Home » Take nutrition studies with a grain of salt

Take nutrition studies with a grain of salt

Comb through enough nutrition research, and you can find a study confirming or rebutting nearly every belief you may hold about how specific nutrients affect your health. “Meat Increases Heart Risks, Latest Study Concludes,” reported The New York Times in 2020. A year earlier, the Times ran this headline: “Eat Less Red Meat, Scientists Said. Now Some Believe That Was Bad Advice.”

Pick a different food group and find a similar contradiction. “Moderate Drinking Has No Health Benefits, Analysis of Decades of Research Finds,” reported the Times in April 2023. Two months later, Forbes declared: “Light And Moderate Drinking Could Improve Long-Term Heart Health Study Finds—Here’s Why.”

These headlines were not misrepresentations. Nutritional epidemiology is, by and large, what Stanford University biostatistician John Ioannidis calls a “null field”: one where there is nothing genuine to be discovered and no genuinely effective treatments exist.

“I think almost all nutrition studies that pertain to the effects of single nutrients on mortality, cancer, and other major health outcomes are null or almost null,” says Ioannidis. “Even the genuine effects seem to have very small magnitude in the best [and] least biased studies.”

When it comes to public policy, most nutritional epidemiologists are unclothed emperors ordering the rest of us around or lobbying lawmakers to do it for them.

This doesn’t mean you can eat an entire pizza, a quart of ice cream, and six beers tonight without some negative health effects. (Sorry.) It means nutritional epidemiology is a very uncertain guide for how to live your life and it certainly isn’t fit for setting public policy.

In short, take nutrition research with a grain of salt. And don’t worry: Even though the World Health Organization (WHO) says “too much salt can kill you,” the Daily Mail noted in 2021 that “it’s not as bad for health as you think.”

Nutritional Epidemiology’s Original Sin

Back in 2019, Ioannidis called nutritional epidemiology “a field that’s grown old and died. At some point, we need to bury the corpse and move on to a more open, transparent sharing and controlled experimental way.” He expressed particular concern that nutritional research findings are largely derived from observational studies, which are essentially surveys. In other fields of health science, hypotheses are tested with strictly supervised randomized controlled trials that are designed to filter out the inherent noise in observational data.

Drawing firm conclusions from weak data is the original sin of nutritional epidemiology. Legendary American physiologist Ancel Keys more or less launched the suspicion that eating steaks and hamburgers caused heart disease during the 1950s. Keys and his colleagues hypothesized that cardiovascular diseases were becoming more common because the saturated fats found in red meat and dairy products were boosting levels of serum cholesterol. In his 1957 article “Diet and the Epidemiology of Heart Disease,” Keys recommended the “exclusion of saturated fats (in butterfats and meat fats)” as a way to lower serum cholesterol levels. He conversely noted vegetable fats such as corn oil and cottonseed oil had the beneficial effect of reducing serum cholesterol.

Keys based his conclusions on observational data including a positive correlation (reported in his 1953 article, “Atherosclerosis: A Problem In Newer Public Health“) between estimates of the amount of fats consumed per capita in six countries and their rates of diagnoses of “degenerative heart disease.” He also pointed to studies that reported a correlation between high levels of serum cholesterol and the presence of atherosclerosis. In addition, Keys cited the results of randomized controlled trials he and his colleagues conducted using cohorts of schizophrenic men in Hastings State Hospital. The subjects were, during periods lasting between three and six months, fed diets of varying levels and types of fats. They reported in 1957 that saturated fats in meat and milk boosted their subjects’ overall cholesterol levels whereas vegetable oils, corn oil in particular, tended to lower them.

As with many examples of “bad” science, Keys’ claims had some basis in fact. There are different types of fats. Some fats are “saturated” with hydrogen atoms; others have one double bond of carbon atoms (monounsaturated) or more (polyunsaturated) in their structure. Generally, saturated fats are solid (lard, cheese, and butter) at room temperature, whereas unsaturated fats (canola, olive, soy, and corn oils) are liquid.

