Case Study 1 — The Folder: Why Both Headlines Were Right

A clinical case. Theo Vasquez and the Riverbend Nutrition Clinic are illustrative composites.

Setup

At the end of Theo's first appointment I asked if I could keep the folder for a week.

He hesitated — I think he'd expected me to wave it away, the way a few clinicians had, or to pick a side and tell him the other articles were junk. Instead I told him I wanted to look up the actual studies behind four of his pairs, and that I'd walk him through what I found.

He came back the following Thursday. I'd printed the source papers, or the closest thing to them I could get, and laid them out next to his clippings.

"Before we start," I said, "I want to make a prediction. I think when we're done, you're going to be more annoyed, not less. Because the answer isn't that one side was lying."

He said that sounded like something a person says right before they fail to explain something.

Fair.

The pair we spent the most time on

Theo's most-highlighted pair was about saturated fat.

Clipping A: "Saturated fat is the primary dietary driver of heart disease. Health authorities continue to recommend limiting intake to under 10% of calories."

Clipping B: "Meta-analysis finds no significant association between saturated fat intake and cardiovascular mortality."

Theo had written in the margin of B, in blue pen: "SO WHICH IS IT."

Here's what we found.

What clipping B's study actually did

The meta-analysis pooled a set of prospective cohort studies — people surveyed about their habitual diet, then followed for years, with deaths recorded. It asked: among these people, did those eating more saturated fat die of cardiovascular disease at higher rates than those eating less?

And the pooled answer was: not to a degree that reached statistical significance.

That is a real finding. It was not fabricated, not funded by the beef industry, not misreported by the journalist. The paper says roughly what the headline says.

But look at the question it asked. It compared people eating more saturated fat to people eating less saturated fat in their ordinary lives. And here is the thing Theo hadn't considered, and that almost nobody considers:

What were the low-saturated-fat people eating instead?

In most of these populations, over the decades these cohorts ran, people who cut saturated fat largely replaced it with refined carbohydrate — because that's what the food supply offered them. Low-fat cookies. Fat-free yogurt with added sugar. White bread. The entire low-fat era, which Theo himself participated in back in 2016.

So the study's real question was closer to: "Is saturated fat worse than refined carbohydrate?"

And the honest answer to that is: not clearly, no.

What clipping A's recommendation is based on

The guidance to limit saturated fat rests on a different body of evidence, asking a different question — largely controlled trials and substitution analyses where saturated fat was replaced with a specified alternative, usually polyunsaturated fat.

Those studies ask: "Is saturated fat worse than polyunsaturated fat?"

And that answer is more consistent: replacing saturated fat with unsaturated fat tends to lower LDL cholesterol and, in the trial evidence, is associated with fewer cardiovascular events.

So both were right

I drew this table for Theo on the back of one of his own printouts.

He looked at it for a long moment and said: "So the second article isn't wrong. It's just answering a question nobody asked."

Close. It's answering a question the researchers were careful to specify and the headline dropped. The paper almost certainly discusses the substitution problem somewhere in its limitations section. The headline could not carry it.

The other three pairs, briefly

Once Theo had the substitution frame, the rest went quickly.

Eggs. Clipping A was reporting an observational association in a population where high egg intake travelled with other things — the classic confounding problem. Clipping B was reporting on the weakening of the dietary-cholesterol-to-blood-cholesterol mechanism. Different questions. Both defensible. Neither headline said which question it was answering.

Red meat. Clipping A reported IARC's classification of processed meat, which is a statement about confidence that it causes cancer, not about how much risk. Clipping B reported a recommendation panel that looked at the same evidence and reached a different conclusion about what to advise people to do, largely because they weighted certainty of evidence differently when making a recommendation. That is not a factual disagreement. It's a values disagreement about how to act under uncertainty, reported as though it were a factual one.

Theo found this one the most annoying, which I think is correct.

Seed oils. Clipping A was, when we traced it, downstream of mechanistic and animal work. Clipping B was reporting human substitution evidence. Different rungs of a ladder Theo hadn't been shown yet. We deferred this one to Chapter 19 territory.

Analysis

Four pairs of contradictory headlines. Zero instances of dishonesty. What actually produced the contradiction, in every case:

1. The substitution question was unspecified. Every dietary comparison is versus something, and the headline never says versus what. This alone accounted for two of the four pairs.
2. Different study designs answer different questions, and the headline never says which design.
3. A statement about evidential confidence (IARC Group 1) was read as a statement about magnitude of risk.
4. A disagreement about how to act under uncertainty was reported as a disagreement about facts.

Notice that none of these require a bad actor. They require only that headlines are short.

Theo's conclusion — "so which one of you is lying?" — was the natural inference from what he'd been given. It was also the inference that had cost him eight years and five diets, because a person who concludes that everyone is lying stops updating on evidence altogether, and then eats according to whichever confident voice reached them most recently.

What I actually gave him wasn't an answer about saturated fat. It was a question: compared to what? He left with that written on an index card. He told me later he'd stuck it to the inside of a kitchen cabinet, and that it had made about half the content in his feed unreadable in a way he described as "extremely irritating and probably good."

Discussion Questions

1. The chapter argues that "misleading-but-true is more dangerous than false." Which of the four pairs best illustrates this, and why?

2. Suppose you're the journalist writing about the meta-analysis in clipping B. You have 900 words and a headline of about eight. Write a headline that is both accurate about the substitution problem and likely to be read. Is this possible? If you conclude it isn't, what follows?

3. The red meat pair involved a disagreement about how to act under uncertainty rather than about facts. Can you think of a non-nutrition example of the same thing — where experts agreed on the evidence and disagreed on the recommendation? What made the disagreement look factual from outside?

4. Theo had highlighted clipping B in two colors and written "SO WHICH IS IT" in the margin. What does the effort he put into the folder tell you about the common assumption that people eat badly because they don't care?

5. I told Theo he'd end up more annoyed rather than less. Was that the right thing to tell a patient at the start of a consultation? What's the argument against it?

Your Turn

Take one contradictory pair of nutrition headlines you've personally encountered — or find one, which takes about four minutes on any search engine.

For each side, answer three questions:

• What study design is behind it? (Guess if you must, and say you're guessing.)
• What was the comparison? Versus what, exactly?
• Is the disagreement about facts, interpretation, or what to recommend under uncertainty?

If you find, as Theo did, that both sides are defensible answers to different questions, write down what the unified honest headline would have to say. It will be too long. Notice how long.