Potential Mediators of Efpeglenatide's Cardiovascular Benefit: An Exploratory Analysis of the AMPLITUDE‐O Trial

ABSTRACT

Aims

In the AMPLITUDE O trial, efpeglenatide reduced major cardiovascular events by 27% in 4076 individuals with Type 2 diabetes and established cardiovascular or kidney disease compared to placebo during a median follow‐up of 1.81 years. This exploratory mediation analysis evaluates the degree to which changes in measured clinical variables and biomarkers during the trial might explain the observed reduction in MACE.

Materials and Methods

Measured variables were considered potential mediators if they were significantly changed by the intervention. The hazard (95% CI) per 1‐unit increase in each time‐updated mediator (change from baseline or updated mean) was estimated using Cox models adjusted for the same variables as in the main AMPLITUDE‐O results. Mediators whose hazard‐ratio CIs excluded 1.0 were then included in the MACE Cox model to estimate how much of efpeglenatide's effect on MACE could be statistically explained by its effect on that mediator in uni‐ and multivariable analyses.

Results

Change in HbA1c, weight, pulse pressure, LDL cholesterol, urinary albumin: creatinine ratio (UACR), eGFR, heart rate, lipase and amylase were modified by efpeglenatide. Of these, univariable analyses showed that only changes in HbA1c, LDL cholesterol, UACR and amylase accounted for 14%, 11%, 16% and 10%, respectively, of the effect of efpeglenatide on MACE. In the multivariable analysis, LDL cholesterol together with UACR accounted for 29.6% mediation of this effect.

Conclusion

In this post hoc analysis from the AMPLITUDE‐O trial, changes in LDL cholesterol together with UACR but not weight were estimated to statistically account for a modest portion of the reduction in risk of MACE with efpeglenatide.

1. Introduction

Cardiovascular outcomes trials have shown that glucagon‐like peptide‐1 receptor agonists (GLP‐1RA) reduce the risk of cardiovascular outcomes in people with Type 2 diabetes and cardiovascular disease [1, 2]. Potential mechanisms include direct anti‐atherosclerotic effects, small vessel effects mediated by improved glycaemia and reductions of body weight, blood pressure and albumin excretion rate [3]. The relative contribution of these pathways (with some overlap) may differ by outcome with direct tissue effects potentially predominating in major cardiovascular event (MACE) prevention [3].

In the placebo‐controlled AMPLITUDE O trial, efpeglenatide (a long‐acting exendin‐4‐based GLP‐1RA) reduced MACE by 27% in 4076 individuals with Type 2 diabetes and established cardiovascular or kidney disease during a median follow‐up of 1.8 years. These effects were consistent across subgroups [4].

The mechanisms underlying the cardiovascular benefits of efpeglenatide remain unclear. In an attempt to gain insight into these beneficial effects, we performed a mediation analysis. Mediation analyses can help explore possible mechanisms by scrutinising the relationship between changes in measured variables in response to efpeglenatide and incident outcomes. They can also quantify the potential contribution of biological pathways to the treatment effect. Mediation analyses from prior trials with GLP‐1RAs, including LEADER with liraglutide and REWIND with dulaglutide, have identified changes in glycated haemoglobin (HbA1c) and to a lesser extent UACR as potential modest mediators of MACE benefit (Table 1) [5, 6]. None have identified weight change as a mediator.

TABLE 1.

Mediation analysis of cardiovascular outcome trials with glucagon‐like peptide‐1 receptor agonist.

