The actual China Study and longevity: Meat, cholesterol and SFA looking pretty good; plant-based not so much ?

in stats •  last year

The book "The China Study", a best selling book written by the venerable T. Colin Campbell
and his physician son, makes a strong case against the consumption of animal products. The book links the consumption of cholesterol, saturated fat and 'animal' protein to a wide range of chronic illnesses and in doing so argues against the consumption of animal products. The book derives its name from the equally named large scale epidemiological study. A study that it zooms into on occasion. Most of the argumentation in the book however draws on other sources of knowledge. In other words, there is surprizingly little China Study (The actual study) to be found in The China Study (the pop-sci book). With the data set available however, we can look at that the China Study data on its own, and see if the message conveyed by the China Study book is backed up by the raw data.

While the book does a great job at zooming in to different chronic illnesses , pointing out associations as it goes along, the sheer number of illnesses and food stuffs that can be zoomed into raises the serious issue of spuriousness. Remember the birthday paradox from math class? Well epi data sets like these will contain loads of spurious associations as a result of this phenomenon. So let us try to avoid the birthday paradox issues with spuriousness and zoom out to a level that should reflect a true main concern that is more fundamental then individual diseases : longevity.

So what longevity attributes can we find in the data set? Well, the data set contains "all cause" mortality for different age groups within the regions that make up the data set. If we combine these mortality numbers, combining age groups starting at zero and ending at 80 years old, we can calculate a metric that denotes the "probability to live to an age of 80 years old".

Remember, we are zooming out to look at the overall picture. To avoid assigning too much value to individual associations, we decide to only look at the top most positively correlated and the top most negatively correlated food stuffs. To do this, there is an other little adjustment we do to the data. As input variables, we shall look only at those variables expressed in grams per day, and we shall adjust these variables by normalizing them for BMI. So what does the data tell us?

Well, to be fair, whatever the data is telling us, it is all a bit disappointing as our R squared only once ends up higher than 0.1. So basically, there isn' t that much of association between food stuffs and longevity. Remember though, the China Study, the book, made rather powerful claims regarding certain food stuffs, so while the associations arn' t that strong, it is all cause mortality after all we are looking at, that might attenuate the results we are seeing a bit, alt least the result should largely fit the claims made in the book with regards to the sign, positive or negative, of the association with longevity.

Well, that, it turns out isn' t the case. Lets start with the top negative associations between food stuffs and longevity. The stuff associated with the lowest probability of reaching the ripe old age of 80 years. We shall limit ourselves to only those variables with an R squared of at least 0.05:

  • Spice intake: 0.23
  • Plant food intake: 0.09
  • Total food intake: 0.06
  • Potassium intake: 0.06
  • Starchy tuber consumption: 0.05
  • Plant protein intake: 0.05
  • Glutamine intake: 0.05

Now the other side of the spectrum. The food stuffs associated with higher probability of living to the age of 80 years old.

  • Meat: 0.09
  • Cholesterol: 0.09
  • Saturated fat: 0.09
  • Omega 3: 0.08
  • Animal protein: 0.07
  • Mono-unsaturated fat: 0.07

Well, talk about a big picture. Turns out of all the weak associations in the china Study data set between food stuff and longevity, the least weak associations not only don' t support the vegan gospel that the book claims is based on this data set, the data might even be seen to support the opposite message. That is, while there are other studies showing for example how a western plant based diet might contribute to longevity, when contrasted against a western diet that is higher in animal products, the China Study epi data most certainly does not strengthen that evidence, it weakens it.

Yes, it is likely there are confounders , but remember this is the data set that gave birth to the Vegan Bible.
So to answer the question. Does the actual data for the "Vegan Bible" suggest we should eat MORE meat and animal based products? Well, no. The amount of meat and animal products consumed is not really that high even in regions that consume the most, and assuming a linear relationship is folly, even if we were to put trust in these clusters of relatively weak associations. It does however hint rather strongly that the claims made in the book and the stance made in the book for consuming less meat animal protein, saturated fat, and cholesterol and eating a more plant based diet might not be supported by the actual epi data.

I've interacted with quite a few Campbell acolytes and their canned response to any criticism of his work basically boils down to an appeal to authority. But think, the data is available and the above analysis is trivial. So trivial that anyone who has spent more than a day with R or IPython should be able to duplicate it in less than an hour. And than, even for those who put great value in the process of peer review, it is not as if a pop-science book is a peer scientific paper. The China Study is a wonderful book that is written quite eloquently. A book that raised interesting theories at the time it was written. A book however that took its name and much of its fame from the equally named study, that quite frankly does not support the dietary advise that flows from the conclusions posed in the book.

Authors get paid when people like you upvote their post.
If you enjoyed what you read here, create your account today and start earning FREE STEEM!