But Keys’ campaign—and those it inspired—treated poorly tested hypotheses as settled science. In 1979, the surgeon general recommended Americans eat “less saturated fat and cholesterol, less salt, less sugar,” and “less red meat.” As recently as July 2023, the WHO issued guidelines warning against consuming saturated fatty acids “because high levels of intake have been correlated with increased risk of CVDs [cardiovascular diseases].”

Today’s datasets are possibly even noisier than those of the 1950s. A quick series of searches in Google Scholar combining the terms “red meat,” “dairy,” and “eggs” with “cardiovascular” finds more than 68,200, 308,000, and 154,000 studies, respectively, and they don’t all say the same thing. You can easily turn up numerous studies on either side of the question for each of those foods.

Is the picture clearer with meta-analysis? Yes and no.

A meta-analysis is a study of past studies. By aggregating studies, the ambitious epidemiologist hopes to tease out a real effect. Often meta-analyses clarify what the data say, and sometimes they simply tell us we can’t trust the data.

For example, a controversial 2019 meta-analysis published in Annals of Internal Medicine “found low- to very-low-certainty evidence that reducing unprocessed red meat intake by 3 servings per week is associated with a very small reduction in risk for cardiovascular mortality, stroke, myocardial infarction (MI), and type 2 diabetes.” It concluded that reduced consumption of processed meats had similarly equivocal effects on cardiovascular health. A companion meta-analysis of just randomized controlled trials by some of the same researchers “found only low- to very-low-certainty evidence that diets lower in red meat compared with those higher in red meat have minimal or no influence on all-cause mortality, cancer mortality, cardiovascular mortality, myocardial infarction, stroke, diabetes, and incidence of gastrointestinal and gynecologic cancer.”

The Annals of Internal Medicine meta-analysis concluded that “findings from our review raise questions regarding whether—on the basis of possible adverse effects on cardiometabolic outcomes—the evidence is sufficient to recommend decreasing consumption of red and processed meat.”

Naturally, the contrarian Annals study was immediately challenged. “It’s the most egregious abuse of data I’ve ever seen,” Harvard nutritional epidemiologist Walter Willett told Medical Daily.

But in 2020, the Cochrane Library issued a systematic review of studies assessing the health effects of reducing saturated fats—that is, replacing animal fats and hard vegetable fats with plant oils, unsaturated spreads, or starchy foods. This review reported that “reducing saturated fat in-take probably makes little or no difference” to all-cause mortality, cardiovascular mortality, nonfatal myocardial infarction, and coronary heart disease mortality. The authors nevertheless concluded that the studies they analyzed “provide moderate-quality evidence that reducing saturated fat reduces our risk of cardiovascular disease.”

Other researchers have gone further. In 2022, scientists associated with the University of Washington’s Institute for Health Metrics and Evaluation unveiled some techniques they had developed to correct for the uncertainties and biases in the studies being evaluated. Their study, published in Nature Medicine, reported “weak evidence of associations between unprocessed red meat consumption and colorectal cancer, breast cancer, IHD [ischemic heart disease] and type 2 diabetes.” No association with strokes was identified. The evidence was so uncertain that they concluded, “While there is some evidence that eating unprocessed red meat is associated with increased risk of disease incidence and mortality, it is weak and insufficient to make stronger or more conclusive recommendations.”

If nothing else, those researchers agree on one thing: The available evidence is insufficient to recommend reducing meat consumption. But not everyone even agrees about that: In 2023, Critical Reviews in Food Science and Nutrition published a meta-analysis concluding that “unprocessed red and processed meat might be risk factors for IHD [ischemic heart disease]. This supports public health recommendations to reduce the consumption of unprocessed red and processed meat intake for the prevention of IHD.”

Note that word might, followed by a much more confident assertion that public health Cassandras should continue to warn people away from meat. A charitable interpretation of the study would be the authors recommend a cautious approach to meat not entirely supported by the evidence because meat might be bad, even if they can’t prove it. It’s not an immoral decision per se, but it’s also not science—and it certainly doesn’t justify anti-meat public policy.

