Sodium/Glucose Cotransporter 2 Inhibitors and the Risk of Diabetic Ketoacidosis

An Example of Complementary Evidence for Rare Adverse Events

Wajd Alkabbani; Ryan Pelletier; John-Michael Gamble

Disclosures

Am J Epidemiol. 2021;190(8):1572-1581. 

In This Article

Results

Following screening of 1,451 unique citations, we identified 18 RCTs[12–29] and 8 cohort studies[30–37] meeting our inclusion criteria (Table 1), 24 of which were of good quality (Table 2 and Web Table 2). Most RCTs were placebo-controlled (12/19)[10–21] and reported a single DKA event (11/19). A total of 92 DKA events (84 events within placebo-controlled trials) were reported across all RCTs, which randomized a total of 55,807 patients (Table 2). All cohort studies used an active comparator and evaluated 2,663 DKA events, of which 1,316 occurred in 580,036 users of SGLT-2 inhibitors. One cohort study was excluded from the meta-analysis because of poor quality and an ambiguous study design.[32]

Based on pooled data from RCTs, there was almost a 2-fold increased risk of DKA (fixed-effects model: relative risk (RR) = 2.21, 95% confidence interval (CI): 1.40, 3.48; random-effects model: RR = 2.08, 95% CI: 1.28, 3.40; I 2 = 0%) among patients randomized to receive an SGLT-2 inhibitor compared with placebo (Figure 1A). Using a cumulative meta-analysis for the placebo-controlled RCTs, a significantly increased risk of DKA was not observed until the Dapagliflozin Effect on Cardiovascular Events (DECLARE)–Thrombolysis in Myocardial Infarction (TIMI) 58 Trial was carried out in 2019 (Web Figure 1). No significant effect was observed when the analysis was limited to an active comparator (Figure 1B).

Figure 1.

Relative risk (RR) of diabetic ketoacidosis among users of sodium/glucose cotransporter 2 inhibitors or controls in a meta-analysis of placebo-controlled randomized controlled trials (2013–2019) (A), active-comparator randomized controlled trials (2013–2019) (B), and cohort studies (2017–2020) (C) reporting 1 or more diabetic ketoacidosis events. Heterogeneity and P value: A) I 2 = 0% (95% confidence interval (CI): 0, 50), P = 0.60; B) I 2 = 0% (95% CI: 0, 4), P = 0.94; C) I 2 = 93% (95% CI: 87, 96), P < 0.01. Bars, 95% CIs.

An increased risk of DKA was also found when data were pooled from cohort studies for both unadjusted (fixed-effects model: RR = 1.68 (95% CI: 1.54, 1.84); random-effects model: RR = 1.53 (95% CI: 0.95, 2.47); I 2 = 93% (Figure 1C)) and adjusted (fixed-effects model: hazard ratio = 1.64 (95% CI: 1.41, 1.89); random-effects model: hazard ratio = 1.74 (95% CI: 1.28, 2.38)) effect estimates.

processing....