In my last post, I explored what the China Study data says about meat and disease—which turns out to be a far cry from what Campbell reports in his book of the same name. In a nutshell, meat has no statistically significant correlations with any diet-related disease, and actually has a negative correlation with death from all causes and death from all cancers. That means the populations that ate more meat generally had fewer chronic diseases than the populations that ate less of it. While it’s impossible to tell from the China Project alone whether this is because meat was protective of illness or simply corresponded with other helpful factors (like better health care), it does undermine Campbell’s assertion that animal product consumption always went hand-in-hand with disease in the China Project.

(If you’re not sure what the China Study is or why I’ve suddenly made it my life’s purpose to examine every modicum of its data, take a gander at the previous entry for an explanation.)

Of course, the “meat” category doesn’t include fish, eggs, or dairy—so these foods aren’t out of the hot seat yet. In this post, I’ll be looking at fish. Sushi lovers, listen up.

Fish intake in rural China

Among the 65 counties studied in the China Project, fish consumption ranged from zero grams per day to 119 grams per day—a bit over a quarter of a pound. Unfortunately, all forms of seafood (finned fish, shell fish, roe, and so forth) are lumped under the same category, even though these foods tend to have much different nutritional profiles.

Surprisingly, fish had no statistically significant association with mercury intake (correlation of -8).

One thing striking about fish consumption in the China Study is that fish-eating regions have quite a few things in common: These counties tend to have a high population density (correlation +46), citizens overwhelmingly work in industry rather than agriculture (correlation +52 for industry and -53 for agriculture), and folks eat a great deal of processed sugar and starch along with the fish (correlation +58) as well as loads of soybean oil, sesame oil, peanut oil, and cotton seed oil (correlation +61). Infant mortality is low, literacy is high, and beer and liquor are quite popular beverages—especially among men. Although occurrence of heart disease is sparse, the average cholesterol levels are high compared to non fish-eating regions (+45 for total cholesterol, +38 for HDL, and +36 for non-HDL)—which doesn’t exactly support the cholesterol-heart disease connection, but I’ll get to that later.

Unlike meat, fish is quite obviously associated with more industrialized coastal regions—which makes things a little tricky when it comes to untangling variables. Putting nutrition aside for a second, we can see that industry-dominated regions are riddled with chronic diseases that rural regions generally don’t exhibit. According to the China Project data, the variable “Percentage of employed population who are in industry” has the following correlations with disease:

Male lung cancer: +62
Leukemia: +53
Female lung cancer: +47
Liver cancer: +47
Colon cancer: +41
Other heart disease: +40
Death from all cancers: +31
Colorectal cancer: +31
Stomach cancer: +25
Breast cancer: +24
Brain cancer: +21

(On the bright side, they have low rates of infectious diseases, pneumonia, parasites, tuberculosis, and other ailments common in the agricultural regions—indicating better living conditions and perhaps better health care.)

In other words, the folks who eat lots of fish typically work in industry, and the folks who work in industry typically have high disease rates. But is the fish to blame? Or is it one of the other factors associated with industry work, like processed starch and sugar intake or occupational hazards? Let’s find out. Here are the straight-up correlations, not adjusted for the variables that could be skewing the trends.

For the sake of science, I’ll only be focusing on the statistically significant correlations in this post.

NEGATIVE CORRELATIONS (more fish= fewer of these diseases)

Rheumatic heart disease: -43**
Diseases of blood and blood forming organs: -35**
Death from all non-cancer causes: -30*

If we prune away the variables that coincide with fish consumption but that negatively impact health—like processed sugar and starch intake, liquor consumption, soybean oil use, and so forth—I wouldn’t be surprised if some of the other correlations would  reach statistical significance:

Neurological diseases: -23
All diseases: -20
Oesophageal cancer: -19
Cervix cancer: -18
Myocardial infarction and coronary heart disease: -15
Stroke: -15
Hypertensive heart disease: -14

Given the non-China-Project research we have regarding fish, this isn’t too surprising: The omega-3 fats in fish tend to benefit cardiovascular function and work as blood thinners, so it’s no shocker that fish-eating regions have low rates of heart disease. In fact, in 2003, Campbell himself co-authored a paper on the protective effects of fish and DHA as revealed by the China Study data: Fish consumption, blood docosahexaenoic acid and chronic diseases in Chinese rural populations. (Too bad he forgot to include that in his book.)

