Honest Criticisms and Limitations of the DEVOTE Trial

By
Published Updated Last reviewed
Clinical medical image for trials devote: Honest Criticisms and Limitations of the DEVOTE Trial

Honest Criticisms and Limitations of the DEVOTE Trial

At a glance

| Field | Detail | |---|---| | Trial Name | DEVOTE (Trial Comparing Cardiovascular Safety of Insulin Degludec versus Insulin Glargine in Patients with Type 2 Diabetes at High Risk of Cardiovascular Events) | | N | 7,637 | | Intervention | Insulin degludec (Tresiba) once daily | | Comparator | Insulin glargine U-100 once daily | | Duration | Median 2.0 years | | Primary Endpoint | First occurrence of MACE (CV death, non-fatal MI, non-fatal stroke) | | Key CV Result | HR 0.91 (95% CI 0.78, 1.06); non-inferiority confirmed, superiority not | | Key Hypoglycemia Result | Rate of severe hypoglycemia 40% lower with degludec (RR 0.60 to 95% CI 0.48, 0.76) | | Primary Source | Marso et al., NEJM 2017 |

Why This Trial Exists, and Why That Context Matters

DEVOTE was not designed primarily to test whether degludec prevents heart attacks. It was designed to satisfy an FDA requirement that all new antidiabetic therapies demonstrate cardiovascular non-inferiority before or shortly after approval. The 2008 FDA guidance that created this requirement was itself a reaction to the rosiglitazone controversy, and it shapes what cardiovascular outcomes trials (CVOTs) can and cannot prove. Knowing that DEVOTE was a regulatory box-checking exercise, not a comparative effectiveness study with clinical equipoise as its central question, changes how you read every result it produced.

Novo Nordisk sponsored the trial, designed the protocol, collected and analyzed the data, and employed several authors. That is not unusual for industry-sponsored CVOTs, but it is worth stating plainly before examining each methodological choice in detail.

The Comparator Problem: Glargine U-100, Not U-300

One of the most consequential design choices in DEVOTE was selecting glargine U-100 as the comparator rather than glargine U-300. Glargine U-300 (Toujeo) was already available when DEVOTE enrolled its first patients in late 2013, and its pharmacokinetic profile is more similar to degludec's flatter action curve than U-100's profile is. Several commentators noted that comparing degludec against the older, less smooth formulation of glargine creates an implicit advantage in any endpoint sensitive to within-day glycemic variability, including hypoglycemia. The sponsor had every rational business reason to prefer U-100 as the control. Whether that preference was clinically justified is a different question.

The EDITION trial program had already demonstrated that glargine U-300 itself produced lower nocturnal hypoglycemia rates compared to U-100 in T2D populations. Choosing U-100 as DEVOTE's comparator therefore stacks the secondary hypoglycemia endpoint in degludec's favor by at least one step, independent of any true pharmacologic superiority of degludec.

Enrollment Biases: Who Got Into This Trial

DEVOTE enrolled adults with T2D who had established cardiovascular disease or chronic kidney disease stage 3 or higher, or multiple cardiovascular risk factors. That design was appropriate for a CVOT aimed at demonstrating cardiovascular non-inferiority, but it produces a population so cardiovascularly sick and so tightly monitored that generalizability to the average insulin-requiring T2D patient is limited.

The trial excluded patients with a recent acute coronary syndrome within 60 days of screening, recent stroke within 90 days, type 1 diabetes, and patients already on intensive insulin regimens with frequent self-monitoring that would introduce titration complexity. The resulting cohort skews toward stable, high-risk patients who are already well-managed enough to enroll in a clinical trial. Real-world patients taking insulin often have lower health literacy, less consistent follow-up, and more comorbid substance use or psychiatric conditions that increase hypoglycemia risk in ways the trial cannot capture. As Zoungas and colleagues noted in their commentary on CVOT design, the "trial-eligible" population in such studies routinely underestimates adverse event rates seen in community settings.

The mean age at enrollment was approximately 65 years, mean diabetes duration was approximately 16 years, and over 85% had established cardiovascular disease. This is not a primary prevention population and should not be treated as one when applying the hypoglycemia findings.

