3

The Unusual Complexity of Nutrition Research

THE EFFECTS OF FUNDING BY DRUG COMPANIES, AS WE HAVE just seen, are measurable. Drug studies also are easier to design. To find out whether a drug is safe and effective, you give people one medication—a single product—and see what it does in comparison to giving nothing or to taking an alternative drug. Food is more complicated. We eat an enormous variety of foods, and diets also vary enormously—from day to day and from one person to another. Everything else we do also varies. Humans make terrible experimental animals. We cannot be locked in cages and fed controlled diets, at least not long enough to learn anything useful. All of this forces studies of diet and health to be largely observational rather than experimental and, therefore, exceptionally vulnerable to biases in design and interpretation.

These complexities also make the effects of industry funding more difficult to research. Studies of food-industry influence are more recent, fewer in number, and more complicated to interpret than studies of drug-industry effects. Much of my particular fascination with nutrition research lies is its inherent complexity. Adding to that complexity are matters pertaining to the distinction between nutrition research and food science, the cost of nutrition research involving human subjects, and the particular difficulties of measuring the effects of food-company sponsorship.

Nutrition Research Differs from Food Science

A bit of history helps explain this distinction. Since at least the time of Hippocrates, scientists have attempted to study how food keeps us warm and alive.¹ But the involvement of the food industry in these kinds of studies dates only to the early 1900s, when scientists began to identify vitamins, the thirteen distinctly different substances that must be obtained from food. They also began to identify minerals needed in extremely small amounts. The vitamin discoveries, which took place from 1913 (vitamin A) to 1948 (vitamin B₁₂), thrilled researchers eager to discover which foods contained them and what they did in the body. Food and drug companies were equally thrilled; they could sell vitamins and minerals not only to improve the quality of animal feed or food products but also as supplements. In those heady days of nutrient discovery, research partnerships between scientists and food or drug companies made perfect sense.

Following World War II, American companies such as General Foods, Kraft, and Quaker Oats built research facilities and recruited scientists to work in them, as did Nestlé and Unilever in Europe. Science directors of food companies had large budgets with enough discretionary funds to conduct basic research along with studies aimed at product development. Unilever scientists, for example, studied fundamental properties of fats and oils that might or might not apply to margarines. But by the mid-1980s, in response to shareholder pressures for higher and more immediate returns on investment, American food companies shifted their research enterprises to focus more directly on commercial goals. As companies merged and consolidated and as short-term financial objectives became ever more pressing, most food companies in the United States closed their in-house research facilities and contracted out their research needs to university scientists. Today, Nestlé (in Switzerland) and Unilever (in the Netherlands) are unique in maintaining large research operations that still do basic research, although they too partner with researchers at universities.²

In the United States, the Morrill Act of 1862 set the stage for future collaborations between university scientists and food companies. This act gave land to states to establish colleges of agriculture for promoting “the liberal and practical education of the industrial classes in the several pursuits and professions in life.” The new “land-grant” colleges created departments of animal, poultry, and dairy sciences and recruited faculty to conduct research explicitly aimed at helping expand animal agriculture in their states. In New York State, where I live, Cornell is the designated land-grant university; it established animal-science departments in 1902. Its dairy department, for example, recruited faculty to teach courses about dairy chemistry and microbiology but also about the processing and manufacture of dairy products. This university created its own dairy business to train students, and the current department still manages a dairy herd and produces Cornell ice cream, yogurt, and cheese.³

Food processing hardly existed in the early twentieth century but quickly became important to New York’s economy. In 1956, in response to a request from the New York State Canners and Freezers Association, Cornell’s dairy department set up a Food Science and Technology program to support the state’s food processors. Eventually, these units were combined into today’s Department of Food Science. This department’s stated mission is to prepare students for leadership roles in the food industry, academia, and government; to expand information about the properties of foods and beverages; and to transfer this new information to food systems.⁴ The purpose of food science is to support the food industry by training students for jobs in that industry and by conducting research to support industry goals and practices. Because food science developed as an arm of the food industry, the missions of food scientists and industry sponsors are closely aligned, minimizing possibilities for conflicts of interest.