Some of the authors of the Annals of Internal Medicine meta-analysis of red meat and cardiovascular mortality have also examined potential links between red and processed meats and cancer. Unlike most nutritional epidemiological studies, this one helpfully translates the relative risks reported into absolute risks.

They calculate that a weekly reduction of three servings of unprocessed meat will reduce a person’s overall lifetime population risk of cancer from 105 per 1,000 to 98 per 1,000. Parsing three breast cancer studies, they calculate that a person’s overall lifetime population risk will fall from 46 per 1,000 to 40 per 1,000. For prostate cancer (drawing on two studies), the absolute risk falls from 38 per 1,000 to 37 per 1,000. For colorectal cancer (five studies), they find that there is no absolute risk reduction. They also estimate that cutting the consumption of processed meats by three servings per week will reduce the absolute lifetime risk of cancer by roughly the same amount. These findings track those reported in the 2022 Nature Medicine study cited above.

The team concludes: “Our systematic review and meta-analysis of cohort studies supports the association between red and processed meat intake and increased risk for cancer. The magnitude of red meat’s effect on cancer over a lifetime of exposure was, however, very small, and the overall certainty of evidence was low or very low.”

But nutritional epidemiologists are nothing if not dogged in the pursuit of uncovering tiny effects. A 2021 meta-analysis in the European Journal of Epidemiology found that eating red meat and processed meats was positively associated with risk of breast, colorectal, colon, rectal, and lung cancers. But the relative risks for each were not much different than those reported in the Annals of Internal Medicine meta-analysis.

When a 2021 meta-analysis in the journal Nutrients looked at cancer risks, it found that “while relative effects for red and processed meat may be positive and statistically significant, absolute effects are small (less than 1%).” It concluded that “the recommendation to reduce the consumption of processed meat and meat products in the general population seems to be based on evidence that is not methodologically strong.”

Researchers Flip on Dairy

With meat, the concessions have been gradual and reluctant. With dairy, the about-face has been far more dramatic.

For years, nutritional epidemiologists condemned dairy foods and eggs for their high saturated fat contents. For example, the doyen of nutritional epidemiology, Walter Willett, wrote in Science in 1994 that butter and other dairy fats boosted cholesterol, thus probably increasing the risk of coronary heart disease. Therefore, he argued, “saturated fats, particularly those from dairy sources, should be minimized.”

Just 20 years later, based on an extensive meta-analysis of saturated fat studies in Annals of Internal Medicine, food writer Mark Bittman famously declared “butter is back.” The researchers found “a possible inverse association” between consuming dairy products and coronary disease. In other words, drinking milk and eating butter actually tended to reduce the risk of heart disease.

Since 2014, the majority of nutritional epidemiological studies have found that consuming dairy products is at worst neutral and more likely slightly protective. For example, a March 2022 meta-analysis in Advances in Nutrition reported, “Total dairy consumption was associated with a modestly lower risk of hypertension, CHD [coronary heart disease], and stroke.” A 2023 conference summary in the Proceedings of the Nutrition Society concluded: “The association between dairy foods and CVD [cardiovascular diseases] is generally neutral despite many of the dairy foods being the major source of SFA [saturated fatty acids] in many diets. This leads to substantial doubt concerning the validity of the traditional diet-heart hypothesis.” Of course, one can turn up more recent studies, such as a 2022 meta-analysis in Critical Reviews in Food Science and Nutrition, that still say consuming high-fat dairy products is associated with cardiovascular disease risk.

