Inside the SUSTAIN-6 Methodology: What Most Summaries Skip

At a glance

Field	Detail
Trial Name	SUSTAIN-6
N	3,297
Intervention	Semaglutide 0.5 mg or 1.0 mg SC once weekly
Comparator	Volume-matched placebo (0.5 mg or 1.0 mg)
Duration	104 weeks (2 years)
Primary Endpoint	3-point MACE: CV death, non-fatal MI, non-fatal stroke
Key Result	HR 0.74 (95% CI 0.58, 0.95); 26% relative risk reduction
Primary Source	Marso et al., NEJM 2016

Why Regulatory Context Determines Everything About This Design

SUSTAIN-6 was not conceived as a superiority trial. It was conceived to satisfy the 2008 FDA guidance requiring cardiovascular outcomes trials (CVOTs) for new antidiabetic drugs, a requirement born from post-hoc concerns about rosiglitazone's cardiac signal. The agency required that the upper bound of the 95% confidence interval for the hazard ratio stay below 1.8 during development and below 1.3 at approval. That is the statistical bar SUSTAIN-6 was engineered to clear.

This regulatory framing matters enormously. The primary analysis was a non-inferiority test. Superiority was a secondary, pre-specified analysis. When Marso et al. reported in the NEJM that semaglutide achieved superiority (HR 0.74, p = 0.02 for superiority), that result was valid and pre-specified, but it sat one step below the primary non-inferiority conclusion in the statistical hierarchy. Many clinical summaries invert that hierarchy and present the superiority result as the headline without that caveat.

Randomization and Allocation Concealment

Patients were randomized 1:1 to semaglutide or placebo using an interactive voice or web response system, stratified by two factors: current insulin use and cardiovascular risk category (established CV disease vs. CKD stage 3 or above as the sole high-risk qualifier). Stratified randomization is standard for CVOTs because baseline insulin use substantially alters background glucose-lowering intensity, and unbalanced distribution across arms could confound the glucose-lowering contribution to any observed outcome difference.

The allocation was concealed at the site level. Investigators could not access the randomization sequence before enrollment, which is appropriate and reduces selection bias. What the trial could not conceal, however, was the glycemic and weight signal. At 104 weeks, semaglutide 1.0 mg reduced HbA1c by roughly 1.1 percentage points more than placebo, and body weight fell by about 4.5 kg more. Any site clinician familiar with GLP-1 receptor agonist pharmacology could reasonably infer treatment assignment from metabolic markers. That inference risk is an acknowledged limitation of all GLP-1 CVOTs.

Blinding: What "Double-Blind" Actually Meant Here

The trial used volume-matched placebo injections, meaning the injection volume was identical between semaglutide and placebo syringes across both dose levels. Patients, site staff, and outcome adjudicators were blinded to assignment. The adjudication committee reviewed outcomes without knowledge of treatment allocation, which protects the integrity of the primary endpoint.

The practical limitation is the GI side-effect profile. Nausea and vomiting, which Marso et al. documented at rates of 20% and 5% for semaglutide 1.0 mg vs. 5% and 2% for placebo, are a clinical fingerprint of GLP-1 receptor agonism. A patient or nurse who observed these symptoms could reasonably suspect active drug. This form of unblinding does not typically invalidate a hard-endpoint trial where outcomes are adjudicated independently, but it should temper confidence in secondary endpoints like patient-reported outcomes or symptom scales.

Inclusion and Exclusion Criteria: A High-Risk Population by Design

Enrollment required type 2 diabetes plus one of two risk categories. The first was established cardiovascular disease (prior MI, stroke, or peripheral arterial disease). The second was CKD stage 3 or above in patients aged 60 or older without prior CV events. The minimum age was 50 for the CV disease group and 60 for the CKD group.

Several practical consequences follow from these criteria. First, the trial population was older and sicker than typical early-stage T2D cohorts. Mean baseline HbA1c was approximately 8.7%, mean duration of diabetes was 14 years, and roughly 83% had established cardiovascular disease at baseline. Second, background therapy was highly heterogeneous: metformin, sulfonylureas, insulin, and SGLT2 inhibitors were all permitted and present in meaningful proportions. That heterogeneity makes isolating semaglutide's independent contribution difficult.

Third, and critically for generalization, patients with a recent acute coronary syndrome within 90 days or a planned coronary revascularization were excluded. The population that generates the most MACE events in clinical practice, those recently destabilized, was systematically absent. This is a design feature shared across all GLP-1 CVOTs and limits extrapolation to patients in the acute or peri-procedural phase.

