Case Study 2 — The Red Meat Recommendations: When Everyone Reads the Same Evidence

A methodological and public-controversy case. The 2019 episode described is real; the reconstruction of the disagreement is my own reading of it.

Setup

In 2019 a group of researchers published a set of systematic reviews of the evidence on red and processed meat, together with a recommendation: that adults continue their current consumption.

The reaction was extraordinary. Major nutrition departments issued rebuttals. Public health organizations objected publicly, some before publication. There were calls for retraction. There were disputes about a lead author's prior disclosures. It was, by the standards of nutrition science, an enormous fight.

Here's what makes it the perfect case study for this chapter, and what almost none of the coverage conveyed:

The panel and its critics were largely looking at the same studies and did not substantially disagree about what those studies found.

They disagreed about something else entirely — and if you can see what, you understand something most science journalists don't.

What both sides agreed on

Strip away the noise and the common ground is large:

• Observational cohort studies consistently show associations between higher red and processed meat intake and higher risk of colorectal cancer and cardiovascular disease.
• Those associations are small in absolute terms — on the order of a small number of additional cases per thousand people.
• The evidence is almost entirely observational, subject to confounding, and rests on dietary measurement of known poor precision.
• There are essentially no long-term randomized trials of red meat consumption with hard endpoints, and there probably never will be, for the reasons in Chapter 1's §1.2.
• Therefore the certainty of the evidence, by formal assessment, is low.

Read that list again. It contains no disputed factual claim. Both camps would sign it.

Where they diverged

The methodological disagreement

The panel applied a formal evidence-grading framework — GRADE, widely used in clinical medicine — which systematically downgrades certainty for observational evidence, imprecision, and risk of bias. Applied to nutrition cohort data, GRADE reliably outputs "low" or "very low" certainty, because nutrition cohort data is, by GRADE's criteria, low certainty.

Critics argued that GRADE was designed for evaluating drug trials, where randomization is feasible, and that mechanically applying it to nutrition guarantees a verdict of "we don't know" for everything — including findings supported by decades of convergent evidence. By that standard, they pointed out, you'd also have to conclude we don't know much about smoking.

The panel's response, in effect: if the evidence really is low-certainty, saying so is honesty, not a flaw in the method.

Neither position is stupid. This is a real, unresolved methodological problem, and it recurs throughout nutrition.

The recommendation disagreement — the actual crux

Here's the deeper split, and it isn't about evidence at all.

Given low-certainty evidence of a small harm, what should you tell people?

That is a disagreement about values, and about the purpose of a recommendation — not about what the studies showed.

It was reported, near-universally, as a factual controversy. "Scientists say red meat is fine." "Experts slam flawed meat study." Readers came away believing the evidence had changed, or that one side was corrupt. Neither was true.

The disclosure question

One further thread, because it matters and because handling it well is itself a skill.

Questions were raised about whether a lead author had adequately disclosed prior industry-related research funding. This became a significant part of the public argument.

How should you weigh that?

What it justifies: heightened scrutiny of the work. Chapter 2 §2.8 says funding effects are real and operate through design and framing choices, not fraud. A disclosure question is a legitimate reason to examine the methodology closely.

What it does not justify: concluding the findings are wrong. The methodology is public. Anyone can check whether GRADE was applied correctly. If the reviews are sound, they are sound regardless of who funded the author's previous work — and if they're unsound, that can be demonstrated directly, which is a much stronger argument than an ad hominem.

And the symmetry test: several of the loudest critics have their own long-standing public positions, institutional identities built on dietary pattern research, and in some cases their own funding relationships. Chapter 1's rule applies without exception: follow the money in every direction, or don't invoke it at all.

Analysis: what this case teaches

1. Most public nutrition disagreements are not about facts. They are about how to act under uncertainty, disguised as factual disputes because that's the only register science coverage has. Once you can spot this, a startling share of nutrition controversy resolves into two people with different risk tolerances shouting past each other.

2. "Low-certainty evidence" is not "no evidence," and it isn't "the claim is false." It's a statement about how much you should update. The panel and its critics both understood this. The coverage did not.

3. Formal evidence frameworks encode values. GRADE isn't neutral machinery; the decision to downgrade observational evidence steeply is a judgment about what counts as knowing. Choosing a framework is choosing a set of priors, and reasonable people choose differently.

4. Precautionary and evidentiary standards genuinely conflict. If you demand high-certainty evidence before recommending, you will under-recommend on real harms. If you recommend on low-certainty evidence, you will sometimes be wrong in public — and the beta-carotene story in Case Study 1 is what that costs. There is no setting of this dial that avoids both errors.

5. Absolute risk is where the argument actually lives. Both sides agreed the absolute effect was small. Whether "small" means "not worth mentioning" or "large at population scale" is the entire dispute, and it's §2.6 doing all the work.

So what should you eat?

You now have enough to answer this for yourself, which is the point.

The evidence indicates that higher processed meat intake carries a small, real increase in colorectal cancer risk, with a dose-response relationship, of roughly the magnitude worked out in §2.6 — on the order of one additional case per hundred people over a lifetime at 50 g/day. Unprocessed red meat's association is weaker and less consistent. The evidence is observational, low-certainty by formal assessment, and unlikely ever to be resolved by trials.

What you do with that depends on things this book cannot know about you: how much you enjoy it, what you'd eat instead, your family history, your budget, whether environmental impact factors into your decisions, and how you personally weigh a small risk of a serious outcome.

That's not a cop-out. That's what an honest recommendation looks like when the evidence is low-certainty and the effect is small — and noticing that this feels less satisfying than "red meat causes cancer" or "red meat is fine" is itself the most useful thing in this case study.

Discussion Questions

1. Both camps agreed on what the studies found. Why was this reported as a factual controversy? What would accurate coverage have looked like, and would anyone have read it?

2. Is GRADE the right tool for nutrition evidence? If applying it rigorously means concluding "low certainty" for nearly every dietary question — including ones we're intuitively confident about — is that a flaw in GRADE, or an accurate reflection of nutrition's evidence base?

3. Construct the strongest case for the panel's position. Then the strongest case for the critics'. Which do you find more persuasive, and — importantly — is your answer about evidence or about your own risk tolerance?

4. The disclosure question: how much should an author's prior funding affect your assessment of a methodologically transparent systematic review? Does your answer change if the funding is from a supplement company rather than a food industry group?

5. Compare this case to Case Study 1. In beta-carotene, acting on low-certainty evidence caused harm. Here, the panel declined to act on low-certainty evidence and was widely condemned. Can you state a principle that gets both cases right? (This is genuinely hard. If your principle is easy, check it against both cases again.)

Your Turn

Find a current nutrition controversy where credentialed experts publicly disagree — there are many; sodium, saturated fat, and ultra-processed food definitions are all live.

For each side, determine:

• What evidence do they cite? Is it the same evidence?
• Where does the disagreement actually sit: facts, methods, interpretation, or what to recommend under uncertainty?
• What risk tolerance does each side's position imply?
• Would they still disagree if they agreed on all the facts?

Most of the time the answer to that last question is yes — which tells you the argument was never going to be settled by more studies.