But that’s only the positive stuff. Now let’s look at the more incriminating side of the fishy coin.

POSITIVE CORRELATIONS (more fish = more of these diseases)

Nasopharyngeal cancer: +56***
Diabetes: +41**
Liver cancer: +34**
Leukemia: +25*

Jeepers! That doesn’t look too good, does it? Positive correlations with several cancers and diabetes, all statistically significant. Could fish be a bona fide health-wrecker, even if it does spare your heart? Let’s take a look at these correlations one by one.

Nasopharyngeal cancer

Nasopharyngeal cancer—or NPC, so I don’t have to keep typing that horrible word—is a type of cancer that occurs at the back of the nose, towards the base of the skull. It’s pretty rare in most of the world: Only about 7 out of every million people get it in the US. The main folks with high rates of NPC are some Chinese populations, the Inuit, and some tribes from North Africa, according to the American Cancer Society.

In the counties studied in the China Project, men generally had between two and twelve times as much NPC as women—indicating that this is gender-biased disease. Since diet didn’t differ severely between the sexes, it’s possible that NPC occurrence is related to a lifestyle practice that Chinese men engaged in more than women (such as drinking or smoking). And indeed, NPC has a correlation of +50 with daily liquor consumption, +40 with daily alcohol consumption, +33 with daily beer consumption, +24 with total tobacco consumption, and +22 with homemade cigarette use.

In looking at NPC factors, something else jumps out at me: The heavy metal cadmium is strongly associated with NPC (correlation of +42). This particular metal is known to contribute to a variety of cancers, including NPC. Although cadmium doesn’t occur naturally in food, eating plants grown in cadmium-contaminated soil—or eating animals that consume said plants—is a surefire way to get it stuck in your body for a few decades. Is it possible the coastal, industry-dominated regions that ate the most fish also had cadmium-contaminated soils? Is it possible certain aquatic regions were also polluted with cadmium, which then accumulated in fish?

First, here’s a graph showcasing the original +56 correlation, using all of the data points for fish intake and NPC.

And here’s a graph using only counties that ate less than 25 micrograms of cadmium per day, as determined by the food composite survey for heavy metals. Some of the now-omitted regions were consuming over 90 micrograms per day, which is considered pretty toxic.

Woosh!

Simply removing cadmium from the equation, we’ve bumped a correlation of +56 down to +14, below the threshold of statistical significance. And get this: That highest data point on the chart, near the “12” grid line, is for the county with the absolute highest use of homemade cigarettes—which not only correlates independently with NPC, but is likely to be a source of cadmium exposure. Coincidence? Perhaps not.

It’s still not clear if cadmium is actually a factor, though, and we’ll never know for sure based on China Study data alone. The American Cancer Society cites infection with the Epstein-Barr virus, consumption of salt-cured foods, and exposure to formaldehyde and wood dust as potential causes of NPC. The China Project data also shows a strong correlation between NPC and tuberculosis, parasitic infection, and peptic ulcers. How are these things related—if they’re related at all? Beats me. I’m no nasopharyngeal cancer expert. But given the prevalence of this disease only in a smattering of places in the world, it’s probably due more to specific food contaminants, certain disease infections, or simply genetics.

Diabetes

Without peeling away any confounding variables, fish has a correlation of +41 with diabetes. That’s pretty high. But given that fish-eating regions also ate above-average amounts of processed sugar and starch (correlation of +58) and also tended to be heavier than most (correlation of +20), we might have a pretty obvious explanation for this sky-high rate. Here’s what our graph looks like using only counties that escaped the ravages of these nutritionally devoid foods, eating 0.5 grams or less of processed sugar and starch per day.

