Diabetic Ketoacidosis After Sodium–Glucose Cotransporter Inhibitor Initiation Under Advanced Hybrid Closed-Loop Therapy in Type 1 Diabetes

Introduction

Sodium–glucose cotransporter inhibitors (SGLTi) are oral antidiabetes medications, recommended for the treatment of type 2 diabetes due to their beneficial effect on glucose control, and cardiovascular and renal outcomes. ¹ Several randomized controlled trials have also shown improvement in glycemic control with SGLTi as adjunct therapy in type 1 diabetes (T1D), along with weight loss and reduction in systolic blood pressure. ²

In Europe, the European Medicines Agency has approved low dose dapagliflozin and sotagliflozin for people with T1D and a body mass index (BMI) >27 kg/m². However, despite clinical benefits and regulatory approval, there is reluctance to SGLTi use in T1D, due to an increased risk of developing diabetic ketoacidosis (DKA). ³ In this study we report a 23-year-old woman with T1D treated with advanced hybrid closed-loop (aHCL) therapy, who developed DKA 2 weeks after initiation of SGLTi.

Case Report

The patient was diagnosed with T1D at age 12 years. As a teenager, her glycemic control was characterized by high glycated hemoglobin (HbA1c) values between 9.0% and 10.7% (74 and 94 mmol/mol), and complicated with one severe DKA episode due to noncompliance with insulin injections. Diabetes control improved in the past 6 months since start of aHCL (Medtronic MiniMed™ 780G system), with an HbA1c between 7.3% and 8.0% (56 and 64 mmol/mol). She struggled with overweight for years, with a BMI reaching 42.1 kg/m², for which her diabetologist recommended SGLTi in the form of empagliflozin 12.5 mg/day.

Two weeks after initiation of SGLTi, the patient presented at the emergency department of a regional hospital because of acute dyspnea on awakening. She complained of polydipsia and polyuria and had measured a capillary blood ketone level of 6.5 mM at home as instructed according to the international recommendation. ³ She also said she had lost 8 kg in the past few weeks, partly due to reduced intake of carbohydrates. There were no abnormalities on physical examination, except for sinus tachycardia (127 per minute), and an elevated blood pressure (154/101 mmHg).

Arterial blood gas was consistent with metabolic acidosis (reference value): pH 7.14 (7.35–7.45), pCO₂ 1.2 kPa (4.3–6.0), pO₂ 16.5 kPa (11.1–14.4), bicarbonate 3.1 mM (22–29), base excess 23 (−2.5 to +2.5), and lactate 2.1 mM (<1.2). Plasma glucose levels were 19.6 mM (353 mg/dL). Dipstick revealed ketonuria (ketones 4+). The diagnosis of DKA was made, with no clear evidence of contributing factors (kinked cannula, infection, etc.). Treatment with fluids and intravenous insulin was started with full recovery. The patient was discharged from the hospital 3 days later with resumption of aHCL therapy, but without SGLTi.

Careful assessment of the evolution of our patient's daily insulin administration and carbohydrate intake in relation to SGLTi use reveals interesting mechanisms that contributed to the onset of DKA (Fig. 1). Within a few days after start of SGLTi, the aHCL control algorithm reduced autobasal and autocorrection doses (Fig. 1A, B). Meal bolus insulin doses were already reduced in the week before initiation of SGLTi (Fig. 1C), as a result of lower carbohydrate intake by our patient (Fig. 1D). After start of SGLTi, meal bolus insulin doses remained at a lower level, not only due to lower carbohydrate intake, but also due to frequent activation of the “safe meal bolus” (“S” in Fig. 1C).

OPEN IN VIEWER

FIG. 1.

Graphical representation of daily autobasal (A), autocorrection (B), and meal bolus doses (C), carbohydrate intake by our patient (D), mean sensor-glucose (E), and time in ranges (F), in relation to initiation of sodium–glucose cotransporter inhibitor (SGLTi) and onset of DKA. Gray bars and lines represent the period before start of SGLTi, black bars and lines represent the period after start of SGLTi. In (C), “S” indicates activation of the “safe meal bolus” feature during one or more meal boluses on that specific day. All data are based on daily ambulatory glucose profiles as reported in CareLink™. Sensor usage was on average >70%, SmartGuard™ usage was on average >95%. DKA, diabetic ketoacidosis; SGLTi, sodium–glucose cotransporter inhibitor.

This specific feature of the MiniMed™ 780G system reduces the calculated meal bolus when the algorithm predicts an excessive risk of postbolus hypoglycemia if the full calculated bolus was administered. In our case, the safe meal bolus feature frequently reduced meal bolus doses by 13.6% to 42% on 10 out of 14 days after start of SGLTi, compared with only 5.5% to 6.4% on 3 out of 14 days before start of SGLTi. Taken together, there was a significant 49% reduction in total daily insulin dose in the 2 weeks after start of SGLTi compared with 14 days before.

In terms of glycemic control, glucose levels decreased immediately in the first days after start of SGLTi (Fig. 1E), resulting in an increase in time in range (Fig. 1F). However, from day 4 after start of SGLTi, glucose levels and time in range gradually returned to levels observed before start of SGLTi, but without a subsequent increase in autobasal, autocorrection, or meal bolus doses.

Discussion

In this case report, we describe in detail the deleterious insulin dose reduction by a Medtronic MiniMed™ 780G aHCL algorithm after SGLTi initiation in a person with T1D leading to DKA. Reporting this potential DKA risk is clinically relevant, as the MiniMed™ 780G system is increasingly used in Europe and will enter the U.S. market in Q2 2022, whereas use of SGLTi in people with T1D is not uncommon.

Our case resembles a previously published case report describing DKA in a patient on a Medtronic MiniMed™ 670G system in whom concomitant use of SGLTi was accompanied by an automatic reduction in basal insulin, ⁴ but several interesting mechanisms specific to the response of the MiniMed™ 780G to the introduction of SGLTi were revealed. The autonomous decrease in both basal and bolus insulin doses resulted in almost halving of the total daily insulin dose, leading to the development of DKA due to insulin doses below the minimum needed to prevent ketone formation. This is especially dangerous when additional risk factors for DKA (e.g., infection, insulin pump failure, patient characteristics, and low carbohydrate diet) are present.

Although SGLTi might help in decreasing postprandial hyperglycemia when using automated insulin delivery, ⁵ our case warns against the use of SGLTi in people with T1D on the MiniMed™ 780G system. In addition, until more is known about the influence of SGLTi on aHCL algorithm functioning, we recommend caution with SGLTi use in people with T1D on aHCL systems to avoid increased DKA risk. As recently reviewed by Pasqua et al., ⁶ clear understanding of the effects of SGLTi on the algorithm should be obtained before considering SLGTi in people on aHCL systems. This should allow for specific recommendations to avoid DKA development on top of careful patient selection and education on self-measurement of capillary blood ketones (specifically beta-hydroxybutyrate) as advised by the international consensus. ³