Supported Open-Source Automated Insulin Delivery for Management of Type 1 Diabetes in Pregnancy

Abstract

Background:

The tight glycemia required to optimize type 1 diabetes (T1D) pregnancy outcomes is difficult to achieve with standard insulin therapies. Automated insulin delivery (AID) offers an avenue to improve glycemia, but most available systems are not configurable to tight pregnancy glucose targets. Open-source AID may meet the needs of some pregnant women with T1D, but available data on its efficacy and safety in pregnancy are limited.

Methods:

This single-center retrospective study describes the glycemic and obstetric outcomes of pregnancies in which supported open-source AID (SOSAID) was used. Included patients had a pregnancy managed on SOSAID at BCDiabetes between January 2023 and October 2024 and consented for inclusion of their clinical data. Charts were reviewed to obtain comprehensive glycemic data, obstetric outcomes, and adverse events.

Results:

Ten patients, mean age 33 years, had a mean pre-pregnancy A1c of 6.7% (range 5.8%-8.0%). There were no episodes of DKA or severe hypoglycemia. Mean time-in-range (TIR_{63-140 mg/dL}) was 68% in trimester 2 and 70% in trimester 3. Seven patients commenced SOSAID during pregnancy, with their median 14-day TIR rising from 52% pre-SOSAID to 71% immediately after commencing SOSAID. There were no perinatal deaths or congenital anomalies. Pre-term delivery occurred in 1/10 and hypertensive disorders of pregnancy occurred in 2/10 women. Birthweight above 4 kg was present in 3/10, and neonatal hypoglycemia occurred in 4/10.

Conclusions:

SOSAID systems represent a promising tool for managing T1D in pregnancy and were successful in reaching target pregnancy glycemia in this single-center cohort.

Introduction

Tight glycemia mitigates much of the risk of adverse pregnancy outcomes in women with type 1 diabetes (T1D). ¹ The recommended continuous glucose monitoring (CGM) metrics include time-in-range (TIR)_{63-140 mg/dL} >70% with time-below-range (TBR)_{<63 mg/dL} <4%. ² The standard of care for T1D in pregnancy in most resource-rich settings continues to be non-automated insulin delivery, comprising either multiple daily injection (MDI) insulin with CGM, or sensor-augmented pump therapy (SAPT). With these approaches, the TIR target was only achieved by 6%, 11% and 44% of CGM users in the first, second and third trimesters respectively in the CONCEPTT trial. ³

The ideal insulin delivery method for T1D in pregnancy would achieve the tight glycemia required for optimal pregnancy outcomes, while minimizing hypoglycemia risk and patient burden. This remains a challenge despite the increasing availability of automated insulin delivery (AID) systems that have revolutionized T1D care in the non-pregnant population. ⁴ Most commercial AID (C-AID) systems do not offer sufficiently-tight glucose targets for pregnancy, and most are not approved by regulators for use in pregnancy. ⁵

AID systems are beneficial in pregnancy when the system used is configurable to tight pregnancy targets. This was demonstrated in a randomized controlled trial (RCT) comparing the CamAPS FX hybrid closed-loop system with standard care, in which those on AID had a mean pregnancy TIR of 68.2% compared with 55.6% in the standard-care group. ⁶ Among C-AID systems which do not support tight targets for pregnancy, mixed results have been shown. In an RCT of women with tight baseline glycemia, the use of the MiniMed780G system did not improve overall TIR; however, it did improve TBR, overnight TIR, and improved treatment satisfaction. ⁷ In a RCT comparing the earlier generation MiniMed 670G system with SAPT, AID did not improve TIR, and the average sensor glucose in the third trimester was lower in the SAPT comparator group. ⁸ Real-world use of several C-AID systems has appeared beneficial in some case series and case reports^9-11 but failed to demonstrate glycemic benefits in a larger prospective cohort utilizing various systems ¹² and a retrospective study of Omnipod 5. ¹³