Mace outcomes
Trial (drug)	Primary outcome	Analysis type	Mediators analysed	Mediators explaining part of effect	Model specifications
Amplitude‐O (efpeglenatide)	↓ MACE (HR 0.73, 95% CI 0.58–0.92)	Traditional mediation: Cox PH models (change from baseline, updated mean)	HbA1c, body weight, PP, LDL, UACR, eGFR, HR, lipase, amylase	Updated mean: HbA1c 14%, LDL 11%, UACR 16%, amylase 10% Baseline change: NS mediators Multivariable: LDL and UACR 30%	Time‐dependent covariates; adjusted for baseline + mediator values; multivariable Cox for combined mediators
REWIND (dulaglutide)	↓ MACE (HR 0.88, 95% CI 0.79–0.99)	Traditional mediation: Cox PH models (change from baseline, updated mean)	HbA1c, UACR, body weight, waist‐to‐hip ratio, SBP, LDL	Updated mean: HbA1c 36%; UACR 28% Multivariable: HbA1c and UACR 65% Baseline change: HbA1c 17%; UACR 25% Multivariable: HbA1c and UACR 42%	Time‐dependent covariates; adjusted for baseline + mediator values; multivariable Cox for combined mediators
LEADER (liraglutide)	↓ MACE (HR 0.87, 95% CI 0.78–0.97)	Traditional mediation: Cox PH models (change from baseline, updated mean); causal mediation: Vansteelandt method	HbA1c, body weight, UACR, hypoglycemia, insulin use, sulfonylurea use, SBP, LDL	Updated mean: HbA1c 41%; UACR 29% Baseline change: HbA1c NS; UACR 22% Vansteelandt: HbA1c 82%; UACR 33%	Adjusted for baseline mediator values; Vansteelandt allowed adjustment for post‐baseline confounders; bootstrap CIs
EXSCEL (exenatide QW)	↓ All‐cause mortality (HR 0.86, 95% CI 0.77–0.97)	Causal mediation: Cox regression and counterfactual framework	HbA1c, BP, HR, LDL, HDL, TG, eGFR, body weight, BMI	NS mediators for all‐cause mortality HbA1c small effect (11% at 6 m, NS)	Adjusted for treatment, age, sex, diabetes duration; assessed 6 and 12 month changes

Abbreviations: ACM, all‐cause mortality; BP, blood pressure; CI, confidence interval; Cox PH, Cox proportional hazards; DBP, diastolic blood pressure; eGFR, estimated glomerular filtration rate; Hb, haemoglobin; HbA1c, glycated haemoglobin; HDL, high‐density lipoprotein cholesterol; HR, hazard ratio; LDL, low‐density lipoprotein cholesterol; MACE, major adverse cardiovascular events (composite of cardiovascular death, non‐fatal myocardial infarction and non‐fatal stroke); NS, not significant; PP, pulse pressure; RAAS, renin‐angiotensin‐aldosterone system; SBP, systolic blood pressure; TG, triglycerides; UACR, urinary albumin‐to‐creatinine ratio; WBC, white blood cell count; WQ, once‐weekly.

We therefore conducted an exploratory mediation analysis of the AMPLITUDE‐O trial to quantify the relative contribution of changes in clinical variables and biomarkers to the observed reduction in MACE with efpeglenatide.

2. Materials and Methods

2.1. Study Design and Participants

The AMPLITUDE‐O study design and eligibility criteria have been published previously [4]. The trial was led by an international steering committee, sponsored by Sanofi. All trial data was analysed at the Population Health Research Institute in Hamilton, Canada. Ethics approval was obtained at all sites and all participants provided written informed consent. Individuals with Type 2 diabetes were eligible if they were at least 18 years of age, had a glycated haemoglobin level (HbA1c) greater than 7% with prior cardiovascular disease or if they were at least 50 years (if male) or 55 years (if female) with kidney disease and at least one additional cardiovascular risk factor. Key exclusion criteria included prior GLP‐1RA or DPP‐4 inhibitor use within 3 months, pancreatitis, severe gastrointestinal disorders or proliferative retinopathy. Participants were recruited at 344 sites across 28 countries and randomly assigned (1:1:1) to receive weekly subcutaneous injections of efpeglenatide 4 mg, 6 mg or placebo, on top of standard of care. Randomisation was stratified according to baseline or planned use of sodium–glucose cotransporter‐2 (SGLT2) inhibitors. The study drug or placebo was continued until trial end, death or withdrawal, with all medications provided in identical prefilled syringes to maintain blinding. The planned sample size of 4000 participants provided 90% power to demonstrate non‐inferiority (margin 1.3) of the combined 4‐ and 6‐mg dose groups versus placebo over up to 3 years of follow‐up. No separate sample size calculation was performed for this post hoc exploratory analysis.