In contrast, the development of nutrition departments took a decidedly different path. To again use Cornell as an example: its faculty interested in the human diet were dispersed among programs in nutrition and home economics; other faculty were interested in the nutritional value of animal feed. In 1974, Cornell established its present Division of Nutritional Sciences by uniting these faculty. The division’s missions are to generate knowledge through scientific research; to educate and mentor the next generation of scholars, researchers, nutrition professionals, and responsible citizens; to improve nutrition and human health; and to inform nutrition policy and practice through public engagement.⁵ These goals differ from those of food science. The purpose of nutrition science is to improve public health. But because not all food products promote health, the goals of nutrition science are not necessarily aligned with those of food companies—creating the possibility of conflicted relationships.

The way I look at it, food science is the food industry. Cornell’s food science students take classes in the PepsiCo Auditorium. But as with everything else about food and nutrition, the distinctions are more complicated and the boundaries more porous. Faculty in either department can be working on similar problems. Many food scientists now investigate matters of human physiology and health, thereby raising the possibility of conflicts of interest in their field as well. But to understand why researchers in either field might welcome food-industry funding, it helps to appreciate what it costs to study diet and health in human subjects.

Human Nutrition Research Is Expensive

If researchers welcome industry funding, it is surely because government and foundation grants are scarce. Because observational studies are especially subject to misinterpretation, scientists consider the gold standard of experimental nutrition research to be randomized, controlled clinical trials. These examine the effects of consuming certain diets, foods, or nutrients on one or more measures of health. Such trials cost fortunes. I learned this during the years I was involved with the Women’s Healthy Eating and Living (WHEL) study, a randomized, controlled clinical trial to find out whether women who survived breast cancer would have fewer recurrences if they ate more fruits and vegetables.⁶ This project began in 1993 as a result of a grant from John and Christie Walton (of the Walmart family), who had a personal interest in diet and cancer prevention. Their foundation gave $5 million to cancer researcher John Pierce and his team at the University of California, San Diego, to pay for a feasibility study and the start-up costs of what turned out to be a lengthy investigation.

When the feasibility results showed that women who enrolled in the study would follow its protocols, Pierce applied to the National Cancer Institute (NCI) for funding for a multicenter trial to measure the dietary intake, health, and survival of three thousand women who had been treated for breast cancer—for an average of seven years. Eventually, the NCI granted $17 million to cover the study’s costs from 1997 to 2002 and, later, an additional $15 million for expenses through 2007. No, these are not misprints. The NCI grants really did add up to $32 million on top of the Walton Foundation’s $5 million. But, Pierce reminded me, those costs need to be viewed in context: “At the time, it was approximately the same as a single piece of innovative technological equipment in the hospital.”⁷

The funding paid for staff at seven locations in several states, equipment and supplies, laboratory tests, office visits, the storage of tissue and blood samples (very expensive), meetings, and juicers and other incentives for participation. It also paid my travel expenses to San Diego for about ten years, first for the initial planning meetings and later for my participation in the trial’s data-management committee. In the end, the WHEL results were disappointing. The study found no overall survival benefit from eating extra servings of fruits and vegetables, although one subgroup of women who were physically active survived longer. Nevertheless, WHEL investigators were able to publish 110 scientific reports about how to conduct such trials and on the health benefits of dietary improvements.⁸

Funding for the WHEL trial came from a wealthy foundation and a government agency. Companies producing fruits or vegetables do not have that kind of money, nor would they—or any other food company—want to invest in a study so costly, lengthy (fourteen years from concept to completion), and unpredictable in outcome. Instead, food companies have more pragmatic interests.

I occasionally receive letters from the California Table Grape Commission’s research director soliciting proposals for “$35,000 grants toward research on any relevant health issue in which grape consumption may have a beneficial impact.”⁹ A similar announcement for $30,000 grants comes from Yogurt in Nutrition for “research on the health benefits associated with yogurt consumption.”¹⁰ The California Strawberry Commission’s request for proposals is even more explicit: “The primary goal of the CSC nutrition research program is to establish the scientific evidence to support a vascular health claim under EFSA [European Food Safety Authority] or FDA criteria.”¹¹ The wording of these requests indicates that these groups are not asking open-ended, basic research questions about the health effects of specific foods. Instead, they are asking for studies designed specifically to establish their products’ benefits as a basis for health claims. As we will see, such research is so useful for marketing purposes that a great many companies and trade associations eagerly invest in these kinds of studies.

Funding Effects Are Complicated

I have argued so far that food science is mainly, though not exclusively, about finding technical solutions to problems related to food products, whereas nutrition science is mainly about the effects of nutrients, foods, and diets on human physiology and health. But marketing research has a decidedly different purpose: to create and sell food products. Food companies have always funded research aimed at product development but are now even more interested in research to demonstrate the health benefits of their products or to discredit evidence to the contrary.

