Inside the SUSTAIN-7 Methodology: What Most Summaries Skip

At a glance

| Field | Detail | |---|---| | Trial name | SUSTAIN-7 (NCT02648204) | | N | 1,201 randomized | | Intervention | Semaglutide 0.5 mg or 1.0 mg SC once weekly | | Comparator | Dulaglutide 0.75 mg or 1.5 mg SC once weekly | | Background therapy | Metformin ≥1 to 500 mg/day (stable ≥30 days) | | Duration | 40 weeks | | Primary endpoint | Change in A1C from baseline to week 40 | | Key secondary endpoint | Change in body weight from baseline to week 40 | | Headline result | Semaglutide superior to dulaglutide on A1C and weight at both dose levels | | Publication | Pratley RE et al. Lancet Diabetes Endocrinol. 2018;6(4):275-286 |

Why a Head-to-Head Trial Matters Here

Most GLP-1 receptor agonist trials in the SUSTAIN program compared semaglutide to placebo or to non-GLP-1 agents. SUSTAIN-7 did something different: it pitted semaglutide directly against dulaglutide, the most prescribed once-weekly GLP-1 RA at the time of enrollment. That design choice gives clinicians something placebo-controlled data cannot: a direct efficacy comparison between two drugs a patient might actually be choosing between.

But direct comparisons carry their own methodological baggage. The dose pairings, the open-label design, the statistical estimand, the titration schedule. Each decision bakes assumptions into the result. This page unpacks them.

Randomization and Treatment Arms

Participants were randomized 1:1:1:1 into four parallel arms:

| Arm | Drug | Weekly dose | Titration | |---|---|---|---| | A | Semaglutide | 0.5 mg | 0.25 mg x 4 wk, then 0.5 mg | | B | Semaglutide | 1.0 mg | 0.25 mg x 4 wk, 0.5 mg x 4 wk, then 1.0 mg | | C | Dulaglutide | 0.75 mg | No titration (started at 0.75 mg) | | D | Dulaglutide | 1.5 mg | 0.75 mg x 4 wk, then 1.5 mg |

Randomization was stratified by country and prior diabetes treatment (metformin alone vs. metformin plus one other oral agent). The interactive web-response system assigned treatment, not site investigators. This is standard but worth noting because the open-label design meant allocation concealment was the primary safeguard against selection bias at enrollment.

A key asymmetry: semaglutide arms used a slower titration ramp (8 weeks to reach 1.0 mg) compared to dulaglutide's 4-week step to 1.5 mg. This was driven by each drug's approved labeling, not by trial design preference. The clinical consequence is that semaglutide patients spent more weeks at sub-therapeutic doses. If anything, this asymmetry worked against semaglutide in a 40-week window.

The Open-Label Problem

SUSTAIN-7 was open-label. Participants and investigators knew which drug was administered. The trial used open-label prefilled pens identical to those used in clinical practice, which is a pragmatic choice reflecting real-world prescribing. Still, open-label design introduces measurable bias risks:

Behavioral changes. Patients who know they are on a "newer" agent may adhere more carefully or change diet and exercise habits. The trial did not collect granular diet or physical activity data, so this channel of bias is unquantifiable.

Investigator reporting. Adverse event documentation and rescue-medication decisions were not blinded. Investigators who expected semaglutide to cause more nausea may have recorded GI events more diligently in those arms (or vice versa).

Objective endpoints as partial mitigation. A1C is a lab value. Body weight is measured on a scale. Neither depends on subjective symptom reporting. This limits (but does not eliminate) the damage open-label design does to the primary and key secondary endpoints. Patient-reported outcomes and adverse-event profiles are more vulnerable.

The 2020 ADA Standards of Medical Care cite SUSTAIN-7 when positioning semaglutide among GLP-1 RAs, suggesting the field accepted the open-label design as sufficient for these endpoints.

Inclusion and Exclusion Criteria: Who Was Actually Studied

The enrolled population was relatively typical for a GLP-1 RA registration trial, but the specifics matter for generalizability:

• A1C range: 7.0% to 10.5% at screening
• Age: ≥18 years
• BMI: not capped (mean baseline ~33 kg/m²)
• Background therapy: metformin ≥1 to 500 mg/day, stable for ≥30 days; one additional oral antidiabetic agent was allowed
• Excluded: insulin use, eGFR <60 mL/min/1.73 m², history of pancreatitis, personal or family history of medullary thyroid carcinoma or MEN2

The metformin-only (or metformin-plus-one) background means SUSTAIN-7 tested a relatively early-in-disease population. Patients on triple therapy or basal insulin were excluded. Clinicians extrapolating these results to patients with longer disease duration or more complex regimens should note that gap.

