Book updates and The China Study

Note - this is a long and wordy post. Consider yourself warned!

Between holidaying and general life-ing, I've been rather belated in updating my book list. It's finally up-to-speed though, with recent additions including:

• The Clan of the Cave Bear by Jean Auel
• The Valley of the Horses by Jean Auel
• Daniel Isn't Talking by Marti Leimbarch
• The World Beneath by Cate Kennedy

I also just finished The China Study (T. Thomas Campbell & Thomas M. Campbell II) and feel the need to organise my thoughts about it in blog form. I am hoping others may have read the book too, and would be curious to hear your perspective if you have.

Before starting, I think it deserves note that this was probably not an easy book to market. The authors needed to translate scientific research into lay terms, and make it punchy enough to sell. They weren't offering rapid weight loss and they were challenging some popular dieting beliefs. They were also challenging the viewpoints and interests of several high-profile research groups and organisations.

The fact that they made a best-selling book, which reads well and can be understood without formal research training, is thus an impressive feat. That they have also inspired the Forks Over Knives documentary (which I really, really hope comes to Australia!) is frankly amazing.

The first author is also extraordinarily well established in the scientific literature, and holds the post of Emeritus Professor of Nutritional Biochemistry at Cornell University. Clearly, he is more qualified than I.

Despite all of this, I can't be completely positive about the book. I'd like to be, but I can't!

Source

In summary

The China Study is one of the largest epidemiological studies ever done. It involved a survey of diet, health and mortality in over 2,400 Chinese counties, and was the culmination of more than 20 years work across Cornell University, Oxford University, and the Chinese Academy of Preventive Medicine.

The book The China Study presents findings from that research project, along with results from other nutritional studies. It focuses on links between diet and cancer, cardiovascular disease, obesity, Type I Diabetes and other autoimmune disorders, as well as life expectancy.

The overall message, and one which is convincely painted, is that intake of animal products (including meat, dairy and eggs - although little data is actually presented regarding eggs) is associated with a higher incidence all of the above diseases. Once diseases are present, intake of animal products is also linked to disease progression and health deterioration.


Conversely, a plant-based diet is linked to a lower incidence of those diseases and a longer life expectancy.

As an interesting side note, fish isn't given a lot of discussion, and the authors advocate eating it in moderation rather than avoiding it altogether (this is the only animal product where this is the case). But that's not my focus today!

The Positives

1. This is a comprehensive summary of some of the benefits of vegetarian and vegan eating.

As someone who believes in those benefits, it is nice to see some supporting data and science laid out in clear, accessible terms.

2. I found the explanations around the research to be well written and I admired the translation of scientific language into understandable terms.

This really is a good example of how academic research can be translated for general consumption, and I think everyone would benefit if we had more books that did this. Not everyone works in science, but we all deserve to know about key scientific findings.

3. The book was interesting, well set out, and certainly highlighted some of the significant problems inherent in the American education, health, and research systems.

When funding is provided by meat / milk / food / pharmaceutical industries with clearly biased interests; school-based nutrition education is guided by the same organisations; and pills are given greater merit (and are often more accessible) than sensible advice about nutrition, something is not quite right.

The authors highlight these issues in America (and by association, other Western countries), even when they know their message isn't popular. This also takes me to (4)...

4. The authors are obviously passionate about their research and mesage.

I think they genuinely believe what they write and the problems I identify below stem, in all likelihood, from that passion - just as we can all be clouded by things we feel strongly about and believe in.

Even if some aspects of their message can be questioned, they present what is largely sensible science and advice relating to food.

The problems

(in my specific, and not necessarily correct, opinon)

I debated writing this post, because I know my perspective on the book may not be matched by others. I don't feel informed enough to fully comment on all aspects, I'm not a nutritionist, and I am not intimately acquainted with this particular body of literature. As such, I have tried to state only things I can back up.

1. The authors tend to dismiss studies that show null effects - that is, studies that don't link diet with positive or negative outcomes.

Specifically, they state that if half (or so) of the studies out there show a significant link between animal product intake and health problems, it doesn't much matter that the other studies show no relationship. Their reasoning is that the studies showing no relationship don't show a negative association between animal product intake and health problems, and thus don't really count in countering out the other findings.

Fair enough. But this isn't actually how science works. If you hypothesise that X is associated with Y, you don't need to find that X is associated with Z to refute that. You just need to find that X isn't associated with Y.

I agree that if 50% of studies link animal consumption to negative health outcomes, that's worth reporting. It would (and does) put me off eating animal foods. But it isn't fair or accurate to say that the other 50% of studies don't count just because they don't link animal consumption with health!