This kind of research is easy to spot. Whenever I see a study suggesting that a single food (such as pork, oats, or pears), eating pattern (having breakfast), or product (beef, diet sodas, or chocolate) improves health, I look to see who paid for it. This is possible because most professional journals now require scientific articles to include special sections where authors must disclose who paid for their study and whatever financial arrangements they might have with the funder or a similar company.

Some years ago, I began posting particularly entertaining examples of sponsored research with self-serving results on my website, FoodPolitics.com. By March 2015, I was running across so many such studies that I began posting summaries of them, five at a time. I found these studies in publications I read routinely, and colleagues who knew I was collecting them sent me others. Although I pleaded with readers to tell me about industry-funded studies with conclusions contrary to sponsors’ interests, hardly anyone did. I continued posting these research summaries for an entire year. By then, I had collected 168 studies sponsored by food companies or conducted by investigators with financial ties to food companies. Of these, 156 reported results favoring the sponsor’s interests; only 12 did not.

This was a casual collection, a “convenience” sample. Because I was not collecting the studies in any systematic way, the results permit only one scientific conclusion: it is much easier to find industry-funded studies with results favorable to the sponsor’s interests than those with unfavorable results. Still, the collection illustrates some useful points. The most obvious: many different kinds of companies fund research potentially useful for marketing purposes. My haphazard collection included research sponsored by the makers and sellers of, among other products, artificial sweeteners, breakfast cereals, chewing gum, canned foods, chocolate, coffee, corn, dairy foods, dietary supplements, garlic, infant formula, lentils, nuts, orange juice, potatoes, soft drinks, soy products, and sugar.

Industry contributions to the research ranged from minimal (providing supplements or other products to be tested in the study), to more significant (paying partial or total costs). Many investigators who disclosed industry funding of their research stated that they had no conflicts of interest. Apparently, they did not view such funding as conflicting. When studies found no statistically significant results, industry-funded researchers tended to interpret the results as favorable to the sponsor. In my collection, favorable spins were especially noticeable in studies sponsored by supplement companies.

These results, casually obtained as they were, are reasonably consistent with the more scientific results of systematic studies. “Systematic” means attempts to ensure scientific validity by setting criteria for analysis in advance, casting a wide net for studies likely to meet the criteria, evaluating each study on the basis of those criteria, and using validated methods of meta-analysis to summarize and interpret the evidence from the studies chosen for evaluation. “Meta-analysis” is the term used to describe a process of combining the results of multiple studies of the same phenomenon to increase the statistical power of their conclusions.

Gary Sacks and his colleagues at Deakin University in Melbourne, for example, used systematic methods to find out how much nutrition research is funded by food companies or conducted by researchers with financial ties to such companies. They examined every peer-reviewed research article published in the fifteen most-cited nutrition journals in 2014. Their as yet unpublished results show that of more than four thousand studies, the great majority were funded by government agencies or foundations. Only 14 percent disclosed food-company funding or financial ties. But of that cohort, more than 60 percent reported results favorable to the sponsor, whereas only 3 percent came to unfavorable conclusions (the rest were not applicable to the sponsor’s interests).¹² Funding effects appear in nutrition research as well as drug research.

But in contrast to the thousands of examinations of drug-industry funding published over the years, only a few studies have examined the effects of food-industry funding on the outcome of research on nutrition and health. By August 2018, I had identified precisely eleven published studies, although from doing journal peer reviews I know that others were in the pipeline.

And while I was working on this book, Professor Ralph Walton, now retired from his position as chair of the Department of Psychiatry at Northeast Ohio Universities College of Medicine, sent me what is surely the first attempt at such studies: a never-published review of research on the safety of the artificial sweetener aspartame that he had prepared for a televised interview with Mike Wallace in 1996. Walton had seen that the conclusions of studies on the safety of aspartame were highly contradictory and wondered whether study outcome correlated with funding source. It did. NutraSweet, the maker of aspartame, funded seventy-four studies; all concluded that the sweetener was safe. But among ninety-two independently funded studies, eighty-four—more than 90 percent—questioned its safety.¹³

The first published funding-effect study of nutrition research appeared only in 2003—decades after the earliest studies of drug-industry effects. I can think of several explanations for the delay. Food is more complicated than drugs, and research on it is more difficult. But beyond that, we must eat to live, and food companies make products we love to eat. It hardly seems possible that food companies would need to be as deliberately manipulative as drug or cigarette companies. But the few studies to date suggest close parallels.