Mean baseline A1C was approximately 8.2%, and mean diabetes duration was about 7 years. This is a population where GLP-1 RAs are a common second- or third-line addition, making the comparison clinically relevant for a large segment of real-world patients.

Primary Endpoint Definition and the Estimand Framework

The primary endpoint was change in A1C from baseline to week 40. But the estimand, the precise causal question the trial answers, matters as much as the endpoint label.

SUSTAIN-7 used two pre-specified estimands, reported side by side in the primary publication:

Treatment policy estimand (ITT-like). This assessed all randomized patients regardless of whether they discontinued treatment or started rescue medication. Data collected after treatment discontinuation or rescue initiation were included. It answers: "What happens in a population assigned to semaglutide vs. dulaglutide, including those who stop or switch?"

Trial product estimand (on-treatment). This assessed data only while patients were on the randomized drug and not on rescue medication. It answers: "What is the drug effect while a patient actually takes it?"

Both estimands showed superiority for semaglutide. The treatment policy estimand is more conservative and more relevant to prescribing decisions (because it includes real-world discontinuation). The trial product estimand inflates effect sizes slightly because it removes dropouts, who tend to be non-responders. Regulators and guidelines appropriately lean on the treatment policy estimand.

Missing data were handled using multiple imputation under a "missing at random" assumption for the treatment policy estimand and a mixed model for repeated measures (MMRM) for the trial product estimand.

The Comparator Dose Question

This is the most common methodological criticism of SUSTAIN-7, and it deserves honest treatment.

The trial compared semaglutide 0.5 mg to dulaglutide 0.75 mg (low vs. low) and semaglutide 1.0 mg to dulaglutide 1.5 mg (high vs. high). These were the approved dose ranges for each drug at the time. But "approved doses" does not mean "equipotent doses." Dose-response curves differ between molecules with different receptor binding kinetics, half-lives, and albumin-binding profiles.

Semaglutide 1.0 mg and dulaglutide 1.5 mg were each the maximum approved dose for their respective drugs. From a clinical-practice standpoint, comparing maximum-to-maximum is the fairest framing: it asks which drug performs best when pushed to its ceiling. From a pharmacologic standpoint, there is no guarantee these maximal doses sit at the same point on their respective dose-response curves.

Since SUSTAIN-7 was published, dulaglutide gained approval for higher doses (3.0 mg and 4.5 mg) based on the AWARD-11 trial. Those higher dulaglutide doses narrow the A1C gap seen in SUSTAIN-7, though no head-to-head trial has tested semaglutide 1.0 mg against dulaglutide 4.5 mg directly.

Statistical Design and Multiplicity Control

The trial used a hierarchical (gate-keeping) testing strategy to control the family-wise type I error rate at 5% across the two dose-level comparisons and across A1C and weight endpoints. The testing order was:

1. A1C: semaglutide 0.5 mg vs. dulaglutide 0.75 mg (non-inferiority, margin 0.4%)
2. If passed: A1C: semaglutide 1.0 mg vs. dulaglutide 1.5 mg (non-inferiority)
3. If passed: A1C superiority at both dose levels
4. If passed: body weight superiority at both dose levels

Every gate was cleared. This hierarchical approach is conservative, and the fact that all four steps achieved statistical significance strengthens the finding. The pre-specified non-inferiority margin of 0.4% for A1C is consistent with FDA guidance for diabetes drug development.

Sample size was calculated to provide >90% power for the non-inferiority test. With 1,201 patients (approximately 300 per arm), the trial was well-powered. Observed treatment differences for A1C were large enough (approximately -0.40% at the low-dose comparison and -0.41% at the high-dose comparison) that superiority was comfortably achieved.

Key Results in Context

| Comparison | A1C change (semaglutide) | A1C change (dulaglutide) | ETD (95% CI) | Weight change (semaglutide) | Weight change (dulaglutide) | ETD (95% CI) | |---|---|---|---|---|---|---| | 0.5 mg vs 0.75 mg | -1.5% | -1.1% | -0.40% (-0.55 to -0.25) | -4.6 kg | -2.3 kg | -2.26 kg (-3.02 to -1.51) | | 1.0 mg vs 1.5 mg | -1.8% | -1.4% | -0.41% (-0.57 to -0.25) | -6.5 kg | -3.0 kg | -3.55 kg (-4.32 to -2.78) |

ETD = estimated treatment difference; treatment policy estimand; Pratley et al., 2018

The weight separation is striking. At the high-dose comparison, semaglutide produced more than double the weight loss of dulaglutide. This is consistent with findings from the broader SUSTAIN program and subsequent real-world data, suggesting the weight effect is reproducible outside trial conditions.

