Was LEADER a superiority trial or a noninferiority trial?

Both, sequentially. The pre-specified primary analysis tested noninferiority (HR upper bound <1.30) first. Only after that threshold was met did the protocol allow testing for superiority, which also reached statistical significance (P = 0.01). The superiority finding is valid but was conditional on the noninferiority gate.

How many patients discontinued liraglutide during LEADER?

Approximately 27% of patients in the liraglutide arm discontinued the study drug before the trial ended, primarily due to gastrointestinal adverse events. The intention-to-treat analysis counted all randomized patients regardless of discontinuation.

Does LEADER apply to patients without established cardiovascular disease?

Only marginally. About 81% of enrolled patients had established CVD. The subgroup of patients with risk factors only (but no established CVD) was smaller, and the confidence intervals for that subgroup crossed 1.0. Guidelines appropriately limit the CV-benefit recommendation to patients with established atherosclerotic disease.

Who funded the LEADER trial?

Novo Nordisk, the manufacturer of liraglutide (Victoza), funded the trial. Company employees participated in study design, data collection, analysis, and manuscript writing. An independent data safety monitoring board oversaw the trial.

What was the number needed to treat in LEADER?

The NNT was approximately 53 over 3.8 years. This means roughly 53 patients needed to take liraglutide for nearly four years to prevent one MACE event compared with placebo plus standard care.

Did liraglutide reduce heart attacks and strokes individually?

Nonfatal MI showed a non-significant trend toward benefit (HR 0.88, P not significant). Nonfatal stroke was similar (HR 0.89, not significant). The composite result was driven primarily by the reduction in cardiovascular death (HR 0.78, P = 0.007).

Has LEADER been replicated by other GLP-1 trials?

SUSTAIN-6 (semaglutide, 2016) showed a similar MACE reduction. REWIND (dulaglutide, 2019) demonstrated benefit in a lower-risk population. SELECT (semaglutide 2.4 mg, 2023) showed CV benefit in patients with obesity but without diabetes. The class effect appears consistent, though each trial has its own limitations.

Did LEADER raise safety concerns about pancreatitis?

There were numerically more pancreatitis events in the liraglutide group (23 vs. 18), but the difference was not statistically significant. The FDA label for Victoza continues to list pancreatitis as a warning, and post-marketing surveillance is ongoing.

Should clinicians still prescribe liraglutide based on LEADER?

LEADER provides legitimate evidence for cardiovascular risk reduction with liraglutide in high-risk T2D patients. The limitations discussed here do not negate the findings but should inform shared decision-making, particularly regarding cost, injection burden, side-effect tolerance, and whether the patient matches the enrolled population.

Honest Criticisms and Limitations of the LEADER Trial

Q: How many patients discontinued liraglutide during LEADER?

Approximately 27% of patients in the liraglutide arm discontinued the study drug before the trial ended, primarily due to gastrointestinal adverse events. The intention-to-treat analysis counted all randomized patients regardless of discontinuation.

Q: Does LEADER apply to patients without established cardiovascular disease?

Only marginally. About 81% of enrolled patients had established CVD. The subgroup of patients with risk factors only (but no established CVD) was smaller, and the confidence intervals for that subgroup crossed 1.0. Guidelines appropriately limit the CV-benefit recommendation to patients with established atherosclerotic disease.

Q: Who funded the LEADER trial?

Novo Nordisk, the manufacturer of liraglutide (Victoza), funded the trial. Company employees participated in study design, data collection, analysis, and manuscript writing. An independent data safety monitoring board oversaw the trial.

Q: What was the number needed to treat in LEADER?

The NNT was approximately 53 over 3.8 years. This means roughly 53 patients needed to take liraglutide for nearly four years to prevent one MACE event compared with placebo plus standard care.

Q: Did liraglutide reduce heart attacks and strokes individually?

Nonfatal MI showed a non-significant trend toward benefit (HR 0.88, P not significant). Nonfatal stroke was similar (HR 0.89, not significant). The composite result was driven primarily by the reduction in cardiovascular death (HR 0.78, P = 0.007).

