How to read an obesity-drug trial: a 10-point checklist
Guide
Prepared by: Research Desk. Sources are linked for verification.
A single weight-change percentage often appears on its own in headlines, investor notes and social posts. Without its context, it says very little. This checklist sets out the ten questions we ask of every obesity-drug paper before a number goes into our status tables for each molecule. Each point shows where to find the answer in a published paper or a public trial record. The page reports how trials are designed and reported. It does not describe how any drug is used.
Why headline percentages mislead
Two trials of two different drugs can report figures that look directly comparable when they are not. They may have enrolled different people, run for different lengths of time, used a different statistical estimand, or compared the drug with placebo in one case and with another active drug in the other. A smaller number from a harder test can be more informative than a bigger number from an easier one. The ten points below help locate each headline figure before it is compared with anything else.
1. Who was enrolled
Population is the first filter. Developers often run separate trials in adults with and without type 2 diabetes, and results from those trials can differ. They answer different questions and should not be pooled.
Worked example: orforglipron's trial in adults with obesity or overweight, ATTAIN-1, is registered as NCT05869903 [2]. Our pipeline guide summarises a separate phase 3 orforglipron trial run only in adults with type 2 diabetes. REDEFINE 5, by contrast, enrolled adults with overweight or obesity with or without type 2 diabetes [1]. A mixed population makes its headline figure harder to line up against a trial that enrolled only one group.
Check: inclusion criteria, diabetes status, baseline body-mass index and baseline HbA1c. In a ClinicalTrials.gov record, these sit in the eligibility section.
2. How long the trial ran
Average weight change in these trials often continues for many months, so trial length shapes the headline. A result read at week 40 and a result read at week 70 are not the same measurement, even for the same drug.
Check: the timepoint of the primary endpoint, which is not always the end of follow-up. Then check whether the weight curve in the paper's main figure had flattened by that point or was still falling.
3. Which estimand the headline figure uses
An estimand defines exactly what quantity a trial is estimating. In obesity trials, two are common:
Treatment-regimen (treatment-policy) estimand: counts everyone randomized, whether or not they kept receiving the study drug. It reflects real-world discontinuation.
Efficacy estimand: estimates the effect had everyone stayed on the study drug. It is usually the larger number.
Worked example: if a press summary and a journal abstract quote different figures for the same trial, the first question is whether one uses the efficacy estimand and the other the treatment-regimen estimand. Neither is wrong; they answer different questions.
Check: the abstract or the first results table. Most papers state the estimand next to the primary endpoint.
4. The gap versus placebo or an active comparator
The raw percentage matters less than the difference from the control group.
Placebo-controlled trials: subtracting the placebo mean from the drug mean gives a rough gap, but it is only an approximation, not the paper's estimated between-group difference. Use the figure the authors report, with its confidence interval.
Active-comparator trials: REDEFINE 5 compared cagrilintide plus semaglutide against semaglutide alone, in an active-controlled design [1]. The gap in a trial like this measures what the added amylin component contributed on top of an already active drug. That is a harder test than beating placebo, so the gap will look smaller than a placebo gap for the same combination.
Check: the comparator arms, listed in the "arms and interventions" section of a registry record such as REDEFINE 1, NCT05567796 [3], and the stated estimated difference in the paper.
5. Adverse events and discontinuations
Efficacy without the safety table is half a paper. Look for three figures per group: the share of participants with any adverse event, the share with serious adverse events, and the share who stopped the study drug because of adverse events.
An active-controlled design changes how these figures read. In REDEFINE 5, both arms received semaglutide [1], so similar adverse-event rates between arms would be expected rather than reassuring or alarming. In a placebo-controlled trial, the placebo arm gives the baseline.
Trial safety tables are a controlled record from a defined population over a fixed period. After approval, the FDA collects adverse-event and product-quality reports on marketed products through its MedWatch program [5], which is a different kind of evidence.
Check: whether adverse-event rates are reported per group, and how discontinuations were handled in the estimand (point 3).
6. Where the trial ran
Geography affects baseline body size, diet, background diabetes care and which regulators will review the data.
Worked example: REDEFINE 5 was run in Japan and Taiwan [1]. Our pipeline guide notes that mazdutide's published phase 3 data come from trials in China. Regional trials are informative, but neither on its own shows how a drug performs in a global population; they typically sit alongside larger multinational programs.
Check: the country list (the "locations" section of a registry record), baseline weight, and whether the trial is part of a wider program.
7. Registration and pre-specified endpoints
A registry record sets out the trial's design and primary outcome on a public page. ATTAIN-1's record, NCT05869903 [2], and REDEFINE 1's record, NCT05567796 [3], are both public on ClinicalTrials.gov.
