Class Effect for Sodium Glucose-Cotransporter-2 Inhibitors in Cardiovascular Outcomes

Sodium glucose-cotransporter-2 inhibitors (SGLT-2i) act independently of pancreatic β-cell function and inhibit glucose uptake in the proximal renal tubule, resulting in increased glycosuria and a resultant decrease in blood glucose. SGLT-2i lead to reduction in blood pressure and weight, at least initially, and cause little hypoglycemia because glycosuria diminishes with decreased filtered glucose load. Marketed compounds in the class include dapagliflozin, canagliflozin, and empagliflozin, with others available in Japan and more in development. Cardiovascular outcome trials have been completed for 2 agents in this class, namely, empagliflozin (EMPA-REG OUTCOME)¹ and canagliflozin (CANVAS Program [Canagliflozin Cardiovascular Assessment Study]),² with further trials expected in 2019: CREDENCE ([Canagliflozin and Renal Endpoints in Diabetes with Established Nephropathy Clinical Evaluation] canagliflozin), DECLARE ([Dapagliflozin Effect on Cardiovascular Events] dapagliflozin), and VERTIS ([Evaluation of Ertugliflozin Efficacy and Safety] ertugliflozin).

In both EMPA-REG OUTCOME and the CANVAS Program, SGLT-2i led to reduction in the 3-component major adverse cardiovascular event outcome (MACE-3; cardiovascular death, nonfatal myocardial infarction, or nonfatal stroke). Fewer heart failure hospitalizations (Figure 1B) and improved renal outcomes were also demonstrated, although for the CANVAS Program they were not formally statistically significant because of the prespecified hierarchical testing plan. However, discrepancies between some trial outcomes raise questions about whether the treatment effects are drug-specific or generalizable to the drug class.

• Download figure
• Download PowerPoint

The most remarkable difference between the trials is the significant reduction in cardiovascular mortality with empagliflozin, not seen with canagliflozin, a finding echoed in the more objective outcome of all-cause mortality (Figure 1B). Critical evaluation of the trial designs, populations, and conduct is important to understand whether discrepant outcomes may be attributable to trial features rather than to the randomized therapy (Figure 1A). EMPA-REG OUTCOME was conducted exclusively in a secondary cardiovascular prevention population over a median follow-up of 3.1 years. In contrast, the CANVAS Program was a combined analysis of 2 separate studies (median follow-up of 2.4 years), enrolling patients with a mixture of cardiovascular risk factors (34%) or established cardiovascular disease (66%). Although both CANVAS studies had identical entry criteria, follow-up times differed markedly: CANVAS had a mean follow-up of 5.7 years, and CANVAS-R had a mean follow-up of 2.1 years. With a higher cardiovascular risk population, EMPA-REG OUTCOME enrolled higher proportions of men and of those treated with antithrombotic agents and renin-angiotensin system blockers. MACE-3 and all-cause mortality rates in the placebo group of EMPA-REG OUTCOME were considerably higher than in the CANVAS Program, reflecting the higher cardiovascular risk population. It is interesting to note that MACE-3 rates in the CANVAS Program placebo group with cardiovascular disease (41 per 1000 person-years) were similar to those in EMPA-REG OUTCOME (44 per 1000 person-years). Subgroup analysis in the CANVAS Program showed a hazard ratio (HR) for MACE-3 of 0.82 (95% confidence interval [CI], 0.72–0.95) for patients with prior disease compared with 0.98 (95% CI, 0.74–1.30) in patients without cardiovascular disease (all-cause mortality subgroups were not reported). Consideration of the differential follow-up in CANVAS and CANVAS-R shows that in the short-term CANVAS-R trial, there were 99/2907 deaths in the canagliflozin group versus 106/2905 in the placebo group (HR, 0.92; 95% CI, 0.70–1.21), and in the much longer follow-up CANVAS trial, the respective numbers were 301/2888 compared with 175/1442 (HR, 0.84; 95% CI, 0.70–1.01). Therefore, even in the long-term CANVAS trial, the effect of canagliflozin remained smaller than that obtained with empagliflozin. The combination of a significant proportion of primary prevention patients in the CANVAS Program and short-term treatment in half the population (CANVAS-R) might partially explain a smaller effect of canagliflozin compared with empagliflozin.

Previously established adverse event risks (eg, genital infections, volume depletion) appear similar between the 2 SGLT-2i trials, but the CANVAS Program identified new safety concerns for amputations—collected as an adverse event of special interest (HR, 1.97; 95% CI, 1.41–2.75) and fractures (any fracture: HR, 1.26; 95% CI, 1.04–1.52; low trauma fracture: HR, 1.23; 95% CI, 0.99–1.52). An increased fracture risk was not identified for empagliflozin.³ The amputations in the CANVAS Program, primarily of toes or metatarsals, occurred more often in those with a prior history of amputation or peripheral arterial disease, but the relative risk for canagliflozin was similar across these subgroups. Similar data have not yet been presented solely from EMPA-REG OUTCOME, which collected amputations via routine adverse event reporting, but pooled empagliflozin safety data show similar overall rates of amputation (1.8%) versus placebo (1.9%) and toe amputation occurred in 1.3% versus 0.9% of participants, respectively.³ A plausible biochemical mechanism for direct SGLT-2i effects on either bone or vascular biology has not been identified, but SGLT-2i-associated hemoconcentration or reduced blood pressure to already compromised vascular beds (leading to amputation) or orthostasis (leading to fall risk) have been hypothesized. However, a differential reduction in blood pressure or an increase in blood viscosity between canagliflozin and empagliflozin has not been demonstrated.

In summary, comparison between trials is always difficult and should be done with caution. Important differences between EMPA-REG OUTCOME and the CANVAS Program, other than in the specific treatment used, could at least partially explain differences in cardiovascular outcomes (eg, primary versus secondary prevention populations, perhaps with short-term treatment in the former). DECLARE, with a large primary prevention cohort, will shed further light on this issue. The differences in adverse events may be more difficult to explain, although the greater proportion of women in CANVAS could account for higher fracture rates, as could differences in SGLT-2 receptor selectivity. Likewise, it is difficult to define a unifying hypothesis for the imbalances in amputation rates. It would be surprising that 1 mechanism (vascular) could produce so clear a signal for a procedure (amputation) that represents the culmination of several distinct pathologies (eg, neuropathy, trauma, vascular disease, infection). Nevertheless, it would be inappropriate to dismiss the possibility of a difference between canagliflozin and empagliflozin, especially with further data pending for canagliflozin, dapagliflozin, and ertugliflozin. Each will add information about all-cause mortality, cardiovascular death, fractures, and amputation. Enhanced regulatory surveillance for amputation may also clarify whether an increased risk exists, show whether it differs between agents in the class, and give insight to an underlying mechanism.

Footnotes