Risk of recurrent diabetic ketoacidosis following initiation of sodium‐glucose co‐transporter 2 inhibitors in patients with an insulin‐deficient diabetes and prior diabetic ketoacidosis: An observational cohort study

Sodium-glucose co-transporter 2 inhibitors (SGLT2i) are central agents in the care of type 2 diabetes mellitus (T2DM), chronic kidney disease, and heart failure.1-4 Despite their well-established benefits in T2DM, the use of SGLT2i in type 1 diabetes mellitus (T1DM) has been approached with caution due to concerns over an increased risk of diabetic ketoacidosis (DKA).1, 5-8 In a meta-analysis of 14 clinical trials of SGLT2i versus placebo in patients with T1DM, lasting 4–52 weeks, the DKA rate was 0.7% versus 3.4% in the placebo versus SGLT2i arm. However, while a prior episode of DKA (ever) was not an exclusion criterion in most of these trials, data was not specifically reported for this population.5 Thus, our present work focuses on the excess risk of DKA in patients with a prior history of DKA prescribed SGLT2i. Anonymized data were obtained from Clalit Health Services (CHS), Israel's largest integrated payer/provider healthcare organisation, covering approximately 4.5 million members. The study included data from January 01, 2015 (when SGLT2i were first introduced in Israel) to December 31, 2022 for initiation of SGLT2i, with follow-up until December 31, 2023. The study population consisted of patients with diabetes aged 18–70 in 2019 who met the following criteria: age at diabetes diagnosis ≤50 years, body mass index (BMI) ≤27 kg/m2, haemoglobin A1c (HbA1c) ≥8.5% (for insulin-naïve patients), or any HbA1c value for patients previously treated with insulin. These criteria were designed to identify patients with a high likelihood of having an insulin-deficient phenotype. As this cohort included patients with a history of DKA, which occurred without use of SGLT2i, it is reasonable to assume many had T1DM. Nonetheless, as the diagnostic codes of type 1 diabetes in our database are unreliable, and antibody testing is not routinely performed, these demographic and laboratory criteria were applied to identify patients with a severe insulin deficiency phenotype. Of note, although SGLT2i are off-label for T1DM in Israel, they are occasionally prescribed to improve glycaemic control, or in the setting of chronic kidney disease or heart failure. Exclusion criteria were treatment with SGLT2i prior to January 01, 2015, baseline estimated glomerular filtration rate (eGFR) <30 mL/min/1.73 m2, history of kidney transplant or diagnosis of polycystic kidney disease, membership in CHS for less than 6 months prior to study entry, or lack of creatinine lab results prior to the index date. The cohort building process has been previously described.6, 7 After matching SGLT2i and control groups for baseline characteristics, patients with a prior episode of DKA were selected for future analysis. As DKA risk differed by prior DKA history, the cohort was restricted to individuals with prior DKA. This restriction disrupted baseline balance between treatment groups. Therefore, we applied inverse probability of treatment weighting using the previously calculated propensity scores, which re-established the balance and yielded estimates of the average treatment effect in the population. A prior diagnosis of DKA was identified from hospital, emergency room (ER), or community medical records. DKA event as an outcome was defined as an ER visit or hospital admission due to DKA. Patient outcomes were censored at the end of membership at CHS, death, or on December 31, 2023. Time-to-DKA was analysed using Kaplan–Meier curves and compared between groups with bivariate and multivariate Cox proportional hazards models. Statistical analyses were conducted using SAS Enterprise Guide 8.3. Overall, 478 patients were identified with a history of DKA, 151 were prescribed SGLT2i. Their baseline characteristics are shown in Table S1, Supporting Information, with standardized mean difference values calculated. The mean age of the cohort was 48.7 ± 11.6 years, BMI 25.0 ± 3.2 kg/m2 and HbA1c 9.2 ± 1.8%; 91.1% had been using insulin. Compared to non-users, SGLT2i users had a higher BMI, greater comorbidity score, and were more likely to be using GLP1-RA and DPP-4 inhibitors. The median follow up was 3.4 (interquartile range [IQR] 1.9–5.8) years, 3.2 (IQR 1.6–5.2) among SGLT2i users and 3.5 (IQR 