Sodium-Glucose Cotransporter-2 Inhibitors and the Risk of Abnormal Serum Potassium Level

Sodium-glucose cotransporter-2 inhibitors (SGLT2i) impede glucose reabsorption in kidney proximal tubules, which in turn can stimulate osmotic diuresis, natriuresis, and the renin-angiotensin system. This mechanism may enhance potassium secretion via increased distal sodium and water delivery and renin-angiotensin system activation. However, the US Food and Drug Administration’s prescription label for canagliflozin describes a higher risk of hyperkalemia in patients with an eGFR of 45–60 ml/min per 1.73 m² (1). This was on the basis of early studies, including interim data from the CANVAS. Contrastingly, in two randomized trials (completed CANVAS and Canagliflozin and Renal Events in Diabetes with Established Nephropathy Clinical Evaluation [CREDENCE]), the rates of hyperkalemia did not significantly differ between participants randomized to canagliflozin and placebo (2,3). Hyperkalemia has not been demonstrated with other medications in the class (4). We conducted a population-based study to evaluate the risk of both hyperkalemia and hypokalemia associated with SGLT2i, using alternate antidiabetics as comparators.

We used integrated tertiary health system data from Geisinger (central and northeast Pennsylvania) to accrue adults aged ≥18 years with diabetes who received the first outpatient prescription for SGLT2i, dipeptidyl peptidase-4 inhibitor (DPP4i), or glucagon-like peptide-1 receptor agonist (GLP1RA) between January 1, 2015 and December 31, 2018. We included patients with (1) ≥1 hemoglobin A1c, serum creatinine, and potassium measurements in the year before drug prescription, and (2) ≥1 potassium measurement in the year after the prescription. We excluded those with ESKD. We used multiple imputation by chained equation to impute missing data and performed 1:1 propensity score matching across 30 imputed datasets for a valid comparison of SGLT2i versus DPP4i and SGLT2i versus GLP1RA. Hyperkalemia and hypokalemia were defined as serum potassium >5.5 mEq/L and <3.5 mEq/L within 1 year of drug initiation, respectively. We estimated odds ratios (ORs) associated with SGLT2i from each imputed dataset using logistic regression and pooled the ORs using Rubin’s rules (5). The study protocol was reviewed and approved by the institutional review boards of Geisinger Medical Center and Johns Hopkins University.

The median size of the propensity-matched cohorts of SGLT2i (canagliflozin [n=435, 26%], dapagliflozin [n=67, 4%] and empagliflozin [n=1170, 70%]) and DPP4i recipients was 3344. Mean age was 58.4 years, 46% were women, 92% were White, and mean eGFR was 87 ml/min per 1.73 m² (12% with eGFR <60 ml/min per 1.73 m²). The median size of the propensity-matched cohorts of SGLT2i and GLP1RA recipients was 2526; these cohorts had slightly younger mean age (56.5 years), more women (52%), and higher proportion of eGFR <60 ml/min per 1.73 m² (13%). Both cohorts had balanced baseline demographics; Charlson comorbidity index; history of CKD, acute kidney injury, hyperkalemia, and hypokalemia; eGFR; serum potassium; hemoglobin A1c; concomitant medications including insulin, other oral antidiabetics, antihypertensives, diuretics, and potassium supplements (all standardized differences <10%).

There was no difference in 1-year risk of hyperkalemia between SGLT2i and DPP4i (2% versus 3%; OR, 0.76; 95% confidence interval [95% CI], 0.47 to 1.23) or between SGLT2i and GLP1RA (2% versus 3%; OR, 0.60; 95% CI, 0.36 to 1.01) groups (Table 1). Similarly, there was no difference in 1-year risk of hypokalemia between SGLT2i and DPP4i (6% versus 7%; OR, 0.92; 95% CI, 0.67 to 1.27) or between SGLT2i and GLP1RA (6% versus 7%; OR, 0.95; 95% CI, 0.68 to 1.33) groups. The results for both hyperkalemia and hypokalemia were consistent when stratified by eGFR (<60 ml/min per 1.73 m² versus ≥60 ml/min per 1.73 m²) and coprescription of nonpotassium-sparing diuretics (yes versus no).

Table 1.

Risk of hyperkalemia and hypokalemia with initiation of sodium-glucose cotransporter-2 inhibitors versus dipeptidyl peptidase-4 inhibitor or glucagon-like peptide-1 receptor agonist in the propensity-matched analysis

