The real findings of the China study

About the book
The controversy

Real correlations in the China study dataset

Diet recommendations

The China Study is a book published in 2004 by T. Colin Campbell, a professor of Nutritional Biochemistry at Cornell University, and his son, Thomas M. Campbell II, a physician. The book examines the relationship between the human diet and chronic civilization illnesses such as cancer, diabetes, heart disease, obesity, autoimmune disease, osteoporosis or degenerative brain disease. The China Study of the title is taken from the China-Cornell-Oxford Project, a 20-year study that began in 1983 and was conducted jointly by the Chinese Academy of Preventive Medicine, Cornell University, and the University of Oxford. The study examined mortality rates from 48 forms of cancer and other chronic diseases from 1973-75 in 65 counties in China, and correlated them with 1983-84 dietary surveys and bloodwork from 6,500 people, 100 from each county. The study was conducted in those counties because they had genetically similar populations that tended, over generations, to live in the same way in the same place, and eat diets specific to those regions.

The controversy

The book had sold 500,000 copies as of January 2011, making it one of America's best-selling books about nutrition. It concluded that Chinese counties with a high consumption of animal-based foods in 1983-84 were more likely to have had higher death rates from chronic civilization diseases as of 1973-75, while the opposite was true for counties that ate more plant foods in 1983-84.

While the public - especially vegetarian community - had accepted the book and its recommendations without a hesitation, there were several statistics enthusiasts who actually took a look at the survey data. The dataset is publicly available here, or in a printed form, book "Diet, life-style, and mortality in China: A study of the characteristics of 65 Chinese counties [Chen J]".

China study dataset

The survey data, when properly analyzed by relevant statistical methods, show different correlations than the ones described by Campbell. It seems that he was right about protective properties of vegetables, but not so right about harmful effects of animal protein. Probably the most thorough analysis with step-by-step commentary was performed by 24-years old student Denise Minger, who acquired several thousand followers practically overnight, after publishing her analysis.

Campbell's book was commented, analyzed and corrected by many other nutrition and statistics enthusiasts (list of articles here). We have researched those articles and summarized the "real" findings of the China study in the following section.

Summary of real correlations in the China study dataset

Following are the correlations identified in the China study dataset (using proper statistical methods):

Cancer

  • There is no statistically significant (only slightly negative) correlation between any type of cancer and animal protein intake. If meat has any effect in relation to the cancer, it is a protective effect.
  • Plant protein has almost three times as many positive correlations with various cancers as the animal protein.
  • For heart disease and stroke, plant protein has a positive correlation while animal protein has negative or nearly neutral correlations.
  • Sugar, soluble carbohydrates, and alcohol all have correlations with cancer mortality about seven times the magnitude of that with animal protein.
  • As the only carbohydrate/plant protein exception, rice was statistically associated with reduced cancer mortality.
  • Total fat intake is negatively correlated with cancer mortality.
  • Green vegetables were statistically associated with reduced cancer mortality.

Heart disease

  • No statistically significant associations were observed for total protein/animal protein intake and mortality from coronary heart disease. If there is any relationship, meat actually seems protective of heart attacks and coronary heart disease.
  • There is no significant correlation between animal food consumption and blood cholesterol levels.
  • Plant protein actually correlates fairly strongly with heart attacks and coronary heart disease.
  • Wheat consumption is powerfully associated with higher insulin levels, higher triglycerides, coronary heart disease, stroke and hypertensive heart disease - far more so than any other food.
  • As the only exception, rice consumption had statistically significant negative correlation with heart disease risk and stroke mortality.
  • No statistically significant associations were observed for saturated fat / total fat and mortality from coronary heart disease.

Overall mortality and morbidity

  • Animal protein doesn't correspond with more disease, even in the highest animal food-eating counties of the study. Total protein intake has a strong negative association with overall mortality.
  • Egg consumption was negatively associated with all-cause mortality.
  • Soy sauce (not soy products) showed a significant decrease in overall mortality risk.
  • The purely vegetarian counties included in the study had the greatest mortality and morbidity risk.

Diet recommendations based on the proper analysis of the China study dataset

It is important to note that the China study was performed as an epidemiological study, which means that it doesn't contain control variables and therefore it is impossible to draw actual conclusions from the data (correlation doesn't equal causation). However, if we were to design a set of recommendations based on the identified correlations, these would be:

  1. Base your daily caloric intake on animal protein and fat.
  2. Eat lots of (green) vegetables.
  3. Do not eat plant protein, especially wheat. If you really have to, choose rice.
  4. Avoid carbohydrates of all kinds (sugar, starch, alcohol).

To sum it up: these recommendations seem to be nearly identical with basic rules of paleolithic diet (more info here).