Discussion
In this post hoc analysis of the CREDENCE study, we demonstrated that the glycemic effects of canagliflozin were modified by baseline eGFR, with a progressively lesser reduction and lower likelihood of achieving a >0.5% decrease in HbA1c across CKD stages G1-2 (eGFR >60 mL/min/1.73 m2), stage G3A (eGFR 45–60 mL/min/1.73 m2), and stage 3B (eGFR 30–44 mL/min/1.73 m2). Looking at the impact of HbA1c reduction on end points during the trial, there was no evidence that adjustment for the postbaseline change in HbA1c reduction (assessed at week 13) led to an attenuation in the effect of canagliflozin on the primary composite outcome or on secondary kidney and cardiovascular outcomes. Similar results were observed when we used a time-varying adjustment that took into account the level of glycemic control throughout the study period.
Glycemia-independent effects of SGLT2 inhibition have been previously inferred from the study of cardiovascular and kidney benefits of canagliflozin that are observed despite a modest lowering of HbA1c when kidney function is impaired. Indeed, the mean decrease of 0.16% in HbA1c that we observed with canagliflozin compared with placebo in this analysis is considerably smaller than that seen in either short-term (approximately 0.7%–1.0%) or longer outcome trials (0.4%–0.6%).16–20 We have already reported that baseline HbA1c level did not predict the impact of canagliflozin on study outcomes in the CREDENCE trial.4 The current, analyses extend these findings by demonstrating that neither the initial HbA1c, the extent of initial reduction in HbA1c at 13 weeks, nor the time-averaged HbA1c during the whole study period, influenced the clinical benefits associated with randomization to canagliflozin as compared with placebo. Furthermore, reduction in cardiovascular and kidney risks were robust regardless of background diabetes therapies. These results therefore support the hypothesis that the cardiovascular and renoprotective effects of canagliflozin in this population of patients with both type 2 diabetes and CKD are largely independent of changes in glycemic status and suggest that non-glycemic mechanisms are operative. This analysis is consistent with findings from the CANagliflozin cardioVascular Assessment Study in which markers of glycemic control did not explain kidney outcomes in a mediation analysis, while albuminuria, hemoglobin, and hematocrit did, suggesting that anti-inflammatory effects, antifibrotic effects, effect on volume status, improvements in renal hypoxia, or other mechanisms of action are more important in explaining clinical benefits.21 This conclusion is supported by a recent prespecified analysis of the DAPA-CKD study, which recruited 4304 participants with CKD of whom 738 had normoglycemia, 660 had prediabetes, and 2906 had type 2 diabetes at baseline. The effects of dapagliflozin on the primary outcome (a composite of doubling of serum creatinine, end-stage kidney disease, or death due to a renal or cardiovascular cause) were consistent in those with normoglycemia (HR 0.62 (95% CI 0.39 to 1.01)), prediabetes (HR 0.37 (0.21 to 0.66)), and type 2 diabetes (HR 0.64 (0.52 to 0.79)) with no effect modification on other outcomes (including all-cause mortality) when adjusted for baseline glycemic status.22 This analysis, however, did not examine the impact of the level of glycemic control during the study on these clinical outcomes, although there were no differences in HbA1c between treatments arms in participants with normoglycemia or prediabetes at baseline. Trials of SGLT2 inhibitors in patients with heart failure have also indicated that the cardiovascular benefits of these drugs, particularly on hospitalization for worsening heart failure, are also observed in participants with heart failure and reduced ejection fraction who do not have type 2 diabetes at the time of randomization.23 24
Our analysis of the CREDENCE dataset thus supports the hypothesis that non-glycemic mechanisms explain many of the benefits of canagliflozin, similar to other SGLT2 inhibitors, on cardiorenal outcomes. A number of alternative mechanisms have been proposed. For example, metabolic changes that result from glycosuria (even in the absence of HbA1c reductions) may activate gluconeogenesis, ketogenesis and fatty acid oxidation, promoting autophagy (a process that allows cells to clear dysfunctional organelles). The consequences of autophagy could include reductions in oxidative stress which may in turn protect residual kidney (or myocardial) function.25 Other pathways proposed to explain kidney protection include suppression of inflammation and fibrosis, mediated via inhibition of the renin-angiotensin-aldosterone system with consequent reductions in ischemia in the kidney.26 27 Non-metabolic mechanisms involving changes in osmotic and non-osmotic sodium handling and intrarenal hemodynamic have also been proposed to explain the cardiovascular benefits of SGLT2 inhibitors.28 Although the precise mechanism of clinical benefit have not been fully elucidated, the evolving clinical data support the use of SGLT2 inhibitors to improve cardiorenal end points in patients with both CKD and chronic heart disease.
There are a number of limitations to this post hoc analysis of the CREDENCE study. First, postrandomization data were incomplete—HbA1c was available in most (4198) participants at week 13 with a baseline HbA1c, and in 4125 participants at week 26 but only in 3990 at week 52, thus limiting our mediation analysis. Second, the CREDENCE study recruited patients with an eGFR >30 mL/min/1.732 and although 174 patients with an eGFR below this level were randomized (because of reductions in kidney function between the date of screen and randomization),29 patients with stage 4 CKD were not well represented. Third, although our results support the hypothesis that the primary benefits of SGLT2 inhibition are derived from mechanisms other than improvement in glycemia, our study cannot prove causality since the decrease in HbA1c is a postrandomization variable with potential confounders and biases, some of which may not be adjusted for with the analytic approaches used. Lastly, variability in the precision of HbA1c measurement and the imperfect correlation of HbA1c with glucose lowering within individuals could have attenuated our power to detect mediation of the effects on clinical outcomes by the glucose-lowering effects of canagliflozin.
In conclusion, this post hoc analysis of the CREDENCE study demonstrates that in the context of CKD, although the glycemic effects of canagliflozin are attenuated at lower levels of eGFR, the clinical benefits on kidney and cardiovascular outcomes are preserved. Good glycemic control or a lack of reduction in HbA1c after starting canagliflozin in a patient with type 2 diabetes and CKD should not discourage ongoing use of the drug. Patients with good glycemic control and those who do not show HbA1c reductions on canagliflozin are still likely to gain renal and cardiovascular benefits from the medication.