Safety and Tolerability Signals

GI adverse events (nausea, vomiting, diarrhea) were the most common in both groups. Rates were numerically higher with semaglutide (approximately 43% vs. 33% at the high-dose pairing), though the open-label design complicates interpretation of these numbers. Discontinuation due to adverse events was low and similar across arms (approximately 5-8% per group).

No cases of pancreatitis were confirmed during the trial. Injection-site reactions were infrequent with both drugs. The semaglutide prescribing information carries the standard GLP-1 RA class warnings for thyroid C-cell tumors (based on rodent data) and pancreatitis risk.

Limitations the Authors Acknowledged

The primary publication is transparent about several constraints:

1. Open-label design. Already discussed above. The authors note it as the first limitation.
2. 40-week duration. Long enough for A1C and weight to reach near-plateau, but insufficient for cardiovascular or renal outcome assessment.
3. No cardiovascular endpoint. SUSTAIN-6 (the dedicated CVOT for semaglutide) addressed that question separately. SUSTAIN-7 was not designed or powered for MACE.
4. Dose-pairing criticism. The authors frame the comparison as "approved dose ranges" but do not claim pharmacologic equipotency.
5. Population homogeneity. The metformin-background requirement limits generalizability to patients on more complex regimens.

What Changed After Publication

Several developments have recontextualized SUSTAIN-7:

• Higher dulaglutide doses approved (2020). AWARD-11 showed dulaglutide 3.0 mg and 4.5 mg provide additional A1C and weight benefit beyond 1.5 mg, partially closing the gap SUSTAIN-7 documented.
• Oral semaglutide (Rybelsus). Approved in 2019, giving patients a non-injectable semaglutide option. The PIONEER program showed oral semaglutide 14 mg is roughly comparable to injectable semaglutide 0.5 mg on A1C.
• Higher semaglutide doses (Wegovy, 2.4 mg). Approved for obesity, not diabetes, but the dose-response curve for semaglutide extends well beyond the 1.0 mg ceiling tested in SUSTAIN-7.
• Tirzepatide (Mounjaro). A dual GIP/GLP-1 agonist that has shown A1C reductions exceeding semaglutide 1.0 mg in the SURPASS-2 trial, resetting the competitive benchmark entirely.

Bottom Line for Clinicians

SUSTAIN-7 demonstrated that semaglutide outperformed dulaglutide on glycemic control and weight reduction at matched approved-dose tiers. The statistical approach was rigorous, the hierarchical testing cleared every gate, and results held across both estimands. The open-label design and dose-equivalency questions are real limitations but do not invalidate the core finding: at the doses tested, semaglutide produced clinically meaningful additional A1C lowering (roughly 0.4%) and substantially more weight loss (2-3.5 kg) over 40 weeks.

For a patient choosing between these two specific GLP-1 RAs at their originally approved dose ranges, SUSTAIN-7 provides the most direct evidence available. For patients considering higher dulaglutide doses (now available) or newer agents like tirzepatide, the competitive picture has shifted since 2018.

Frequently asked questions

›

›

›

›

›

›

›

›

›

›

References

1. Pratley RE, Aroda VR, Lingvay I, et al. Semaglutide versus dulaglutide once weekly in patients with type 2 diabetes (SUSTAIN 7): a randomised, open-label, phase 3b trial. Lancet Diabetes Endocrinol. 2018;6(4):275-286. PubMed
2. Frias JP, Davies MJ, Rosenstock J, et al. Tirzepatide versus semaglutide once weekly in patients with type 2 diabetes. N Engl J Med. 2021;385(6):503-515. PubMed
3. Ludvik B, Giorgino F, Jodar E, et al. Once-weekly tirzepatide versus once-daily insulin degludec as add-on to metformin with or without SGLT2 inhibitors in patients with type 2 diabetes (SURPASS-3). Lancet. 2021;398(10300):583-598. PubMed
4. Novo Nordisk. Ozempic (semaglutide) prescribing information. FDA. Label
5. Eli Lilly. Trulicity (dulaglutide) prescribing information. FDA. Label
6. American Diabetes Association. Standards of Medical Care in Diabetes, 2020. Diabetes Care. 2020;43(Suppl 1). PubMed