I say “suggest” because this research is difficult to discuss in any coherent way. The few studies vary in funding sources, degree of funding, and whether authors had financial ties to the funder. They examine the effects of widely different products, and they differ in the health outcomes measured, in the methods of examination and analysis, and in their findings. In Table 3.1, I have summarized the differing elements to emphasize how difficult it is to draw simple conclusions from just eleven studies.

Credit for the first published study of food-industry influence goes to investigators at Teachers College, Columbia University. They were interested in whether funding source had anything to do with conclusions about the safety and effectiveness of Olestra, a no-calorie fat substitute created by Procter & Gamble (P&G). I had a special interest in this study. In 1998, I had written an account of P&G’s $500 million, twenty-five-year campaign to convince the FDA to approve Olestra as an ingredient in snack foods, despite concerns about its tolerability, interference with vitamin A absorption, and ability to help people lose weight.¹⁴

TABLE 3.1. Studies Examining Food-Industry Influence on Nutritional Health Research, 2003–2018

NOTE: Source notes are located in the “Notes to Tables” here.

The Columbia investigators examined sixty-seven articles of various types: research reports, reviews, commentaries, and letters. Of those arguing for the safety and utility of Olestra, 80 percent were sponsored by P&G or other food companies. But among articles expressing doubts, 89 percent were funded by nonindustry groups. Among studies by authors who reported employment or consulting affiliations with P&G, every single one favored Olestra. If you have never heard of Olestra, it is because it did poorly on the market, and few products still contain it; in 2010, Time ranked it as one of the world’s fifty worst inventions.¹⁵

By the early 2000s, weight gain and obesity had become widely recognized as national public health problems. In looking for causes, researchers singled out soft drinks for their high sugar content and lack of nutritional value. They began to examine links between sugar-sweetened beverages and increased risks for obesity, especially among children.¹⁶ To counter these findings, Coca-Cola and the American Beverage Association began funding their own studies. This made it possible to compare study outcome by funding source.

Six of the eleven funding-effect studies deal with the effects of sugar-sweetened beverages (SSBs). The first appeared in 2007. Its authors reviewed more than two hundred studies of the effects of sugary drinks on health. Studies funded by industry were eight times more likely to produce favorable conclusions than those funded by nonindustry sources. That same year, a systematic review and meta-analysis of eighty-eight studies linked sugary drinks to higher calorie intake and related health risks, but studies funded by food companies reported smaller effects than those funded by nonindustry sources. Three other studies confirm these findings.¹⁷ A fourth, published in 2016, concluded, “This industry seems to be manipulating contemporary scientific processes to create controversy and advance their business interests at the expense of the public’s health.”¹⁸ The most recent, which observed systematic underreporting of adverse health effects in industry-funded studies, argues that such research “hindered the pursuit of scientific truth about the health effects of SSBs, and may have harmed public health.”¹⁹

My interpretation: these results are generally consistent with what we know about the effects of drug-industry funding, but with a bit less certainty. In 2016, Lisa Bero and her colleagues attempted to resolve the uncertainty by performing a systematic analysis of every study they could find on nutrition-funding effects. Of an initial pool of 775 studies, only 12 met their criteria for inclusion. Their cautious analysis: “Although industry-sponsored studies were more likely to have favorable conclusions than non-industry sponsored studies, the difference was not significant,” meaning that this result could have occurred by chance.²⁰

What are we to make of this? Reviews of reviews, no matter how systematic, have inherent sources of error. The small number of studies lacks the statistical power of the hundreds or thousands of studies examining drug-company influence. The Bero group’s review included the studies listed in Table 3.1 except for the three published in 2016 or later, which appeared after its analysis. But this review also included four studies omitted from Table 3.1. One dealt with genetically modified foods, not health.²¹ The other three also did not look at health outcomes. Instead, they examined the effects of funding source on the quality of the science. This is a separate issue worth its own discussion, not least because those three studies, all finding no effect of industry funding on scientific quality, were either sponsored by a food-industry front group or conducted by authors with ties to that group or to food companies.

