Inside the STEP-5 Methodology: What Most Summaries Skip

At a glance

Detail	Value
N	304 (semaglutide 203, placebo 101), as reported
Intervention	Semaglutide 2.4 mg subcutaneous, once weekly
Comparator	Matched placebo injection + lifestyle counseling
Duration	104 weeks (2 years)
Primary endpoint	Percent change in body weight from baseline to week 104
Key result (treatment-policy estimand)	Roughly −15% with semaglutide vs a much smaller loss with placebo; exact percentage-point difference should be confirmed against the primary paper

The direct answer

STEP-5 is a phase 3 randomized controlled trial testing whether semaglutide 2.4 mg's weight-loss effect, established over roughly one year in earlier STEP trials, persists over two years. The trial's core methodological feature is its estimand framework: the headline result is a treatment-policy estimate, meaning it reflects the average outcome across everyone randomized, including people who stopped the drug early. That is a deliberately conservative and more clinically honest number than a per-protocol or on-treatment-only estimate would produce, because it captures real-world discontinuation as part of the effect rather than excluding it.

Why the number is less interesting than how it was built

A large weight-loss percentage over two years is easy to repeat and easy to misuse. What determines whether that number generalizes to a specific patient is who was enrolled, how dropouts were handled statistically, and which analytic population the percentage describes. Each of those choices can shift an effect estimate by several percentage points, and most summaries of STEP-5 skip past all three.

Randomization and blinding

STEP-5 randomized participants 2:1 to semaglutide or placebo, a ratio commonly used in obesity trials so the sponsor collects more granular safety and efficacy data on the active arm without dramatically increasing total enrollment. Both arms received identical-appearing pre-filled injection pens, and neither participants nor investigators knew the assignment.

Randomization is generally described as having used an interactive web-response system with stratification by baseline BMI category. Stratifying by BMI matters because response to weight-loss pharmacotherapy can vary by starting weight; without it, chance imbalances between arms could confound the primary endpoint. The exact BMI-category cutoffs used for stratification should be verified against the primary paper before being repeated as precise figures.

A detail that rarely gets discussed: both arms received identical lifestyle counseling, structured around a calorie deficit and a physical activity target consistent with standard obesity-trial protocols. That means the placebo arm's own weight loss reflects the effect of sustained counseling plus the ritual of injection, not an inert comparison. The gap between the two arms is what isolates the drug's pharmacological contribution from behavioral and expectation effects, not the semaglutide arm's raw number in isolation.

Who was actually studied

Eligibility generally required adults with a BMI in the obesity range, or a lower BMI threshold plus at least one weight-related comorbidity such as hypertension, dyslipidemia, obstructive sleep apnea, or cardiovascular disease, along with a history of at least one unsuccessful dietary weight-loss attempt.

Type 2 diabetes was excluded. That matters because a separate STEP trial (STEP-2) studied semaglutide specifically in people with type 2 diabetes and obesity, whose glucose-mediated appetite and satiety pathways differ. STEP-5's result therefore describes the population most often seeking weight-loss pharmacotherapy in primary care, not people with diabetes.

The population was reported to be predominantly female and overwhelmingly white. That narrow demographic composition is a genuine limitation. The FDA label for Wegovy does not restrict the indication by race or ethnicity, which means clinicians using this drug in more diverse populations are extrapolating from a study population that does not reflect the group they are treating.

Dose escalation most summaries skip

Semaglutide 2.4 mg is never started at 2.4 mg. Per the FDA-approved prescribing information, the dose is titrated upward over roughly 16 weeks through several intermediate doses before reaching the 2.4 mg maintenance dose, specifically to reduce dose-dependent gastrointestinal side effects such as nausea, vomiting, and diarrhea, which are most common during escalation.

This has a direct interpretive consequence: a meaningful fraction of the 104-week trial period was spent at sub-therapeutic doses, so the weight-loss curve during escalation is expected to be shallower than during the maintenance phase. In real-world use, patients who cannot tolerate escalation may never reach the target dose at all, an attrition pathway that a fixed-duration trial captures only partially, since trial participants are supported through titration in ways routine clinical follow-up may not replicate.

The estimand framework: the part almost every summary skips

This is the single most consequential methodological feature of STEP-5, and it is rarely discussed outside the primary paper and its statistical appendix.

Long-term obesity trials generally report results under more than one estimand, or analytic definition of "the treatment effect":

Treatment-policy estimand. This counts all randomized participants at the final time point regardless of whether they discontinued the drug, switched to rescue therapy, or stopped injecting altogether. Missing data is typically handled through multiple imputation. In plain terms: if a participant stopped semaglutide partway through and regained weight by the end of the trial, that regain is counted against the drug rather than excluded. This is the more conservative estimand and the one that produces the headline result most commonly cited for STEP-5.

Trial-product (on-treatment) estimand. This estimates the effect only in participants who remained on treatment for the full duration, censoring data after discontinuation. This estimand generally produces a larger apparent effect, because it removes people who stopped the drug (often for tolerability or access reasons) from the denominator.

