EXSCEL misses: weekly exenatide is not superior to placebo

Results from EXSCEL, the largest cardiovascular outcomes trial of a GLP-1 receptor agonist by enrollment, showed that once-weekly extended-release exenatide was safe for the heart but did not clear the bar for proving heart benefit ^[1]. The hazard ratio for the main outcome was 0.91, with a 95% confidence interval of 0.83 to 1.00 and a P value of 0.06 for superiority ^[1].

What the trial tested

Researchers randomly assigned adults with type 2 diabetes, with or without previous cardiovascular disease, to subcutaneous injections of extended-release exenatide at 2 mg once weekly or a matching placebo ^[1]. The drug was added to usual care, and the trial had two co-primary hypotheses: that weekly exenatide would be noninferior to placebo on safety and superior to placebo on efficacy ^[1].

In all, 14,752 patients took part, and 10,782 of them — 73.1% — had known cardiovascular disease at the start ^[1]. Median follow-up was 3.2 years, with an interquartile range of 2.2 to 4.4 years ^[1]. The primary outcome was the first occurrence of death from cardiovascular causes, nonfatal heart attack, or nonfatal stroke ^[1].

The numbers

A primary outcome event happened in 839 of 7,356 patients in the exenatide group (11.4%, or 3.7 events per 100 person-years) and in 905 of 7,396 patients in the placebo group (12.2%, or 4.0 events per 100 person-years) ^[1]. That worked out to a hazard ratio of 0.91 (95% CI, 0.83 to 1.00) ^[1].

The intention-to-treat analysis showed weekly exenatide was noninferior to placebo on safety, with P<0.001 for noninferiority ^[1]. But it was not superior on efficacy, with P=0.06 ^[1]. In plain terms, the trial saw fewer events with the drug, but the difference was small enough that it could not be separated from chance under the trial's own rules.

On individual outcomes, the rates of cardiovascular death, fatal or nonfatal heart attack, fatal or nonfatal stroke, hospitalization for heart failure, and hospitalization for acute coronary syndrome did not differ significantly between groups ^[1]. Neither did the incidence of acute pancreatitis, pancreatic cancer, medullary thyroid carcinoma, or serious adverse events ^[1]. The trial also tracked effects on A1C, body weight, systolic blood pressure, and heart rate, though the specific figures are not given in the abstract ^[1].

The trial was funded by Amylin Pharmaceuticals and is registered as NCT01144338 ^[1].

Why it matters for patients

GLP-1 drugs are often discussed as a single group, but EXSCEL is a reminder that cardiovascular evidence is drug-specific. A result like this does not show that weekly exenatide harms the heart — the safety analysis was clearly met ^[1] — but it also does not establish a heart benefit, which is a different and higher standard. When doctors and insurers compare GLP-1 options for someone with type 2 diabetes and existing heart disease, the strength of each drug's outcomes data is part of that comparison.

The near-miss also illustrates how much weight a single P value can carry. A hazard ratio of 0.91 with an upper confidence limit of exactly 1.00 is the statistical definition of a coin flip at the edge ^[1]. Trials are designed in advance with a threshold, and this one landed just outside it.

What happens next

The results were published on September 14, 2017 ^[1]. Independent commentaries followed, including an Annals of Internal Medicine note on December 19, 2017, concluding that exenatide did not reduce major cardiovascular outcomes in type 2 diabetes, and correspondence in the New England Journal of Medicine on December 21, 2017 ^[1]. What regulators did with this evidence, and how it affected the product's label or sales, is not addressed in the material reviewed here.