The Primary Endpoint: How MACE Was Defined and Adjudicated

The primary endpoint was time to first occurrence of a 3-point MACE composite: cardiovascular death, non-fatal myocardial infarction, or non-fatal stroke. This is now the near-universal CVOT composite following FDA guidance and allows cross-trial comparisons with LEADER (liraglutide), EMPA-REG OUTCOME (empagliflozin), and DECLARE-TIMI 58 (dapagliflozin). A blinded, independent clinical events committee adjudicated all suspected events using pre-specified criteria, and those criteria were set before any unblinding.

The choice to use a time-to-first-event analysis rather than total event count is worth noting. Because SUSTAIN-6 ran only 104 weeks, patients who had a second or third MACE were essentially invisible to the primary analysis after their first event. A recurrent-events analysis would capture more of the total disease burden, but such analyses were not standard in CVOTs at the time of design. The 2020 ACC/AHA guidelines on antidiabetic therapies have since recommended interpretation of CVOTs in the context of total event burden, not just first events.

The Comparator Choice: What Placebo Really Means in Practice

The comparator was placebo, not an active glucose-lowering agent. This is ethically permissible and scientifically cleaner for isolating the drug's effect, but it introduces glucose-control asymmetry over a two-year period. Placebo patients required more frequent background-therapy intensification to maintain HbA1c within acceptable clinical bounds, which the protocol permitted. If those intensifications (often insulin additions) carried their own cardiovascular risks or benefits, the net treatment effect could be confounded.

Investigators tracked background-therapy changes and reported them as safety data. Insulin use increased more in the placebo group, and sulfonylurea use was similar across arms. This suggests most of the HbA1c gap reflects genuine semaglutide pharmacodynamics rather than background-therapy suppression in the active arm, but the asymmetry remains a recognized confound.

Statistical Architecture: Non-Inferiority First, Superiority Second

The trial used a hierarchical testing procedure, which controls the familywise error rate. Superiority was only formally tested after non-inferiority was established, protecting against inflated type I error. The power calculation assumed a 5% annual MACE rate in the placebo group and targeted at least 122 primary events for the non-inferiority analysis. The trial accumulated 254 primary events, giving it substantially more power than originally required, though this also reflects an unexpectedly high event rate in a relatively short trial.

One underappreciated feature is that semaglutide was tested at two doses (0.5 mg and 1.0 mg), and both doses were pooled for the primary analysis. The 1.0 mg arm showed a more pronounced point estimate for MACE reduction (HR approximately 0.68) compared with the 0.5 mg arm (HR approximately 0.82), though the interaction test was not statistically significant. This pooling is reasonable given the regulatory purpose, but it means the approved 1.0 mg label indication rests partly on evidence from a lower dose that may not perform identically.

What the Trial Did Not Measure: Gaps That Matter Clinically

SUSTAIN-6 ran for two years. The biological mechanisms proposed for GLP-1 cardiovascular benefit, including anti-atherosclerotic plaque effects and endothelial function improvements, are generally thought to operate over years to decades. The LEADER trial with liraglutide ran 3.8 years and showed a similar but slightly attenuated hazard ratio of 0.87, suggesting longer exposure may modulate effect size. Two years is enough to see a signal, particularly for stroke, but likely insufficient to observe the full trajectory of benefit.

Heart failure hospitalization was a secondary endpoint in SUSTAIN-6, and it showed a non-significant trend favoring semaglutide (HR 1.11, 95% CI 0.77, 1.61). This is the opposite direction from what SGLT2 inhibitors show and from what subsequent semaglutide data in the SELECT trial suggested. The heart failure signal in SUSTAIN-6 should not be over-interpreted in either direction given the confidence interval width, but it does underscore that GLP-1 receptor agonists and SGLT2 inhibitors appear to address different cardiovascular phenotypes.

Renal outcomes were included as secondary endpoints, and semaglutide showed a significant reduction in new or worsening nephropathy (HR 0.64, 95% CI 0.46, 0.88). This was an early signal later explored in dedicated renal outcome trials and is now reflected in ADA Standards of Care recommendations for GLP-1 use in CKD.

Retinopathy: A Safety Finding That Deserves More Attention

One of the most clinically consequential findings from SUSTAIN-6 was an increase in diabetic retinopathy complications in the semaglutide group (HR 1.76, 95% CI 1.11, 2.78). This was a pre-specified secondary safety endpoint, and the signal was statistically significant. The proposed mechanism is rapid HbA1c reduction triggering transient worsening of retinopathy, a phenomenon previously described with insulin intensification and known as "early worsening." Patients with pre-existing proliferative retinopathy had the highest relative risk.

This finding influenced the Ozempic prescribing label, which includes a warning about retinopathy complications, and it remains a factor in ophthalmologic monitoring recommendations for patients initiating semaglutide with advanced baseline retinopathy.

Frequently asked questions

›Was SUSTAIN-6 designed to prove cardiovascular benefit?