The gap between these two numbers is not noise. It is a direct, quantifiable estimate of the "adherence penalty" between an idealized trial population and the average patient a clinician will actually prescribe to. Because the exact percentage-point difference reported for STEP-5 under each estimand needs confirmation against the primary paper, we describe the direction and mechanism of the gap here rather than repeating a specific number without a verified source.

Why this distinction should change how a clinician counsels a patient

For counseling purposes, the treatment-policy estimate is the more honest one to lead with, since it already accounts for the reality that some patients discontinue due to side effects, cost, or access disruption.

Statistical significance and multiplicity control

Trials like STEP-5 generally pre-specify a hierarchical (fixed-sequence) testing procedure across co-primary and secondary endpoints to control the overall false-positive rate. In a hierarchical design, if an earlier endpoint in the chain fails to reach significance, later endpoints are reported only as descriptive or nominal findings, not as confirmed statistical results, regardless of their own p-values. Whether every endpoint in STEP-5's hierarchy reached formal significance, and the exact alpha-spending structure used, should be verified against the primary paper rather than assumed from secondary summaries.

Larger-than-anticipated treatment effects in an adequately powered trial also produce tighter confidence intervals than a marginally powered trial would, which can create an impression of more precision than a smaller replication trial might show. This is a general statistical property of overpowered trials, not a criticism specific to STEP-5, but it is worth keeping in mind when a single trial's confidence interval is quoted as though it were the final word on the drug's effect size.

Discontinuation: the number that matters most for real-world planning

STEP-5 is generally reported to have had meaningfully higher drug discontinuation in the placebo arm than the active arm, which is expected since perceived lack of benefit is a common reason participants drop out of a placebo arm. The semaglutide-arm discontinuation rate, commonly cited as roughly one in eight to one in ten participants over two years, is the more clinically important number, because it estimates how many patients on the actual drug could not or chose not to continue it even in a supported trial setting.

The published trial literature does not appear to separate cost- or insurance-related discontinuation from personal-preference discontinuation. That distinction matters enormously outside a trial, where cost and coverage disruption are frequently the dominant reason patients stop GLP-1 therapy in routine practice, a pattern documented in analyses of real-world prescription and claims data rather than in the trial itself. Readers should treat the trial's discontinuation rate as a floor, not a ceiling, on real-world discontinuation.

The weight-loss trajectory is not linear

Weight loss with semaglutide in long-duration trials like STEP-5 characteristically follows a curve: relatively rapid loss during the first year, gradual deceleration afterward, and approximate stabilization by the end of the second year, rather than continued proportional loss. Placebo-arm weight typically shows modest early loss followed by partial regain over time.

This has a direct counseling implication: patients who expect continuous, linear weight loss indefinitely are likely to be disappointed around the one-year mark and may mistakenly conclude the drug has "stopped working." Clinical guidance on obesity pharmacotherapy generally frames stabilization, not indefinite continued loss, as the expected long-term outcome for GLP-1 receptor agonist therapy, and clinicians should set that expectation before, not after, a patient reaches the plateau.

Limitations, including one the published paper does not emphasize

Commonly acknowledged limitations of STEP-5 include a study population concentrated in North America and Europe, limited racial and ethnic diversity, and a protocol that did not include a post-treatment withdrawal phase, meaning STEP-5 itself cannot answer what happens to weight after semaglutide is stopped (a separate STEP trial, STEP-4, used a withdrawal design for that specific question).

A limitation worth naming even though it is rarely discussed: the 2:1 randomization ratio means the placebo comparator arm is roughly half the size of the active arm. Smaller comparator arms produce wider confidence intervals around the placebo estimate, meaning the placebo group's trajectory is generally estimated with less precision than the drug group's, even though the overall trial is well powered for the primary comparison.

A second limitation applies to essentially the entire obesity pharmacotherapy trial literature, not just STEP-5: the intensive counseling infrastructure of a clinical trial (regular visits, dietitian support, structured accountability) is not representative of routine primary care, where follow-up may occur every few months rather than continuously. Both arms in STEP-5 likely lost more weight from counseling alone than a typical patient in routine care would, which affects how the placebo comparison, and by extension the drug-attributable effect, should be interpreted outside a trial setting.

Where STEP-5 fits in the broader evidence

STEP-5 belongs to a family of phase 3 trials of semaglutide 2.4 mg for weight management, distinguished mainly by its two-year duration compared to the roughly 68-week duration of the first trial in the program. Its specific contribution to the evidence base is durability: it is the primary direct evidence that semaglutide's weight-loss effect is sustained beyond one year in a population without type 2 diabetes, rather than an assumption extrapolated from shorter trials. The FDA label for Wegovy draws on data across the full trial program, but STEP-5 is the trial doing the specific work of supporting a durability claim.

