How much of an oral GLP-1 medicine’s value depends on weight loss? OUTSTAND-1 offers a useful case study: safiglipron improved glucose control at every studied dose, but its lowest dose did not establish greater weight loss than placebo. Those are separate findings, even though they came from the same participants.
Published in Nature Medicine on 29 September 2026, this phase 3 trial studied people early in the course of type 2 diabetes who were managing it with diet and exercise alone. Its results help explain why a diabetes trial should be read on its own terms, rather than as an audition for the next obesity medicine.
Who was studied—and what “without medication” means
The original OUTSTAND-1 report describes 284 adults randomized at 46 sites in China. Median diabetes duration was 1.6 years, average HbA1c was 7.95%, and average BMI was 27.8. Everyone had gone without glucose-lowering medication for at least two months before screening; that does not mean everyone had never used it. About one-third had previously received such treatment.
Participants received once-daily oral safiglipron targeting 30, 60 or 90 mg, or placebo, for a 32-week core period. These are research regimens, not prescribing instructions. The sponsor-submitted registry, NCT06672172, identifies the drug as HRS-7535 and corroborates the randomized, double-blind design, enrollment and primary HbA1c endpoint at 32 weeks. It has no posted results; the paper supplies the outcome numbers.
Glucose improvement did not require a large weight change
At week 32, estimated HbA1c reductions were 1.40, 1.38 and 1.63 percentage points in the ascending safiglipron dose groups, versus 0.18 points with placebo. The largest placebo-adjusted difference was therefore 1.45 percentage points, not a 1.45% relative reduction in glucose. Its 95% confidence interval ranged from 1.14 to 1.75 points more reduction than placebo.
The corresponding average weight reductions were 1.87%, 3.45% and 4.78%, versus 1.22% with placebo. Weight-loss superiority was confirmed for the 60- and 90-mg groups in the prespecified testing sequence. For 30 mg, the estimated advantage was only 0.65 percentage points, with a confidence interval that included no difference.
That lowest-dose result is instructive: glucose control improved without an established additional weight benefit. It does not prove weight has no role in glucose changes, identify an ideal dose, or show that higher doses are preferable for a particular person. It does show why a single “works” or “doesn’t work” verdict misses what the trial measured. These figures and comparisons come from the paper’s primary and confirmatory secondary results.
Keep the analysis and the calendar attached to each number
All glucose and weight figures above use the treatment-policy analysis: it sought to estimate effects regardless of treatment discontinuation, insufficient exposure or prolonged additional glucose-lowering medication. Missing observations still required statistical assumptions. A separate supportive analysis asked a hypothetical question about outcomes without those intervening events; its estimates should not be substituted into the same comparison. Our guide to trial estimands explains the distinction.
The study continued for another 20 weeks, but participants initially assigned to placebo switched to safiglipron at week 33. Consequently, the 52-week findings are not a year-long placebo comparison. This matters when judging durability: later changes cannot establish what would have happened over the same period without active treatment.
Tolerability changes the interpretation
During the 32-week controlled period, gastrointestinal adverse events occurred in 44.3%, 60.0% and 73.6% of the ascending safiglipron dose groups, compared with 15.3% on placebo. Most were mild or moderate. Adverse events led to treatment discontinuation in 1.4%, 2.9% and 6.9%, respectively, versus none on placebo.
The report’s safety findings also include serious adverse events and one acute pancreatitis case in the 30-mg group. An event occurring during a trial does not by itself establish that the medicine caused it. With 284 participants, this study cannot settle uncommon risks. Symptom frequency and stopping treatment also answer different questions, as our tolerability explainer discusses.
Where this leaves oral GLP-1 evidence
OUTSTAND-1 differs from OUTSTAND-2, which compared safiglipron with dapagliflozin on background metformin. Neither comparison can rank safiglipron against orforglipron or other medicines tested in different populations. Nor does this early-diabetes population establish the expected weight loss in people with obesity without diabetes.
All participants were Chinese, and the trial was designed around glucose control rather than rare harms or heart and kidney outcomes. The sponsor funded and helped conduct and analyze the study; several authors were employees and some disclosed pending patents. We inspected the open-access main report and registry, but not the supplementary files or separate protocol.
A completed trial and a journal publication are also different from marketing authorization. Neither inspected record supplies a regulator’s approval decision or a weight-management label. The useful next evidence would include longer controlled follow-up, results in the intended weight-management population, and any actual regulatory decision. For now, OUTSTAND-1 answers a narrower question well: glucose and weight outcomes can diverge, and readers need both to understand the result.