Q: Were women adequately represented in LEADER?

Women made up 36% of the trial population, which is consistent with many cardiovascular outcome trials but limits confidence in the effect estimate for female patients. Sex-specific subgroup analyses did not show a statistically different treatment effect, but they were underpowered.

Q: Has LEADER been replicated by other GLP-1 trials?

SUSTAIN-6 (semaglutide, 2016) showed a similar MACE reduction. REWIND (dulaglutide, 2019) demonstrated benefit in a lower-risk population. SELECT (semaglutide 2.4 mg, 2023) showed CV benefit in patients with obesity but without diabetes. The class effect appears consistent, though each trial has its own limitations.

Q: Did LEADER raise safety concerns about pancreatitis?

There were numerically more pancreatitis events in the liraglutide group (23 vs. 18), but the difference was not statistically significant. The FDA label for Victoza continues to list pancreatitis as a warning, and post-marketing surveillance is ongoing.

Q: Should clinicians still prescribe liraglutide based on LEADER?

LEADER provides legitimate evidence for cardiovascular risk reduction with liraglutide in high-risk T2D patients. The limitations discussed here do not negate the findings but should inform shared decision-making, particularly regarding cost, injection burden, side-effect tolerance, and whether the patient matches the enrolled population.

By HealthRX.com Medical Team

Published May 25, 2026Updated May 25, 2026Last reviewed May 25, 2026

Clinical image for Honest Criticisms and Limitations of the LEADER Trial Image: HealthRX.com clinical image

At a glance

| Parameter | Detail | |---|---| | N | 9,340 | | Intervention | Liraglutide 1.8 mg daily (subcutaneous) | | Comparator | Placebo (both added to standard of care) | | Duration | Median 3.8 years | | Primary endpoint | First occurrence of 3-point MACE (CV death, nonfatal MI, nonfatal stroke) | | Key result | HR 0.87 (95% CI 0.78, 0.97; P = 0.01 for superiority) |

Why This Page Exists

The LEADER trial was published in the New England Journal of Medicine in 2016 and became the first GLP-1 receptor agonist study to demonstrate cardiovascular superiority over placebo in type 2 diabetes. That result changed prescribing patterns, updated ADA Standards of Care, and built the commercial case for liraglutide as more than a glucose-lowering drug.

But a landmark result deserves landmark scrutiny. What follows is a detailed accounting of the trial's limitations, drawn from the publication itself, subsequent editorials, and independent statistical commentary.

Enrollment Bias: Who Actually Got Into LEADER

LEADER enrolled patients aged 50 or older with established cardiovascular disease, or aged 60 or older with at least one cardiovascular risk factor. This threshold guaranteed a high-event-rate population, which is standard for cardiovascular outcomes trials (CVOTs) mandated by the FDA's 2008 guidance on diabetes drugs. The trade-off is clear: the results speak most directly to patients who look like the ones randomized.

Roughly 81% of participants had established cardiovascular disease at baseline. Mean age was 64. Mean diabetes duration was nearly 13 years. Mean HbA1c was 8.7%. This is not the average patient a primary care physician starts on a GLP-1 agonist today. Younger patients with shorter disease duration, lower baseline risk, or no prior cardiovascular events were largely absent from the data.

The sex distribution compounded this. Only 36% of participants were women. While male-predominant enrollment is common in cardiology trials, it limits confidence in the effect estimate for female patients, particularly given known sex differences in cardiovascular event presentation and GLP-1 pharmacokinetics.

The Absolute Risk Reduction Was Small

The 13% relative risk reduction (HR 0.87) received the headlines, but the absolute numbers tell a different story. The primary endpoint occurred in 608 of 4,668 patients (13.0%) in the liraglutide group versus 694 of 4,672 (14.9%) in the placebo group. That yields an absolute risk reduction of approximately 1.9 percentage points over 3.8 years, corresponding to a number needed to treat (NNT) of roughly 53.

For context, statin therapy in secondary prevention trials typically delivers NNTs in the range of 20 to 30 over five years. LEADER's NNT of 53 is clinically real but modest, especially when weighed against the cost of branded liraglutide (Victoza) and the daily injection burden.

