Why It's So Hard to Get to the Bottom of Links Between Food and Cancer

Drop pretty much everything you're eating if you want to live. A recent study published in the American Journal of Clinical Nutrition -- featuring the scary title, "Is everything we eat associated with cancer?" -- found that somewhere around 80 percent of ingredients it randomly pulled from a hallowed cookbook was associated with the chronic disease that kills millions a year.

Wait, scratch that. Those alarming links frayed the further the study's researchers crunched the data. Welcome to cancer's uncertainty principle.

"We undertook this study because the media and the public focus intently on published research that links various foods to cancer," Harvard University radiation oncologist Jonathan Schoenfeld told AlterNet. Schoenfeld co-authored the study with Stanford University Prevention Research Center director John Ioannidis, author of "Why Most Published Research Findings Are False," the open-access journal Public Library of Science Medicine's most downloaded study ever.

"But less effort has been spent trying to confirm initial findings, and less attention has been dedicated to studies with negative results that find no relationships between various foods and cancer," Schoenfeld said. "The results of our study suggest that many studies that claim links between different foods and cancer are based on relatively weak evidence, and when additional research is performed, it often fails to reproduce the initial findings. Our results also suggest that consumers, industry and decision-makers should not focus too much on any one study to guide decisions related to diet and cancer."

That's hard medicine to swallow for patients searching for elusive answers and hope in scientific and medical studies, which often "focus on the new and exciting," Schoenfeld added. Or consumers of the quite healthy cancer cookbook market, whose titles like Anticancer: A New Way of Life and The Cancer Cookbook: Food For Life offer recipes and advice on health and wellness for those with and without cancer.

There's "uncertainty about the strength of the connection between many cookbook and nutritional claims and conclusive evidence," a spokesperson for the National Cancer Institute, the federal government's principal agency for cancer research, told AlterNet. He deferred to the NCI's fact sheet on nutrition and cancer for further illumination on the risk and reward factors of particular foods, but noted that most of its links were epidemiological, which is to say not based on controlled clinical trials. Which is to say, unreliable.

"You should be careful about epidemiological studies," warned Center For Science in the Public Interest executive director Michael Jacobson, when AlterNet recently reported a study linking soda and prostate cancer. "Epidemiologic studies are insensitive and inexact means of detecting causation of cancer, or other health problems."

Causation was the goal of Schoenfeld and Ioannidis' research, which used the respected Boston Cooking School Cookbook -- written by Fannie Merritt Farmer and published in 1896 -- to "randomly select the ingredients for our study, most of which are commonly found in the U.S. diet," Schoenfeld said. "Fannie Farmer was an early advocate of the importance of diet in disease. The research on potential links between food and cancer has been growing, and we found that four out of every five ingredients selected at random had been previously analyzed."

"However, very few of these studies’ results were confirmed when results from multiple studies were combined, and often these various studies’ results contradicted one another," he added.

In other words, the primary culprit in Schoenfeld and Ioannidis' study ended up being not cancer itself, but rather publishing about cancer.

"In many of the studies that we looked at, the significance was just strong enough to meet the criteria to get published, which really isn't very convincing for most individual studies, and was even more dubious when we saw the same pattern over and over again," Schoenfeld explained. "We also compared the results of these individual studies to results of meta-analyses, studies where results from many individual studies are combined and analyzed together."

"In stark contrast to the individual studies," he added, "very few meta-analyses confirmed links between food and cancers. And when they did, the risks or benefits associated with the foods were almost always reduced compared to the individual studies' findings."

Schoenfeld and Ioannidis did not, however, set out to evaluate whether claims for any particular foods were overblown. Instead, they decided to look for larger trends, and when they found them, they often led inexorably to further skepticism. Part of that uncertainty is perhaps derived from meta-analysis itself, which can sometimes be skewed by publication and agenda bias. Or the fact that, in the end, no original study is actually undertaken. But the kind of causation these studies seek is often out of reach.

"The most convincing evidence generally comes from randomized trials, where the researchers designing the study specify what they expect to find ahead of time and have limited ability to change the design of the study afterwards," said Schoenfeld. "But randomized studies are very expensive and often not feasible, especially when it comes to evaluating the links between food and cancer."

The great takeaway of Schoenfeld and Ioannidis' study seems to be that skepticism is as trustworthy as science, especially when it comes to the shark tank of scientific publishing. "Skepticism is an important part of science," Schoenfeld said. "The incentives that reward overstating findings are widespread."

What consumers are left with are resilient gaps where their medical and nutritional certainties should be, which is sadly to say back where they were in the first place. Arguing that the links between particular foods and forms of cancer are specious seems to be going against the medical mainstream, which Schoenfeld noted is seeing more links as cancer rates predictably rise.

"As a radiation oncologist, I have witnessed many cancer patients’ search for answers about what caused their disease," Schoenfeld concluded. "They're eager to do anything that may help increase chances of long-term survival. But our research shows that drastically changing behavior based on a single study is likely not a good idea, and may even distract from more proven ways to preserve health, such as eating a balanced diet, maintaining a healthy weight and avoiding smoking."