Obesity Trials Are Optimized to Win Approval, Not to Predict Real-World Use

Open Nature Medicine’s 2026 analysis on obesity drug development and you will find a number that should stop every protocol designer in the field: 16.6%. That is the permanent discontinuation rate in the semaglutide arm of the SELECT trial, compared to 8.2% in placebo, driven overwhelmingly by gastrointestinal adverse events. SELECT enrolled patients with established cardiovascular disease and overweight or obesity, and semaglutide still produced landmark cardiovascular outcomes. But nearly one in six patients could not stay on the drug. That gap, between what a trial measures and what patients can actually tolerate over time, is the fault line running through every obesity drug program currently in Phase 3.

The puzzle is this: the field has never been better at hitting its endpoints, and it has never been less clear that hitting those endpoints predicts durable clinical value.

The Machinery Behind the Measurement Gap

To understand why, you have to understand how the FDA defines success for obesity therapeutics. The agency’s January 2025 draft guidance sets the threshold plainly: a new obesity drug must demonstrate at least a 5% decrease in body weight compared to control after one year of treatment, and that difference must be statistically significant. Meet that bar, and you have a regulatory foundation for approval. The problem is that 5% at one year is a snapshot, not a trajectory. It captures where the drug takes patients, not whether patients stay on the drug long enough to get there, or stay there afterward.

The pharmacology makes this worse in a specific, mechanistic way. A 2025 Bayesian network meta-analysis published in Frontiers in Pharmacology, covering 48 randomized controlled trials and 27,729 participants, showed that GLP-1 receptor agonist-induced gastrointestinal adverse events operate through two distinct pathways: peripheral inhibition of gastric emptying and central activation of GLP-1 receptors in the brainstem. These are not transient annoyances that resolve after titration. They are structural features of the mechanism. The same receptor activity that suppresses appetite and drives weight loss is also responsible for the nausea, vomiting, and gastrointestinal distress that push patients off therapy.

This creates a design trap. Trials are optimized around dose escalation schedules that maximize efficacy at the cost of tolerability. Faster titration gets patients to therapeutic dose faster, which compresses trial duration and produces cleaner weight loss curves at the primary endpoint. But faster titration also concentrates the GLP-1 receptor activation that triggers gastrointestinal side effects. The dropout events that result are distributed across the early and middle portions of a trial, before the weight loss data matures. Intention-to-treat analysis can absorb these dropouts statistically, but it cannot tell a prescriber what will happen to a patient in month eight of real-world use when there is no study nurse and no protocol-mandated follow-up call.

Call this the Efficacy-Persistence Divergence. The metric a trial is built to optimize, mean weight loss from baseline to week 52 or 68, moves in the opposite direction from the metric that predicts real-world value, the proportion of patients still on therapy at 18 or 24 months. A drug can post an impressive hazard ratio in a cardiovascular outcomes trial and still be unusable for a third of the population that most needs it. That divergence does not appear in the labeling. It barely appears in the published trial data.

The placebo-corrected weight loss hierarchy tells a similar story. A 2024 analysis comparing GLP-1 receptor agonists placed liraglutide at approximately 5% placebo-corrected weight loss, semaglutide at approximately 12%, and tirzepatide at approximately 18%. Each step up in efficacy corresponds to higher receptor potency or dual-agonist activity. Each step also brings a steeper tolerability gradient. The field has been running a linear optimization on weight loss while ignoring the nonlinear relationship between dose, efficacy, and attrition.

When the Goalposts Were Always in the Wrong Field

The counterintuitive claim worth making here is this: the most rigorous obesity trials in history may be producing the least useful information for clinical practice. Not because the science is wrong, but because the endpoints were never designed to answer the question that matters to a prescribing physician or a health system payer. The question on the label is “how much weight does this drug produce?” The question in the clinic is “for how long, in whom, at what tolerability cost, and does the benefit persist after discontinuation?”

A 2024 analysis in PMC reviewing FDA obesity approval criteria made exactly this critique: focusing solely on weight loss as the primary efficacy measure can produce systematically inaccurate assessments of medication effectiveness, particularly in populations with differential tolerability profiles. Older patients, patients with gastroparesis risk factors, patients on polypharmacy that slows gastric motility: these groups appear in Phase 3 trials but are not the populations around whom endpoint selection is built. Their dropout curves are averaged away.

The SELECT trial data sharpens this point. In a trial designed to demonstrate cardiovascular benefit, where investigators had every incentive to keep patients on drug, 1,461 out of 8,803 semaglutide-arm patients still discontinued permanently due to adverse events. In a commercial setting, without the infrastructure of a Phase 3 trial, without the protocol-mandated monitoring, without the financial and logistical support that trial participation provides, that discontinuation rate almost certainly rises. The SELECT discontinuation figure is a floor, not a ceiling.

What Protocol Designers Can No Longer Afford to Ignore

The Nature Medicine analysis lands at a moment when the FDA itself is reconsidering its evidentiary frameworks. In February 2026, FDA Commissioner Dr. Marty Makary and Deputy Director Dr. Vinay Prasad published a New England Journal of Medicine commentary signaling a shift toward a one-trial default for drug approval, emphasizing a single robust pivotal trial over the traditional two-study requirement. That shift increases pressure on each individual trial to carry more evidentiary weight. A single pivotal trial built entirely around a weight loss endpoint now becomes the totality of the evidence base. If that trial captures nothing about persistence, tolerability-adjusted response, or patient-level heterogeneity in gastrointestinal burden, there is no second trial to fill the gap.

Protocol designers operating under these conditions have a narrowing window to act. Embedding patient-reported outcome instruments that capture gastrointestinal burden as a co-primary or key secondary endpoint is no longer optional if sponsors want their data to survive payer scrutiny. Building in pre-specified subgroup analyses for patients who discontinue due to adverse events, then tracking their weight trajectory post-discontinuation, generates the rebound data that health technology assessment bodies in the UK, Germany, and Canada are already demanding before granting reimbursement. None of this requires a new regulatory framework. It requires sponsors to stop treating tolerability as a safety table footnote and start treating it as a primary scientific question.

The FDA’s January 2025 draft guidance on obesity drug development creates the 5% threshold, but it does not prohibit sponsors from designing trials that exceed it in informational richness. The agency cannot force sponsors to measure persistence. Sponsors have to decide that the commercial and scientific cost of not measuring it is higher than the cost of the more complex protocol. Given that every major payer in every major market is now constructing reimbursement decisions around real-world adherence and durability data, that calculation is no longer difficult.

The next obesity drug that reaches an FDA advisory committee with 18-month persistence data, tolerability-stratified responder analyses, and post-discontinuation weight rebound curves will not just have a stronger label. It will have made every competitor’s existing approval look like it was asking the wrong question all along.

References

1. Nature Medicine — “Shifting the goalposts in obesity drug development” (2026)
2. NEJM — SELECT Trial: Semaglutide discontinuation data (16.6% vs 8.2%)
3. Frontiers in Pharmacology — “GLP-1RA gastrointestinal adverse event mechanisms: Bayesian network meta-analysis of 48 RCTs, 27,729 participants” (2025)
4. SOCHOB/2024 Review — “GLP-1 receptor agonists for obesity: weight loss outcomes, tolerability, side effects and risks”
5. GWU STOP Obesity Alliance — FDA January 2025 Draft Guidance on obesity drug development: 5% weight loss threshold criteria
6. PMC — Review of FDA obesity approval criteria and limitations of weight-loss-only endpoints (2024)
7. Cromos Pharma — FDA one-trial default commentary by Makary and Prasad, NEJM (February 2026)