The HealthRX.com LEADER Limitation-Severity Framework

Not all trial limitations carry equal weight. The table below grades each identified concern by its potential to alter clinical interpretation, scored from 1 (minor) to 3 (could change the conclusion).

| Limitation | Severity | Rationale | |---|---|---| | Population skewed to established CVD | 3 | Directly limits applicability to primary prevention | | Small absolute risk reduction | 2 | Effect is real but NNT ~53 demands cost-benefit conversation | | Open-label background therapy imbalance | 2 | Differential statin/antihypertensive use could confound | | High discontinuation rate | 2 | ~27% stopped liraglutide; ITT analysis preserves validity but dilutes signal | | Manufacturer funding and authorship | 2 | No evidence of data manipulation, but incentive structure is real | | Short median follow-up (3.8 yr) | 1 | Adequate for MACE; unclear for long-term organ protection | | Geographic enrollment imbalance | 1 | Heavy European representation; limited Asian and African data | | CV death driving composite | 1 | Raises biological plausibility questions but is not a flaw per se |

Statistical Considerations That Deserve Attention

LEADER used a hierarchical testing procedure. The trial first tested for noninferiority (upper bound of HR <1.30), and only after passing that gate did it test for superiority. Both tests succeeded, and the primary publication reported P = 0.01 for superiority. This design is methodologically sound but worth understanding: the superiority P value was not the pre-specified primary hypothesis. It was a conditional secondary test.

The Kaplan-Meier curves for the primary endpoint separated gradually. Early separation was minimal, with visible divergence appearing only after 12 to 18 months. Some commentators, including Nissen in an invited editorial, noted that this slow separation pattern is consistent with a modest treatment effect that accumulates over time rather than a dramatic early benefit.

The primary composite was driven predominantly by cardiovascular death (HR 0.78, 95% CI 0.66, 0.93). Nonfatal MI showed a non-significant trend toward benefit (HR 0.88, 95% CI 0.75, 1.03), and nonfatal stroke showed no meaningful difference (HR 0.89, 95% CI 0.72, 1.11). When a composite endpoint is driven by a single component, the clinical interpretation changes. Liraglutide's cardiovascular story in LEADER is primarily a story about CV death, not about preventing heart attacks or strokes.

Discontinuation and Its Consequences

Approximately 27% of patients randomized to liraglutide discontinued the drug before the trial ended, compared with about 21% in the placebo group. Gastrointestinal side effects (nausea, vomiting, diarrhea) were the most common reason.

The trial correctly used intention-to-treat analysis, meaning every randomized patient was counted in their assigned group regardless of whether they continued the drug. This preserves the integrity of randomization but dilutes the observed treatment effect. The "true" biological effect of liraglutide in patients who tolerate it may be larger than the ITT estimate, or it may not. Per-protocol analyses were not reported with the same prominence.

High discontinuation also raises practical questions. If roughly one in four patients stops the drug due to side effects in a clinical trial (where monitoring and motivation are high), real-world adherence is likely worse. The Victoza prescribing label lists nausea in up to 28% of patients at the 1.8 mg dose.

Background Therapy Imbalances

LEADER was a placebo-controlled trial layered on top of standard-of-care diabetes and cardiovascular management. Investigators were free to adjust background medications. By trial end, there were measurable differences between arms in the use of insulin, sulfonylureas, and other glucose-lowering drugs, because clinicians in the placebo arm needed to intensify therapy to reach glycemic targets.

This differential treatment intensification is expected in any placebo-controlled diabetes CVOT. But it means the comparison is not purely "liraglutide vs. nothing." It is "liraglutide vs. more of other glucose-lowering drugs." If those other drugs carry their own cardiovascular risks (sulfonylureas have long been suspected of adverse CV effects per the UKPDS and ADVANCE data), the observed benefit could partly reflect harm in the comparator arm rather than protection in the liraglutide arm.

The HbA1c difference between groups was about 0.4% at 36 months, favoring liraglutide. Whether that glycemic gap alone explains any of the MACE difference is debatable, but it cannot be excluded as a confounder.