Evidence-boundary statement

Established: Semaglutide 2.4 mg produces substantially greater weight loss than placebo over two years when added to structured lifestyle counseling, in adults with obesity or overweight plus a weight-related comorbidity, without type 2 diabetes. The trial used a treatment-policy estimand as its primary analysis, which is a more conservative and more clinically realistic approach than an on-treatment-only analysis.

Plausible but requiring primary-source confirmation: The exact numeric size of the treatment-policy versus trial-product gap, the precise discontinuation percentages by arm and by reason, and the specific hierarchical testing outcome for every secondary endpoint. These figures are commonly repeated in secondary summaries of STEP-5 but should be checked against the original published trial report before being used in any clinical or patient-facing document.

Not established by STEP-5 alone: What happens to weight after semaglutide is stopped (addressed by a separate withdrawal-design trial), how the drug performs in more racially and ethnically diverse populations, and how closely trial-level counseling intensity and adherence support translate into typical primary-care practice.

A framework for reading (and re-reading) this trial before using it clinically

Use this sequence when a patient, a colleague, or a marketing summary cites the STEP-5 "15% weight loss" figure, to decide what claim is actually being made and whether it applies to the person in front of you.

1. Which estimand is being quoted? If a source states a single weight-loss percentage without saying whether it is treatment-policy or on-treatment, treat the number as underspecified. Ask, or assume, it is the treatment-policy (more conservative, more realistic) figure unless stated otherwise.
2. Does the patient resemble the trial population? Type 2 diabetes, prior bariatric surgery, or recent GLP-1 use are exclusion criteria; a patient with any of these is outside the population STEP-5 directly studied. Race and ethnicity outside the trial's predominantly white sample also fall outside directly observed data, even though the FDA label does not restrict use by these factors.
3. Is the patient being counseled about the escalation phase or the maintenance phase? Early weeks on a low, sub-therapeutic dose should not be compared to week-104 maintenance-dose results; expecting maintenance-phase loss during titration sets an unrealistic bar.
4. Is the plateau being explained in advance? Set the expectation that weight loss decelerates and stabilizes rather than continuing linearly, before the patient reaches that point around one year in, not after they raise concern that the drug "stopped working."
5. Is discontinuation risk part of the conversation? Roughly one in eight to one in ten trial participants on the drug stopped it before the two-year mark, mostly for gastrointestinal tolerability; real-world discontinuation (often cost- or access-driven) is plausibly higher and was not fully separated from other reasons in the trial.
6. Would a specific number in this discussion change a clinical decision? If yes, verify it against the primary published trial and its statistical appendix rather than a secondary summary, including this one, before using it to set dosing expectations or make coverage-related arguments.

Common questions

Frequently asked questions

›What was the primary endpoint in STEP-5?

Percent change in body weight from baseline to week 104, analyzed under a treatment-policy estimand that accounts for participants who discontinued treatment before the end of the trial.

›What is a treatment-policy estimand and why does it matter?

It is an analysis that includes every randomized participant's outcome regardless of whether they stayed on the assigned drug, rather than only analyzing people who completed treatment as planned. It produces a more conservative and arguably more clinically realistic effect estimate than an on-treatment-only analysis, because it reflects the reality that some patients stop treatment.

›Were patients with type 2 diabetes included in STEP-5?

No. Type 2 diabetes was an exclusion criterion in STEP-5. A separate trial in the same program, STEP-2, studied semaglutide 2.4 mg specifically in people with type 2 diabetes and obesity.

›How long is the dose-escalation period for semaglutide 2.4 mg?

Per the FDA-approved prescribing information for Wegovy, the dose is titrated upward over roughly 16 weeks through several intermediate doses before the 2.4 mg maintenance dose is reached, primarily to reduce gastrointestinal side effects.

›Does weight loss continue indefinitely on semaglutide?

No. Long-duration trial data generally show a characteristic pattern of relatively rapid early loss followed by deceleration and approximate stabilization rather than continued proportional loss. Clinicians should set this expectation with patients before they reach the plateau.

›Did STEP-5 study what happens after stopping semaglutide?

No. STEP-5 did not include a post-treatment withdrawal phase. A separate trial in the program, STEP-4, used a withdrawal design specifically to study weight regain after stopping the drug.

›How diverse was the STEP-5 study population?

Limited. The population was reported to be predominantly white, which is a genuine limitation for generalizing the trial's precise results to other racial and ethnic groups, even though the drug's FDA label does not restrict use by race or ethnicity.

References

• Wegovy (semaglutide) prescribing information. U.S. Food and Drug Administration. Accessed via FDA label archive. FDA label
• STEP-5 trial (Garvey et al., two-year effects of semaglutide in adults with overweight or obesity) and related STEP-program publications are referenced throughout this article by description rather than by specific PubMed identifier. The PMID links carried in earlier drafts of this page could not be independently verified against the claims they were attached to and have been removed. Editors should confirm the primary published trial report and its statistical appendix before this page is finalized, particularly for the exact numeric values quoted for the treatment-policy versus trial-product gap and arm-specific discontinuation rates.