Can we go beyond surrogates?

Two years ago, data presented at the annual American Diabetes Association (ADA) meeting in New Orleans showed a marked decrease in deaths, especially those due to cardiovascular disease, with the use of empagliflozin. Two major questions have been asked: (i) was the result a fluke; and (ii) was it a class effect, or was it specific to the agent used? The hope that both questions would be answered by a second study has been answered: the conclusions of EMPA-REG were not an anomaly and it is a class effect, not one caused by a specific drug. Importantly, do these studies require us to alter our algorithms for the treatment of type 2 diabetes? The CANagliflozin cardioVascular Assessment Study (CANVAS) was designed similarly to EMPA-REG, enrolling individuals who either had known cardiac disease or were at high risk for cardiac disease. In fact, CANVAS involved two stages, CANVAS and CANVAS R, which can be analyzed together. There were 10 142 patients in the combined trial followed for a mean of 3.6 years. The average age was 63.3 years, 35.8% were women, the mean duration of diabetes was 13.5 years, and 65.6% had known cardiovascular disease. In EMPA-REG, 7020 patients were followed for 3.1 years, 28.5% were women, and all had established cardiovascular disease. The primary endpoint of both studies was the composite of cardiovascular death, non-fatal myocardial infarction, and non-fatal stroke. In both studies, the components of the composite endpoint were studied along with other endpoints, including hospitalization for heart failure, all-cause mortality, and progression of renal disease. The primary endpoint was identically positive for superiority with hazard ratio of 0.86 for empagliflozin versus placebo and for canagliflozin versus placebo. Some differences were observed: both death from cardiovascular disease and all-cause mortality occurred significantly less frequently with empagliflozin in EMPA-REG, but the 13% lower total and cardiovascular mortality seen with canagliflozin in CANVAS was not statistically significant. Both studies showed a lower likelihood of hospitalization for heart failure with the intervention, even in patients without known heart failure at baseline, suggesting either that that diagnosis is often missed or that de novo heart failure may be prevented by these agents. Perhaps the strongest answer to the initial two questions comes from the CVD-REAL Study. Rather than being a randomized controlled clinical trial, that study analyzed medical claims, primary care or hospital records, and national registries from the US, Norway, Denmark, Sweden, Germany, and the UK, including 309 056 patients started on a sodium-glucose cotransporter 2 (SGLT2) inhibitor or on other glucose-lowering drugs using a propensity score matching algorithm. Canagliflozin was used by 53% of those receiving SGLT2 inhibitors, dapagliflozin was used by 42% and empagliflozin was used by 5%; those using these agents had hazard ratios for hospitalization for heart failure and for mortality of 0.61 and 0.49, respectively, with no differences either by country or by drug. The interesting hypothesis that this and the two controlled trials show the harm of alternative treatments of diabetes rather than the benefit of SGLT2 inhibitors has not been confirmed by subset analyses of drug use in the control groups. There are several potentially important differences between the studies. Strokes increased with empagliflozin in EMPA-REG (not statistically significant). There was a significantly increased likelihood of lower extremity amputation in CANVAS, although these events may not have been fully studied in EMPA-REG. Finally, the benefit of SGLT2 inhibitor therapy were seen within 3 months in EMPA-REG, while not appearing as rapidly in CANVAS. Furthermore, in neither study was there a reduction in non-fatal myocardial infarction. These considerations suggest that the benefit of SGLT2 inhibitors is not due to improvement in the underlying atherosclerotic process. It is likely that all SGLT2 inhibitors will receive an indication for secondary prevention of heart failure; whether the agents should be used in primary prevention is a much more difficult question, because it would require a very large study of patients without heart disease. The introduction of HbA1c measurements in the late 1970s radically and significantly improved our treatment of diabetes. However, our treatment of diabetes is ultimately not to lower HbA1c, but to prevent the complications of diabetes. Values of HbA1c are only a surrogate measure of diabetes. Currently, there is little agreement on the treatment algorithm for diabetes after metformin therapy. One may now argue that an SGLT2 inhibitor should be the preferred second-line therapy.