Open-source AID systems (OSAID), also known as DIY AID, are potential candidates for achieving tight glycemia in T1D pregnancies. OSAID systems comprise free-to-use algorithms developed by and for the T1D community, which can be connected with commercially-available pump and CGM hardware. They undergo iterative developments to continually improve their functionality. Several popular OSAID systems, including Loop, Android Artificial Pancreas System (AAPS) and iPhone Artificial Pancreas System (iAPS), have features that are suited to management of T1D in pregnancy. These features include the option to select tighter glucose targets and a wide array of custom options to tailor and boost insulin delivery in the setting of rising insulin resistance and problematic post-prandial hyperglycemia. Evidence for OSAID systems in pregnancy to date comprises individual case reports/series of patients who self-initiated and self-managed their systems in pregnancy. ¹⁴

BCDiabetes is a publicly-funded diabetes center in Vancouver, British Columbia. BCDiabetes has been supporting people with T1D to install and use OSAID since 2020, with a total of 1760 current patients using these systems. The technical, logistical and clinical aspects of this program have been described previously. ¹⁵ With this level of clinical support, the term supported open-source AID (SOSAID) is used. Given the inadequacy of non-automated insulin strategies to achieve target glycemia for pregnancy in most patients, and the lack of any C-AID system approved for pregnancy in Canada, several patients have opted to use SOSAID during their pregnancy. These patients were supported extensively by an endocrinologist and diabetes case manager, with reviews (predominantly virtual) every 1 to 4 weeks throughout pregnancy depending on clinical need. Reviews included titration of insulin settings and education regarding optimization of OSAID system features to address glycemic patterns at each stage of pregnancy. We present the glycemic and obstetric outcome data of those opting to use SOSAID in pregnancy here. They represent the first published cohort of OSAID users in pregnancy who were fully clinic-supported for their AID installation and management.

Methods

We undertook a single-center, nonanalytical (case-only), open, clinical, retrospective cohort analysis of women with T1D who used SOSAID during pregnancy to evaluate their glycemic and obstetric outcomes. The study was approved by the University of British Columbia Clinical Research Ethics Board (Number H24-00521; protocol approved May 13, 2024). Patients consented for the use of their data in this study via a dedicated written consent process.

Patients were eligible for inclusion if they had T1D, were age 19 to 50 years, were using an OSAID (either Loop, AAPS or iAPS) with clinical support through BCDiabetes, and had a pregnancy between 1 January 2023 and 1 October 2024.

Patient medical records were reviewed on the BCDiabetes database to determine the glycemic and clinical outcomes of pregnancies during which SOSAID was used. Baseline (pre-pregnancy) clinical characteristics were collected, including current age, duration of diabetes, diabetes complications, gravidity/parity, BMI, type of insulin pump and CGM, type of SOSAID, laboratory A1c and CGM parameters. Glycemic data from the three trimesters of pregnancy was collected, including laboratory A1c and CGM parameters (TIR, TBR, mean sensor glucose, standard deviation), with the CGM data being directly available from a CGM cloud platform. For TIR, the international standard for pregnancy of 63 to 140 mg/dL was used throughout. Also recorded were any major clinical events related to the insulin delivery system including pump failure/AID system failure, diabetic ketoacidosis (defined as positive serum or urinary ketones and evidence of metabolic acidosis with a venous blood pH <7.30 and/or bicarbonate <15 mmol/l) or severe hypoglycemia (defined as a state of an altered mental and/or physical functioning that needed assistance from another person for recovery from hypoglycemia). Obstetric outcome data were collected, including gestation at delivery, mode of delivery, adverse pregnancy outcomes and birthweight. The BCDiabetes database is not linked to the inpatient maternity facilities at which the patients received obstetric care, so the obstetric and neonatal outcomes were recorded based on patient self-report during clinical visits. Patient-reported outcome measures were routinely documented in clinical practice pre- and post-SOSAID start at BCDiabetes, so these data were collected if available for any patients who newly commenced SOSAID in pregnancy. The measures included Diabetes Distress Scale (DDS) and Diabetes Impact Device Satisfaction (DIDS).^16,17

Statistical analyses were performed using SPSS. Continuous variables were expressed as mean (standard deviation) for normally distributed data or median (interquartile range) for non-normally distributed data. Categorical variables were summarized using frequencies or proportions.