2.2. Statistical Analysis

Any measured clinical or biochemical variable whose value was significantly affected by random assignment to efpeglenatide (either 4 or 6 mg subcutaneously weekly) versus placebo was considered a potential mediator of the effect of the drug on MACE. All analyses were restricted to participants for whom there was a baseline and at least one follow‐up mediator value available in the database. Potential mediators that were not normally distributed were natural logarithm transformed before analyses. Two general approaches were used to identify potential mediators of the effect of efpeglenatide on the MACE outcome (a composite of nonfatal myocardial infarction, nonfatal stroke or death from cardiovascular or unknown causes). The first one focused on the change in value from baseline and the second focused on the updated mean. As previously published, the change in value of the potential mediators from baseline to either the first MACE or study end was estimated as the difference in the current value to baseline value [7]. The updated mean of the value of the potential mediator was calculated from all available values before the occurrence of the outcome of interest as previously summarised [7].

The hazard (95% CI) of either outcome per 1 unit increase in the value of the change from baseline or the updated mean of the potential mediator was estimated using Cox models. All models were adjusted for the same variables used to estimate the effect of efpeglenatide on MACE in the main report of the AMPLITUDE‐O paper (i.e., geographic region and stratification factor). Potential mediators for which the confidence interval of their hazard ratios excluded 1.0 were then analysed to estimate the degree to which the effect of efpeglenatide on MACE could be statistically explained by (i.e., accounted for) the effect of efpeglenatide on the potential mediator. This was done by including the baseline value and either the change from baseline or the updated mean value of the potential mediator as an independent covariate in the adjusted Cox model. The degree to which the hazard ratio for efpeglenatide's effect changed after including the potential mediator in the model was estimated by calculating the percent mediation using the following formula:

$$\frac{100\times \left(\ln \mathrm{HR}\left[\text{without the potential mediator}\right]-\ln \mathrm{HR}\left[\text{with the potential mediator}\right]\right)}{\ln \mathrm{HR}\left[\text{without the potential mediator}\right]}$$

where HR is the hazard ratio and ln is the natural logarithm.

Mediators identified using the above univariable analyses were then assessed in a multivariable analysis. First, all the mediators that were identified from the analysis using change from baseline were included in an adjusted Cox model that included efpeglenatide assignment, baseline value and change from the baseline value, as a time‐dependent covariate. Then a backward selection approach was used to identify only those mediators for which the p‐value for the beta coefficient of the change from the baseline was < 0.05/the number of mediators included. Interactions between mediators were also assessed. The remaining mediators were then used to estimate the overall effect of the independent mediators and the percent mediation for this overall effect was estimated in the same way as noted above. The same approach was repeated using the updated mean. All analyses were performed in SAS using version 9.4 or higher.

3. Results

Of the 4076 participants (mean age 64.5 ± 8.2), 33% were women, 89.6% had established cardiovascular disease and 31.6% had chronic kidney disease. Mean diabetes duration was 15.4 years, mean HbA1c 8.9% (74 mmol/mol) and 15.2% were on an SGLT2 inhibitor. During a median follow‐up period of 1.81 years (interquartile range, 1.69–1.98), incident MACE occurred in 189/2717 (7.0%) participants assigned to receive efpeglenatide and 125/1359 (9.2%) participants assigned to receive placebo (HR 0.73, 95% CI 0.58, 0.92) [4].

As noted in Table 2, HbA1c, weight, pulse pressure, low‐density lipoprotein (LDL) cholesterol, UACR, eGFR, heart rate, lipase and amylase were significantly affected by efpeglenatide compared to placebo. These variables were therefore scrutinised as potential mediators of the efpeglenatide benefit on MACE.

TABLE 2.

Potential mediators of the effect of efpeglenatide (4/6 mg) versus placebo on MACE: univariable analysis.