No positive correlation there. Without the influence of processed sugar and starch, the correlation between fish intake and diabetes nosedives from +41 to -5. We’re still in the wobbly land of epidemiological correlations, so we can’t say anything for sure, but it looks like fish itself doesn’t contribute to diabetes. (Unless your fillet of salmon is encrusted with Twinkies.)

Liver cancer

In China, liver cancer—like NPC—is extremely gender biased. In nearly every county, the mortality rates for liver cancer were between two and seven times higher for men than women. Since diet didn’t differ much—if at all—between the sexes, this is a red flag that liver cancer is something related to gender-specific lifestyle habits rather than diet alone.

Indeed, this is one disease that, in the China Study populations, has very little to do with diet: It’s linked pretty blatantly to hepatitis B infection across all counties. This is nothing revelatory, though. The World Health Organization already figured it out:

Hepatitis B is endemic in China and other parts of Asia. Most people in the region become infected with HBV during childhood. In these regions, 8% to 10% of the adult population are chronically infected. Liver cancer caused by HBV is among the first three causes of death from cancer in men, and a major cause of cancer in women.

Dur. Not surprisingly, fish-eating regions not only exhibited high rates of liver cancer, but also had large chunks of the population testing positive for the hepatitis B virus—a correlation of +35, a smidgen higher than the correlation with liver cancer. In other words, it looks like fish-eating regions had a lot of HPV-induced liver cancer, but diet wasn’t the cause.

Leukemia

According to the raw data, fish had a correlation of +25 with leukemia. Is this one legit?

For starters, let’s look at factors that correlate with leukemia in the China Project data. The most significant variable both fish consumption and leukemia have in common is industry work: +52 for fish and +53 for leukemia. (They both share a inverse correlation of -53 with agricultural work.) Could something about industry labor itself be a factor in leukemia occurrence? My spidey senses are tingling.

And look what we have here:

A retrospective cohort study was carried out in 1982–1983 among 28,460 benzene-exposed workers (15,643 males, 12,817 females) from 233 factories and 28,257 control workers (16,621 males, 12,366 females) from 83 factories in 12 large cities in China. … Risk of leukemia rose as duration to exposure to benzene increased up to 15 years, and then declined with additional years of exposure. Leukemia occurred among some workers with as little as 6 to 10 ppm average exposure and 50 ppm-years (or possibly less) cumulative lifetime exposure (based on all available measurements for the exposed work units).

The years cited for this study, by the way, are right around the time the China Study data was collected. How convenient!

Indeed, it seems benzene exposure was a significant problem for some industrial workers in China—the same industrial workers who ate a lot of fish. So is seafood to blame for higher leukemia rates, or could it be occupational hazards like benzene?

Unfortunately, the China Project doesn’t record benzene exposure. And we’re a few decades too late to go track down that data now (if you have a time machine, please contact me). What we can do is clarify the picture by looking at fish consumption and leukemia in counties with low levels of industrial employment. Here’s a graph showing exactly that—using only regions where less than 10% of the population worked in industry:

So there you have it. With industrial work out of the equation, fish no longer bears a statistically significant correlation with leukemia.

In a nutshell:

• Based on the China Project data, fish may have a protective effect against some forms of heart disease, diseases of the blood, and diseases of blood-forming organs.
• Fish intake has raw correlations with several cancers and diabetes, but these trends appear to be a result of other factors that accompany fish consumption rather than the fish itself (cadmium exposure, processed sugar and starch intake/high body weight, hepatitis B infection, and benzene exposure). When we tease out these confounding factors, fish loses or reverses its positive correlation with these diseases.
• No matter what you eat—vegan, vegetarian, pescetarian, or otherwise—be  cautious about cadmium intake, since this metal does appear to cause some cancers and disease.

That’s all for the fish, folks. Next up: eggs and dairy.