SURMOUNT-5 confidence intervals: arm intervals come from the abstract, the between-group interval from the registry
What WPF says today
In SURMOUNT-5, a 72-week open-label randomized trial in adults with obesity, tirzepatide produced a larger average weight loss than semaglutide (-20.2% against -13.7%). Other weight-loss results for the two drugs come from separate trials and cannot be subtracted.
Claim check: tirzepatide and semaglutide in the same trial (tirzepatide-semaglutide-same-trial)
Claim check: weight-loss results comparable (weight-loss-results-comparable)
Flagship result: SURMOUNT-5, week 72 (surmount-5-weight-72w)
What this change did to the conclusion
Source attribution corrected. WPF named the wrong source for a figure. The figures and the verdicts did not change; only where each one comes from.
Verdict: unchanged (Overstated (weight-loss check); Supported within scope (same-trial check))
The correction changed source attribution only. Every estimate and every verdict stayed the same. The intervals were already correct; the records now say that the arm intervals come from the publication abstract and the between-group interval from the registry results, so a reader who goes to either source finds the figure where WPF says it is.
Before and after
Earlier: version 1.4.0 (the same wording appeared in 1.1.0 to 1.3.0)
- Same-trial check: "Result -20.2% vs -13.7%, 95% CI -21.4 to -19.1 and -14.9 to -12.6; registry NCT05822830 results, primary outcome measure 1." The registry was named as the place for the intervals; the registry does not post them.
- Weight-loss check: "Result: -20.2% with tirzepatide vs -13.7% with semaglutide (least-squares means; P<0.001); registry NCT05822830 results, primary outcome measure 1." The arm means are posted in the registry and are also stated in the abstract; the P value is from the abstract only, because the registry posts none. This check never stated the arm confidence intervals.
Now: version 1.5.0
- Tirzepatide arm: -20.2%, 95% CI -21.4 to -19.1. Semaglutide arm: -13.7%, 95% CI -14.9 to -12.6. Each interval is for one arm's own mean and is stated in the publication abstract (with P<0.001).
- Between the arms: a least-squares mean difference of -6.5 percentage points, 95% CI -8.1 to -4.9. It is posted in the registry results and is not in the abstract.
- The registry also posts each arm's mean with a standard error of 0.59, and no arm confidence intervals.
Exact records. Earlier: Same-trial claim record, version 1.4.0 (sha256 277f6820…) · Weight-loss claim record, version 1.4.0 (sha256 1bf530a1…). Updated: Same-trial claim record, version 1.5.0 (as published 2026-11-16) (sha256 fbadc61d…) · Weight-loss claim record, version 1.5.0 (as published 2026-11-16) (sha256 79db0f05…) · Flagship record, SURMOUNT-5 (unchanged by this correction).
Where to verify it
- NEJM abstract, PMID 40353578: Abstract, Results: "-20.2% (95% confidence interval [CI], -21.4 to -19.1) with tirzepatide and -13.7% (95% CI, -14.9 to -12.6) with semaglutide (P<0.001)"; no between-group difference is stated in the abstract
- ClinicalTrials.gov NCT05822830, results: Primary outcome measure 1, percent change in body weight to week 72: arm means -20.2 and -13.7 (standard error 0.59 each); statistical analysis row, LS mean difference -6.5, 95% CI -8.1 to -4.9, two-sided
- Flagship Weight Management, SURMOUNT-5 result: Result surmount-5-weight-72w; its uncertainty field already attributed the arm intervals to the abstract
Dates
- Publication posted online (NEJM, PMID 40353578): . Print issue 2025-07-03.
- Registry results first posted (NCT05822830):
- WPF verified the source for this change:
- WPF published the corrected version:
Limits and review status
- The NEJM full text returned HTTP 403 to automated access and was not read (no workaround was attempted). The abstract and the registry were read instead, so WPF does not know whether the paper's own tables report the between-group difference.
- SURMOUNT-5 was open-label, and the abstract does not name the estimand; the match to the analysis plan was made by method (see the Flagship record).
Review status: AI-assisted source verification. Documented human review: none recorded. This is not scientific review and not attribution to a named reviewer.
Formal record: Correction Log entry; dated clarification: Correction Log entry.
Cite this change
World Peptide Foundation. "SURMOUNT-5 confidence intervals: arm intervals come from the abstract, the between-group interval from the registry". Change record surmount-5-confidence-interval-attribution, published 2026-11-16; WPF verified the source on 2026-11-16. Affected records: tirzepatide-semaglutide-same-trial, weight-loss-results-comparable, surmount-5-weight-72w. Change type: Source attribution corrected. Versions: 1.4.0 to 1.5.0. AI-assisted source verification; documented human review: none recorded. https://www.worldpeptidefoundation.org/research/changes/surmount-5-confidence-interval-attribution