Results

Ten patients met inclusion criteria and all consented to inclusion of their data in this study. Baseline characteristics of the ten patients are shown in Table 1. The mean age was 33 years, mean diabetes duration 19 years, and seven of the ten women were nulliparous. Mean pre-pregnancy A1c was 6.7%. Four of the ten women had pre-pregnancy A1c levels at the Diabetes Canada-recommended target of ≤6.5%, including the three who were already established on SOSAID pre-pregnancy (Cases 1, 9, and 10). Median pre-pregnancy BMI was 24 (IQR 21.9-26.6).

Table 1.

Baseline Characteristics Prior to the Pregnancy in Which SOSAID Was Used.

Case	Age (years)	Diabetes duration (years)	A1c pre-pregnancy (%)	Diabetes complications	Parity a	BMI pre-pregnancy	Insulin therapy & CGM pre-pregnancy	Gestational age at first pregnancy visit (weeks) b
1	30	20	6.3	Retinopathy (mild)	0	21.1	SOSAID	5
2	39	30	7.1	None	0	44.6	MDI + CGM	6
3	31	17	6.0	None	0	21.9	MDI + CGM	8
4	40	27	6.7	Retinopathy (treated)	2	21.3	SAPT	4
5	35	23	8.0	Retinopathy (treated)	0	26.6	SAPT	4
6	34	15	6.8	None	1	25.2	SAPT	5
7	29	21	6.8	None	0	23.1	SAPT	6
8	35	7	6.9	None	0	22	MDI + CGM	7
9	27	21	5.8	Retinopathy (mild)	0	32.6	SOSAID	21
10	34	7	6.5	None	1	26.6	SOSAID	16
Mean (SD)	33 (4.2)	19 (7.6)	6.7 (0.6)

Abbreviations: CGM, continuous glucose monitor; MDI, multiple daily injection; SAPT, sensor-augmented pump therapy (without automated insulin delivery); SOSAID, supported open-source automated insulin delivery.

^aParity refers to parity at commencement of the pregnancy in which SOSAID was used.

^bThe first pregnancy visit with BCDiabetes. Several patients were under shared care with a local endocrinologist in addition to care with BCDiabetes. All patients with a first visit later than 8 weeks gestational age were seeing a local endocrinologist in early pregnancy.

Dedicated preconception counseling regarding T1D was confirmed to have occurred in four cases, was likely absent in three cases, and in the remaining three was performed to varying extents with the patient’s local endocrinologist (with BCDiabetes acting as a remote clinical support for OSAID and pregnancy glycemic management). CGM use was universal at the first pregnancy visit with BCDiabetes, in both the pump-users and the MDI users.

Seven patients not already established on SOSAID pre-pregnancy opted to commence SOSAID during pregnancy (including Case 4, who planned to commence preconception, but in retrospect was peri-conception at the time of SOSAID commencement). The reasons reported for changing therapy during pregnancy by these seven women were primarily related to either (1) a desire to improve glycemia beyond what they anticipated achieving in pregnancy without automation or (2) a pre-existing plan to change to SOSAID, with the onset of pregnancy expediting their desire to use SOSAID. All seven had improvements in their TIR within the first week of SOSAID therapy, and there was no patient who experienced any loss of glycemic control with changing therapy during pregnancy. Their CGM parameters for the 14 days pre-SOSAID and the 14 days immediately after SOSAID commencement are shown in Table 2.

Table 2.

Short-Term Glycemic Control Pre-SOSAID and After SOSAID, for Those Commencing SOSAID During Pregnancy.

Abbreviations: TIR, time-in-range_{(63-140 mg/dL)}; TBR, time-below-range _{(<63 mg/dL).}

No patients ceased SOSAID during pregnancy and all provided generally positive feedback about their experience with SOSAID. Formal patient-reported outcome measures were available pre- and post-SOSAID for four of the patients who newly commenced in pregnancy (Cases 2, 3, 6, and 8). Among this small sample, all had numerically lower diabetes distress (median DDS trending down from 3.6 to 2.5) and higher device satisfaction (median DS subscale of DIDS trending up from 6.1 to 6.9) approximately 3 months after commencing SOSAID.