	HbA1c	Weight	PP	LDL	Ln UACR	eGFR	Heart rate	Ln lipase	Ln amylase
N a	3971	3973	3977	3647	3963	3967	3977	3967	3967
HR (95% CI) of efpeglenatide on MACE b	0.72 (0.56–0.92)	0.73 (0.57–0.93)	0.73 (0.57–0.93)	0.78 (0.60–1.01)	0.73 (0.57–0.93)	0.71 (0.56–0.91)	0.73 (0.57–0.93)	0.71 (0.56–0.91)	0.71 (0.56–0.91)
LSM (95% CI) change from baseline of variable c	−1.25 (−1.3, −1.2)	−2.6 (−2.9, −2.3)	−2.1 (−2.7, −1.5)	−0.08 (−0.13, −0.03)	0.71 (0.67, 0.75)	0.8 (0.2, 1.4)	3.8 (3.4, 4.3)	1.23 (1.20, 1.26)	1.15 (1.13, 1.16)
Change from baseline vs. outcome (HR, 95% CI) d	0.97 (0.90–1.04)	1.02 (1.00–1.04)	1.00 (1.00–1.01)	1.15 (0.99–1.33)	1.06 (0.95–1.19)	1.00 (0.99–1.01)	1.00 (0.99–1.01)	0.90 (0.71–1.13)	0.85 (0.56–1.31)
Adjusted HR (95% CI) efpeglenatide e	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A
Percent mediation f	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A
Updated mean (SD)	8.0 (1.4)	90.4 (19.7)	56.2 (11.1)	2.04 (0.88)	1.32 (1.65)	70.5 (21.7)	75.7 (8.8)	3.73 (0.53)	4.14 (0.45)
Updated mean vs. outcome (95% CI) g	1.13 (1.04–1.23)	1.00 (1.00–1.01)	1.01 (1.00–1.02)	1.36 (1.20–1.55)	1.27 (1.19–1.36)	0.99 (0.99–1.00)	1.00 (0.99–1.02)	0.97 (0.78–1.21)	0.69 (0.53–0.90)
Adjusted HR (95% CI) efpeglenatide h	0.75 (0.56–1.01)	N/A	N/A	0.80 (0.62–1.04)	0.76 (0.60–0.98)	N/A	N/A	N/A	0.74 (0.57–0.96)
Percent mediation i	14.1	N/A	N/A	10.8	16.2	N/A	N/A	N/A	10.0

Abbreviations: CI, confidence interval; eGFR, estimated glomerular filtration rate; HR, hazard ratio; LDL, low‐density lipoprotein cholesterol; LSM, least squares mean; MACE, major adverse cardiovascular events; N/A, not applicable because the CI of the HR for relationship between the potential mediator and incident outcome include 1; PP, pulse pressure; UACR, urine albumin: creatinine; all efpeglenatide effects are adjusted for geographic region and randomisation stratification factor.

^a Participants with a baseline and ≥ 1 follow‐up value before MACE or last visit.

^b Effect of efpeglenatide vs. placebo on MACE in the N people in row 1.

^c Estimated from available values until the outcome or study end.

^d Relationship between potential mediator change from baseline and MACE.

^e Effect of efpeglenatide versus placebo on MACE adjusted for baseline value and change from baseline of the potential mediator.

^f [100 × (ln HR [unadjusted] − ln HR [adjusted])/ln HR [unadjusted]], where HR [unadjusted] is from row 2 and HR [adjusted] is from row 5.

^g Relationship between potential mediator updated mean and MACE.

^h Effect of efpeglenatide versus placebo on MACE adjusted for baseline value and updated mean of the potential mediator.

ⁱ [100 × (ln HR [unadjusted] − ln HR [adjusted])/ln HR [unadjusted]], where HR [unadjusted] is from row 2 and HR [adjusted] is from row 9.

When using the value of the change from baseline, none of the variables were significantly associated with MACE. Conversely, when the updated mean value was assessed, only HbA1c, LDL cholesterol, UACR and amylase, were significantly associated with MACE, with hazard ratios of 1.13 (95% CI 1.04–1.23) for HbA1c, 1.36 (95% CI 1.20–1.55) for LDL cholesterol, 1.27 (95% CI 1.19–1.36) for UACR and 0.69 (95% CI 0.53–0.90) for amylase (Table 2).