Check: compare the registered primary outcome and timepoint with what the paper headlines. The record's history of changes shows whether outcomes were edited and when. A headline built on a secondary or post-hoc endpoint deserves more caution than one built on the pre-specified primary.
8. Trial phase and regulatory status
Phase 2 trials explore which strengths to carry forward. Phase 3 trials are the confirmatory evidence regulators rely on. REDEFINE 5 is described as phase 3a [1].
Phase is not the same as approval. According to a May 2026 industry summary, orforglipron was approved for weight management, an FDA decision on cagrilintide plus semaglutide was expected in December 2026, and decisions on retatrutide and survodutide were listed as 2027 events [4]. That is a secondary summary, not an FDA record, so dates should be rechecked against official announcements. A strong phase 3 result for an investigational drug does not make it legal to sell; see our explainer on the approved, compounded and unapproved categories.
Check: phase, the date of the status claim, and whether the source is an official record or a secondary summary.
9. Who sponsored and ran the trial
Most pivotal obesity trials are funded by the drug's developer. That is normal and does not invalidate the results, but it shapes design choices, such as which comparator and which estimand lead the headline.
Check: the funding statement and author affiliations, usually at the end of the abstract or paper, and the "sponsor" field in the registry record. Also check whether an independent data monitoring committee is described.
10. Peer-reviewed paper or press release
Topline results usually appear first in a company announcement. The peer-reviewed paper arrives later with full methods, safety tables, subgroup analyses and a supplementary appendix. Numbers can shift between the two, for example when a different estimand is emphasized.
Check: whether a quoted figure comes from a journal article, a conference abstract, a registry record or a press release. On this page, trial design details come from a published paper and from trial registries, and regulatory dates come from a dated industry summary [4], labelled as such.
Where each answer lives: a worked lookup
Question | REDEFINE 5, from the published paper [1] | Where to look in a registry record (e.g. REDEFINE 1 [3]) |
|---|---|---|
Population | Overweight or obesity, with or without type 2 diabetes | Eligibility section |
Comparator | Semaglutide alone (active-controlled) | Arms and interventions |
Geography | Japan and Taiwan | Locations |
Phase | Phase 3a | Study details, phase field |
Design | Multicentre, randomised | Study design section |
Primary endpoint | Stated in the abstract | Outcome measures, with change history |
Sponsor | Funding statement in the paper | Sponsor field |
The headline weight figures for each program are collected in our pipeline guide. Read with this table, two trials of the same combination can differ on population, comparator and geography. The only fair statement is that each trial answered its own question; a head-to-head trial would be needed for a direct comparison.
Printable checklist
Who was enrolled, and what was their diabetes status?
How long did the trial run, and when was the primary endpoint measured?
Which estimand does the headline use?
What is the estimated difference versus placebo or the active comparator?
What were the adverse-event and discontinuation rates per group?
Where did the trial run?
Does the headline match the registered primary endpoint?
What phase is it, and what is the dated regulatory status?
Who funded and ran it?
Is the figure from a peer-reviewed paper or a press release?
Tools and further reading
A checklist covers the questions; a good book explains the reasoning behind them. Four kinds of title are worth comparing:
Type of book | Best suited to | What to look for |
|---|---|---|
Plain-language statistics guides | General readers new to percentages, averages and uncertainty | Clear explanations of confidence intervals and why averages mislead |
"How to read a paper" guides | Readers who want to work through journal articles section by section | Chapters on methods, results tables and appendices |
Clinical trial design textbooks | Researchers and analysts who want the full method | Coverage of randomization, blinding, missing data, estimands and multiplicity |
Drug-development books for investors | Readers following company pipelines | Explanations of trial phases, regulatory review and how topline results are reported |
Across all four types, we compare titles on neutral criteria:
Accuracy and currency: methods for handling dropouts and defining estimands have changed in recent years, so check the edition date.
Worked examples: real published studies dissected step by step teach more than invented data.
Level: whether the book assumes algebra and some statistics, or none.
Format: print and ebook editions reproduce tables well; audiobooks suit concept-heavy introductions better than technical texts.
Price transparency: list price, used availability, and whether a library or ebook edition exists.
A practical pairing is one "how to read a paper" guide with one trial design text, since together they cover both the published article and the decisions behind it.
Any commissioned link in this section is labelled, and how the site earns money is set out in the affiliate and advertising disclosure on our about page.
Book choices are made on the criteria above. Advertisers and affiliate partners have no say over which titles appear or how evidence is described on this site.
Last updated: 18 October 2026.
Not medical advice. This page reports published research and regulatory records for information only.
Research-only notice: Content on this site is provided for research and educational purposes only and is not intended for human use. It is not medical advice.
Translation notice: Some content on this site may have been machine-translated using Google Translate and may contain errors.
Sources
Reviewed by
Research Desk