Every other trial in the semaglutide obesity program stopped at 68 weeks, which left one question open that a person paying monthly actually needs answered: does a second year of injections buy a second year of weight loss. STEP 5 is the trial built to settle it, and the answer it returned is not the one the phrase “two-year data” suggests. Read against the one-year trial, the more useful way to describe STEP 5 is as a durability experiment that happened to run twice as long.
A small trial with a long clock
Between 5 October 2018 and 1 February 2019, 304 adults were randomly assigned 1:1 to once-weekly subcutaneous semaglutide 2.4 mg (n = 152) or placebo (n = 152), both plus a behavioral intervention, for 104 weeks. Eligibility was obesity, or overweight with at least one weight-related comorbidity, without diabetes.[1]
Three hundred and four is a small trial by the standards of this program. The sample was sized to give at least 96% power for the two coprimary endpoints and only at least 43% power for the confirmatory secondary endpoints, which were tested in a predefined hierarchy.[1] A secondary result from STEP 5 is therefore a result from a trial that was not built to detect it, which matters when one is quoted as though it carried the same weight as the headline.
The cohort was 236 (77.6%) female and 283 (93.1%) white, with a mean age of 47.3 years (SD 11.0), a body-mass index of 38.5 (6.9) and a body weight of 106.0 kg (22.0). Of the 304, 282 (92.8%) completed the trial, 272 (89.5%) had a body-weight measurement at week 104, and 243 (79.9%) were still on treatment at the end. The authors named the narrow racial composition and the preponderance of women as limitations of their own trial.[1]
Both endpoints, and the number that did not change
The coprimary endpoints were the percentage change in body weight and the achievement of at least 5% weight loss at week 104, assessed among all randomized participants regardless of treatment discontinuation or rescue intervention. Mean weight change was −15.2% with semaglutide against −2.6% with placebo, an estimated treatment difference of −12.6 percentage points (95% CI, −15.3 to −9.8; P < 0.0001). A loss of at least 5% was reached by 111 of 144 (77.1%) against 44 of 128 (34.4%), an odds ratio of 5.0 (3.0 to 8.4).[1]
The interesting figure sits in the supportive secondary table. At week 52, halfway through, the same trial reported −15.6% against −3.0%, an estimated treatment difference of −12.6 percentage points (95% CI, −14.5 to −10.7).[1]
The gap between the arms at one year and at two years is the same number. What the second year changed was the confidence interval around it, which widened from a span of 3.8 points to one of 5.5 as participants left treatment and the estimate rested on fewer measurements. The paper states the shape directly: weight reduction plateaued after approximately week 60 and was maintained for the remainder of the study.[1] The general phenomenon is described in the plateau article.
What the plateau is not
The obvious explanation for a flattening curve is that the drug stops suppressing appetite. STEP 5 has a companion analysis that tests exactly that, and it does not support the explanation.
A 19-item Control of Eating Questionnaire was administered at weeks 0, 20, 52 and 104 in a subgroup. Semaglutide separated from placebo on the Craving Control and Craving for Savory domains at all three post-baseline visits, including week 104 (p < 0.01). At week 104 the scores for desire to eat salty and spicy food, cravings for dairy and starchy foods, difficulty resisting cravings and control of eating were all significantly reduced against placebo (all p < 0.05).[2]
Two domains behaved differently. Positive Mood and Craving for Sweet separated at weeks 20 and 52 but not at 104, and hunger and fullness separated only at week 20.[2] So the appetite effect is neither uniform nor gone: craving control persisted for two years while the hunger and fullness signals faded early, and the weight curve flattened anyway. Whatever ends the descent is not a simple loss of the drug's effect on wanting food.
The same shape at four years, in a far larger trial
A prespecified analysis of the SELECT cardiovascular outcomes trial followed 17,604 adults with preexisting cardiovascular disease, overweight or obesity and without diabetes. Weight loss continued over 65 weeks and was then sustained for up to four years. At 208 weeks the mean reduction was −10.2% with semaglutide against −1.5% with placebo, with waist circumference falling 7.7 cm against 1.3 cm (P < 0.0001 for all comparisons).[3]
Two things travel from that. The inflection near week 60 to 65 is not an artifact of a 304-person trial; it reproduces in a cohort fifty-eight times the size. And the absolute figure is lower, because SELECT enrolled a different population with cardiovascular disease and included participants below a body-mass index of 30. A two-year percentage is not a floor that a fourth year improves on. The longer view is collected in the long-term article.
The safety column runs against the expectation
Gastrointestinal disorders were the most frequent adverse events and were reported by 125 of 152 (82.2%) on semaglutide against 82 of 152 (53.9%) on placebo. Nausea affected 53.3% against 21.7%, vomiting 30.3% against 4.6% and constipation 30.9% against 11.2%. Those events led to permanent treatment discontinuation in six participants (3.9%) on semaglutide and one (0.7%) on placebo, and any adverse event ended treatment in nine (5.9%) against seven (4.6%).[1]
Then the reversal. Serious adverse events were reported by 12 of 152 (7.9%) in the semaglutide group and 18 of 152 (11.8%) in the placebo group.[1] One death occurred, an acute myocardial infarction in the semaglutide group, adjudicated as unrelated to the trial product.
SELECT points the same way at scale, reporting lower rates of serious adverse events with semaglutide in every body-mass index category, while also reporting higher rates of trial-product discontinuation that increased as body-mass index class decreased.[3] Frequent unpleasant events and fewer serious ones is an unusual combination, and a page that publishes only the first half of it quotes real percentages while describing the wrong risk.
What two years bought that a scale does not show
Among participants who entered with prediabetes, 59 of 74 (79.7%) on semaglutide had returned to normoglycemia at week 104, against 20 of 54 (37.0%) on placebo. No participant on semaglutide had progressed to type 2 diabetes, against two (3.7%) on placebo.[1]
Waist circumference fell 14.4 cm against 5.2 cm (estimated treatment difference −9.2; 95% CI, −12.2 to −6.2), and C-reactive protein fell 56.7% against 7.8%, a relative difference of −53.1% (95% CI, −63.2 to −40.0).[1] These are the endpoints the trial was underpowered for, so they are consistent with the weight result rather than independently established by it.
What STEP 5 does not settle
It never stopped the drug. Nobody in STEP 5 was withdrawn to placebo, so the trial says nothing about what happens after two years of treatment end — the question that the withdrawal trial was built for, and whose consequences are summarized in the stopping article. It counted no cardiovascular events, excluded diabetes, and ran a behavioral intervention of counseling every four weeks by a dietitian or similarly qualified professional, which is lighter than the program in the STEP 3 article and heavier than nothing.
It also cannot describe the individual. A mean of −15.2% in a group where 77.1% reached 5% means that roughly one participant in four with a week-104 measurement did not, after two years of injections. The distribution matters more than the mean when the decision is whether to keep paying, which is what the cost calculator is for.
What was in the syringe
Every figure belongs to branded, FDA-approved semaglutide at a labeled dose, supplied without interruption for two years at trial sites. Compounded semaglutide, sold by most services on the semaglutide board, is not FDA-approved and is not reviewed by the FDA for safety, efficacy or quality before it is dispensed — the gap described in the compounded-versus-brand article. A 104-week supply that never lapses is itself part of what STEP 5 measured, and it is the part a subscription is least able to guarantee. How each figure here was established is set out in the methodology.