Univariable mediation analyses of these variables showed that UACR had the largest mediation effect. Thus, when accounting for the change in mean UACR, the observed hazard ratio of the effect of efpeglenatide on MACE changed from 0.73 to 0.76 (95% CI 0.60–0.98) corresponding to a mediation effect of 16.2%. Further, when accounting for the change in mean HbA1c, LDL cholesterol and amylase, mediation effects of 14.1%, 10.8% and 10.0% respectively were estimated. When these mediators were included in a multivariable analysis with backward selection, only LDL cholesterol and UACR had p‐values below 0.05/number of mediators and were therefore included in the final multivariable Cox model. In this model, the observed hazard ratio of efpeglenatide vs. placebo on the MACE outcome changed from 0.79 (95% CI 0.61–1.02) to 0.85 (95% CI 0.65–1.10) resulting in a mediation effect of 29.6%. There was no evidence of an interaction between LDL cholesterol and UACR in this model (Table 3).

TABLE 3.

Independent mediators of the effect of efpeglenatide (4/6 mg) versus placebo on MACE.

	Based on change from baseline	Based on updated mean
Univariable mediators included in the multivariable model	None	HbA1c, LDL, ln UACR, ln amylase
Independent mediators of the effect of efpeglenatide on MACE a	N/A	LDL, ln UACR
N with data for the independent mediators	N/A	3639
HR (95% CI) of efpeglenatide b	N/A	0.79 (0.61–1.02)
Adjusted HR (95% CI) efpeglenatide c	N/A	0.85 (0.65–1.10)
Percent mediation d	N/A	29.6

Abbreviations: CI, confidence interval; HR, hazard ratio; LDL, low‐density lipoprotein cholesterol; ln, natural logarithm; MACE, major adverse cardiovascular events; N/A, not applicable; UACR, urine albumin: creatinine; all efpeglenatide effects are adjusted for geographic region and randomisation stratification factor.

^a After application of backward selection to the univariable mediators, with P for inclusion in the multivariable model < 0.05/number of univariable mediators.

^b Effect of efpeglenatide versus placebo on MACE in the N people in row 3.

^c Effect of efpeglenatide versus placebo on MACE adjusted for the variables in row 2.

^d [100 × (ln HR [unadjusted] − ln HR [adjusted])/ln HR [unadjusted]], where HR [unadjusted] is from row 4 and HR [adjusted] is from row 5; P for interaction of LDL and ln UACR = 0.59.

Since a dose‐dependent effect (4 and 6 mg efpeglenatide) on clinical outcomes has been reported, we applied mediation analysis to these two doses [8]. Findings broadly consistent with the analysis were noted for any dose of efpeglenatide (Tables S1 and S2).

4. Discussion

In this exploratory mediation analysis of AMPLITUDE‐O, a modest proportion of efpeglenatide's cardiovascular benefit was statistically explained by changes in traditional risk factors measured as updated means. Thus, the effect of efpeglenatide on mean HbA1c, LDL cholesterol, UACR and amylase accounted for 14%, 11%, 16% and 10% of the observed MACE risk reduction, respectively, with no evidence for an effect of changes in weight or blood pressure. In the multivariable analysis, LDL cholesterol, together with UACR, statistically accounted for 29.6% of the observed MACE reduction. These results suggest that efpeglenatide's cardiovascular benefits likely largely operate through mechanisms independent of changes in these variables.

Our findings broadly echo prior GLP‐1RA trial mediation analyses (Table 1). In LEADER, HbA1c explained 36%–41% and UACR 22%–29% of liraglutide's effect on MACE when using traditional Cox models. Subsequent analysis using a causal mediation framework estimated stronger mediation by HbA1c, 82% and UACR, 33%, highlighting how analytical framework influences estimated proportions [5]. In REWIND, dulaglutide's effects were also partially mediated by HbA1c and UACR jointly explaining 42%–65% of the risk reduction in MACE. A post hoc mediation analysis of the EXSCEL trial assessed risk factors that could explain the effect of exenatide once weekly vs. placebo on all‐cause mortality. The mediation analysis was done on risk factors that changed from baseline to 6 and to 12 months. HbA1c was the only statistically significant variable and explained 10.7% of the effect from baseline to 6 months [9].