The role of eggs with respect to cardiovascular disease is contested. A 2019 Journal of the American Medical Association meta-analysis concluded that “each additional half an egg consumed per day was significantly associated with higher risk of incident CVD and all-cause mortality.” A 2021 cohort study in PLOS Medicine similarly found that “intakes of eggs and cholesterol were associated with higher all-cause, CVD, and cancer mortality.” Contrariwise, a 2021 cohort analysis in BMJ reported that “no association was found between egg consumption and cardiovascular disease risk among US cohorts, or European cohorts, but an inverse association was seen in Asian cohorts.” A May 2023 evaluation of recent evidence in Current Atherosclerosis Reports said that “most studies assessing egg consumption and CVD risk factors found a reduced risk or no association.”

Demon Rum and the J-shaped Mortality Curve

The cacophony of murky findings coupled with strong recommendations is not limited to solid foods. Earlier this year, the WHO declared “no level of alcohol consumption is safe for our health.” It based the claim on studies that suggest drinking in any amount is associated with higher risks of various cancers.

A team of Italian statisticians contradicted the organization’s proclamation in a July 2023 working paper. Their dive into the literature on alcohol’s health effects found the field rife with methodological problems, including a huge bias toward positive results and a probably enormous underreporting of actual consumption in surveys of drinkers. They conclude that “given the methodological limitations in detecting the effects of modest alcohol quantities, from a scientific point of view it is incorrect to claim that ‘there is no safe level.’ We should rather say that ‘we are unable to determine if there is a safe amount’ and, likely, we will never be.”

Nevertheless, since the 1980s, numerous epidemiological studies identified a U- or J-shaped curve—a graphical representation showing the risks for heart disease and overall mortality were lower for light to moderate drinkers than for nondrinkers and heavy drinkers. A June 2023 BMC Medicine study comparing nondrinkers and drinkers reconfirmed the existence of the J-shaped curve. “Compared with lifetime abstainers, current infrequent, light, or moderate drinkers were at a lower risk of mortality from all causes, CVD, chronic lower respiratory tract diseases, Alzheimer’s disease, and influenza and pneumonia,” it reported. But heavy and binge drinkers had a “higher risk of mortality from all causes, cancer, and accidents.”

In a 2022 editorial in European Heart Journal Supplements, Andrea Poli, president of the Nutrition Foundation of Italy, highlighted the health tradeoffs between alcohol’s cardiovascular benefits and cancer risks. The association of moderate consumption “with a reduced cardiovascular risk,” Poli wrote, “seems to prevail over the increase in [cancer] risk, with the consequence that all-cause mortality is reduced as compared to abstainers.” A 2015 study in Drug and Alcohol Review investigated the question of whether industry funding has biased studies of the protective effects of alcohol on cardiovascular disease. The researchers found “no evidence of funding effects for cardiovascular disease mortality, incident coronary heart disease, coronary heart disease mortality and all-cause mortality.”

Tradeoffs are an underdiscussed concept in the public health literature, which generally fails to recognize that we are all entitled to balance a desire for a long life with a desire to enjoy living. So how does one weigh the cancer risks of drinking? The lifetime population risk of colorectal cancer is 22.5 per 1,000 people. (Of course, this includes people who drink alcohol, but let’s use it as a baseline anyway.) A 2014 British Journal of Cancer article reported that moderate to heavy drinking increased the relative risk of colorectal cancer by 1.17, yielding a 17 percent increase in risk over nondrinkers. That suggests that moderate to heavy drinking increases the lifetime risk of colorectal cancer from 22.5 to 26.3 per 1,000 people.

Interestingly, a 2020 meta-analysis in the International Journal of Cancer identified a J-shaped relationship in which light, moderate, and even heavy drinking was actually associated with a lower risk of colorectal cancer compared to nondrinkers and very heavy drinkers.

A Grain of Salt

“Salt,” an unknown wit once said, “is what makes things taste bad when it isn’t in them.” The Centers for Disease Control and Prevention advises that “most Americans should consume less sodium” because “excess sodium can increase your blood pressure and your risk for a heart disease and stroke.” Most of the sodium Americans consume comes in the form of sodium chloride, otherwise known as table salt. The Dietary Guidelines for Americans recommends that adults limit sodium intake to less than 2,300 mg per day—about 1 teaspoon of table salt. The American Heart Association’s “ideal limit” of sodium intake for most adults is “less than 1,500 mg a day.” Instead, Americans consume an average of 3,400 mg of sodium per day.

