Canagliflozin, serum magnesium and cardiovascular outcomes—Analysis from the CANVAS Program

Abstract Background Patients with type 2 diabetes (T2D) are predisposed to derangements in serum Magnesium (Mg), which may have implications for cardiometabolic events and outcomes. In clinical trials, participants with T2D randomized to sodium‐glucose co‐transporter 2 (SGLT2) inhibitors have shown mild to moderate increases in serum Mg from baseline levels. This post hoc analysis assesses the relation between serum Mg with cardiovascular outcomes in 10,140 participants of the Canagliflozin Cardiovascular Assessment Study (CANVAS) Program. Methods We evaluated the association of baseline serum Mg with the primary composite end point of death from cardiovascular causes, non‐fatal myocardial infarction, and non‐fatal stroke, and tested whether this association is modified by baseline serum Mg. Using mediation analysis, we determined whether change in serum Mg post‐randomization mediates the beneficial effect of canagliflozin on cardiovascular outcomes. Results Mean serum Mg levels at baseline were 0.77 ± 0.09 mmol/L in both canagliflozin group and placebo groups. The canagliflozin group experienced an average increase in serum Mg by 0.07 mmol/L (95% CI, 0.065–0.072 mmol/L; p < .001) for the duration of the trial. We found no association between baseline serum Mg levels and the primary composite end point, and no evidence of effect modification by baseline Mg levels. Change in serum Mg post‐randomization was not a mediator of the effects of canagliflozin on cardiovascular outcomes. Conclusions In participants of the CANVAS Program, baseline and post‐randomization serum Mg levels are not associated with cardiovascular outcomes.


| INTRODUC TI ON
Magnesium (Mg) is the second most abundant intracellular cation in the body and serves as a cofactor for over 300 enzymatic reactions. 1 Mg modulates vascular tone and cardiac rhythm and plays an important role in maintaining cardiovascular health. 2 Overall Mg balance depends on an interplay between intestinal absorption, bone exchange, and faecal and urinary excretion, with the kidney serving as the primary regulator of serum Mg homoeostasis. 2 Abnormalities in serum Mg levels, especially hypomagnesaemia, have been implicated in various cardiometabolic outcomes. 3 In mouse models, hypomagnesaemia leads to cardiac and renal inflammation and fibrosis. 4 In humans, hypomagnesaemia has been associated with atrial fibrillation, 5 coronary heart disease, 6 cardiovascular death 6,7 including sudden cardiac death, 7,8 and all-cause mortality. 9 Patients with type 2 diabetes (T2D) are particularly predisposed to hypomagnesaemia due to poor intestinal absorption and Mg wasting through the kidneys. 10 Canagliflozin is a sodium-glucose co-transporter 2 (SGLT2) inhibitor, which lowers serum glucose by blocking renal glucose reabsorption in the proximal tubule to induce glucosuria. Beyond improvement of glycaemic control and cardiovascular and kidney outcomes, these agents have also been shown to increase serum Mg levels in participants with T2D. [11][12][13][14][15][16] The Canagliflozin Cardiovascular Assessment Study (CANVAS) Program found that participants randomized to canagliflozin versus placebo showed a 14% risk reduction in major adverse cardiovascular events (MACE), which included cardiovascular death, non-fatal myocardial infarction and non-fatal stroke. 12 This post hoc analysis leverages data from the CANVAS program to determine whether baseline serum Mg levels associates with or modifies the effect of canagliflozin, on cardiovascular outcomes. In addition, given prior data that SGLT2 inhibitors increase serum Mg levels and the association between serum Mg and cardiovascular outcomes, we performed a mediation analysis to assess whether a post-randomization change in serum Mg mechanistically plays a role in the attenuation of cardiovascular risk with canagliflozin versus placebo.

| Study design
The CANVAS Program pooled data from CANVAS and CANVAS-R trials, which together randomized 10,142 participants, ages ≥30 years with T2D at high risk for cardiovascular events, to canagliflozin versus placebo. Participants in CANVAS were assigned to canagliflozin 100 mg, canagliflozin 300 mg and placebo in a 1:1:1 fashion, while participants in CANVAS-R were assigned to canagliflozin at an initial dose of 100 mg with option to increase to 300 mg at week 13 (of which 71.4% of participants in the canagliflozin did so), versus placebo. Mean follow-up time was 188 weeks. Major exclusion criteria included type 1 diabetes, not being on a stable antihyperglycaemic regimen for at least 8 weeks, cardiovascular events within 3 months of screening or planned revascularization, and eGFR <30 ml/min/1.73 m 2 . Further details regarding study design, participant inclusion and exclusion criteria, and main results can be found in the original manuscript. 12