Prior mediation analyses of many potential mediators of GLP‐1RA's treatment effect also identified HbA1c and UACR as potential mediators, suggesting that these mediators may reflect pathways relevant to GLP‐1 receptor agonism. In contrast, both our uni‐ and multivariable analyses suggest that LDL cholesterol may mediate 14% and together with UACR, 29% of the effect on MACE, respectively. These mediation estimates should not be interpreted as the proportions of the treatment effect transmitted through a single causal pathway, but rather as the degree to which the treatment effect is attenuated after accounting for changes in candidate mediator(s). Because candidate mediators may capture overlapping cardiometabolic processes, separate univariable mediation estimates may reflect shared underlying pathways rather than independent causal mechanisms. Thus, these estimates are not mutually exclusive. Taken together, these findings suggest that changes in UACR and LDL cholesterol may lie on or mark, pathways related to cardiovascular benefit but are unable to identify or establish causality.

These findings should also be interpreted in the broader context of GLP‐1RA cardiovascular outcome trials [10]. Efpeglenatide is an exendin‐4–based GLP‐1RA and AMPLITUDE‐O extended cardiovascular and kidney benefits to a long‐acting agent of this structural type. The UACR finding supports evidence that albuminuria is a marker of microvascular injury and that improvements in UACR likely reflect beneficiary GLP‐1RA class‐related effects on kidney and systemic vascular‐risk pathways [11]. By contrast, LDL cholesterol has not been established as a potential GLP‐1RA class‐related mediator. Although LDL cholesterol lowering is a proven means of reducing cardiovascular risk, the modest LDL cholesterol reduction is unlikely to have meaningfully affected MACE risk. Thus, its apparent mediation likely reflects correlation with other treatment‐related or background risk‐factor changes, rather than a clearly efpeglenatide‐specific or GLP‐1RA class‐related mechanism.

Although HbA1c statistically accounted for 14.1% of the treatment effect in the univariate analysis, it did not significantly account for the effect after adjusting for UACR and LDL in the multi‐variable model. Possible explanations for why this differs from previous mediation analyses of GLP‐1RA outcomes trials include the possibility that efpeglenatide's effect on HbA1c was unrelated to its effect on the outcome, the shorter duration of the AMPLITUDE‐O trial or the possibility that changes in HbA1c were captured by changes in UACR (leading HbA1c to be dropped from the multivariable model).

The proportion of effect explained by clinical variables differs between traditional and causal mediation analytic approaches. The Cox proportional hazards approach used in this paper is consistent with prior exploratory GLP‐1RA cardiovascular outcome trial analyses. It also provides a clinically interpretable method for exploring whether treatment‐induced changes in candidate mediators statistically explain part of the observed treatment effect. The approach is limited by the assumption that: (a) the effect of the intervention (efpeglenatide) on the outcome did not vary by its effect on the potential mediator and (b) there is no unmeasured confounding variable that affects both the outcome and the mediator(s). By contrast, causal mediation analyses, such as those proposed by Vansteelandt et al., provide estimates of direct and indirect effects of the intervention through specified pathways under specific assumptions, but require more complex modelling [12].