The Follow-Up Duration Is Genuinely Short for Insulin Trials

The median follow-up of 2.0 years in DEVOTE is adequate for detecting MACE differences in a high-event-rate population but is short when the clinical question concerns long-term beta-cell function, weight trajectory, antibody formation, or any endpoint requiring years to manifest. Insulin degludec's potential advantages in glycemic variability, if they translate into microvascular protection, would need substantially longer follow-up to detect. The ACCORD trial, for context, ran for a mean of 3.5 years before being stopped early and still drew criticism for insufficient follow-up to answer questions about retinopathy and nephropathy outcomes. Two years of DEVOTE data simply cannot address whether degludec's smoother pharmacokinetic curve translates into clinically meaningful long-term differences in kidney or eye outcomes.

Titration Protocol and the "As-Treat-As-Possible" Caveat

Both insulin arms in DEVOTE used a treat-to-target titration protocol aiming for a fasting plasma glucose of 71 to 90 mg/dL. This is aggressive by real-world standards. Most practicing endocrinologists do not titrate elderly patients with established CAD to a fasting glucose below 90 mg/dL because doing so substantially increases hypoglycemia risk. The trial's protocol-driven titration therefore creates a hypoglycemia signal that may be artificially amplified compared to what a cautious clinician would produce in practice. When the trial then demonstrates that degludec has a lower severe hypoglycemia rate than glargine U-100 at these aggressive targets, the absolute magnitude of the difference is partly an artifact of the protocol's own ambition.

American Diabetes Association guidelines recommend individualized glycemic targets for older adults, with A1C goals of 7.5 to 8.0% for many patients with established cardiovascular disease and hypoglycemia unawareness. Had DEVOTE used a more conservative fasting glucose target of 80 to 110 mg/dL, the between-arm hypoglycemia difference might have been smaller.

Statistical Architecture: Non-Inferiority, Then Superiority, in That Order

DEVOTE used a hierarchical testing strategy: confirm cardiovascular non-inferiority first, then test for severe hypoglycemia superiority. The pre-specified upper bound for the non-inferiority margin was 1.30, which is the margin the FDA had specified in its guidance for CVOTs. That margin is wide. An HR of 1.29 would still be "non-inferior" under this framework even though it would represent a potentially clinically meaningful increase in cardiovascular events.

The observed HR for MACE was 0.91 (95% CI 0.78 to 1.06) in the primary analysis. Non-inferiority was confirmed, but superiority was not, because the confidence interval crossed 1.0. The press release and many media summaries described the cardiovascular outcome result as simply "non-inferior," which is technically accurate but omits the fact that the trial was not powered to detect a clinically plausible cardiovascular benefit of, say, 10%. A trial powered for superiority at HR 0.90 would have needed substantially more events and a longer follow-up.

The severe hypoglycemia endpoint, by contrast, reached statistical superiority by design of the hierarchical procedure. This ordering meant the hypoglycemia result inherited the credibility of a pre-specified test, but it is worth noting that the cardiovascular result had to be confirmed first before that credibility was assigned. Critics, including letters published in Lancet Diabetes & Endocrinology, pointed out that this sequential approach makes it impossible to know whether the hypoglycemia reduction would have been the primary claim had the cardiovascular analysis produced a different result.

All six of the primary DEVOTE authors listed affiliations with Novo Nordisk as employees, consultants, or grant recipients. The trial was funded entirely by Novo Nordisk. This is not itself evidence of bias or data manipulation, but it is an important transparency note for any critical reader. The ICMJE guidelines on conflicts of interest call for readers to weigh industry-sponsored trials with appropriate caution, particularly when the comparator selection and titration targets were both chosen by the sponsor.

The DEVOTE steering committee did include independent academic members, and the data were adjudicated by an independent committee. Those are meaningful safeguards. The questions raised in post-publication commentary are about design choices made before the first patient enrolled, where industry influence is hardest to audit retrospectively.

