Most of what the public knows about semaglutide 2.4 mg comes from trials run largely in North America and Europe, in cohorts that were heavier than this one and mostly women. STEP 6 was run in Japan and South Korea, and almost every structural fact about it is different: the participants were lighter, most of them were men, a second semaglutide dose was tested alongside the licensed one, and a subset had their abdominal fat measured by computed tomography rather than inferred from a tape measure. The headline percentage landed close to the one from the pivotal obesity trial. Almost nothing underneath it did. The other east Asian trial in the program, which ran a shorter schedule and a busier placebo arm, is covered in the STEP 7 article.
What was run
STEP 6 was a randomized, double-blind, double-dummy, placebo-controlled phase 3a superiority trial at 28 outpatient clinics in Japan and South Korea. Eligible adults — 18 or older in South Korea, 20 or older in Japan — needed a body-mass index of at least 27.0 with two or more weight-related coexisting conditions, or 35.0 or above with one, where one condition had to be hypertension, dyslipidemia, or, in Japan only, type 2 diabetes. Everyone had to report at least one unsuccessful attempt to lose weight by diet.[1]
Between 21 January and 4 June 2019, 437 were screened and 401 randomized in a 4:1:2:1 ratio to once-weekly subcutaneous semaglutide 2.4 mg (n = 199), matching placebo, semaglutide 1.7 mg (n = 101) or matching placebo, all with lifestyle recommendations for 68 weeks. The two placebo groups were pooled for analysis, giving 101.[1] The double-dummy design exists because the two semaglutide doses use different pen presentations; pooling the placebos is what the 4:1:2:1 ratio was built to permit.
Two endpoints were preregistered as coprimary: the percentage change in body weight from baseline to week 68, and the proportion of participants achieving a reduction of at least 5% of baseline weight at week 68. Change in abdominal visceral fat area, measured by computed tomography in a subset, was a supportive secondary endpoint rather than a primary one.[1] The registry entry for NCT03811574 lists the same two primaries in the same order.[2]
The cohort is the finding
The trial’s patient-reported-outcomes report states the baseline characteristics for all 401 participants: mean body weight 87.5 kg, mean body-mass index 31.9, mean waist circumference 103.2 cm, mean age 51.[3] A body-mass index of 31.9 sits barely over the line that defines obesity at all, and the absolute numbers show what that does to the result: weight fell 11.3 kg on 2.4 mg against 1.7 on placebo here,[2] against 15.3 kg and 2.6 in the pivotal trial of the same drug at the same dose over the same 68 weeks.[4] Nearly the same percentage, four fewer kilograms, because there was less of a person to take it from.
That is not an accident of recruitment. It follows from the entry criterion, which starts at a body-mass index of 27.0 with two coexisting conditions, and from the clinical reality the trial was designed around: metabolic disease appears at lower body-mass index in east Asian populations than the thresholds written for European and North American ones. The trial is testing the drug where the disease actually sits in that population, not where a global cutoff says it should.
The other cohort fact is sex. The registry records 253 men and 148 women among the 401 — 63% men.[2] The pivotal trial ran the other way: 1,453 of its 1,961 participants were women, 74%.[5] Anyone reading STEP 6 as a confirmation of that result should notice that it is also the closest thing the program has to a male-majority cohort, and that nothing in the trial was powered to test whether sex changed the answer.
Two accounts of the same endpoint
The abstract and the posted results describe the same primary endpoint with different summaries, and both belong to the trial. The abstract gives model-estimated means with standard errors: weight change of −13.2% (SEM 0.5) at 2.4 mg and −9.6% (SEM 0.8) at 1.7 mg against −2.1% (SEM 0.8) on placebo, with estimated treatment differences of −11.1 percentage points (95% CI, −12.9 to −9.2) and −7.5 (95% CI, −9.6 to −5.4), both at P < 0.0001.[1]
The registry posts observed means with standard deviations: −13.4% (SD 8.6) and −9.9% (SD 7.8) against −1.9% (SD 5.9), with treatment-policy estimand differences of −11.06 (95% CI, −12.88 to −9.24) and −7.52 (95% CI, −9.62 to −5.43).[2] The differences agree to two decimal places because they are the same estimand; the arm means differ because one set is modeled and the other observed. Quote one set or the other, never a mean from each.
On the second coprimary endpoint, 160 of 193 evaluable participants (83%) on 2.4 mg and 71 of 98 (72%) on 1.7 mg reached a 5% reduction, against 21 of 100 (21%) on placebo — odds ratios of 21.7 (95% CI, 11.3 to 41.9) and 11.1 (95% CI, 5.5 to 22.2).[1] Those intervals are enormous because the placebo arm has 100 people in it and 21 events; an odds ratio built on that denominator is a direction, not a precise quantity.
The dose below the licensed one
STEP 6 is the only trial in the obesity program that carried a 1.7 mg arm alongside 2.4 mg, and the gradient is real: 9.9% against 13.4% observed, 7.52 against 11.06 points of treatment difference, 72% against 83% reaching 5%, and 41 of 101 against 117 of 199 reaching 10%.[2] More drug did more.
What the trial does not contain is a preregistered test between the two doses. Both were compared against pooled placebo; neither was compared against the other, and no interval is posted for the gap between them. A reader can see that 2.4 mg outperformed 1.7 mg in this cohort. A reader cannot get a confidence interval for that difference out of STEP 6, and should not treat the two arms as a randomized head-to-head. How the ladder is meant to be climbed, and what each rung is licensed for, is in the dose article.