Glycemic control during pregnancy is shown in Table 3. Mean TIR in trimester 2 was 68%, and that in trimester 3 was 70%.

Table 3.

Glycemic Control During Pregnancies in Which SOSAID Was Used.

Case	SOSAID app	Trimester 1				Trimester 2				Trimester 3				AID system glucose target/s used in pregnancy (mg/dl) b
		TIR (%)	TBR (%)	MSG (mg/dL)	A1c (%)	TIR (%)	TBR (%)	MSG (mg/dL)	A1c (%)	TIR (%)	TBR (%)	MSG (mg/dL)	A1c (%)
1	Loop	82	2.4	110	5.9	83	3.4	106	5.0	84	3.8	103	nd	86-88
2		61 a	6.0 a	128 a	5.8 a	73	2.7	119	5.9	75	3.2	117	nd	90-99
3		54 a	6.4 a	137 a	5.8 a	65	5.8	124	5.5	76	4.1	112	5.6	90-99
4		57	1.4	144	6.0	59	2.1	137	nd	66	2.0	128	6.5	90-99
5		57	0.8	146	6.2	62	1.5	139	6.3	65	1.9	131	6.1	86-99
6	AAPS	46 a	2.9 a	158 a	5.7 a	61	3.0	133	5.8	61	2.2	135	5.8	81-99
7		66	2.0	128	5.8	63	1.4	131	5.7	64	1.6	131	6.1	81-90
8		60 a	5.6 a	130 a	6.3 a	72	2.4	122	5.9	71	2.8	121	6.0	90-99
9	iAPS	75	3.2	115	5.4	71	3.6	119	5.4	69	3.0	122	nd	81-90
10		61	1.4	137	nd	68	1.6	128	5.8	65	0.8	130	nd	90-99
Mean (SD)		61.9 (10.3)	3.2 (2.0)	133 (14)	5.9 (0.3)	67.7 (7.3)	2.8 (1.3)	126 (10)	5.7 (0.4)	69.6 (7.0)	2.5 (1.0)	123 (10)

Mean insulin total daily dose at the end of each trimester respectively was 49 (±14) units, 79 (±25) units, and 99 (±37) units.

Abbreviations: A1c, glycosylated hemoglobin; MSG, mean sensor glucose; nd, not done; TIR, time-in-range_{(63-140 mg/dL)}; TBR, time-below-range _{(<63 mg/dL).}

^aGlycemic data from trimesters when the patient was not on SOSAID for the majority of the trimester.

^bThe upper and lower bounds of the glucose targets show the range of the glucose targets used at different gestational ages and/or for different time segments of the day.

Maternal and neonatal outcomes are shown in Table 4. Most deliveries were at 37 to 39 weeks’ gestation and mean birthweight was 3.6kg (SD 0.7 kg). One patient (Case 2) developed pre-eclampsia and delivered pre-term at 35 weeks and 5 days gestation, on a background of concomitant risk factors of age 39 years and BMI 44.

Table 4.

Maternal and Neonatal Outcomes of Pregnancies in Which SOSAID Was Used.

Case	Gestation at delivery (weeks+days)	Birthweight (kg)	Birthweight percentile category a	Mode of delivery	Adverse pregnancy outcomes (maternal)	Adverse pregnancy outcomes (neonatal)	Requirement for NICU >24 hours	Maternal weight gain (kg)
1	38+4	3.0	AGA	EmCS	Gestational hypertension	-	No	na
2	35+5	2.7	AGA	EmCS	Pre-eclampsia	RDS	No	1
3	38+3	4.6	LGA	EmCS	-	NH	No	12.3
4	37+4	4.5	LGA	ElCS	-	Possible respiratory issues	No	19
5	37+1	3.5	AGA	ElCS	-	-	No	24
6	37+5	3.8	LGA	EmCS	-	NH	No	27
7	38+5	3.5	AGA	Vaginal (instrumental)	-	NH	No	16
8	38+1	2.6	SGA	Vaginal (instrumental)	-	Neonatal jaundice	No	9.6
9	37+3	4.3	LGA	EmCS	Pulmonary embolism following CS in labor	NH & feeding difficulties	Yes	22
10	38+0	3.2	SGA	Vaginal	-	RDS & jaundice	Yes	13
Mean (SD)	37+7 (6 days)	3.6 (0.7)						16 (8)

