Is cardioprotection with GLP-1 receptor agonists a class effect?

This editorial refers to ‘Cardiovascular effectiveness of human-based vs. exendin-based glucagon like peptide-1 receptor agonists: a retrospective study in patients with type 2 diabetes’, by E. Longato et al., pp. 22–29.

The addition of glucagon like peptide 1 receptor agonists (GLP-1 RA) to the therapeutic armamentarium represents a welcomed improvement in our possibility to manage patients with type 2 diabetes in both primary and secondary prevention. This possibility has been verified in major, prospective, randomized, placebo-controlled clinical trials as recently summarized in a meta-analysis by Kristensen et al.¹ An ample number of such compounds is available, among them Lixisenatide,² Exenatide,³ Liraglutide,⁴ Semaglutide,^5,6 and Dulaglutide⁷ with further moieties under investigation. It should not come as a surprise that the question if the effect of GLP-1 RA can be considered as a class effect has been raised.

What is a class effect?

Before answering the question whether cardioprotection with GLP-1 RA is a class effect or not it is reasonable to look for a definition of this concept. There does not, however, seem to be a universal definition, but a useful attempt to delineate one was made by the US Food and Drug Administration in 2018. They declared that a class effect characterizes ‘A group of active moieties that share scientifically documented properties defined on the basis of any combination of three attributes: mechanism of action, physiologic effect and chemical structure’.⁸ If these were the only prerequisites, we could already now state that there are reasons to accept that different GLP-1 RA can be considered as compounds within the same pharmacological class. But the relevant question is whether compounds fulfilling these three requirements by necessity are interchangeable from a clinical point of view. Then aspects on evidence of benefit, safety and possibly also costs must be taken in consideration as addressed by Furberg et al.⁹ in an interesting viewpoint published 1999 in the Lancet. They claim, with all rights, that scientific data addressing all these aspects are rarely available for individual drugs within a chemical class. The only way to obtain truly valid information of clinical importance would be by means of head-to-head comparisons with acceptable, i.e. not surrogate, endpoints of different compounds within the class. Such trials must be performed in relevant study populations, which by necessity must be large, thereby costly. Taken together this makes the incentive for pharmaceutical companies to invest in such trials neglectable. For them it would be much better to compensate for lack of data for specific agents by claiming that if one drug from a class is shown beneficial, all drugs of that class should be deemed beneficial. The obvious problem with such attitude is, as underlined by Furberg et al., that ‘when harm is suspected efforts proliferate to distinguish other similar agents from the offending class member associated with the safety issue’. In conclusion, it is easy to agree with these authors that a poorly tested drug of any class is an unproven compound.

On the impact of GLP-RA

The impact of different GLP-1 RA on a composite of cardiovascular death and non-fatal myocardial infarction or stroke is shown in Figure 1. The overall benefit is a 12% reduction [hazard ratio (HR) 0.88; confidence interval (CI) 0.82–0.94], but it varies from trial to trial from +2% to −26%, indicating that there may be differences between one compound and another. There may be several explanations for this as presented in Table 1. Among them are e.g. differences in study populations with varying proportion of patients with cardiovascular disease, from 100% with a recent acute coronary syndrome in the Evaluation of Lixisenatide in Acute Coronary Syndrome (ELIXA)² to 31% in Researching cardiovascular events with a weekly incretin in diabetes (REWIND)⁷; differences in glycaemic control, as reflected by a HbA1c 8.7% (72 mmol/mol) in Liraglutide effect and action in diabetes: evaluation of cardiovascular outcome (LEADER)⁴ compared to 7.4% (57 mmol/mol) in REWIND; and the use of non-study antihyperglycaemic medication, with 59% of the participants on insulin at study start in Albiglutide and cardiovascular outcomes in patients with type 2 diabetes and cardiovascular disease (HARMONY)¹⁰ compared to 24% in REWIND; and varying times of follow-up, from 1.3 years in Peptide innovation for early diabetes treatment (PIONEER 6)⁶ to 5.4 in REWIND.

