Four cardiovascular outcome trials have now published for the glucagon-like peptide (GLP)-1 receptor agonist class, with another three due to report in 2018. But the results so far are rather conflicting, with two trials – LEADER and SUSTAIN-6 – reporting a cardioprotective effect, despite SUSTAIN-6 being designed only as a noninferiority trial, one – ELIXA – being neutral, and the last – EXSCEL – a near miss.
EXSCEL: Clarifying the situation or muddying the water?
In EXSCEL, the hazard ratio for the primary endpoint in type 2 diabetes patients taking exenatide versus placebo was 0.91, with a 95% confidence interval of 0.83 to 1.00 and a p value of 0.06. Individual and secondary endpoints leaned the same way, and all-cause mortality could be regarded as statistically significant were it not for a technicality: the trial protocol stated that secondary endpoints would be considered nonsignificant if the primary endpoint failed to achieve significance.
“Just try and explain that to a layperson with diabetes, to say, well, it reduced death but that doesn’t count because of what we wrote down on a piece of paper,” says John Petrie, Professor of Diabetic Medicine at the University of Glasgow, UK, who was chief investigator for the REMOVAL trial and a global expert panel member for LEADER.
I think, in a sense, it’s a bit of spin – that if you had a different result you might not have played it the same way.
John Petrie, Professor of Diabetic Medicine at the University of Glasgow, UK, was chief investigator for the REMOVAL trial and a global expert panel member for LEADER.
If you put the three trials that used the same primary endpoint (ie, excluding ELIXA) together in a meta-analysis, as EXSCEL principal investigator Rury Holman (Director of the Diabetes Trials Unit, University of Oxford, UK) did, you get an overall positive effect for the GLP-1 receptor agonists, with a 12% reduction in the risk for major adverse cardiovascular events and for all-cause mortality. This could be construed as supporting a class effect, assuming of course that the contradictory findings arise from differences in trial design and patient characteristics, rather than from genuine differences in efficacy between the medications.
“I think, in a sense, it’s a bit of spin – that if you had a different result you might not have played it the same way,” says Petrie. Nonetheless, he describes the EXSCEL trial as “unlucky,” being somewhat a victim of the widely accepted but arbitrary significance threshold of p less than 0.05 and, in the case of the all-cause mortality endpoint, the need for clearly predefined trial protocols.
“You can’t just say: well actually it was very nearly a positive trial, which is really the truth,” he observes.
Also siding with this view is Kevin McConway, Emeritus Professor of Applied Statistics from The Open University in the UK. “One of the strengths of [randomized controlled trials] is the protocol and the fact that people can see if you don’t stick to it,” he says. “But it’s also a bit of a straitjacket.”
Plausibility isn’t everything, but it’s a good place to start.
He stresses that the absence of a statistically significant effect “might mean that there is no effect, or a tiny one that we’re not interested in, or it might mean there is an effect but the study wasn’t powerful enough to find it. And you can’t tell which. You can never tell for sure.”
But looking at the findings of the four trials published to date, McConway asks: “Is there really strong evidence that they differ?”
“On the whole I’d say there isn’t,” he says, noting that this lends plausibility to Holman’s meta-analysis and claim for a likely class effect. “Plausibility isn’t everything, but it’s a good place to start.”
During the EXSCEL presentation at the EASD 2017 meeting, Holman highlighted the four variables he considered most likely to have contributed to the conflicting results, potentially preventing the trials from showing a consistent positive benefit. These were the duration of exposure to the study medication and of overall follow-up; baseline glycated hemoglobin (HbA1c), which was lower in ELIXA than the other trials; and the proportion of patients without pre-existing cardiovascular disease, which ranged from 0% in ELIXA to around 25% in EXSCEL. ELIXA was the only inarguably neutral trial, but also the only study to enroll solely patients with recent (72 days previously, on average) unstable cardiovascular disease.
Heterogeneity was low in Holman’s meta-analysis, at 22% for the primary composite endpoint, meaning “there’s no strong statistical evidence for major differences” in patient and other variables between the trials, says McConway, “and therefore it makes sense to consider them together.”
However, he also notes that the very few clear differences mean that comparing patient and trial variables “is a bit of a waste of time, you could argue.”
“Or rather, you’re not going to get an answer – you’re not going to get beyond ‘it might be this or it might be that’.”
McConway stresses that EXSCEL being a near miss and broadly similar to LEADER and SUSTAIN-6 in design and outcome should not lead clinicians to assume that exenatide is definitely cardioprotective. “That’s going too far in the wrong direction.”
“But you can’t say it definitely doesn’t work,” he adds. “And there is a lot of what you might describe in a different context as circumstantial evidence that it does.”
The rocky road from trial to clinic
Such uncertainty around the cardiovascular outcome trials is unhelpful to doctors considering which medications to prescribe their patients. And added to this is the fact that trial findings don’t necessarily have the anticipated influence on prescribing practice, with other factors such as the timing of a medication’s entry onto the market, the popularity of the injection device, and – certainly in countries such as the USA – the marketing efforts of the pharmaceutical companies all having an effect.
In the UK, for example, Petrie notes that although daily GLP-1 receptor agonists are the current standard of care, once-weekly dulaglutide is “kind of getting traction even though it doesn’t have cardiovascular data.”