2. The casein link suggesting animal protein is associated with cancer is taken out of context.

One of the most compelling cited studies, when taken at face value, was the research that found rats fed a diet of 5% casein protein (a form of milk protein) did not develop cancer, whilst rats fed a diet of 20% casein did so at dramatic levels. These findings were reported in conjunction with those from studies that found high levels of wheat and soy protein did not produce similar effects - thus, Campbell asserts, it is animal protein that is problematic.

However.

In a study that Campbell did not cite, he and his co-authors found that wheat protein showed similarly carcinogenic properties when combined with the amino acid lysine and delivered in high levels [1].

Lysine is an amino acid that typically occurs with wheat its natural form. Examining wheat protein in an isolated state, as done in the studies where carcinogenic effects were not identified, is an error of the sort Campbell criticises - viewing specific food components in isolation, instead of in the forms they typically occur.

What is more, whey, another milk protein, has been found to demonstrate anti-cancer properties, even when consumed at very high levels [2-4].

Given that milk products will typically be consumed in a form that combines casein and whey, it is inappropriate to separate out one and call it carcinogenic, when the other half of the product exerts anti-cancer properties. Again, as Campbell himself states, we need to look at the big picture and not specific food components in forms they do not naturally occur.

Campbell and colleagues also found (again, this was not cited in the book) that wheat flour intake is significantly related to cardiovascular disease, with a correlation of 0.67 [5]. If two things are perfectly related (they always occur together), the correlation would be 1.00. The correlation between animal protein intake and cardiovascular disease was 0.26.

Clearly, these results are quite different to those presented in the book. Effectively, these findings suggest that wheat flour intake is moderately associated with cardiovascular disease, whilst animal protein intake is only slightly associated with cardiovascular disease. The authors present some other data that may partially explain this finding (wheat flour intake is associated with greater intake of various other things, including salt, for instance) [5], but it still deserves note - and isn't mentioned in the book.

3. The China Study made use of a hugely impressive data set. However, because the data set was so big, and because so many possible associations were examined, the chances of finding "false" findings increase.

On page 40, it is noted that The China Study yielded "over 8,000 statistically significant correlations". Statistical significance means that these finding are unlikely, in theory, to have occurred by chance. The standard cut point for significance testing means that the probability of the identified relationship being due to chance is <5%.

In the book, this <5% probability is (appropriately) referred to as statistically significant and <1% is referred to as highly statitsically significant.

However, it was not noted whether adjustments were made for the extraordinarily large number of analyses and comparisons that were conducted.

Given this, I went back to check some of the original articles in the scientific journals in which they were published [e.g., 5]. These confirmed that adjustments were not made to account for the many correlations examined. I suspect today some objection would be made to that, but at the time of publishing (late 1990s) it was probably less common to adjust for multiple tests.

To grasp why this is a problem, consider that the 5% cut point means that if you examine a relationship 100 times, you should only have 5 findings (at most) due to chance. 

If you use 5% when you conduct 8000 (plus) correlations, you could find 400 significant associations due to chance. 

What is more, significant correlations are more likely to occur in large samples. This is because, in a large sample, there is more variation in behaviour / diet / health. In addition to exploring thousands of possible relationships, The China Study used data from thousands of people. This has to be accounted for when interpreting the findings. Unfortunately, though, it isn't.

--

For a more detailed critique of some other methodological / interpretational flaws, see an extensive online review by Denise Minger.  For crazy criticism from people with no apparent grasp on science, just google The China Study :)

I will repeat that I liked this book, and the problems I felt deserved mention don't detract from the overall significance of the work. I just don't think the results are quite as cut and dry as the authors reported, and I would have liked less sensationalism and perhaps more consideration of alternative perspectives.

If you have got this far (congratulations!), have you read The China Study? Any thoughts? Or read any other scientific reflections on meat vs. vegetarian eating?

References:

1. Schulsinger DA, Root MM, Campbell TC (1989). Effect of dietary protein quality on development of aflatoxin B1-induced hepatic preneoplastic lesions. Journal of the National Cancer Institute, 81, 1241 - 1245.

2. Parodi PW (2007). A role for milk proteins and their peptides in cancer prevention. Current Pharmaceutical Design, 13, 813 - 828.

3. Bounous G, Batist G, Gold P (1991). Whey proteins in cancer prevention. Cancer Letters, 57, 91 - 94.

4. Bounous G. (2000). Whey protein concentrate (WPC) and glutathione modulation in cancer treatment. Anticancer Research, 20, 4785 - 4792.

5. Campbell TC, Parpia B, Chen J (1998). Diet, lifestyle, and the etiology of coronary artery disease: The Cornell China Study. The American Journal of Cardiology, 82, 18 - 21.

Tweet

Pin It