No. The trial's primary purpose was to satisfy FDA non-inferiority requirements showing semaglutide did not increase cardiovascular risk. Superiority, meaning proof of actual benefit, was a pre-specified secondary analysis. The distinction matters for how you weight the result in clinical decision-making.

›Why was the trial only two years long?

The two-year duration reflected the regulatory requirement for CVOTs during the drug development phase rather than a scientifically optimal study length. Longer trials like LEADER (3.8 years) have since provided more durability data for GLP-1 CV effects.

›Who was included in SUSTAIN-6?

Adults with type 2 diabetes and either established cardiovascular disease (age ≥50) or CKD stage 3 or above without prior CV events (age ≥60). Roughly 83% had established CV disease, making this a high-risk secondary-prevention population.

›What does the 26% relative risk reduction actually mean in absolute terms?

The absolute risk reduction was approximately 2.3 percentage points (6.6% placebo vs. 6.0% semaglutide... wait, the actual event rates were 8.9% placebo vs. 6.6% semaglutide over two years), corresponding to a number needed to treat of roughly 43 patients over two years to prevent one MACE event.

›Why were two doses of semaglutide pooled for the primary analysis?

Both the 0.5 mg and 1.0 mg doses were pooled to increase event count and statistical power for the non-inferiority test, which was the trial's primary purpose. The dose-specific interaction test was non-significant, supporting pooling, but the 1.0 mg arm showed a numerically stronger point estimate.

›How was blinding maintained if GLP-1 side effects are recognizable?

Volume-matched placebo syringes preserved injection-level blinding. Outcome adjudication was conducted by an independent committee without access to treatment assignment. However, patients and some site staff may have inferred treatment from nausea and weight loss, which is a recognized limitation in all GLP-1 CVOTs.

›What happened with the retinopathy finding?

Semaglutide was associated with a 76% relative increase in diabetic retinopathy complications (HR 1.76). The most likely explanation is early worsening from rapid glycemic correction, a known phenomenon. This finding is now reflected in the FDA-approved prescribing label and guides ophthalmologic monitoring in clinical practice.

›Does SUSTAIN-6 support using semaglutide in heart failure?

Not strongly. The heart failure hospitalization endpoint trended in the wrong direction (HR 1.11) and was not statistically significant. SGLT2 inhibitors carry the clearer heart failure evidence. Later data, including the SELECT trial in people without diabetes, offer a broader look at semaglutide's cardiometabolic profile.

›How do SUSTAIN-6 results apply to patients without prior cardiovascular disease?

They do not directly apply. The trial enrolled almost exclusively patients with established CV disease or high-risk CKD. Extrapolating the MACE reduction to lower-risk primary-prevention populations requires caution and reliance on indirect or modeling data.

›How does SUSTAIN-6 compare to LEADER for liraglutide?

LEADER showed an HR of 0.87 (95% CI 0.78, 0.97) for MACE over 3.8 years in a similarly high-risk population. SUSTAIN-6 showed HR 0.74 over two years. Direct cross-trial comparison is complicated by different durations, patient populations, and background therapies, but both support CV benefit for the GLP-1 class in high-risk T2D.

References

1. Marso SP, Bain SC, Consoli A, et al. Semaglutide and Cardiovascular Outcomes in Patients with Type 2 Diabetes. N Engl J Med. 2016;375(19):1834-1844. https://pubmed.ncbi.nlm.nih.gov/27633186/

2. Marso SP, Daniels GH, Brown-Frandsen K, et al. Liraglutide and Cardiovascular Outcomes in Type 2 Diabetes (LEADER). N Engl J Med. 2016;375(4):311-322. https://pubmed.ncbi.nlm.nih.gov/27295427/

3. Lincoff AM, Brown-Frandsen K, Colhoun HM, et al. Semaglutide and Cardiovascular Outcomes in Obesity without Diabetes (SELECT). N Engl J Med. 2023;389(24):2221-2232. https://pubmed.ncbi.nlm.nih.gov/37952131/

4. FDA. Ozempic (semaglutide) Prescribing Information. U.S. Food and Drug Administration. https://www.accessdata.fda.gov/drugsatfda_docs/label/2017/209637lbl.pdf

5. American Diabetes Association. Standards of Medical Care in Diabetes. Diabetes Care. 2023;46(Suppl 1). https://pubmed.ncbi.nlm.nih.gov/36507631/

6. Das SR, Everett BM, Birtcher KK, et al. 2020 Expert Consensus Decision Pathway on Novel Therapies for Cardiovascular Risk Reduction in Patients With Type 2 Diabetes. J Am Coll Cardiol. 2020;76(9):1117-1145. https://www.ahajournals.org/doi/10.1161/CIR.0000000000000938