| Exposure: Serum Mg
Serum Mg was collected as part of the Serum Chemistry Panel, and measured at baseline, post-randomization weeks 6, 13, 18, 26, 39, 52, and then at 26-week intervals per protocol. Testing was performed by a central laboratory, with exception of local laboratory processing in situations where immediate testing was necessary for clinical care.
This analysis includes only participants who had a baseline and postbaseline Mg measurement up to individual trial completion. We used a SAS ® macro to determine whether the exposure should be modelled as continuous/linear, categorical, quadratic or spline. Baseline serum Mg is modelled as a categorical variable, in quintiles, based on optimal model fit and interpretability of the results. 17

| Outcomes
The primary outcome for the present analysis was MACE, a composite of cardiovascular death, non-fatal myocardial infarction and the primary composite end point, and no evidence of effect modification by baseline Mg levels. Change in serum Mg post-randomization was not a mediator of the effects of canagliflozin on cardiovascular outcomes.

Conclusions:
In participants of the CANVAS Program, baseline and post-randomization serum Mg levels are not associated with cardiovascular outcomes.

K E Y W O R D S
canagliflozin, cardiovascular outcomes, magnesium, type 2 diabetes TA B L E 1 Baseline characteristics of the study cohort overall and stratified by quintile of baseline serum Mg

| Statistical analyses
The mean change from baseline Mg level by treatment group was determined using a mixed-model repeated-measures analysis. A Cox proportional hazards model was used to examine the association of baseline Mg quintile with cardiovascular outcomes, with the third quintile serving as the reference group. Results are reported as hazard ratios (HR) and 95% confidence intervals (CI).

| Baseline serum Mg and cardiovascular outcomes
We found no significant association between baseline quintile of serum Mg and MACE, cardiovascular death, sudden cardiac death, or heart failure in adjusted models ( Table 2). Event rates were low for cardiovascular death, sudden cardiac death and heart failure.
Unadjusted hazard ratios between baseline quintile of serum Mg and cardiovascular outcomes can be found in Table S1. We also found no evidence that baseline serum Mg modified the beneficial effect of canagliflozin versus placebo on MACE (Figure 2, p = .88).

| Change in serum Mg as a mediator
An exploratory mediation analysis was performed to examine pathways whereby canagliflozin leads to risk reduction in MACE.
Mediation analysis seeks to disentangle the total effect of the expo-

| DISCUSS ION
This post hoc analysis of the CANVAS program included 10,140 participants, allowing for a robust analysis of the treatment effect of TA B L E 2 Association between baseline serum Mg quintile and risk of cardiovascular outcomes in adjusted Cox models An inherent limitation of any post hoc analysis includes the possibility that associations found in the study may be spurious; all analyses are exploratory and require further clinical investigation.
We also note limitations specific to this study. Due to incomplete medication information for all participants, there was inability to adjust for specific agents that may decrease serum Mg (eg thiazidetype diuretics, proton pump inhibitors) or medications that may increase serum Mg (eg mineralocorticoid antagonists, amiloride).
There was no information on participants' dietary Mg intake or alcohol use, the latter of which is associated with Mg depletion due to low intake, low body stores and renal wasting. 37, 38 We also did not have information on urinary levels of Mg to determine whether participants experienced changes in fractional excretion of Mg. A final consideration is that serum Mg may not be the most accurate measurement of total body stores. Over 99% of total body Mg is stored in bone, muscles and soft tissue, with <1% contained in the serum and red blood cells. 2 Normal serum Mg levels may be seen in patients with Mg deficiency, typically due to recruitment of intracellular stores (rapidly through muscle stores, and over weeks through bone stores). [39][40][41] However, serum Mg measurements remain the most commonly used test by clinicians due to relative ease and cost. 41

| CON CLUS ION
In the CANVAS Program, participants randomized to canagliflozin showed a mean increase in serum Mg levels during the course of the study; however, there was no association between baseline serum Mg levels and cardiovascular outcomes, and the beneficial effect of canagliflozin does not vary by baseline serum Mg. Changes from baseline serum Mg levels in the canagliflozin group during the trial did not mediate the beneficial effects of canagliflozin on cardiovascular outcomes. Given that this investigation is exploratory, further physiologic or mechanistic studies specifically designed to elucidate the relationship between serum Mg and cardiovascular outcomes may be beneficial.

ACK N OWLED G EM ENTS
We thank the participants, investigators and all who contributed to the successful completion of the CANVAS Program trial.