We found a consistently increased risk of DKA among users of SGLT-2 inhibitors on the basis of pooling results from either RCTs or cohort studies. Beyond consistency, the evidence from RCTs and cohort studies is complementary, which is illustrated by the 5 following points.
First, the RCTs primarily compared the risk of SGLT-2 inhibitors with placebo, whereas the cohort studies used active comparators. Second, given the rare occurrence of DKA, especially in type 2 diabetes, the large study population sizes that are typical of observational studies are an advantage for measuring precise associations. Third, large pragmatic RCTs are possible; however, they are uncommon and may capture fewer events than most observational studies. This was evident with the recent EMPEROR-Reduced Trial (Empagliflozin Outcome Trial in Patients With Chronic Heart Failure With Reduced Ejection Fraction), which was captured by our search but was excluded because there were no events in either arm of the trial. Additionally, there were only 8 events across all included RCTs with an active comparator, in comparison with more than 2,600 events within included cohort studies. Fourth, a significantly increased DKA risk was not observed until after 10 trials and approximately 7 years after the first SGLT-2 inhibitor was marketed globally. This begs the question: Would more intensive active pharmacosurveillance, using cohort signal detection methods or large-scale phase IV pragmatic RCTs have decreased this timeline? Fifth, our comprehensive quality assessment illustrates how RCTs and observational studies counterbalance each other in terms of internal and external validity.
This analysis was limited to 1 example of a rare outcome, and broader analyses for a range of safety outcomes are necessary to generalize these findings.
In conclusion, this analysis showed that irrespective of the evidentiary source, there is a consistently increased risk of DKA in patients using SGLT-2 inhibitors compared with those receiving placebo or an active comparator. Importantly, this analysis illustrates that, when used in concert, RCTs and observational studies with robust methodology provide more comprehensive evidence for the risk of rare harm.
CI, confidence interval; DECLARE, Dapagliflozin Effect on Cardiovascular Events; DKA, diabetic ketoacidosis; RCT, randomized controlled trial; RR, relative risk; SGLT-2, sodium/glucose cotransporter 2; TIMI, Thrombolysis in Myocardial Infarction.
Additional data are available from the authors upon request.
Am J Epidemiol. 2021;190(8):1572-1581. © 2021 Oxford University Press