Given that the EXSCEL findings, using weekly exenatide, were at least neutral and with a probable cardioprotective signal, whereas dulaglutide does not yet have cardiovascular outcomes data, he says: “I think if we lived in a rational world we’d probably see a switch towards exenatide long-acting, away from dulaglutide.”
In theory, national guidelines should steer healthcare providers towards the medications with the strongest evidence base, but these vary in how user-friendly and up-to-date they are. The UK’s NICE guidelines, for example, offer a very specific evidence-based management algorithm, but present the information in a densely footnoted chart, which Petrie says is unpopular with many users. Another issue with the NICE guidelines is the relative infrequency of updates, requiring them to issue separate guidance as new information appears. “So it gets very confusing for people,” observes Petrie.
The ADA guidance, by contrast, is clearly presented and annually updated, he says, “but it basically says do anything you like.”
And irrespective of ease of use, guidelines can vary in the rigor of their development and whether they are at risk for sponsor bias, as laid out in a 2012 review (later expanded to include initially omitted guidelines). The review authors compared published diabetes management guidelines against the Appraisal of Guidelines Research and Evaluation (AGREE) instrument, awarding some as little as 16.7% for developmental rigor and 8.3% for editorial independence.
Prescribing in the real world
Petrie is a developer of one of the guidelines initially omitted from the study, but eventually awarded the joint second-highest score – the Scottish Intercollegiate Guidelines Network (SIGN) diabetes management recommendations. As such, he is very familiar with the healthcare budget constraints always hanging over the process.
I do think that if a proper trial was still possible with metformin [in type 2 diabetes] it probably would show cardiovascular benefit,
But even in an ideal world with no cost considerations, Petrie would still start patients with type 2 diabetes on metformin. “I am a fan of metformin,” he says, noting that the REMOVAL trial showed a hint of cardioprotective benefit in type 1 diabetes patients, despite missing its primary endpoint of reduced progression of common carotid artery intima-media thickness.
“I do think that if a proper trial was still possible with metformin [in type 2 diabetes] it probably would show cardiovascular benefit, so I think metformin should still be up there,” says Petrie.
He would also bring sulfonylureas into play early after diagnosis, for their rapid effectiveness in patients with symptoms, although he cautions against long-term use.
And he sees a role for dipeptidyl peptidase (DPP)-4 inhibitors relatively early in the management process, because of their low cost and very benign safety profile. For patients only slightly above target on metformin, DPP-4 inhibitors can be added without concerns about adverse effects such as genital mycotic infections and diabetic ketoacidosis, he notes. Being oral medications, they also tend to be popular with patients wishing to avoid daily injections, “although when you get to a once-weekly [injection] it starts to become more marginal.”
“But I think after that we shouldn’t delay too long before going on to SGLT2s [sodium-glucose cotransporter 2 inhibitors] and GLP-1s,” says Petrie.
From this point, he says the risk profiles of the individual patients really come into play. For example, evidence from the EMPA-REG OUTCOME and CANVAS trials suggest that SGLT2 inhibitors are beneficial for patients with or at risk for heart failure. Conversely, they are currently contraindicated in patients with impaired renal function, because of concerns about their effect on the kidneys.
“So it comes down to these sorts of patient profiles, which one you choose,” says Petrie. He notes, however, that the evidence from the SGLT2 inhibitor trials to date has not borne out fears of nephrotoxicity, and suggests this may prompt a relaxing of the contraindications. “I think it will happen – sooner rather than later, actually, because there’s lots and lots of trials looking at this.”
Moreover, he suggests that SGLT2 inhibitors and GLP-1 receptor agonists can be considered for patients outside of the high-risk groups in which they were tested in randomized trials. “It comes down to how purist you are about statistics, but if you believe that the same biological processes are in play – they’re just harder to detect in low-risk people – then there should be a benefit for more people.”
We have to be responsible as well and think about the sustainability of the healthcare system.
Overall, Petrie advocates “moving through the classes quite rapidly and extrapolating a little from the evidence.” But he adds that, in the real world, physicians have to bear costs in mind. “We have to be responsible as well and think about the sustainability of the healthcare system.”
Knock-on benefits for type 1 diabetes?
As someone involved in creating diabetes management guidelines, one thing that particularly pleases Petrie is the amount of solid data they now have to work with, partly a result of the regulatory requirement for cardiovascular safety trials, introduced in 2008.
...the amount of evidence we’ve got to support our practice now is enormous.
One of the people responsible for this change, Steven Nissen (Cleveland Clinic, Ohio, USA), said at the 2017 ADA conference that “there was a lot of whining” when it was brought into force, but healthcare professionals working in diabetes are beginning to reap the rewards. “Suddenly you’ve got this rash of outcome trials – the amount of evidence we’ve got to support our practice now is enormous,” says Petrie.
His hope is that the same will happen for type 1 diabetes. “I don’t know what it’s going to take to make it happen, but there’s some glimmers of hope with the SGLT2 inhibitors,” says Petrie, referring to the positive results of the DEPICT 1 and inTandem3 trials. He adds: “I think it will be a double standard if they don’t have cardiovascular outcome trials in type 1 when they can afford to have them in type 2.”
The importance of such trials is highlighted by the inTandem findings, which showed benefit but at the price of more diabetic ketoacidosis. Petrie says: “I think it really makes the case for proper pragmatic outcome trials.”
medwireNews is an independent medical news service provided by Springer Healthcare. © 2017 Springer Healthcare part of the Springer Nature group