Management of Type 1 Diabetes Mellitus Using Open-Source Automated Insulin Delivery During Pregnancy: A Case Series

To the Editor,

Pregnancy in women with type 1 diabetes (T1D) is associated with increased rates of adverse obstetric and perinatal outcomes. Achieving stringent glycemic control during pregnancy is essential. ¹ However, glycemic targets are still unmet and the risk of maternal and fetal complications is constantly high in T1D pregnancy. ²

Artificial pancreas systems (APSs) combine a continuous glucose monitoring (CGM) system, an insulin infusion pump, and a glucose control algorithm to automatically adjust the delivery of basal insulin based on the real-time CGM data. APSs are a promising approach to improve glycemic control during pregnancy. ³ The open-source automated insulin delivery (AID) systems have been developed by people living with diabetes and code is provided on an open-source online platform, outside the usual regulatory environment. ⁴ Open-source AID systems have the potential to help women to meet their treatment goals and improve quality of life during T1D pregnancies. ^5,6

We present a case series on four women with pre-existing T1D using open-source AID at their own risk to manage their diabetes throughout pregnancy since the use of open-source AID is outside of the usual regulatory environment and no AID systems have yet been approved during pregnancies. The aim of our present investigation is to describe a case series of women using open-source AID during pregnancy, to assess the quality of glycemic control, as well as the obstetric and perinatal outcomes of these pregnancies. Data were collected retrospectively from the Nightscout documentation and the medical records of the diabetes outpatient clinic as well as of the obstetrics department. HbA1c was obtained from venous blood and measured at the laboratory of the Clinic Hietzing. People with T1D are screened annually for microvascular complications (neuropathy screening, albumin/creatinine ratio, and fundoscopy by an ophthalmologist). Once during pregnancy, an additional retinopathy screening is performed; microalbuminuria and proteinuria screening are performed at each outpatient visit during pregnancy.

The fetal growth calculator of the World Health Organization (WHO) was used for weight percentile calculations. Baseline characteristics and details on prepregnancy HbA1c, body weight, body mass index and body height, HbA1c, time in range (TIR), time below range (TBR), and time above range (TAR) during the three trimesters as well as a summary of the obstetric medical records and perinatal outcome are indicated in Table 1. The target range of 63–140 mg/dL was defined, along with TIR_{63–140mg/dL} of >70%, TBR_<63mg/dL of <5%, and TAR_>140mg/dL of <25%. Except for the first trimester of case 3, TIR_{63–140mg/dL} was >70% throughout all pregnancies. No severe hypoglycemic event occurred during these four pregnancies. In Case 4, TBR was high throughout the entire pregnancy. However, no severe hypoglycemic event occurred. In gestational diabetes, an increased risk of small for gestational age (SGA) and large for gestational age (LGA) was found when mean blood glucose of 87 mg/dL was not achieved or when blood glucose of 104 mg/dL was exceeded. ⁷ According to the German Diabetes Association, it is justified to transfer these findings also to pre-existing diabetes during pregnancy. To our knowledge, there are no studies discussing the impact of a high TBR on maternal and neonatal outcome. As we only report a small number of highly selected, motivated, and well-educated cases (all women attained high-level education with university degrees ranging from bachelor to master), it is not possible to draw conclusions on a potential causality between TBR and the outcomes for mother and baby. Larger randomized controlled studies are needed in the future to address this aspect.

Mean HbA1c during the first, the second, and the third trimesters was 35.5 ± 5.4 mmol/mol (5.5% ± 0.5%), 32.9 ± 3.0 mmol/mol (5.2% ± 0.3%), and 33.5 ± 2.9 mmol/mol (5.2% ± 0.2%), respectively. In all four cases, no microvascular complications were present before, during, and after the pregnancy. No arterial hypertension was observed during these pregnancies. These women visited the outpatient clinic every two to three weeks and the settings of the AID were adjusted based on the insulin requirements throughout the pregnancies. Retrospectively, it was not possible to obtain the exact settings over time in these pregnancies.

In two cases, cesarean section had to be performed due to pathological cardiotocography (Cases 1 and 3). In Case 2, a prelabor rupture of membranes required induction of labor followed by vaginal delivery. This woman had required eradication therapy for methicillin-resistant Staphylococcus aureus in urine culture at 32 weeks of gestation. Case 4 was an uneventful pregnancy with spontaneous vaginal delivery at 38 + 5 weeks of gestation. The birth weight of two neonates was normal for gestational age, whereas one was SGA and the other was LGA. Medical records state that no perinatal morbidity in this cohort occurred.

So far only short-term hybrid closed-loop trials not covering the whole course of pregnancy were performed reporting both good glycemic control and positive user experiences. ^8,9 When compared with those in the CONCEPTT trial, the women in our case series achieved more TIR with less TAR; all four women achieved HbA1c goals throughout pregnancy. ¹⁰ In contrast, women with T1D under routine conditions achieve HbA1c goals in early and late gestation only in 16% and 40%. ¹¹ Case series of T1D pregnancy managed using a commercially available AID system (670G; Medtronic) off-label in pregnancy showed ambiguous results. One case reported TIR_{63–140mg/dL} of 50–64%, whereas a case series reported TIR_{70–180mg/dL} of 70–97% throughout pregnancy. ^12,13 When comparing our data with a large cohort of 558 people with T1D using open-source AID, even more stringent glycemic control could be achieved [61–88% TIR (63–140 mg/dL) vs. 73 ± 13% TIR (70–180 mg/dL)]. ¹⁴

People with T1D can be trained in the use of commercially available AID systems by means of standardized training programs. Some customizations for the individual user might be necessary that can be done by trained health care professionals. Although commercially available AID systems are certified and officially approved by health authorities for their use in T1D, they are not yet approved to be used in T1D pregnancy due to the nonadjustable target range that is too high for pregnancy. In contrast, settings of open-source AID systems can be adjusted to meet glycemic targets during pregnancy. ^5,6 However, as open-source AID systems are not off-the-shelf products, they require more user input when setting up the systems and thus are probably available only for a limited group of users. Still, there is an engaged user community for all three available open-source platforms who help new users to move along their development pathway.

We observed satisfactory glycemic control with regard to TIR, TBR, TAR, and HbA1c goals when using open-source AID in four T1D pregnancies. Our case series showed that using open-source AID may help women to achieve good glycemic control while avoiding severe hypoglycemic episodes. Open-source AID might not be suitable for every T1D population. However, women with T1D might be more motivated using open-source AID during pregnancy, since currently, no commercially available APS is approved for T1D pregnancies.