A couple different people pointed me to this excellent news article by Gina Kolata and Brad Plumer, who write:
Public health officials for years have urged Americans to limit consumption of red meat and processed meats because of concerns that these foods are linked to heart disease, cancer and other ills.
But on Monday, in a remarkable turnabout, an international collaboration of researchers produced a series of analyses concluding that the advice, a bedrock of almost all dietary guidelines, is not backed by good scientific evidence.
The distinction between truth and evidence. Check.
Kolata continues:
If there are health benefits from eating less beef and pork, they are small, the researchers concluded. Indeed, the advantages are so faint that they can be discerned only when looking at large populations, the scientists said, and are not sufficient to tell individuals to change their meat-eating habits.
The distinction between inference and decision. Check.
I don’t like this quote so much:
“The certainty of evidence for these risk reductions was low to very low,” said Bradley Johnston, an epidemiologist at Dalhousie University in Canada and leader of the group publishing the new research in the Annals of Internal Medicine.
Asking for certainty is typically a mistake.
Kolata then puts this into perspective:
The prospect of a renewed appetite for red meat also runs counter to two other important trends: a growing awareness of the environmental degradation caused by livestock production, and longstanding concern about the welfare of animals employed in industrial farming.
Then comes a clear distinction between aggregate and individual effects:
In each study, the scientists concluded that the links between eating red meat and disease and death were small, and the quality of the evidence was low to very low.
That is not to say that those links don’t exist. But they are mostly in studies that observe groups of people, a weak form of evidence. Even then, the health effects of red meat consumption are detectable only in the largest groups, the team concluded, and an individual cannot conclude that he or she will be better off not eating red meat.
And an expert quote reinforcing a key point:
David Allison, dean of the Indiana University School of Public Health—Bloomington, cited “a difference between a decision to act and making a scientific conclusion.”
It is one thing for an individual to believe eating less red meat and processed meat will improve health. But he said, “if you want to say the evidence shows that eating red meat or processed meats has these effects, that’s more objective,” adding “the evidence does not support it.”
Along with a relevant note:
Dr. Allison has received research funding from the National Cattlemen’s Beef Association, a lobbying group for meat producers.
And another perspective:
Evidence of red meat’s hazards still persuaded the American Cancer Society, said Marjorie McCullough, a senior scientific director of the group.
“It is important to recognize that this group reviewed the evidence and found the same risk from red and processed meat as have other experts,” she said in a statement. “So they’re not saying meat is less risky; they’re saying the risk that everyone agrees on is acceptable for individuals.”
Also some discussion of confounding:
“Do individuals who habitually consume burgers for lunch typically also consume fries and a Coke, rather than yogurt or a salad and a piece of fruit?” asked Alice Lichtenstein, a nutritionist at Tufts University. “I don’t think an evidence-based position can be taken unless we know and adjust for the replacement food.”
But this argument goes both ways. On one hand, sure, in an observational study it’s hard to untangle the difference components of diet and say that one particular food is bad for you. On the other hand, these observational data represent the real world, in which people don’t always eat “red meat” in isolation; rather, they’re eating burgers, fries, and a Coke, so we care about the effect of that too.
Some general concerns:
“I would not run any more observational studies,” said Dr. John Ioannidis, a Stanford professor who studies health research and policy. “We have had enough of them. It is extremely unlikely that we are missing a large signal,” referring to a large effect of any particular dietary change on health.
But, then again, the necessity of making decisions:
Despite flaws in the evidence, health officials still must give advice and offer guidelines, said Dr. Meir Stampfer . . . of the Harvard T.H. Chan School of Public Health. He believes that the data in favor of eating less meat, although imperfect, indicate there are likely to be health benefits.
One way to give advice would be to say “reduce your red meat intake,” Dr. Stampfer said. But then, “People would say, ‘Well, what does that mean?’”
Officials making recommendations feel they have to suggest a number of servings. Yet when they do, “that gives it an aura of having greater accuracy than exists,” he added.
And the conclusion:
Perhaps there is no way to make policies that can be conveyed to the public and simultaneously communicate the breadth of scientific evidence concerning diet.
Or maybe, said Dr. Bier, policymakers should try something more straightforward: “When you don’t have the highest-quality evidence, the correct conclusion is ‘maybe.’”
A fine ending, but we still need to decide what to serve for dinner! I’m left with no clear advice, but I applaud Kolata and Plumer for bringing up so many of the key statistical issues without getting distracted by non-issues such as statistical significance etc. There will undoubtedly be lots more reporting on this and related diet stories, and I hope that future news articles will continue this perspective.
In contrast, I’m less happy with this NPR report by Alison Aubrey, which to my taste puts too much of the focus on a narrow statistical issue of how particular studies are rated and doesn’t get to the larger concerns of aggregate vs. individual effects, truth vs. evidence, and the necessity of decision making. But my point here is not to slam this NPR report; it’s the way that such scientific controversies are typically covered in the news. The above-linked NYT article points to a better way forward.
from Statistical Modeling, Causal Inference, and Social Science https://ift.tt/2mrvXGJ
via IFTTT
Comments
Post a Comment