Figure 1

Forest plot on the impact of different glucagon-like peptide-1 receptor agonist (GLP-1 RA) on three-component MACE comprising the first of CV death, myocardial infarction, and stroke. CI, confidence interval; HR, hazard ratio; NNT, number needed to treat. By permission after Kristensen et al.¹

Open in new tabDownload slide

Another difference between GLP-1 RAs, related to the chemical structure, is their background and duration of action. As depicted in Figure 2, there are two types of GLP-1 RA, exendin-based and human analogues, and in both these groups the compounds may have a short or long duration of action.^11,12 As indicated in the Forest plot in Figure 1 it seems as if the exendin-based drugs lixisenatide, used in the ELIXA trial,² and exenatide, used in the EXSCEL trial,³ are less efficient than the human-based compounds used in the other trials, LEADER,⁴ HARMONY,¹⁰ SUSTAIN 6,⁵ REWIND,⁷ and Pioneer 6.⁶ Further investigations of this potential discrepancy between drugs within the GLP-1 RA class would therefore be of considerable interest.

Figure 2

Some characteristics of glucagon like peptide-1 agonists. BID, twice daily; LAR, long-acting release; OD, once daily; OW, once-weekly.

Open in new tabDownload slide

Exendin-4 vs. human-based GLP-1 RA

As already underlined the incentive for pharmaceutical companies to invest in head-to-head comparisons of different GLP-1 RA may be predicted as low or non existant, and no such trial has been performed or is planned. An alternative is to use observational data from contemporary clinical practice to compare the outcome of patients prescribed either type of GLP-1 RA. An investigation based on such principles is presented by the Italian clinical scientists Longato et al. in this issue of European Journal of Preventive Cardiology (page xx).

In a retrospective analysis of a well-known data base of patients with type 2 diabetes from North-Eastern Italy they identified patients who were prescribed a human- or an exendin-based GLP-1 RA during 2011–2018. A total of 6620 new users of a GLP-1 RA could be identified. After a careful propensity score matching the investigators could allocate 1098 patients in each of two groups, one with those who were started on an exendin-based GLP-1 RA and another with those who were prescribed a human-based GLP-1 RA. The patients were followed during an average of 18 months for a composite of major adverse cardiovascular events (MACE = myocardial infarction, stroke, or all-cause mortality). Among secondary endpoints were the individual MACE components and any hospital admission by cardiovascular reasons. Patients treated with exendin-based GLP-1 RA had a 31% lower risk for MACE (HR 0.61; 95% CI 0.39–0.95), 49% lower risk for myocardial infarction (HR 0.51; 95% CI 0.28–0.94), and a 34% lower risk for hospitalization for cardiovascular causes (HR 0.66; 95% CI 0.47–0.92).

Since the duration of action differs between the two daily GLP-1 RA lixisenatide, which has a short half-life in plasma and therefore offers less exposure time to GLP-1 receptor activation, and liraglutide, the authors performed a separate analysis in which the short-acting lixisenatide was omitted. Since the outcome did not change it was considered unlikely that the differences in efficacy related to the duration of action, leaving the difference in structure to the advantage of human-based GLP-1 RA as the most plausible explanation.

Even if the analyses by Longato et al. were performed according to the highest standards it must be acknowledged that there remains a possibility for bias. Even the best propensity scoring may fail due to variables that were unavailable in the registry or impossible to put a value on. Another drawback was the relatively few primary endpoint events (n = 80) decreasing the power of the observations. A strength is that the presented data are based on patients representative of everyday practice. These issues are well discussed by the authors that were contained in their conclusion: ‘In the absence of comparative trials and in view of the limitations of retrospective studies, this finding provides a moderate level of evidence to guide clinical decision’. A statement that it is easy to subscribe to.

In summary, the Italian study contains clinically useful information. It also underlines the notion that the use of the term class effect must be restricted and that each drug intended to be used in clinical practice should be tested in well planned and conducted clinical trials before being introduced in guidelines recommendations.

Conflict of interest: LR reports grants from The Swedish Heart and Lung Foundation, Stockholm County Council (ALF) and Boehringer-Ingelheim and consultant and personal fees from Bayer AG, Boehringer-Ingelheim, Eli Lilly and Novo-Nordisk all outside this article. GF does not have any conflicts of interest to report. LM reports personal fees from Novo Nordisk, Sanofi Aventis, Astra Zeneca, MSD, Boehringer-Ingelheim and Amgen outside this article.

The opinions expressed in this article are not necessarily those of the Editors of the European Journal Preventive Cardiology or of the European Society of Cardiology.

References