In other words, the official nutrition scolds want your food to taste bad, or at least bland.

Remember that the recommendation to cut back on salt is intended to apply populationwide. But more recent research shows individuals exhibit a range of responses to various doses of salt. By some estimates, about 25 percent of people are salt-sensitive, meaning that higher salt intakes tend to increase their blood pressure. Another 15 percent of the population is inverse salt-sensitive, meaning that low intakes of salt conversely increase their blood pressure. A 2023 study in the Journal of Hypertension tested the effects of 7-day low- and high-sodium diets on subjects with normal blood pressure. It found that about 13 percent were salt-sensitive, 11 percent were inverse salt-sensitive, and 76 percent were salt-resistant—that is, consuming salt did not significantly increase or decrease their blood pressures.

Unfortunately, no widely accessible clinical tests have been devised for establishing a “personal salt index” for individuals to let them know if they are salt-resistant, salt-sensitive, or inverse salt-sensitive.

Epidemiological studies focused on the health effects of salt consumption come to different conclusions. For example, a 2020 Cochrane Library review of the effects of low-sodium versus high-sodium diets on blood pressure analyzed 195 randomized controlled trials. It found that “a low- versus high-sodium diet in white people with normal blood pressure (BP) decreases BP less than 1%.” Meanwhile, lower sodium intakes led to “a significant increase in plasma cholesterol and plasma triglyceride,” which are associated with higher cardiovascular disease risk.

The upshot: The results did not support the idea “that sodium reduction may have net beneficial effects in a population of white people with normal BP.” On the other hand, if you’re a white person with elevated blood pressure, “sodium reduction decreases BP by about 3.5%, indicating that sodium reduction may be used as a supplementary treatment for hypertension.” Lower-sodium diets did tend to reduce blood pressure a bit more in Asian and black subjects, though there hadn’t been enough studies to reach separate conclusions for those groups.

In 2020, a comprehensive review in the European Heart Journal pointed to the growing evidence that the relation of sodium intake with cardiovascular events is, like alcohol, J-shaped. That is, both deficient and high sodium intakes are associated with greater mortality and cardiovascular disease risks. The authors conclude that at the population level, moderate sodium consumption—about 1 to 2 teaspoons daily—has been “consistently associated with lower cardiovascular risk, compared to both high and low sodium intake.” A 2021 commentary in the Journal of Hypertension noted that at this point, “the ‘J-shape hypothesis’ cannot yet be either neglected or verified.”

A 2021 study in the European Heart Journal tested the hypothesis that high salt consumption was a risk factor for cardiovascular disease and premature death. The authors found that “daily sodium intake correlates positively with healthy life expectancy at birth and healthy life expectancy after age 60 and inversely with all-cause mortality in 181 countries worldwide.” They concluded that consuming a moderate range of salt (1 to 2 teaspoons daily) is not associated with increased cardiovascular risk. The American average of 3.4 grams of sodium a day is within that range.

The researchers add that their results are population averages, and that individuals will want to tailor their salt consumption to their specific health circumstances. The best evidence is that people with hypertension should cut back on salt, but whether people with normal blood pressure should is not a settled issue.

The Enemy of Your Enemy Is Not Always Right

If we can’t trust the epidemiological establishment, it might stand to reason that we can trust dissenters. Unfortunately, heterodox researchers also have biases.

The controversy over polyunsaturated seed oils is the mirror image of the fight over saturated fats in meat, milk, and eggs. Omega-3 and omega-6 fatty acids are essential fatty acids. These molecules are necessary for health but can’t be synthesized by the body, so we must get them from food. Both fatty acids act as structural components in cellular membranes and modulate inflammatory responses. The principal sources of omega-3 fatty acids are oily fish, flaxseeds, and some nuts. The chief omega-6 fatty acid is linoleic acid. The prime sources of linoleic acid in modern diets are oils derived from soybean, corn, cottonseed, sunflower, canola, safflower, rice bran, and grapeseed.