Funding, Authorship, and Conflicts of Interest

LEADER was funded by Novo Nordisk, the manufacturer of liraglutide. Novo Nordisk employees were co-authors. The company participated in study design, data collection, data analysis, and manuscript preparation. The published declaration lists extensive financial relationships between the academic authors and Novo Nordisk.

This does not mean the data are wrong. The trial was well designed, independently monitored by a data safety monitoring board, and the statistical analysis plan was pre-registered. But pharmaceutical sponsorship introduces structural incentives that independent observers should acknowledge. Publication bias, selective emphasis of favorable secondary endpoints, and framing choices in the manuscript all exist on a spectrum. The LEADER authors made reasonable choices, but readers should note who paid for the evidence and interpret accordingly.

Generalizability Gaps

Several populations were underrepresented or excluded from LEADER:

Patients with eGFR <15 mL/min were excluded. Renal impairment is common in the target population, and the safety and efficacy of liraglutide in advanced CKD remained uncertain after LEADER.
Patients without established CVD or risk factors were a small minority. LEADER cannot support using liraglutide for primary cardiovascular prevention.
Racial and ethnic diversity was limited. Approximately 78% of participants were white, 10% Asian, and 9% Black. The trial enrolled heavily from European sites.
Patients on SGLT2 inhibitors were underrepresented because empagliflozin and canagliflozin CVOTs were contemporaneous. Today, many patients take both drug classes, and LEADER provides no direct data on the incremental CV benefit of adding liraglutide to an SGLT2 inhibitor.

The 2019 ADA/EASD consensus report later recommended GLP-1 agonists with proven CV benefit for patients with established atherosclerotic cardiovascular disease, appropriately narrowing the guideline recommendation to match the enrolled population.

What Post-Publication Commentary Added

Several letters to the editor and invited editorials raised points that the primary publication did not emphasize:

The "noninferiority first" design meant that LEADER was powered to exclude a 30% excess risk, not to detect a specific magnitude of benefit. The observed 13% reduction, while statistically significant, fell at the lower end of what might be clinically decisive for some patients.
Pancreatitis and pancreatic cancer signals received attention. LEADER reported numerically more pancreatitis events with liraglutide (18 vs. 23 cases), though the difference was not significant. Long-term pancreatic safety remains a monitored concern per the FDA label.
The LEADER-extension data did not materially change the primary findings but confirmed that the Kaplan-Meier curves did not continue to diverge after the main trial period, suggesting the benefit plateaus rather than compounds indefinitely.

Placing LEADER in Context

LEADER was the right trial at the right time. Before it, GLP-1 agonists had no cardiovascular outcome data. After it, liraglutide became the first agent in the class with a superiority claim. Subsequent trials (SUSTAIN-6 for semaglutide, REWIND for dulaglutide, SELECT for semaglutide in obesity without diabetes) built on LEADER's foundation.

But context does not erase limitations. The trial enrolled a narrow, high-risk population. The absolute benefit was modest. The composite was driven by one component. The manufacturer funded and co-authored the work. None of these facts invalidate the results. All of them should inform how a clinician uses LEADER at the point of care, particularly when prescribing liraglutide to a patient who does not resemble the average LEADER participant.

Frequently asked questions

›

References

Marso SP, Daniels GH, Tanaka K, et al. Liraglutide and Cardiovascular Outcomes in Type 2 Diabetes. N Engl J Med. 2016;375(4):311-322. PubMed
ADVANCE Collaborative Group. Intensive Blood Glucose Control and Vascular Outcomes in Patients with Type 2 Diabetes. N Engl J Med. 2008;358(24):2560-2572. PubMed
Victoza (liraglutide) Prescribing Information. Novo Nordisk. FDA Label
Buse JB, Wexler DJ, Tsapas A, et al. 2019 Update to: Management of Hyperglycemia in Type 2 Diabetes. Diabetes Care. 2020;43(2):487-493. Diabetes Care
ADA Standards of Care in Diabetes, 2024. Diabetes Care. 2024;47(Suppl 1). Diabetes Care