ultimately constructive, I’ve received a few very fiery emails that have made me realize what a deep nerve diet debates can strike. For those whose lives have been profoundly affected—for better or for worse—by food and nutrition, diet can become a personal issue inextricably bound with identity. And as someone who’s already run through a gamut of eating styles due to allergies, ethical goals, and the pursuit of vibrant health, I know how this goes. I’ve been there. In many ways, I’m still there. For this reason, I can wholly empathize with the emotional response my critique triggered in some readers, and I understand why a backlash is apt to occur.
By the same token, I think it’s important to look at what that impassioned response signifies. Are we trying to be healthy, or are we trying to be right? Are we trying to learn, or do rigid beliefs deafen our ears to new knowledge? Have the open minds that led us to search for the truth in nutrition suddenly slammed shut, clamping tight around an ideology that may or may not truly serve us?
Critical thinking isn’t a privilege reserved for the elite; it’s a birthright. My goal is not to tell people what to think, but to show them how to think. How to sift through the vast expanse of nutritional litter and pull out the gems. How to stop blindly following the advice of so-called authorities who may not have our best interest at heart. How to think independently.
To everyone who’s taken the time to plod through this post and the last, to read, to write, to comment, to think, or to reconsider any limiting beliefs you hold about diet, I extend my deepest gratitude and wish you nothing but health and happiness.
Thank you for reading.
Coronary Artery Disease: The Cornell