These oils have increased in modern diets, and some health and wellness gurus have dubbed them the “hateful eight.” Their main contention is that the modern dietary “balance” between omega-3 and omega-6 essential fatty acids is out of whack, resulting in a host of alleged bad effects on health.

Most recent research has not been kind to these claims. A 2020 meta-analysis in The American Journal of Clinical Nutrition reported that higher linoleic acid intake is “associated with a modestly lower risk of mortality from all causes, CVD, and cancer.” A 2020 narrative review in Atherosclerosis found it likely that “both dietary intake and circulating concentrations of [linoleic acid] inversely correlate with cardiovascular disease risk.” A 2020 review article in The Lancet Diabetes & Endocrinology concluded that plant oils with lots of linoleic acid “seem to be moderately protective” against coronary heart disease, especially myocardial infarction. That same review reported that a several-fold higher omega-6 to omega-3 ratio “has no adverse effects on either multiple markers of inflammation or oxidative stress.” Nor was there any “evidence to suggest an important role of the omega-6 to omega-3 ratio on glucose metabolism.” (The latter is relevant to the risk of developing diabetes.)

As for inflammation, a 2012 review of randomized controlled trials in the Journal of the Academy of Nutrition and Dietetics reported that “virtually no evidence is available from randomized, controlled intervention studies among healthy, noninfant human beings to show that addition of LA [linoleic acid] to the diet increases the concentration of inflammatory markers.” A 2017 meta-analysis of randomized controlled trials in Food & Function concluded that consuming more linoleic acid “does not have a significant effect on the blood concentrations of inflammatory markers.”

And last year, a systematic review in Food Science and Biotechnology concluded that omega-6 fatty acids “have beneficial effects on cancers, blood lipoprotein profiles, diabetes, renal disease, muscle function, and glaucoma without inflammation response.”

‘What Some Call Health’

Nutritional epidemiology as practiced currently is mostly bunk.

“Nutritional epidemiologists valiantly work in an important, challenging frontier of science and health,” Ioannidis generously observes in his 2019 article titled “Unreformed nutritional epidemiology: a lamp post in a dark forest” in the European Journal of Epidemiology. “However, methods used to-date (even by the best scientists with best intentions) have yielded little reliable, useful information.” As an example, Ioannidis specifically cites the prevailing recommendation to eat less red meat as one of the many “‘classics’ of nutritional guidelines” that are based on “mostly weak evidence and small (or null) effects.” As Ioannidis argued in BMJ in 2013, “almost every single nutrient imaginable has peer reviewed publications associating it with almost any outcome.”

In the meantime, the recommendations to “eat this; not that” derived from nutritional epidemiology fervently promoted by nutrition “experts” and the media confuse and frustrate regular folks. They also encourage policy makers and regulators to meddle with what people want to eat. The researchers in the 2019 Annals of Internal Medicine meat study pointed out, “For the majority of individuals, the desirable effects (a potential lowered risk for cancer and cardiometabolic outcomes) associated with reducing meat consumption probably do not outweigh the undesirable effects (impact on quality of life, burden of modifying cultural and personal meal preparation and eating habits).” In other words, the very weak evidence that eating meat might harm their health is most likely counterbalanced by most omnivores’ preferences to continue eating steaks and hot dogs.

Ioannidis concludes that nutritional epidemiology as currently practiced is rife with “fervent allegiance beliefs and group-think.” Consequently, many, if not most, of the observed effects reported by nutritional epidemiologists largely reflect the magnitude of the biases prevailing among the field’s researchers.

So enjoy the pleasures of drink and of the table in moderation, while keeping in mind English poet Alexander Pope’s astute observation: “What some call health, if purchased by perpetual anxiety about diet, isn’t much better than tedious disease.”


October 2023