SGLT2 inhibitors and the risk of diabetic ketoacidosis among adults with Type 2 Diabetes: A systematic review and meta-analysis

INTRODUCTION

Sodium-glucose co-transporter 2 (SGLT2) inhibitors were approved in 2013 for the treatment of adults with type 2 diabetes mellitus.¹ Unlike most other diabetes medications, they have consistently been shown to reduce the risk of myocardial infarction, cardiovascular mortality, heart failure, and renal failure.^2–5 SGLT2 inhibitors are recommended as a second-line agent after metformin by most international diabetes associations.^6,7 The recently published DAPA-HF (Dapagliflozin in Patients with Heart Failure and Reduced Ejection Fraction) trial demonstrated a lower risk of cardiovascular death or hospitalization for heart failure among patients with and without type 2 diabetes who were randomized to an SGLT2 inhibitor as compared with placebo.⁸ Thus, SGLT2 inhibitors may soon be indicated for a wider patient population, which highlights the importance of identifying and quantifying the risk of rare adverse events.⁹

Diabetic ketoacidosis (DKA) is a potential rare adverse event associated with SGLT2 inhibitors in adults with type 2 diabetes.¹⁰ An increased risk of diabetic ketoacidosis was not detected until widespread use in routine care.^11,12 These findings were surprising since DKA is typically a complication of type 1 diabetes mellitus rather than type 2 diabetes,¹³ and is usually characterized by marked hyperglycemia and acidemia due to profound insulin deficiency.¹³ By contrast, initial case reports of DKA among patients who received an SGLT2 inhibitor, typically occurred in the absence of severe hyperglycemia (“euglycemic” diabetic ketoacidosis, blood glucose <200 mg/dL).¹⁰ As case reports accumulated, the US Food and Drug Administration (FDA) released a warning in May 2015 and updated the drug label in December 2015 to list DKA as a potential side-effect of SGLT2 inhibitors.¹⁴

Since this FDA warning, several observational studies have identified an increased incidence of DKA with SGLT2 inhibitors compared to other diabetes medications when used in routine care.^9,15 However, three systematic reviews and meta-analyses of randomized controlled trials (RCTs) found that SGLT2 inhibitors were not associated with an increased risk of DKA.^16–18 There are several reasons why these meta-analyses may not have identified an increased risk of DKA. First, patients included in RCTs are typically healthier than those in routine care.¹⁹ Second, RCTs are not powered to detect rare adverse events.^19,20 Third, these meta-analyses do not include the most recently completed RCTs of SGLT2 inhibitors. Accordingly, we conducted a systematic review and meta-analysis of large clinical trials and observational studies to assess the risk of DKA with SGLT2 inhibitors.

METHODS

RESULTS

We identified 21,715 potential articles, and 12 studies met our inclusion criteria (Figure 1). These consisted of 7 RCTs (23,218 patients on SGLT2 inhibitors and 19,057 on a comparator medication or placebo, Table 1A), with 73 DKA events. Likewise, there were 5 cohort studies (159,318 patients on SGLT2 inhibitors and 159,318 patients on a comparator medication, Table 1B) with 262 DKA events.

• Download figure
• Open in new tab

Figure 1:

Flow diagram of included studies

Among the RCTs, the average age of patients randomized to an SGLT2 inhibitor ranged from 55.4 years to 63.9 years, and the percentage of female patients ranged from 28.8% to 53.9% (Table 1A). All RCTs were international studies and industry funded.^{2–5,28–30} The SGLT2 inhibitor used was canagliflozin in three studies, empagliflozin in three studies and dapagliflozin in one study. ^{2–5,28–30} The comparator was placebo in five studies, glimepiride in one study and placebo/sitagliptin in one study.^{2–5,28–30} All RCTs identified diabetic ketoacidosis through clinical adjudication.

In the observational studies, the average age of participants ranged from 53.1 years to 64.0 years, and the percentage of female patients ranged from 39.0% to 47.3% (Table 1B). Three observational studies were conducted in the US, one in South Korea, and one in Scandinavia.^9,15,31–33 Two of the observational studies were industry funded.^32,33 The SGLT2 inhibitor used was empagliflozin in one study, any SGLT2 inhibitor in two studies and unspecified in two studies.^9,15,31–33 For three studies the comparator medication was a DPP4 inhibitor, one study used GLP1 agonists and one used any diabetes medication.^9,15,31–33 All studies used propensity-score matching and identified diabetic ketoacidosis using validated diagnostic codes (ICD-9 or ICD-10).²⁴

Risk of Diabetic Ketoacidosis Reported in Randomized Trials

Four studies had a DKA event in both the SGLT2 inhibitor and comparator group and were included in the meta-analysis. Of these studies, the duration of follow up ranged from 2.6 to 4.2 years. The absolute rate of DKA among patients randomized to an SGLT2 inhibitor ranged from 0.6 to 2.2 events per 1000 person years, compared to 0.1 to 0.3 per 1000 person years among patients in the control group. This corresponded to a 2.4-fold higher risk of DKA for people receiving an SGLT2 inhibitor compared with placebo/comparator medication (RR⍰=⍰2.46 [95% CI, 1.16-5.21]; I²⍰=⍰0%; P⍰= 0.54, Figure 2).

• Download figure
• Open in new tab

Figure 2.