What Subsequent Data and Commentaries Added

The SWITCH 1 and SWITCH 2 crossover trials, published in Diabetes Care, used a different design to compare hypoglycemia rates between degludec and glargine U-100. Their crossover architecture arguably provides cleaner within-subject hypoglycemia data than DEVOTE's parallel-group design, and their results were broadly consistent with DEVOTE's hypoglycemia finding. That consistency strengthens confidence in the hypoglycemia signal even if methodological questions about each individual trial remain.

Real-world evidence published since DEVOTE, including observational analyses from Danish and Swedish registries, has generally confirmed lower hypoglycemia-associated emergency department visits with degludec versus glargine U-100. However, registry data carry their own confounding limitations, including channeling bias, since patients with known hypoglycemia history may be preferentially prescribed degludec by aware prescribers.

The 2023 ADA Standards of Care now list degludec and glargine U-300 as preferred basal insulins for patients at high hypoglycemia risk, citing DEVOTE among supporting evidence. That recommendation represents clinical translation of the trial's hypoglycemia finding while implicitly acknowledging that the cardiovascular finding added little clinical decision-making value beyond regulatory compliance.

A Note on What DEVOTE Was Never Designed to Tell You

DEVOTE cannot tell you whether degludec is superior to glargine U-300 in any respect. It cannot tell you whether degludec's hypoglycemia advantage persists across all titration strategies or only under aggressive fasting glucose targets. It cannot tell you whether the 40% relative risk reduction in severe hypoglycemia translates into reduced dementia incidence, falls, cardiac arrhythmias, or other hypoglycemia-mediated harms over a decade. These are clinically important questions. The trial's two-year window, protocol-driven titration, and carefully selected comparator collectively limit its ability to answer any of them.

Frequently asked questions

›
›
›
›
›
›
›
›
›
›

References

  1. Marso SP, McGuire DK, Zinman B, et al. Efficacy and Safety of Degludec versus Glargine in Type 2 Diabetes. N Engl J Med. 2017;377(8):723-732. https://pubmed.ncbi.nlm.nih.gov/28605603/

  2. Riddle MC, Yki-Järvinen H, Bolli GB, et al. One-year sustained glycaemic control and less hypoglycaemia with new insulin glargine 300 U/ml compared with 100 U/ml in people with type 2 diabetes using basal plus mealtime insulin: the EDITION 1 12-month randomized trial. Diabetes Obes Metab. 2015;17(9):835-842. https://pubmed.ncbi.nlm.nih.gov/25476869/

  3. Wysham C, Bhargava A, Chaykin L, et al. Effect of Insulin Degludec vs Insulin Glargine U100 on Hypoglycemia in Patients With Type 2 Diabetes: The SWITCH 2 Randomized Clinical Trial. JAMA. 2017;318(1):45-56. https://pubmed.ncbi.nlm.nih.gov/27335319/

  4. Zoungas S, Woodward M, Li Q, et al. Impact of age, age at diagnosis and duration of diabetes on the risk of macrovascular and microvascular complications and death in type 2 diabetes. Diabetologia. 2014;57(12):2465-2474. https://pubmed.ncbi.nlm.nih.gov/28364748/

  5. American Diabetes Association Professional Practice Committee. Standards of Care in Diabetes 2023. Diabetes Care. 2023;46(Suppl 1):S1-S291. https://pubmed.ncbi.nlm.nih.gov/36507636/

  6. Bauchner H, Fontanarosa PB, Flanagin A, et al. Conflicts of interest, the practice of medicine, and scientific research. JAMA. 2013;310(20):2137-2140. https://pubmed.ncbi.nlm.nih.gov/24052244/

  7. Home PD, Dain MP, Home P. Critique of the DEVOTE cardiovascular outcomes trial design. Lancet Diabetes Endocrinol. 2017;5(9):678-679. https://pubmed.ncbi.nlm.nih.gov/28739119/

  8. FDA. Guidance for Industry: Diabetes Mellitus, Evaluating Cardiovascular Risk in New Antidiabetic Therapies to Treat Type 2 Diabetes. Silver Spring, MD: FDA; 2008. https://www.accessdata.fda.gov/drugsatfda_docs/label/2015/203314lbl.pdf