A new phase 3 trial of oral safiglipron offers a useful reminder: a drug can improve blood sugar more than its comparator without establishing a greater weight-loss benefit. Published in Nature Medicine on 5 October 2026, OUTSTAND-2 compared the experimental GLP-1 receptor agonist with dapagliflozin in adults whose type 2 diabetes was inadequately controlled with metformin.
The clearest positive finding concerned glucose control at the highest safiglipron dose. Weight fell modestly in every group, while gastrointestinal symptoms and treatment discontinuation were more frequent with safiglipron. For readers following oral weight-loss medicines, the useful question is what this diabetes trial actually establishes.
A diabetes comparison with an active treatment
The original OUTSTAND-2 report describes 810 adults randomized at 98 sites in China. Participants were aged 18–75, had type 2 diabetes and continued stable metformin. Mean baseline HbA1c was 8.60%; mean body weight was 74.7 kg. This was not a trial restricted to people with obesity without diabetes.
The research arms received once-daily safiglipron targeting 30, 60 or 90 mg, or dapagliflozin 10 mg. Those quantities identify the study regimens; they are not instructions for personal use. The primary comparison occurred at 32 weeks, with a further 20-week treatment extension. Dapagliflozin was an active comparator, so the study asked how two treatments compared rather than how safiglipron compared with an inactive pill.
The ClinicalTrials.gov record, NCT06589765, independently identifies OUTSTAND-2 under the development name HRS-7535. It records 810 participants, a randomized double-blind phase 3 design and a 32-week HbA1c primary endpoint. The registry corroborates the study identity and planned questions; the published paper supplies the numerical results discussed here.
What the glucose finding means
All three safiglipron doses met the prespecified non-inferiority criterion for HbA1c reduction. That means the analysis ruled out worse glucose lowering beyond the trial’s allowed margin of 0.4 percentage points. It does not mean the treatments were identical.
Under that primary analysis, HbA1c fell by an estimated 1.58, 1.50 and 1.68 percentage points with the three safiglipron doses, respectively, versus 1.28 points with dapagliflozin. These were estimates under a hypothetical continued-treatment scenario that addressed permanent discontinuation and prolonged additional glucose-lowering medication.
For superiority, the investigators used a different question: the treatment-policy effect, including outcomes regardless of those events. At 90 mg, safiglipron reduced HbA1c by 0.25 percentage points more than dapagliflozin; the 95% confidence interval ran from 0.05 to 0.45 points more reduction. The 60-mg superiority test did not meet the prespecified threshold, and the ordered testing sequence stopped. Subsequent endpoints therefore lack the same confirmatory statistical status.
Our guide to trial estimands explains why the question behind a number matters. The primary non-inferiority estimates and the treatment-policy superiority result should not be blended into a single claim of benefit across all doses.
Weight loss remained modest
At week 32, the paper’s supportive efficacy analysis estimated average weight changes of −2.35%, −3.55% and −4.17% with safiglipron 30, 60 and 90 mg, compared with −3.60% with dapagliflozin. This analysis assumed continued treatment and adequate drug exposure, with hypothetical handling of specified intervening events.
The estimated difference for 90 mg versus dapagliflozin was −0.56 percentage points, with a 95% confidence interval from −1.37 to +0.24. That interval includes no difference. Together with the stopped testing sequence, it leaves no established weight-loss superiority at the highest dose. A larger numerical average alone cannot settle the comparison.
These findings also cannot rank safiglipron against injectable obesity treatments or other oral GLP-1 medicines studied in different populations. Diabetes status, duration, comparator and analysis assumptions affect the meaning of each result. Orforglipron, another oral small-molecule GLP-1 receptor agonist discussed in the paper, is a separate molecule with its own evidence; its results cannot be transferred to safiglipron.
Tolerability belongs beside efficacy
Over the 52-week treatment period, gastrointestinal adverse events were more frequent with safiglipron, and most were mild or moderate. Adverse events led to permanent treatment discontinuation in 4.0%, 4.0% and 3.9% of the three safiglipron groups versus 1.5% with dapagliflozin. These safety figures cover a longer window than the primary 32-week efficacy comparison.
Calling symptoms mostly mild does not make their practical burden disappear. The tolerability and discontinuation guide explains why symptom frequency, severity and stopping treatment answer different questions. A hypothetical continued-treatment estimate should always be read alongside evidence about whether people remained on therapy.
What remains unresolved
The study provides controlled comparative diabetes evidence, but it does not establish prevention of heart attacks, kidney failure or other long-term clinical events. All sites were in China, follow-up was limited, and the population differs from many obesity-trial cohorts. Several authors were company employees, and some disclosed pending safiglipron patents.
A completed registry entry and a peer-reviewed publication also do not establish marketing authorization. Neither source inspected here supplies a regulator’s approval decision or a weight-management label. Our report on Foundayo’s documented approval illustrates why a regulatory decision is a separate milestone.
The next useful evidence would clarify safiglipron’s weight-management results in the relevant population, longer-term safety and any actual regulator decision. For now, OUTSTAND-2 supports a specific glucose-control finding and a cautious reading of modest weight changes, rather than a new weight-loss ranking.