Outcome	1:1 Matched Cohort of Sodium-Glucose Cotransporter-2 Inhibitors versus Dipeptidyl Peptidase-4 Inhibitor Recipientsa				1:1 Matched Cohort of Sodium-Glucose Cotransporter-2 Inhibitors versus Glucagon-like Peptide-1 Receptor Agonist Recipientsb
	Events, n (%)		Adjusted Odds Ratio (95% Confidence Interval)	P for Interaction	Events, n (%)		Adjusted Odds Ratio (95% Confidence Interval)	P for Interaction
	Sodium-Glucose Cotransporter-2 Inhibitors	Dipeptidyl Peptidase-4 Inhibitor			Sodium-Glucose Cotransporter-2 Inhibitors	Glucagon-like Peptide-1 Receptor Agonist
	(n=1672)	(n=1672)			(n=1263)	(n=1263)
Hyperkalemia c
Overall	33 (2.0)	45 (2.7)	0.76 (0.47 to 1.23)	–	30 (2.4)	42 (3.3)	0.60 (0.36 to 1.01)	–
Subgroups by eGFR, ml/min per 1.73 m2
≥60	25 (1.7)	30 (2.1)	0.85 (0.48 to 1.51)	0.63	21 (1.9)	33 (3.0)	0.58 (0.32 to 1.05)	0.45
<60	8 (4.0)	15 (7.0)	0.57 (0.21 to 1.49)		9 (5.2)	9 (5.4)	0.68 (0.25 to 1.82)
Subgroups by diuretic coprescription
No	25 (2.3)	29 (2.6)	0.95 (0.54 to 1.68)	0.28	21 (2.5)	27 (3.2)	0.69 (0.37 to 1.29)	0.42
Yes	8 (1.5)	16 (3.0)	0.50 (0.19 to 1.34)		9 (2.1)	15 (3.5)	0.44 (0.17 to 1.12)
Hypokalemia d
Overall	96 (5.8)	118 (7.1)	0.92 (0.67 to 1.27)	–	74 (5.9)	86 (6.8)	0.95 (0.68 to 1.33)	–
Subgroups by eGFR, ml/min per 1.73 m2
≥60	80 (5.5)	92 (6.3)	0.98 (0.70 to 1.39)	0.76	62 (5.7)	77 (7.0)	0.91 (0.63 to 1.31)	0.30
<60	16 (8.0)	26 (12.1)	0.69 (0.32 to 1.47)		12 (6.9)	9 (5.4)	1.24 (0.51 to 3.04)
Subgroups by diuretic coprescription
No	53 (4.6)	59 (5.2)	0.95 (0.63 to 1.45)	0.59	42 (5.0)	39 (4.7)	1.08 (0.67 to 1.73)	0.47
Yes	43 (8.2)	59 (11.1)	0.80 (0.50 to 1.29)		32 (7.5)	47 (11.0)	0.83 (0.50 to 1.38)

^aSubgroup analysis by (1) eGFR included 2944 individuals with eGFR ≥60 ml/min per 1.73 m² and 400 individuals with eGFR <60 ml/min per 1.73 m², and (2) diuretic coprescription included 2311 individuals without concomitant prescription and 1033 individuals with concomitant prescription.

^bSubgroup analysis by (1) eGFR included 2185 individuals with eGFR ≥60 ml/min per 1.73 m² and 341 individuals with eGFR <60 ml/min per 1.73 m², and (2) diuretic coprescription included 1673 individuals without concomitant prescription and 853 individuals with concomitant prescription.

^cHyperkalemia was defined as serum potassium >5.5 mEq/L.

^dHypokalemia was defined as serum potassium <3.5 mEq/L.

In our cohort study, SGLT2i initiation was not associated with the risk of hyperkalemia or hypokalemia compared with DPP4i or GLP1RA initiation. The absence of association was consistent in the subgroups with eGFR <60 ml/min per 1.73 m² and with coprescription of nonpotassium-sparing diuretics. Our findings are congruent with those from prior randomized studies of SGLT2i and support the absence of significant risk of hyperkalemia with the drug class in the real-world setting (2–4). Interestingly, we also did not find a significant risk of hypokalemia with SGLT2i, contrary to what might be expected from its mechanism of action.

To our knowledge, this was the first population-based study that examined the association between SGLT2i and abnormal serum potassium levels. Observational studies provide real-world evidence for the safety of new medications, informing postmarketing surveillance. However, one must recognize patients may receive specific medications for reasons related to the outcome of interest. To minimize “treatment-by-indication” bias, we used active comparators and applied rigorous methods including propensity-score matching and multiple imputation. In our study, drug exposures were ascertained by prescription records and thus actual adherence could not be confirmed. Moreover, we performed intention-to-treat analysis and thus the risk estimates were on the basis of initial prescribing decision. Finally, the study was conducted in a geographic region with a predominantly White population. There were relatively few patients with an eGFR <60 ml/min per 1.73 m². Our study had a 1-year follow-up, and thus longer-term changes in serum potassium levels could not be assessed.

In conclusion, we did not find an association between SGLT2i and the risk of hyperkalemia or hypokalemia, including among adults with an eGFR <60 ml/min per 1.73 m². These findings provide additional supportive evidence for the safety of SGLT2i.

Disclosures

A.R. Chang reports employment with Geisinger Health System; consultancy agreements with Novartis; reports receiving research funding for a Novo Nordisk Investigator Sponsored Study; reports receiving honoraria from Reata; reports serving as a scientific advisor or member of Reata and Relypsa; and has other interests/relationships with National Kidney Foundation (NKF) grant support from the NKF Patient Network. J.-I. Shin reports receiving research funding from Merck and the National Institutes of Health. L.A. Inker reports consultancy agreements with Diamtrix and Tricidia; reports receiving research funding from the National Institutes of Health, NKF, Omeros, Reata, and Retrophin; reports serving as a scientific advisor or member of Alport’s Foundation, Diametrix, and Goldfinch; and reports serving as a member of the National Kidney Disease Education Program, AKF, and ASN. M.E. Grams reports receiving honoraria from academic institutions for giving grand rounds and ASN for Young Investigator Award; reports serving as a scientific advisor or member of American Journal of Kidney Diseases, CJASN, JASN Editorial Fellowship Committee, NKF Scientific Advisory Board, Kidney Disease: Improving Global Outcomes (KDIGO) Executive Committee, and United States Renal Data System Scientific Advisory Board; reports other interests/relationships with the NKF, which receives funding from Abbvie, Relypsa, and Thrasos; and reports receiving travel support from Dialysis Clinic, Inc. to speak at the annual meeting and KDIGO for participation in scientific meetings and the executive committee. The remaining authors have nothing to disclose.

Funding

This work was supported by the National Institute of Diabetes and Digestive and Kidney Disease grants R01DK115534 to M.E. Grams (Principal Investigator) and K01DK121825 to J.-I. Shin (Principal Investigator).