The identified mediators accounted for only a modest proportion of the efpeglenatide‐associated reduction in MACE. This residual unexplained effect is biologically plausible. Experimental studies suggest that GLP‐1RAs may exert anti‐inflammatory, anti‐oxidative and anti‐atherothrombotic effects, enhance endothelial nitric‐oxide signalling and improve plaque stability [13]. The present analysis was limited to variables measured in AMPLITUDE‐O and therefore cannot assess these pathways directly. Efpeglenatide's benefits on MACE may therefore largely reflect these direct tissue actions rather than changes in traditional risk factors. The findings also complement a recent conceptual framework proposing that weight loss accounts for more of the incretin‐induced improvements in heart‐failure and kidney outcomes, whereas MACE reduction appears mainly driven by direct tissue and vascular effects [3]. That noted, in the SELECT trial of people with cardiovascular disease and overweight or obesity, semaglutide reduced weight by 8.5% and MACE by 20% versus placebo and subsequent mediation analysis suggested that 33% of this benefit may be mediated by reduced waist circumference; notably, weight reduction per se did not mediate the observed reduction in MACE [14]. Weight reductions in SELECT were well beyond those in AMPLITUDE‐O or prior incretin‐based cardiovascular outcome trials in Type 2 diabetes and the survival curves seemed to separate much faster than seen in prior GLP‐1RA CVOTs. To what extent this apparent faster effect on MACE reduction in SELECT may relate to the greater weight loss seen in that trial remains unclear.

Strengths of this analysis include use of participant‐level data, modelling of both short‐ and long‐term mediator effects and alignment with prior GLP‐1RA mediation studies. Limitations include the exploratory post hoc design and residual confounding between mediators and outcomes. The mediated percentages are model‐dependent and should not be directly compared with estimates from different analytical frameworks. These findings should therefore be interpreted statistically and viewed as hypothesis‐generating rather than causal.

Clinically, these findings further suggest that the cardiovascular benefits of GLP‐1RAs likely extend beyond changes in glycemia and traditional risk factors. Although reductions in LDL cholesterol, HbA1c and UACR are prognostically favourable and remain key targets, they cannot fully account for the MACE reduction observed with efpeglenatide in the AMPLITUDE‐O trial.

Together with prior mediation analyses of other GLP‐1RA, the results suggest that GLP‐1RA exert cardioprotection largely through pathways not captured by standard trial biomarkers. Further mechanistic studies combining causal mediation methods with novel biomarkers (that capture impacts on differing tissues) may be required to disentangle the biological pathways of incretin benefits. Meanwhile, the consistent reduction in MACE events across multiple GLP‐1RA trials remains the most compelling evidence of their clinical value.

Author Contributions

D.R.‐O., N.S. and H.C.G. researched data, contributed to discussion and wrote the first draft of the manuscript. T.K. conducted statistical analyses, reviewed and edited the manuscript. C.R.S., K.R.B., J.R., C.S.P.L., S.D.P., R.D.L. and R.P. reviewed and edited the manuscript. All authors approved the final version. H.C.G. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Funding

The AMPLITUDE‐O trial (Effect of Efpeglenatide on Cardiovascular Outcomes) was funded by Sanofi. The analysis was supported by Hanmi Pharmaceutical.

