Evidence & Research Status
Real-world GLP-1 outcomes rarely match the figures reported in registration trials, and the size of that gap has become one of the more consequential open questions in metabolic research…
Evidence Level reflects the general state of published research for this compound category and is not a claim about the safety, efficacy, or approval status of any product sold on this site. All compounds are offered strictly for research use.
Real-world GLP-1 outcomes rarely match the figures reported in registration trials, and the size of that gap has become one of the more consequential open questions in metabolic research. Understanding why the two data sources diverge changes how any single headline percentage should be read.
The trial numbers
Controlled trials produce the numbers most people have seen. In the Phase 3 TRIUMPH-1 study of retatrutide, reported by Eli Lilly in May 2026 across 2,339 participants, investigators observed average body-weight reductions at 80 weeks of 17.6% in the 4 mg arm, 23.7% in the 9 mg arm and 25.0% in the 12 mg arm, against 3.9% on placebo (The Pharmaceutical Journal). Earlier Phase 2 work on the same triple-receptor agonist was published in the New England Journal of Medicine. Tirzepatide, a dual GLP-1/GIP agonist, has been associated with roughly 20–22% average reductions in comparable obesity populations.
What routine-care datasets report
Observational analyses drawn from ordinary clinical practice describe a different distribution. Summaries of routine-care data circulating in 2026 report average weight reduction in the range of 2–8% after one year, with roughly one third of individuals reaching more than 5% reduction. That is not a small discrepancy — it is a difference of an order of magnitude at the upper end, and it is the single most important caveat attached to any trial-derived percentage.
Four explanations researchers have proposed
Persistence. Trial participants are supported, monitored and reimbursed. In routine care, a substantial share of people stop within the first year. A 12-month average that includes people who discontinued at month three is arithmetically bound to be lower than a per-protocol trial figure.
Escalation that never completes. Trial arms are defined by the amount participants actually reach and maintain. Outside a trial, cost, supply interruption and tolerability mean many never reach the amount that generated the headline number, so the comparison is not like-for-like.
Population differences. Registration trials apply entry criteria — BMI thresholds, excluded comorbidities, excluded concomitant medications. Routine-care cohorts include everyone. Research on obesity subtype and response variation suggests the underlying biology is not uniform across those populations.
Measurement. Trials weigh participants on a schedule. Retrospective chart reviews capture whatever weight happened to be recorded, at irregular intervals, with loss to follow-up concentrated among those doing least well.
Side-by-side
| Dimension | Registration trial | Routine-care dataset |
|---|---|---|
| Population | Screened, criteria-limited | Unselected |
| Follow-up | Scheduled, funded | Opportunistic, incomplete |
| Continuation at 12 months | High | Substantially lower |
| Reported average reduction | ~17–25% | ~2–8% |
| Comparator | Randomised placebo | Usually none |
Why this generalises beyond GLP-1 research
The same asymmetry applies to every compound class discussed in this library. A controlled finding describes what happened under controlled conditions; it does not describe what happens at population scale. That distinction is also why the gap between biomarker movement and clinical outcomes deserves attention, and why evidence gaps in recovery peptide literature matter more than the volume of published papers.
Safety data show the same pattern. A KFF Health News analysis of the FDA Adverse Event Reporting System found medication-error reports involving these compounds rose from just over 2,000 in 2020 to more than 25,000 in 2025 (KFF Health News). Those errors are, almost by definition, invisible inside a monitored trial.
Open questions for researchers
- How much of the gap is explained by discontinuation alone, once analyses are restricted to individuals still engaged at twelve months?
- Do triple-agonist compounds such as retatrutide show a narrower or wider trial-to-practice gap than dual agonists, given differing tolerability profiles?
- Are routine-care datasets systematically under-capturing responders, who have less reason to return for a recorded weight?
- What measurement standard would make observational cohorts genuinely comparable to randomised arms?
Until those questions are addressed, the honest reading of any published percentage is that it is a ceiling observed under specific conditions, not an expectation. Anyone evaluating should hold both numbers in view at once.
How to read a percentage in a press release
Three questions separate a durable finding from a headline. First, what was the denominator — every randomised participant, or only those still enrolled at the end? Intention-to-treat and completer analyses can differ by several percentage points on the same dataset. Second, over what interval was the change measured; an 80-week figure and a 40-week figure are not interchangeable, and the curve is rarely linear. Third, what happened in the comparator arm? A 25% reduction against a 3.9% placebo arm carries different information than the same number reported with no control group at all. Applying those three questions consistently is the difference between reading published research and repeating it.
None of this makes the trial data wrong. Randomised evidence remains the strongest signal available for any of these compounds, and the retatrutide and tirzepatide programmes are unusually large. It does mean that the distance between a controlled result and an observed population average is itself a research finding, and one that is still being characterised.
All compounds referenced in this article are available for research use only. This article is provided for informational and educational purposes. No therapeutic claims are made or implied.