Accounting for non-adherence: A re-analysis of the Liraglutide Effect and Action in Diabetes: Evaluation of Cardiovascular Outcome Results trial

A causal re-analysis of the LEADER trial found that sustained liraglutide exposure was not statistically significantly different from sustained placebo for the primary cardiovascular composite outcome at 3.5 years (risk difference 1.1%, 95% CI -0.4 to 2.6). This was consistent with the original intention-to-treat analysis, which estimated lower risk with liraglutide (11.8%) than placebo (13.3%).

Journal article — Causal re-analysis of a randomised controlled trial. Population: Patients with diabetes from the Liraglutide Effect and Action in Diabetes: Evaluation of Cardiovascular Outcome Results trial. Follow-up: 3.5-year risk horizon. Interventions: Liraglutide.

The intention-to-treat analysis estimated 3.5-year risks of the primary composite outcome of 11.8% (95% CI 10.8 to 12.8) in the liraglutide arm and 13.3% (95% CI 12.3 to 14.3) in the placebo arm, a risk difference of 1.5% (95% CI 0.1 to 2.9). The sustained treatment analysis, accounting for post-baseline confounders, estimated 3.5-year risks of 11.4% (95% CI 10.4 to 12.5) for sustained liraglutide and 12.6% (95% CI 11.5 to 13.7) for sustained placebo, a risk difference of 1.1% (95% CI -0.4 to 2.6). This difference was not statistically significant, and no secondary outcome differed significantly between sustained liraglutide and sustained placebo at 3.5 years. The authors concluded that the results were consistent with the original trial.

Researchers working on liraglutide will find here an example of how to estimate a well-defined per-protocol or sustained-exposure effect from a cardiovascular outcomes trial using modern causal-inference methods. It does not establish a new clinical effect beyond the original trial, and the estimates are only as credible as the causal assumptions underlying them. It provides no dosing or treatment recommendation.

Key findings

• The intention-to-treat analysis estimated 3.5-year primary composite risks of 11.8% in the liraglutide arm and 13.3% in the placebo arm, a risk difference of 1.5% (95% CI 0.1 to 2.9).
• The sustained treatment analysis estimated 3.5-year primary composite risks of 11.4% for sustained liraglutide and 12.6% for sustained placebo, a risk difference of 1.1% (95% CI -0.4 to 2.6).
• The sustained treatment risk difference was not statistically significant.
• Sustained liraglutide showed no statistically significant difference versus sustained placebo for any secondary outcome at 3.5 years.
• The authors reported that the sustained-treatment results were consistent with the original LEADER trial findings.

Limitations

• Re-analysis of existing trial data rather than a new randomised comparison.
• Causal estimates rely on assumptions of the longitudinal targeted minimum loss-based estimator, including no unmeasured confounding and correct modelling of adherence and censoring.
• The sustained-exposure estimand depends on post-baseline adherence definitions and may not reflect real-world patterns of use.
• The sustained treatment risk difference was not statistically significant, and no secondary outcome was significant.

The record

• Journal: Clinical trials (London, England): 17407745261449668
• Published: 2026-06-02
• PMID: 42227113 — DOI: 10.1177/17407745261449668
• Funding: European Union; UK Research and Innovation; european health and digital executive agency
• Author countries: Denmark

Peptide profiles: Liraglutide.

All indexed evidence: Liraglutide trials & papers.

Related studies: Glucagon-like peptide-1 receptor agonist-based regimens compared with metformin-based…, Hypertensive disorders of pregnancy and GLP-1 receptor agonist timing: a systematic…, Relationship Between GLP-1-Based Therapies and Periodontal Health: A Systematic Review…, GLP-1 receptor agonists in multiple sclerosis: a systematic review of preclinical and….