Glycaemic studies: Food for thought and fuel for debate

Just over a decade ago, a controversial meta-analysis ¹ started a series of events that culminated in the launch of cardiovascular (CV) safety criteria for newly released hypoglycaemic agents (https://www.fda.gov). Some argued that these new rules added considerable expense to new agents for relatively little benefit other than reassurance. Indeed, a number of high-profile studies, including dipeptidyl peptidase inhibitors and one of the glucagon-like peptide (GLP)-1 analogues, showed CV non-inferiority but universally failed to demonstrate superiority. ² Moreover, these studies generated unexpected and unexplained results, such as increased heart failure with saxagliptin, creating uncertainty with some agents. However, a major breakthrough came in 2015 with the EMPA-REG results that showed improved CV outcome with empagliflozin, ³ followed less than 1 year later by the LEADER trial demonstrating vascular protection by liraglutide. ⁴

The CV theme of hypoglycaemic agents continued at the recent American Diabetes Association meeting in San Diego (June 2017). Results from three clinical trials were presented: the vascular effects of metformin in type 1 diabetes (T1DM; REMOVAL trial) and the CV safety of insulin degludec and canagliflozin in type 2 diabetes (T2DM; DEVOTE and CANVAS trials, respectively).

By CV trial standards, REMOVAL was a small study randomising patients with T1DM to metformin (n = 219) or placebo (n = 209) in addition to their insulin therapy. ⁵ The primary end point was change in mean carotid intima-media thickness (cIMT) at 3 years, used as a surrogate marker for future vascular events. Comparing metformin with the placebo arm, the study failed to reach the primary end point with a difference in mean cIMT of −0.005 mm/year (95% confidence interval (CI): −0.012, 0.002, p = 0.17), although a reduction in maximal cIMT, a tertiary end point, was documented [−0.013 mm/year (−0.024, −0.003); p = 0.009]. Disappointingly, the study failed to show an effect for metformin on HbA1c, but there was some good news in relation to a 1.7 kg weight reduction, a minor decrease in insulin dose by 2 U/day and a 0.2 mmol/L decrease in low-density lipoprotein (LDL)-cholesterol. Overall, metformin showed a favourable safety profile although discontinuation rates were double that of placebo due to gastrointestinal side effects.

Some will conclude that metformin failed to alter mean cIMT and therefore this agent has no role in vascular protection in T1DM. Others, however, will argue that the effects on weight and cholesterol, as well as maximal cIMT, may prove to be clinically beneficial in the long term. The reality is that we are trying too hard to ‘generalise’ one therapy in a relatively heterogeneous population with varied CV risk profile. Moreover, the study does not represent the typical T1DM population as HbA1c was relatively well controlled, blood pressure control was excellent and over 80% of patients were on statin therapy. This indicates that the trial recruited well-motivated patients, which may have affected the results and ‘prevented’ metformin from showing its true vascular benefits. In light of the data, a large CV outcome trial with metformin in T1DM is unlikely and even if this takes place it would be years before data became available. The clinical implications of study findings remain a matter of debate and perhaps it is best to have a pragmatic approach by individualising metformin therapy in T1DM according to CV risk, weight, insulin dose and, importantly, response to this therapy.

DEVOTE, a non-inferiority trial, investigated the CV safety of insulin degludec (n = 2818) compared with glargine (n = 2819) in high CV risk T2DM patients (more than 85% enrolled had established vascular disease). ⁶ Over a median of 1.99 years, degludec was similar to glargine in relation to three-point major adverse cardiac events [MACE; hazard ratio (HR): 0.91 (CI 0.78, 1.06)] with almost identical results for all-cause mortality [HR, 0.91 (0.76, 1.11)]. While the drop in HbA1c was identical in both groups (from 8.4% to 7.5%), severe hypoglycaemia occurred in 4.9% and 6.6% of degludec and glargine-treated patients, respectively [HR, 0.73 (0.6, 0.89)], and the difference was even more pronounced in the presence of renal impairment. There was no difference between study arms in relation to biochemical/haematological parameters or incidence of neoplasms.

The simple conclusion from this work is that degludec can be ‘safely’ used in T2DM patients at high CV risk. However, the significant reduction in severe hypoglycaemia (including 53% reduction in severe nocturnal hypoglycaemia) should not go unnoticed as this has both clinical and health economic implications. Until guidelines are updated, physicians should consider using degludec in insulin-treated T2DM patients with established CV disease, particularly in the presence of other risk factors for hypoglycaemia such as renal impairment. One obstacle to this approach is the associated expense that rules out routine use of this agent in some countries.

The CANVAS programme evaluated the CV safety of canagliflozin. This study recruited 10,142 patients with T2DM and represented the integration of two studies: CANVAS and CANVAS-R. ⁷ The primary end point [the composite of CV death, non-fatal myocardial infarction (MI) and non-fatal stroke] was reduced by 14% [HR, 0.86 (0.75, 0.96)] in canagliflozin-treated patients demonstrating both non-inferiority (p < 0.001) and superiority (p = 0.016). When each component of the primary end point was analysed separately, none was significant although all showed ‘numerical’ benefit in favour of canagliflozin, including stroke. The study also revealed a 40% reduction in adverse renal outcome and a 33% reduction in hospitalisation for heart failure. However, there was a twofold increase in amputations in the canagliflozin group compared with placebo, including minor [toes, metatarsals; HR, 1.94 (1.31–2.88)] and major [ankle, below and above knee; HR, 2.03 (1.08–3.82)] amputations, independent of a previous history of amputation. In both study arms, a threefold increase in amputation rates was noted in those with a history of peripheral vascular disease (PVD). Paradoxically, however, active treatment had no significant effect on amputations in this subgroup compared with placebo [HR, 1.39 (0.80, 2.40)] and the adverse outcome with canagliflozin was limited to those without a history of PVD [HR, 2.34 (1.53–3.58)]. The mechanisms behind increased amputations remain unclear and warrant investigation before advocating further restrictions in canagliflozin use. The results may simply represent a chance finding (particularly as the difference was not significant in those with PVD) or it may be due to a complex clinical scenario in the presence of co-morbidities. As expected, attempts were made at comparing the results of the CANVAS programme with the highly impressive results of EMPA-REG trial. ³ It can be rightly argued that this approach is scientifically flawed for a number of reasons including the different population of patients studied (EMPA-REG had higher CV risk patients) and differences in study design. Ignoring these reservations, the primary end point was remarkably similar in the two studies although analysis of the three components demonstrated differences in non-fatal stroke in favour of canagliflozin, whereas empagliflozin had more pronounced effects on CV death. Heart failure and renal data also showed similarities while amputation rates were not formally analysed in EMPA-REG.

A highly controversial meta-analysis transformed diabetes research and clinical practice by establishing CV safety outcome trials for new hypoglycaemic agents. These trials are providing an unprecedented wealth of data, creating in the process a fertile ground for new ideas and scientific debates. One criticism of these outcome studies is the lack of a clear hypothesis (unless one regards CV non-inferiority as a hypothesis!) and therefore mechanisms are not investigated. Perhaps the way forward is to combine outcome studies with mechanistic work designed to explain the positive findings and also explore reasons for unexpected results. This in turn will help to target the relevant pathway(s) and refine our treatment strategies to further improve vascular outcome in patients with diabetes.