GLP-1 Non-Responders: New SELECT Analyses and a Nature Report on Why Response Varies

What the new research on GLP-1 non-responders actually says

Nature published a news feature on 24 September 2026 examining a problem that has been visible in clinical practice since GLP-1 receptor agonists became widely used for obesity: the drugs work dramatically for many people and barely at all for others. The piece frames the question as a research quest rather than a settled finding, reporting that understanding why some patients see no benefit from potent anti-obesity medications could point toward new therapies and more personalized weight-loss regimens.

That framing matters because the trial averages most people encounter are means, not promises. A reported average weight change of a given size across thousands of participants is compatible with a wide spread of individual outcomes, including a tail of participants whose weight barely moved. The people in that tail are usually invisible in headline results, and until recently they were largely invisible in the research agenda too.

GLP-1 receptor agonists — semaglutide, marketed as Wegovy and Ozempic, and the dual GIP/GLP-1 agonist tirzepatide, marketed as Zepbound and Mounjaro — act on receptors in the pancreas, gut and brain that influence insulin secretion, gastric emptying and appetite signalling. Why the same receptor engagement produces very different downstream results across individuals is not resolved.

Why do some people not lose weight on semaglutide or tirzepatide?

There is no single validated answer, which is precisely the point of the Nature report. Candidate explanations under investigation span genetics, differences in receptor signalling, variation in how much of the dose is tolerated before side effects force a change, behavioural and environmental factors, and the possibility that obesity is not one disease but several with distinct biology.

One practical complication is that apparent non-response and interrupted exposure look identical on a weight chart. Someone who paused because of supply gaps, nausea or cost will show a flat trend that is indistinguishable, in retrospect, from true pharmacological non-response. Distinguishing the two requires a record of what was actually taken and when — something clinical trials capture carefully and real-world use often does not.

A second complication is timing. Weight loss on these drugs is not linear, and the point at which an individual reaches maximum weight loss varies. A trend that looks like failure at one stage of a titration schedule may look different months later, which is part of why the SELECT analysis of peak weight loss is useful.

Readers reconciling a prescribed schedule with what they actually recorded often start with a semaglutide dosage calculator to keep the arithmetic of their logged weeks consistent.

The SELECT predictors analysis, published in Diabetes Care by Lingvay, Kushner, Kaul and colleagues, examined what was associated with peak body weight loss on semaglutide. the SELECT predictors analysis.

What the SELECT trial analyses add

SELECT was a large cardiovascular outcomes trial that tested weekly semaglutide against placebo in adults with overweight or obesity and established cardiovascular disease who did not have diabetes. Its primary result — a reduction in major adverse cardiovascular events — was the basis for semaglutide's cardiovascular indication. Because the trial followed participants for years and recorded weight throughout, its dataset is now being mined for questions the primary publication did not address.

The first of the two new papers, published in Diabetes Care by Ildiko Lingvay, Robert Kushner, Upendra Kaul and colleagues, is titled "Predictors of Peak Body Weight Loss With Semaglutide in the SELECT Trial." It examines which baseline and on-treatment characteristics were associated with how much weight participants ultimately lost at their individual peak — the analytic complement to asking who does not respond.

The second, in the European Heart Journal from Helen Colhoun, A. Michael Lincoff, Donna Ryan and co-authors, is a mediation analysis: a statistical method that asks how much of an observed treatment effect travels through a specific intermediate variable. Here the question is how much of semaglutide's cardiovascular risk reduction is explained by the weight loss it produces, and how much appears to occur through other pathways.

Because interrupted exposure and true non-response look identical on a weight chart, adherence tracking is what separates the two in a retrospective review.

A mediation analysis in the European Heart Journal from Colhoun, Lincoff, Ryan and co-authors asked how much of semaglutide's cardiovascular risk reduction travels through weight loss itself. a mediation analysis in the European Heart Journal.

Why the mediation question matters to people who lose little weight

If a large share of cardiovascular benefit in SELECT is not statistically mediated by weight change, that implies the drug is doing something for cardiovascular risk that is not simply a consequence of a smaller body. Prior work on this class has pointed toward effects on blood pressure, inflammation, lipids and vascular biology that are not obviously downstream of weight alone.

The practical implication for a reader watching their own scale is a change in what the scale means. Weight is the most visible and most frequently recorded output of a GLP-1 protocol, but it may not be the only endpoint that matters clinically, and it is a poor proxy for everything else the drug is doing. What counts as benefit is a conversation for a prescriber, informed by the full picture rather than one number.

None of this should be read as reassurance that flat weight equals hidden benefit. A mediation analysis describes statistical relationships across a trial population; it does not tell any individual what their own drug is doing.

Judging a plateau also depends on where someone sits in a stepped schedule, which is why it helps to be clear on what titration means before reading a flat trend as failure.

What is being developed for the gaps GLP-1 drugs leave

The research response to variable GLP-1 response is showing up on ClinicalTrials.gov. Alveus Therapeutics is recruiting for NCT07837973, a Phase 2 study of ALV-100 in adults with overweight or obesity who have already achieved at least a 10% body-weight reduction on semaglutide; the stated aim is efficacy, safety and tolerability for maintenance of that loss, with planned enrollment of 220 participants. Maintenance after GLP-1 therapy is a distinct problem from initial response, and it is now being studied as one.

A separate line of work targets body composition rather than body weight. A paper in the Proceedings of the National Academy of Sciences from Thorne, Waldeck, Rose and colleagues describes an RNA thermogenic therapy intended to preserve lean mass and improve metabolic health during GLP-1-associated weight loss. That work is preclinical and should be read as an early-stage research direction, not a treatment.

Comparator trials are also multiplying. A completed Phase 3 study registered as NCT06738979, sponsored by Chia Tai Tianqing Pharmaceutical Group, compared a semaglutide injection against Wegovy in 408 participants with a BMI of 28 or above, using an equivalence design with a prespecified margin for percentage weight loss at 44 weeks.

Pauses driven by tolerability are one of the most common mundane explanations for a stalled trend, and side-effect logging is what makes them visible months later.

What remains unresolved

No biomarker currently identifies a likely GLP-1 non-responder before treatment starts, and the Nature feature does not claim one exists. The SELECT analyses describe associations within a trial population that was selected for established cardiovascular disease and the absence of diabetes, which limits how far their predictors generalize to people taking these drugs purely for weight management.

There is also a definitional problem. "Non-response" is usually operationalized as failing to reach a fixed percentage of weight loss by a fixed time, and that threshold is a convention rather than a biological boundary. Shifting the threshold or the time point changes how many people are counted as non-responders, which complicates comparison across studies.

Expect follow-up work to focus on whether early response predicts eventual peak, whether switching within the class helps, and whether non-response to one incretin mechanism predicts non-response to another. Those questions are open.

In a Nature news feature published on 24 September 2026, reporters described an active scientific effort to explain why potent anti-obesity medications produce no benefit in some people.