Conflicts of Interest

D.R.‐O. has received a travel grant from The Danish Diabetes Foundation. C.R.S. reports honoraria for lectures and consulting from Eli Lilly and Novo Nordisk. K.R.B. reports research support from NIH, Population Health Research Institute, Bayer, Kestra, Novartis, Amgen, Nanox AI and Medic One Foundation and consulting fees from Janssen, Amgen, Kestra and Cleerly. S.D.P. has been a member on or has consulted with Abbott, Dexcom, Eli Lilly & Co, Hoffman La Roche, Innovent, Menarini International, Novo Nordisk, Sun Pharmaceuticals. He has also served on the speakers' bureau of Abbott, AstraZeneca, Boehringer Ingelheim, Eli Lilly & Co, Merck & Co., Menarini International, NovoNordisk, Sanofi. C.S.P.L. has received research grants from the National Medical Research Council of Singapore, Novo Nordisk and Roche Diagnostic; has served in advisory, consulting and trial leadership roles for Alnylam Pharma, AnaCardio AB, Applied Therapeutics, AstraZeneca, Bayer, Boehringer Ingelheim, Boston Scientific, Bristol Myers Squibb, Corteria, CPC Clinical Research, Cytokinetics, Eli Lilly, Impulse Dynamics, Intellia Therapeutics, Janssen Research & Development LLC, Medscape/WebMD Global LLC, Merck, Novartis, Novo Nordisk, Pfizer, Radcliffe Group Ltd., Roche and Us2.ai; has patent PCT/SG2016/050217 pending and patent US Patent No. 10631828 B1; US 10702247 B2; US 11301996 B2; US 11446009 B2; US 11931207 B2; US 12001939; US 12400762 B2; and is a co‐founder and non‐executive director of Us2.ai. R.D.L. reports research grants or contracts from Bristol Myers Squibb, GlaxoSmithKline, Pfizer, Novartis, Ionis Pharmaceuticals Inc., Regeneron Pharmaceuticals Inc., Alnylam, Roche, Bayer, VeraDermics, Priovant Therapeutics Basilea Pharmaceutica, Theravance Biopharma Inc. and Cytokinetics Inc. Funding for educational activities or lectures from Pfizer, Novartis, Novo Nordisk and Bristol Myers Squibb. Funding for consulting from Bayer, Boehringer Ingelheim, Bristol Myers Squibb, Novo Nordisk, Novartis and Pfizer. R.P. declares speaker fees from Abbott, Corcept, Lilly and Novo Nordisk; consulting fees from AbbVie, Amgen, Bayer AG, Bayer HealthCare Pharmaceuticals, Boehringer Ingelheim, Corcept, Endogenex, Gasherbrum Bio, Genprex, Getz Pharma, Intas Pharmaceuticals, Lilly, Novo Nordisk, Pfizer, Sun Pharmaceutical Industries and Verdivia; and research grants (directed to his institution) from AstraZeneca, Boehringer Ingelheim, Carmot Therapeutics, Dompe, Endogenex, Lilly, Novo Nordisk and Sanofi. J.R. has served on scientific advisory boards and received honoraria or consulting fees from: Amgen, Applied Therapeutics, Biomea Fusion, Boehringer Ingelheim, Corcept, Eccogene, Eli Lilly and Company, Novo Nordisk, Oramed Pharmaceuticals, Regeneron, Regor, Sanofi, Scholar Rock, Structure Therapeutics and Terns Pharmaceuticals; and has received grants and research support from: Amgen, Applied Therapeutics, Biomea Fusion, Boehringer Ingelheim, Corcept, Eli Lilly and Company, Hanmi Pharmaceutical, Merck, Novo Nordisk, Oramed Pharmaceuticals, Pfizer, Regeneron, Regor, Sanofi, Structure Therapeutics and Terns Pharmaceuticals. N.S. has consulted for and/or received speaker honoraria from AbbVie, Amgen, AstraZeneca, Boehringer Ingelheim, Carmot Therapeutics, Eli Lilly, Gan & Lee, GlaxoSmithKline, Hanmi Pharmaceuticals, Kailera, Mass Medicines, Menarini‐Ricerche, Metsera, Novo Nordisk, Pfizer, Regeneron, Roche, UCB Pharma and Verdiva Bio; and has received grant support paid to his University from AstraZeneca, Boehringer Ingelheim, Novartis and Roche outside the submitted work. H.C.G. holds the McMaster‐Sanofi Population Health Institute Chair in Diabetes Research and Care. He reports research grants from Novo Nordisk; continuing education grants from Eli Lilly, Abbott, Sanofi, Novo Nordisk and Boehringer Ingelheim; honoraria for speaking from AstraZeneca and Jiangsu Hanson; and consulting fees from Abbott, Eli Lilly, Novo Nordisk, Sanofi and Zealand and holds a patent for the use of GDF15 as a biomarker for metformin intake. The other author declare no conflicts of interest.

Supporting information

Table S1: Potential mediators of the effect of efpeglenatide (4 mg) versus placebo on MACE: univariable analysis.

Table S2: Potential mediators of the effect of efpeglenatide (6 mg) versus placebo on MACE: univariable analysis.

Raaschou‐Oddershede D., Schwarz C. R., Kovalova T., et al., “Potential Mediators of Efpeglenatide's Cardiovascular Benefit: An Exploratory Analysis of the AMPLITUDE‐O Trial,” Diabetes, Obesity and Metabolism 28, no. 9 (2026): 8552–8559, 10.1111/dom.71062.

Data Availability Statement

Research data are not shared.