Forest plot of diabetic ketoacidosis risk with sodium-glucose cotransporter-2 (SGLT2) inhibitors vs. comparator in randomized controlled trials

Legend: CI: Confidence Interval

Risk of Diabetic Ketoacidosis Reported in Observational Studies

Median patient follow up among the observational studies ranged from approximately 6 to 12 months. The absolute rate of DKA among patients taking SGLT2 inhibitors in observational studies ranged from 0.6 to 4.9 per 1000 person years compared to 0.6 to 2.2 per 1000 person years among patients who received an active comparator. This corresponded to a 1.7-fold higher risk of DKA for people receiving an SGLT2 inhibitor compared to the active comparator (RR⍰=⍰1.74 [95% CI, 1.07-2.83]; I²⍰=⍰45%; P⍰= 0.12, Figure 3).

• Download figure
• Open in new tab

Figure 3.

Forest plot of diabetic ketoacidosis risk with sodium-glucose cotransporter-2 (SGLT2) inhibitors vs. comparator in observational studies

Legend: CI: Confidence Interval

DISCUSSION

In this systematic review and meta-analysis that included over 360,000 adults with type 2 diabetes, SGLT2 inhibitors were associated with an increased risk of DKA. This finding was consistent regardless of study design (i.e., randomized trial versus cohort study) or comparator group (i.e., placebo versus active comparator). The risk was approximately 2-fold higher than with placebo/comparator medications, though the absolute rate was low (i.e., 0.6 - 4.9 events per 1000 person years).

Three prior meta-analyses have assessed the risk of DKA with SGLT2 inhibitors, however, none included observational studies.^16–18 One industry-funded meta-analysis included 8 RCTs (N=10,157 patients taking SGLT2 inhibitors) and found that the risk for DKA was similar in adults randomized to SGLT2 inhibitors compared with placebo (OR 1.14 [95% CI 0.45 to 2.88], P = 0.78).¹⁷ A second study assessed the risk of DKA in 26 RCTs (N=9,452 patients taking SGLT2 inhibitors), and found a lower risk with SGLT2 inhibitors compared to placebo (RR 0.66; [95%⍰CI 0.30 to 1.45], I²=0.0%) and incretin treatment (RR 0.43; [95%⍰CI 0.07 to 2.75], I²=0.0%), although with wide confidence intervals that included the possibility of a null effect.¹⁶ The third study (N=14,726 patients taking SGLT2 inhibitors) assessed the risk of DKA among 34 RCTs, and found a higher odds of DKA with SGLT2 inhibitors, albeit with wide confidence intervals which included the possibility of a null effect (OR 1.36; [95% CI 0.39–4.74], P = 0.63, I² = 0%).¹⁸

In contrast, our study identified an increased risk of DKA among patients receiving SGLT2 inhibitors in both randomized controlled trials and observational studies. There are two principal reasons why our results differ from the aforementioned meta-analyses. First, we restricted our meta-analysis to large clinical trials, as smaller trials are likely to be under-powered to detect the rare event of DKA. Second, our study included two RCTs that were recently published, and thus not included in the previous meta-analyses; both were large cardiovascular outcome trials that identified an increased risk of DKA among patients randomized to an SGLT2 inhibitor.^3,5

The mechanism by which SGLT2 inhibitors cause DKA remains uncertain. The SGLT2 receptor is primarily expressed in the proximal tubule of the nephron, which does not explain why it would potentially lead to DKA. However, a recent study identified that SGLT2 was also expressed in the glucagon-secreting alpha cells in the pancreas and directly triggered glucagon secretion.³⁴ Furthermore, by causing persistent glycosuria, SGLT2 inhibitors induce a catabolic state, which indirectly causes glucagon and catecholamine secretion.^35,36 Lastly, SGLT2 inhibitors have been shown to decrease insulin secretion by pancreatic beta-cells.³⁷ The resulting imbalance in insulin and counterregulatory hormones can result in DKA by promoting lipolysis and ketogenesis, which can be exacerbated by any additional cause of volume depletion or catecholamine excess (e.g, concurrent illness, reduced food/water intake, etc.).³⁵

There are several limitations to our study. First, the studies primarily included middle-aged adults in North America and thus it is unknown how our results might apply to other populations. Second, the included studies controlled for baseline confounders, but not time-varying confounders or events that occurred after starting the medication (e.g., surgery). Recent data suggest that patients are particularly prone to DKA during the post-operative period which has led to recommendations to hold SGLT2 inhibitors post-operatively for a minimum of one day and until the patient is eating a normal diet.^38,39 Third, due to the limited number of available studies we were unable to perform planned secondary analyses to identify patient-level factors associated with DKA, or differences in risk between individual SGLT2 inhibitor molecules (e.g., canagliflozin, empagliflozin and dapagliflozin). Lastly, although five of the 12 studies in our meta-analysis were observational in nature, all had identical study designs, comparator groups, statistical analysis, and outcome definitions

DKA is an important, albeit rare, potentially life-threatening adverse event associated with the use of SGLT2 inhibitors. Our data suggest the relative risk of DKA is approximately two times higher than with other antihyperglycemic medications, and the absolute rate is approximately 1 per 1000 person years. Patients should be informed of this risk and counselled to temporarily hold their SGLT2 inhibitors under specific circumstances (e.g., sick day, perioperatively). Providers should be alert to the possibility of euglycemic DKA in patients taking SGLT2 inhibitors. Future studies are needed to identify those at greatest risk.