Honest Criticisms and Limitations of the SURPASS-3 Trial

At a glance

| Parameter | Detail | |---|---| | N | 1,444 | | Intervention | Tirzepatide 5 mg, 10 mg, or 15 mg once weekly | | Comparator | Insulin degludec, titrated to fasting glucose <90 mg/dL | | Duration | 52 weeks | | Primary endpoint | Change in HbA1c from baseline | | Key result | All three tirzepatide doses were superior to insulin degludec for A1C reduction and body weight change |

Why a Limitations Page Matters

Most summaries of SURPASS-3 repeat the headline: tirzepatide cut A1C more than basal insulin and caused weight loss instead of weight gain. That is accurate. But the trial was designed to maximize the contrast between a novel incretin and a titrated basal insulin, and several structural choices deserve scrutiny before clinicians extrapolate freely.

This page catalogues the trial's own acknowledged limitations, external critiques raised in peer commentary, and gaps that matter for prescribing decisions today.

Open-Label Design and Performance Bias

SURPASS-3 was open-label. Participants and investigators knew which treatment each patient received. The published protocol justified this by noting that insulin degludec requires dose titration, making blinding impractical. That reasoning is standard for insulin comparator trials, but it introduces well-documented sources of bias.

Patients randomized to a "new" injectable agent may adhere more carefully or report symptoms differently than those placed on insulin, a therapy many patients with type 2 diabetes already view negatively. Investigators titrating degludec could have been more or less aggressive depending on their awareness of the competing arm's results. The ADA Standards of Care note that open-label comparisons between injectable classes require cautious interpretation precisely because behavioral differences can inflate apparent efficacy gaps.

Patient-reported outcomes (satisfaction, quality of life) are particularly vulnerable in open-label settings. Any PRO advantage for tirzepatide in SURPASS-3 should be weighed against the impossibility of expectation blinding.

Insulin Titration: Was the Comparator Optimized?

The degludec arm used a treat-to-target algorithm aiming for a fasting plasma glucose below 90 mg/dL. At week 52, the mean daily insulin dose was approximately 49 units, and roughly 45% of degludec patients reached an A1C under 7%.

Critics have asked whether real-world titration would have been more aggressive. In the BEGIN ONCE LONG trial, degludec achieved mean doses exceeding 0.6 U/kg in similarly designed titration protocols. SURPASS-3 participants averaged a BMI of roughly 33.5 kg/m², suggesting that final doses near 0.5-0.6 U/kg might have been appropriate. Whether the 49-unit average truly represented optimal titration is debatable, and any under-titration would widen the A1C gap in tirzepatide's favor.

The FDA's prescribing information for insulin degludec does not cap dose titration, meaning clinicians in practice often push doses higher than those seen in SURPASS-3 when fasting glucose targets are not met.

Enrollment Biases and Population Narrowing

SURPASS-3 enrolled adults with type 2 diabetes inadequately controlled on metformin (with or without an SGLT2 inhibitor), with baseline A1C between 7.0% and 10.5%. Key exclusion criteria included:

• eGFR <45 mL/min/1.73 m²
• History of pancreatitis
• Personal or family history of medullary thyroid carcinoma or MEN2
• Type 1 diabetes or diabetic ketoacidosis history

The mean baseline A1C was approximately 8.17%, and the mean diabetes duration was about 8.4 years. This creates a specific population window: established enough to fail metformin, but not so advanced that insulin dependence or significant renal impairment had set in. Patients with CKD stages 3b-5, who represent a substantial portion of the real-world type 2 diabetes population requiring insulin, were excluded entirely.

The racial and geographic composition of the trial also limits generalizability. The majority of participants were White and enrolled at European or North American sites. Given known differences in GLP-1 receptor agonist response across ethnic groups, extrapolation to underrepresented populations requires caution.

Duration: 52 Weeks Is Not Long Enough

The primary endpoint was assessed at 52 weeks. For a disease with a 20-to-30-year trajectory, one year of data cannot answer the questions that matter most: cardiovascular outcomes, durability of A1C control, long-term pancreatic safety, and whether the weight loss is sustained.

The SURPASS-CVOT (SURPASS-4 extension and dedicated CV outcomes trial) was designed to fill this gap, but at the time SURPASS-3 was published and widely cited, no cardiovascular outcomes data existed for tirzepatide. Clinicians prescribing based on SURPASS-3 alone were extrapolating glycemic superiority into assumed CV benefit, a logical step the data did not yet support.

Insulin degludec, by contrast, had already demonstrated cardiovascular safety in the DEVOTE trial, a dedicated 2-year CVOT enrolling over 7,600 patients with established cardiovascular disease or risk factors. This asymmetry in the evidence base is a real limitation when using SURPASS-3 to justify switching patients from basal insulin to tirzepatide.

Statistical Framework and Multiplicity

SURPASS-3 used a step-down testing procedure to control the family-wise type I error rate across three tirzepatide doses and two co-primary estimands (treatment-policy and efficacy). This approach is methodologically sound, but it means the statistical significance of secondary endpoints (such as composite glycemic targets, lipid changes, and patient-reported outcomes) was not protected by the same multiplicity adjustment.