The endpoint that needed a CT scanner
A subset of participants had abdominal visceral fat area measured by computed tomography at baseline and again at week 68. Waist circumference, which almost every weight trial records, is a tape measure around the outside of a torso; this endpoint is a cross-section of what is inside it, and it is preregistered here as a supportive secondary.[1]
It fell by 40.0% (SEM 2.6) on 2.4 mg and 22.2% (SEM 3.7) on 1.7 mg against 6.9% (SEM 3.8) on placebo, an estimated treatment difference at the higher dose of −33.2% (95% CI, −42.1 to −24.2).[1] The registry’s observed means are −41.0% (SD 23.3), −22.3% (SD 31.3) and −7.1% (SD 19.5), which in absolute terms is 67.4 cm² of visceral fat against 13.8.[2]
Visceral fat fell roughly three times as fast, in percentage terms, as total body weight. It is also a secondary endpoint in a subset, with an interval spanning eighteen percentage points — informative, not decisive. What the class does to lean mass and fat mass more generally is in the body composition article. Waist circumference fell 11.2 cm against 1.8, and 10.2 cm against 1.9 when measured by the Japan Society for the Study of Obesity method, which the trial reported separately because the two rulers do not measure at the same place on the torso.[2]
Almost nobody left
The dropout ledger is the most striking table in the trial. Of 401 randomized, 395 completed: 195 of 199 on 2.4 mg, 99 of 101 on 1.7 mg, and 101 of 101 on placebo. Every one of the six non-completions is filed under withdrawal by participant. Nobody was lost to follow-up, and no site closed.[2]
That is 1.5% attrition. The pivotal trial, over the same 68 weeks, lost 66 of 1,306 and 46 of 655 — 5.1% and 7.0%, including 67 participants lost to follow-up and two deaths.[5] STEP 6 lost nobody to follow-up and nobody died. When almost everyone stays, retention stops being a source of doubt: the treatment-policy estimand and the on-treatment estimand converge, and the imputation assumptions that do heavy lifting in larger trials barely move anything here. The estimates from STEP 6 are, in that narrow sense, cleaner than the estimates from trials ten times its size.
Side effects, and which one came first
Adverse events were reported by 171 of 199 (86%) on 2.4 mg, 82 of 100 (82%) on 1.7 mg and 80 of 101 (79%) on placebo. Gastrointestinal disorders, mostly mild to moderate, occurred in 59%, 64% and 30%. Adverse events leading to discontinuation of trial product occurred in five of 199 (3%), three of 100 (3%) and one of 101 (1%).[1] Serious adverse events were 10 of 199 (5.0%) against seven of 101 (6.9%), and no participant died.[2]
The ranking inside the symptom table is not the usual one. In Western trials nausea leads. Here the most frequent gastrointestinal event at 2.4 mg was constipation, 52 of 199 (26.1%) against three of 101 on placebo, ahead of nausea at 35 of 199 (17.6%) against four, diarrhea at 32 (16.1%) against six, and vomiting at 17 (8.5%) against two.[2] Nasopharyngitis was reported by 53 of 199 and 18 of 101 — a background infection rate, not a drug effect, and a reminder of how much of an adverse-event table is simply life over 68 weeks. What these symptoms feel like and how long they last is in the side effects article.
Where STEP 6 and STEP 7 meet, and where they do not
These two trials overlap more than their separate numbering suggests, and less than a shared label would imply. Both enrolled adults with overweight or obesity, with or without type 2 diabetes, described in their own titles as east Asian. Their recruiting countries intersect at South Korea. Both tested the same molecule at the same licensed dose against placebo.
Three things separate them, and all three make the headline percentages non-comparable. STEP 6 ran 68 weeks in Japan and South Korea with every participant recorded as Asian; STEP 7 ran 44 weeks across China, Hong Kong, Brazil and South Korea, where 340 of 375 were recorded as Asian and 31 as White. STEP 6 randomized four ways with two doses; STEP 7 randomized two ways with one. And the cohorts are a decade apart in age — 51 against 41.[2] A shorter trial in a younger cohort produced a smaller treatment difference, and nothing in either design tells you how much of that gap is the schedule and how much is the population.
What STEP 6 did not establish
It ran no active comparator. There is no arm of liraglutide, no arm of tirzepatide, no lifestyle-only arm beyond the counseling both groups received. Nothing in the trial ranks this drug against another.
It ran no cardiovascular or mortality endpoint, and 401 participants over 68 weeks could not have powered one. It ran no off-treatment period, so what happens to 13.4% after the injections stop is not in this trial at all. It ran no formal comparison against a Western cohort; the contrast with the pivotal trial above is a cross-trial observation between different protocols, different years and different countries, and no interval exists for it.
And the imaging endpoint, the most distinctive thing here, was a supportive secondary in a subset. It was not part of the confirmatory testing hierarchy, so it establishes a direction and a rough magnitude rather than a tested claim. A trial of 401 people can tell you that visceral fat moved a lot. It cannot tell you what that means for anyone’s risk of anything.
What was in the injection
Every figure above belongs to branded semaglutide supplied free by the sponsor, titrated under protocol, injected weekly for 68 weeks alongside lifestyle counseling, in a cohort selected by comorbidity as well as by weight and followed at 28 clinics with essentially no attrition. Sellers listed on the compounded semaglutide board dispense compounded preparations, which are not FDA-approved and are not reviewed by the FDA for safety, efficacy or quality before they reach a patient.
The number worth carrying out of STEP 6 is not 13.4%. It is 11.06 points, the part attributable to the drug, with an interval running from 12.88 down to 9.24 — measured in people averaging 87.5 kg, most of them men, almost none of whom quit.