Case Study 1 — The Number That Grew in Transmission

Tracing one figure from a trial report to a headline

Type: Constructed analysis of a real, documented pattern · Tier 1 trial facts, Tier 3 transmission chain · Relevance: §9.4, and Chapter 6 §6.1

The starting material

SURMOUNT-1 reported, for tirzepatide 15 mg over 72 weeks in adults with obesity or overweight with a weight-related comorbidity and without diabetes, with lifestyle support in all arms:

About −21% from baseline on the treatment-regimen estimand. About −22.5% from baseline on the efficacy estimand.

Both figures appear in the trial report. Both are pre-specified. Both are correct. The report also gives the lower doses, the placebo arm, the discontinuation rates, and the adverse effects.

Nothing about that publication is misleading.

The transmission

What follows is a reconstruction of a pattern this book has observed repeatedly rather than a trace of any specific chain. No individual publication or outlet is being described.

HOW A NUMBER GROWS                                  [constructed reconstruction]

  STAGE 0 — THE TRIAL REPORT
  "At 72 weeks, mean percentage change in body weight with tirzepatide 15 mg was
   approximately −21% (treatment-regimen estimand) and approximately −22.5%
   (efficacy estimand) in adults with obesity or overweight with at least one
   weight-related comorbidity, without diabetes, all arms receiving lifestyle
   intervention."
                            │
                            ▼   [the estimand names are dropped as jargon]
  STAGE 1 — THE PRESS RELEASE
  "Tirzepatide produced up to 22.5% weight loss at 72 weeks."
                            │
                            ▼   [the duration goes; "up to" survives as a hedge]
  STAGE 2 — THE NEWS ARTICLE
  "Trial participants lost up to 22.5% of their body weight."
                            │
                            ▼   ["up to" is dropped as awkward]
  STAGE 3 — THE SUMMARY
  "Tirzepatide produces 22.5% weight loss."
                            │
                            ▼   [rounded up; the population goes]
  STAGE 4 — THE SOCIAL POST
  "People are losing almost a quarter of their body weight on this."
                            │
                            ▼
  STAGE 5 — WHAT A READER BELIEVES
  "If I take this, I'll lose about a quarter of my body weight."

  DISTANCE TRAVELLED: from a figure with five qualifiers to a figure with none.
  NUMBER OF FALSE STATEMENTS MADE: arguably zero until stage 3, and even stage 3
  is a true sentence about the efficacy estimand at the highest dose.

What was lost, and where

Six qualifiers, all present in the trial report, none present in what the reader believes.

And note that a reader at stage 5 who goes looking for the source will find stages 3 and 4 much more easily than stage 0. That is Chapter 6 §6.6's ranking problem, operating on a drug with excellent evidence.

⚠️ Hype Check — this happens to good drugs

The instinct on reading the above is that somebody behaved badly. Mostly nobody did.

The press release dropped the estimand names because they are genuinely jargon that most readers cannot use. That is a defensible editorial decision with a systematic consequence.

The news article dropped the duration because a headline has a word limit.

The summary dropped "up to" because it reads as hedging, and hedges are removed by every style guide in existence.

The social post rounded and generalized because that is what conversation does.

Each step is individually reasonable. The aggregate is a systematic exaggeration, and it is systematic rather than random because every step selects in the same direction: shorter, more definite, larger.

The important observation: this is exactly Chapter 6's pipeline, running on a drug with an excellent Phase 3 program, an approval, and real evidence. The mechanism does not care whether the underlying science is good. Which is why the skill has to be applied to compounds you believe in, not only to ones you doubt.

What this case teaches

The estimand is the first qualifier to go, and it is the most consequential. It is unfamiliar, it sounds technical, and it is the one whose loss systematically inflates the number. Every other lost qualifier narrows the population or the conditions; this one changes the arithmetic.

"Up to" is a hedge that survives one step and then dies. It is worth recognizing as a marker: a figure preceded by "up to" is a maximum under favorable conditions, and it will be reported without the "up to" within one or two retellings.

And the countermeasure is a habit, not knowledge. You cannot memorize the qualifiers for every drug. What you can do is reflexively ask, on meeting any effect size: in whom, at what dose, for how long, on which analysis, against what comparator. Five questions, ten seconds, and they reconstruct most of what was lost.

Discussion questions

1. Identify the first stage at which a false statement is made. Is it as early as you expected? What does that suggest about how misinformation is usually produced?

2. The press release dropped the estimand names because they are jargon. Was that the right call? Propose wording that preserves the distinction without using the word "estimand."

3. Every step in the chain selects in the same direction — shorter, more definite, larger. Why is the direction consistent? Is there any step where the incentive would push the other way?

4. A reader at stage 5 who searches for the source will find stages 3 and 4 more easily than stage 0. Whose problem is that, and what would fix it?

5. This case involves a drug with excellent evidence. Does that make the transmission problem better or worse than it would be for a compound with weak evidence? Argue both.

6. Build the countermeasure. Write the five questions you would ask reflexively on meeting any effect size. Then apply them to a health claim you encounter this week and record how many the source answered.