2.0–6.1) years among non-users. During follow up, 82 patients experienced a DKA event: 38 (25.2%) among SGLT2i users and 44 (13.5%) among non-users (hazard ratio [HR] [95% confidence interval (CI)] 2.03 [1.30–3.18]; p = 0.002) (Figure 1). The observed increase in DKA events was consistent across multiple subgroups of interest and interaction analyses showed no evidence of effect modification (Figure S1). In multivariate analysis, SGLT2i use remained an independent predictor of DKA with a 77% increased risk (Table 1). Our work demonstrates a very high risk of recurrent DKA in patients with an insulin deficient phenotype, which is near-doubled with the use of SGLT2i, reaching 25%. The increased risk with SGLT2i remained consistent across multiple subgroups of interest. In multivariate analysis, SGLT2i remained an independent risk factor for DKA alongside poor glycaemic control and a recent episode of DKA. In our recent analysis we demonstrated an increased risk of DKA in patients with diabetes and an insulin deficient phenotype who did not have a history of prior DKA listed in our records (HR 1.39 [95% CI, 1.07–1.79]).7 Our present analysis focuses on patients with a prior history of DKA, which did not occur on SGLT2 treatment; thus, they are likely to have T1DM or another form of severe insulin deficient diabetes. SGLT2i have demonstrated multiple benefits beyond glucose lowering in patients with T2DM,1-3 and these benefits appear to extend to T1DM patients as well.5, 6, 8 Thus, their off-label use in patients with T1DM is expanding,9-11 and physicians may be inclined to prescribe SGLT2i to patients with albuminuria despite previous DKA episodes.12 However, within this insulin-deficient population, the risk of a recurrent episode of DKA was 12.9%, nearly doubling to 25.2% with SGLT2i use, mandating careful appraisal of this approach. These data indicate that in those patients, possibly not originally recognised as insulin deficient, a DKA episode mandates careful consideration of the diabetes phenotype. Notably, 50% of the patients whose prior DKA episode occurred in the 2 years prior to SGLT2i prescription suffered a recurrent DKA episode during follow up. The strengths of our study include its extended follow-up and real-world population, which enhance the generalizability of our findings. Moreover, focusing on an under-represented population of patients with prior DKA elicited their extremely high risk, an observation of significant clinical relevance. However, as in any observational study, causality cannot be established, and residual confounding may influence results. Additionally, the retrospective design precluded access to detailed data on insulin dosing or dose adjustments, which may have affected outcomes. Finally, as antibody testing was unavailable, the diagnosis of T1DM remains indeterminate. In conclusion, during median follow up of 3.4 years, in patients with an insulin-deficient phenotype and a previous DKA event, prescribing an SGLT2i was associated with a near doubling of the risk of recurrent DKA with an absolute risk of 25.2%. While these agents offer significant benefits, the associated DKA risk necessitates a thoughtful, evidence-based approach to their use, with caution exercised in high-risk groups with higher HbA1c or a more recent DKA episode. The authors thank Tamar Zohar MPH, Anna Elgart BSc and Amit Davidpur, BSc, at Clalit Research Services, Kaplan Hospital, Israel, for their support in acquiring and preparing the data from the CHS registries. Anat Tsur has received advisory board and/or honoraria from Novonordisk, Sanofi, Boehringher Ingelheim, AstraZeneca, Elli Lilly. Avivit Cahn has received advisory board and/or honoraria from Novonordisk, Sanofi, Boehringher Ingelheim, AstraZeneca, Elli Lilly, Rafa, Bayer and MedialEarlySign. Omer Hamtzany and Rena Pollack have nothing to disclose. The peer review history for this article is available at https://www.webofscience.com/api/gateway/wos/peer-review/10.1111/dom.70457. Clalit Health Services H.M.O. granted permission to use this dataset for the current analysis. The dataset for our study, the data dictionary, and complete documentation are securely held by Clalit Health Services H.M.O. Due to ethical restrictions, data for the current study cannot be shared. Table S1. Baseline characteristics. Figure S1. Subgroup analysis of the risk for DKA. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.