Jul 20 2010
Over a year ago I wrote about The China Study, a book by T. Colin Campbell and his son based on a huge epidemiologic study of diet and health done in China. The book’s major thesis is that we could prevent or cure most disease (heart disease, cancer, diabetes, autoimmune diseases, bone, kidney, eye and other diseases) by eating a whole foods plant-based diet, drastically reducing our protein intake, and avoiding meat and dairy products entirely.
I noticed a number of things in the book that bothered me. I found evidence of sloppy citations, cherry-picked references, omission of data that contradicted the thesis, and recommendations that went beyond the data. I concluded:
He marshals a lot of evidence, but is it sufficient to support his recommendation that everyone give up animal protein entirely, including dairy products? I don’t think so.
The China Study involved 367 variables and 8000 correlations. I said I would leave it to others to comment on the study design and the statistical analysis, and now someone has done just that. Denise Minger devoted a month and a half to examining the raw data to see how closely Campbell’s claims aligned with the data he drew from; she found many weaknesses and errors.
Plasma cholesterol… is positively associated with most cancer mortality rates. Plasma cholesterol is positively associated with animal protein intake and inversely associated with plant protein intake.
The data do show that cholesterol is positively associated with various cancers, that cholesterol is positively associated with animal protein, and that cholesterol is negatively associated with plant protein. So by indirect deduction they assume that animal protein is associated with cancers and that reducing intake is protective. But if you compare animal protein intake directly with cancer, there are as many negative correlations as positive, and not one of those correlations reaches a level of statistical significance. Comparing dietary plant protein to various types of cancer, there are many more positive correlations and one of them does show strong statistical significance. The variable “death from all cancers” is four times as strongly associated with plant protein as with animal protein. And Campbell fails to mention an important confounder: cholesterol is higher in geographic areas with a higher incidence of schistosomiasis and hepatitis B infection, both risk factors for cancer.
Campbell says breast cancer is associated with dietary fat (which is associated with animal protein intake). The data show a non-significant association with dietary fat, but stronger (still non-significant) associations with several other factors and a significant association with wine, alcohol, and blood glucose level. The (non-significant) association of breast cancer with legume intake is virtually identical to the (non-significant) association with dietary fat. Animal protein itself shows a weaker correlation with breast cancer than light-colored vegetables, legume intake, fruit, and a number of other purportedly healthy plant foods.)
He indicts animal protein as being correlated with cardiovascular disease, but fails to mention that plant protein is more strongly correlated and wheat protein is far, far more strongly correlated. The China Study data show the opposite of what Campbell claims: animal protein doesn’t correspond with more disease, even in the highest animal food-eating counties.
These are just a couple of examples. Minger found many more, which she describes in her long article, complete with impressive graphs. Her exposé is well worth reading in its entirety, if only as a demonstration of how to think about epidemiologic data.
Minger goes on to reveal gaping logical holes in Campbell’s own research on casein, a milk protein that he believes causes cancer. He showed that casein was associated with cancer when given in isolation to lab animals, but he projects those results onto humans and onto all sources of animal protein. Other animal proteins have been shown to have anti-cancer effects, and the results of a normal diet containing multiple protein sources are likely to be very different from his casein-only studies.
I believe Campbell was influenced by his own expectations about animal protein and disease, leading him to seek out specific correlations in the China Study data (and elsewhere) to confirm his predictions.
She is being polite.
This is a cautionary tale. It shows how complex issues can be over-simplified into meaninglessness, how epidemiologic data can be misinterpreted and mislead us, and how a researcher can approach a problem with preconceptions that allow him to see only what he wants to see. The China Study was embraced by vegetarians because it seemed to support their beliefs with strong evidence. Minger has shown that that evidence is largely illusory. The issues raised are important and deserve further study by unbiased scientists. At any rate, one thing is clear: the China Study is not sufficient reason to recommend drastic reductions in protein intake, let alone total avoidance of meat and dairy foods.
58 Responses to “The China Study Revisited: New Analysis of Raw Data Doesn’t Support Vegetarian Ideology”