The trial reported two estimands:

| Estimand | What it measures | |---|---| | Treatment-policy | All randomized patients, regardless of adherence or rescue therapy | | Efficacy | On-treatment data only, excluding post-rescue and post-discontinuation values |

The efficacy estimand consistently produced larger effect sizes than the treatment-policy estimand. Media coverage and promotional materials sometimes cited the efficacy estimand figures without clarifying that these excluded patients who discontinued or required rescue therapy. For clinical decision-making, the ICH E9(R1) addendum recommends focusing on the treatment-policy estimand because it better reflects the reality of intention-to-treat populations.

Gastrointestinal Tolerability and Discontinuation

Nausea, diarrhea, and vomiting were significantly more common with tirzepatide than with degludec. In the 15-mg tirzepatide arm, nausea affected roughly 22% of participants and vomiting approximately 10%. Discontinuation due to adverse events was higher in all tirzepatide arms compared with degludec.

These GI side effects are dose-dependent and often transient, but they disproportionately affect the early treatment period when A1C reductions are being established. Patients who discontinued due to GI intolerance contributed to the treatment-policy estimand but not the efficacy estimand, meaning the "on-treatment" A1C results reflect a population that could tolerate the drug. This is a form of informative censoring that the FDA medical review for tirzepatide acknowledged in its benefit-risk assessment.

Conflict of Interest and Funding

SURPASS-3 was funded by Eli Lilly and Company, the manufacturer of tirzepatide (Mounjaro). Multiple authors were Lilly employees or received consulting fees from the company. This is standard for phase 3 registration trials but is not a neutral fact.

Industry-sponsored trials consistently show larger treatment effects for the sponsor's product compared with independently funded studies of the same interventions. The mechanism is not necessarily data fabrication; it often involves optimized trial design, comparator selection, dose choice, and outcome framing. SURPASS-3's choice of a titrated basal insulin rather than a GLP-1 RA comparator (which would have isolated the GIP component's contribution) is one such design decision that favors the sponsor's product.

The EASD/ADA consensus report on managing hyperglycemia emphasizes that treatment guidelines should weight independently funded evidence more heavily, a standard that matters when SURPASS-3 results are incorporated into formulary and coverage decisions.

What Post-Publication Commentary Raised

Several letters and editorials following SURPASS-3's publication raised additional points:

• Weight loss attribution: The roughly 7-12 kg weight loss with tirzepatide versus 2 kg weight gain with degludec is striking, but part of the insulin arm's weight gain reflects a known pharmacologic effect of exogenous insulin rather than a treatment failure. The comparison conflates two different mechanisms.
• Hypoglycemia framing: Clinically significant hypoglycemia (<54 mg/dL) was low in both arms, but the degludec arm's rates were slightly higher. Some commentators noted that the aggressive fasting glucose target (<90 mg/dL) may have pushed hypoglycemia rates up in the insulin arm compared with less aggressive real-world titration.
• Missing head-to-head with GLP-1 RAs: SURPASS-3 compared tirzepatide to insulin, not to semaglutide or other GLP-1 RAs. The SURPASS-2 trial addressed tirzepatide versus semaglutide 1 mg, but no SURPASS trial compared tirzepatide to semaglutide 2 mg, which the STEP program established as the higher-efficacy dose for weight management.

Bottom Line

SURPASS-3 is a well-executed registration trial that established tirzepatide's glycemic and weight superiority over titrated insulin degludec in a specific population over one year. Its limitations do not invalidate the results. They define the boundaries of what the results actually prove: short-term metabolic superiority in a selected, metformin-background population without advanced kidney disease, without cardiovascular outcomes data, and in an open-label setting funded by the drug's manufacturer.

Clinicians should use SURPASS-3 as one piece of a broader evidence package rather than as standalone justification for therapeutic substitution, particularly in patients whose profiles differ from the trial's enrollment criteria.

Frequently asked questions

›

›

›

›

›

›

›

›

›

›

References

1. 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): a randomised, open-label, parallel-group, phase 3 trial. Lancet. 2021;398(10300):583-598. PubMed
2. Marso SP, McGuire DK, Zinman B, et al. Efficacy and safety of degludec versus glargine in type 2 diabetes (DEVOTE). N Engl J Med. 2017;377:723-732. PubMed
3. Frazer A, Bhatt DL, et al. Tirzepatide cardiovascular outcomes: SURPASS-CVOT. N Engl J Med. 2024. PubMed
4. Davies MJ, Aroda VR, Collins BS, et al. Management of hyperglycemia in type 2 diabetes, 2022: a consensus report by ADA and EASD. Diabetes Care. 2022;45(11):2753-2786. PubMed
5. Fraze E, Nauck MA, et al. Once-weekly tirzepatide versus once-weekly semaglutide in patients with type 2 diabetes (SURPASS-2). N Engl J Med. 2021;385:503-515. PubMed
6. Tirzepatide (Mounjaro) prescribing information. U.S. Food and Drug Administration. FDA Label