The front-group funder was the North American branch of the International Life Sciences Institute (ILSI), an organization that turns up often in this book. ILSI describes itself as an independent scientific think tank, but it was created and is largely funded by the food industry. This makes it, by definition, a front group. But you might not realize this from reading the study authors’ disclosure statement, which describes ILSI as “a public, nonprofit scientific foundation that provides… a neutral forum for government, academic, and industry scientists to discuss and resolve scientific issues of common concern for the well-being of the general public.”²²

ILSI keeps a relatively low public profile but seems never to miss an opportunity to defend the interests of its four hundred or so corporate sponsors. Its 2016 annual report takes four pages and fifteen columns to list industry supporters of its national and international branches; these contribute two-thirds of this group’s nearly $18 million in annual revenues (the rest comes from government or private grants or contributions). ILSI’s board of trustees is about half industry and half academia, all unpaid volunteers. Critics describe ILSI as a “two-level” organization. On the surface, it engages in legitimate scientific activities. But deep down, it provides funders with “global lobbying services… structured in a way which ensures that the funding corporations have majority membership in all its major decision-making committees.”²³

ILSI has a vested interest in defending the scientific quality of industry-funded research. So do investigators with financial ties to food companies. One of the other two studies finding no funding-source effect on scientific quality looked at randomized controlled clinical trials reported in top-tier medical journals. Its senior author, David Allison, disclosed that he “has received grants, honoraria, donations, royalties, and consulting fees from numerous publishers, food, beverage, pharmaceutical companies, and other commercial and nonprofit entities with interests in obesity and randomized controlled trials.” The second study, also with Allison as senior author, found that industry-funded clinical trials had higher scores for the quality of science reporting than those funded by nonindustry sources.²⁴ These studies of scientific quality prove the rule; the conclusions of studies by investigators with industry ties generally tend to favor the sponsor’s interests.

This rule especially holds in these instances because that last study is the only one of which I am aware suggesting that the scientific quality of industry-funded studies is better than that of studies funded by nonindustry sources. How well the studies are conducted is not the issue. Here again, the concerns about nutrition research are more complicated.

FIGURE 3.1. Biases can be introduced into the research process at any stage. Because biases can be unconscious and unintentional and are typically unrecognized, scientists must make special efforts to protect against them at every stage. Adapted from Odierna DH, Forsyth SR, White J, Bero LA, “The cycle of bias in health research,” Accountability Res. 2013, 20(2):127–141. Used with permission, courtesy of Lisa Bero.

Bero and her colleagues explain that investigators can introduce biases—consciously or unconsciously—at every stage of the research process, as shown in Figure 3.1.²⁵ The Bero group examined where funding biases are most likely to appear in research on obesity interventions. Their analysis finds food-industry funding to have little effect on how well the studies are conducted. Instead, the funding effects are more likely to show up in the way the research question is asked and the results interpreted. Obesity trials funded by industry tend to focus on the role of specific nutrients, whereas trials funded by independent sources ask broader and more complicated questions about dietary behavior. This discrepancy, they noted, “could limit the public health relevance of rigorous evidence available for systematic reviews and dietary guidelines.”²⁶

Research funded by food companies has an especially high probability of bias. For example, the requests noted earlier for proposals from the grape, yogurt, and strawberry trade associations expect that studies will yield evidence of health benefits; otherwise, the associations will not fund them. Bero’s group notes the ways researchers can skew studies to demonstrate benefits. They can focus on single nutrients, ingredients, or foods rather than on interactions or overall diets. They can compare the effects of single foods by contrasting diets that include them to diets that lack them. They can design trials without randomization, blinding, or appropriate comparisons. They can focus on obvious or irrelevant effects. And they can give a positive spin to results that show no effect or fail to publish unfavorable results.²⁷

Martijn Katan, a lipid researcher in Amsterdam, points out that food companies have no reason to look for unfavorable effects of their products and much prefer studies that allow them to adjust product dosages to increase the probability of finding benefits or of keeping adverse effects below statistical significance. He also emphasizes that industry funding can influence investigators to overlook unfavorable data, downplay negative results, or avoid publishing them out of reluctance to displease a sponsor.²⁸ Because all of this can happen unconsciously, investigators need to make special efforts to control for such bias. Some do. Others do not.

The ethicist Jonathan Marks notes that biases in industry-funded research are so embedded in the process that inappropriate framing of research questions and spinning of results are predictable. But investigators can favor industry in other ways—explaining obesity as a matter of individual responsibility (rather than as a result of the food environment), emphasizing physical inactivity rather than diet as its cause, or focusing on supplements rather than diets as solutions to health problems.²⁹ Even though not all industry-funded research favors the funder, we can understand why the funding is associated with research outcome. Food companies want to sell products. Researchers want to get grants. Industry funding is by no means the only reason for caution in interpreting research, but it is a big one. The next chapters explain why.