Real, and Below the Threshold

Semaglutide improves patient-reported physical functioning. The improvement is 1.71 points, with a confidence interval from 1.07 to 2.35 — comfortably excluding zero, so it is real. The threshold at which a patient would notice a change is 3.7 points.^[1] Both of those sentences are true at once, and holding them together is what statistical and clinical significance actually mean.

Two different bars

Statistical significance asks whether an effect is distinguishable from chance. Clinical significance asks whether it is large enough to matter to the person experiencing it. They are separate questions with separate answers, and a result can clear one bar and fail the other.

This is that case, stated cleanly. The meaningful within-patient change threshold is a number researchers establish in advance, by asking patients how much a score has to shift before they perceive a difference. For this instrument it is 3.7 points. The drug delivers 1.71.

Definitely there, and less than half of what a person would feel.

The same pattern held on a second, weight-specific instrument across two trials and 2,768 participants: larger effects in people without type 2 diabetes, smaller in people with it, and both below the published thresholds.

The picture this completes

This register has arrived at the same place three times now by different routes, and this makes the fourth.

So the supportable claim is narrow and consistent: there is no good evidence these drugs make people feel better in general, and there is good evidence they relieve the symptoms of specific conditions they treat. Those are different promises, and only one of them has been kept.

Who ran every one of these trials

That matters less than it might, because the finding is a modest one. A sponsor’s own trials reporting an improvement smaller than patients would notice is not the direction bias usually runs. If anything, it makes the result more credible than a larger one from the same source would be.

Heterogeneity was moderate at 52.8%, which is unremarkable for a pooled analysis and worth noting beside the tighter confidence interval.

What to do with it

• Expect physical improvement proportional to the weight lost, and more of it if you lose more — that relationship is consistent across the evidence.
• Do not plan on feeling better in general. If that is the main reason for starting, this is the evidence to weigh, and it is not encouraging.
• The exception is real and specific. In obesity-related heart failure, these drugs relieve the symptoms of that condition by an amount patients notice.
• “Statistically significant” is not a synonym for “you will feel it.” When a paper reports an improvement, look for whether it names a threshold and whether the effect cleared it.

The authors’ own conclusion is that the gains are modest and should be communicated cautiously. That is an unusually plain thing for a paper to say about a positive finding, and it is the right instruction.