Abbreviations: AGA, appropriate for gestational age; ElCS, elective cesarean section; EmCS, emergency cesarean section; LGA, large for gestational age; NH, neonatal hypoglycemia (refers to any diagnosis of Neonatal Hypoglycemia, whether requiring IV treatment or not); NICU, neonatal intensive care unit; RDS, respiratory distress syndrome; SGA, small for gestational age.

^aBirthweight percentile category classifies neonates as AGA if they have a birthweight between the 10th and 90th percentile on the Fetal Medicine Foundation neonatal population weight charts. ¹⁸

Discussion

In this study, SOSAID was used in 10 T1D pregnancies without any serious adverse diabetes-related events. The majority of patients experienced either at-target glycemic parameters by trimesters 2 and 3 or a substantial improvement in glycemic parameters compared to their prior glycemia. Mean TIR_{63-140 mg/dL} in trimester 2 was 68%, and that in trimester 3 was 70% with a mean TBR of 2.8% and 2.5%, respectively. All seven who started OSAID during pregnancy showed a significant improvement in glycemic parameters in the first two weeks, somewhat alleviating concerns that a change to SOSAID during pregnancy may risk temporary loss of tight glycemic control. Notably, no ketoacidosis or severe hypoglycemic events occurred.

The incidence of adverse pregnancy outcomes in this study was similar or lower than that seen in other studies of women with T1D. ¹⁹ Hypertensive disorders of pregnancy occurred in 2/10 women. Birthweight above 4 kg was present in 3/10. Neonatal hypoglycemia occurred in 4/10 neonates, and 2/10 required NICU for >24 hours. In this small group there were no congenital anomalies or perinatal death.

Lower glucose targets and frequent adjustments of insulin are required in women with T1D during pregnancy. These considerations may impact the effectiveness and safety of AID when using algorithms designed for non-pregnant populations. One limitation of many C-AID systems is that the glycemic targets are not adjustable for pregnancy. Although OSAID technology has not been formally studied in pregnant women with diabetes, it provides the necessary flexibility to individualize glucose targets and automatic adjustments to insulin dose when blood glucose levels are anticipated to rise above or fall below prespecified targets. With OSAID the algorithm glucose target can be set to 80 to 90mg/dL if appropriate, whereas most C-AID systems have glucose targets ≥100 mg/dL (with the exception of CamAPS).

There have been three RCTs assessing C-AID in T1D pregnancy. These showed mixed results regarding glycemic benefits. The AiDAPT study investigated CamAPS compared with standard insulin therapy in 124 pregnant women with a mean baseline A1c of 7.7%. ⁶ TIR in the AID group was 68.2%, an overall benefit of 10.5% above standard care, without increased TBR. While not powered to assess pregnancy outcomes, these were generally similar except that the AID group had a numerically lower rate of hypertensive disorders (20% vs 42%), lower gestational weight gain (11.1±6.1 kg vs 14.1 ± 6.1 kg) and lower large for gestational age (LGA) (39% vs 50%). The CRISTAL study included 95 women with mean baseline A1c 6.5%, and compared the MiniMed 780G system with routine care. ⁷ The results showed no improvement in overall TIR, with TIR 66.5% in the AID group compared with 63.2% in the standard insulin therapy group. However there was improved overnight TIR (adjusted mean difference: 6.58 percentage points, P = .0026) and reduced TBR (adjusted mean difference: –1.34 percentage points, P = .0020) with AID. While not powered for pregnancy outcomes, the AID group had a lower excess gestational weight gain (32.6% vs 56.5%, P = .033), lower NICU admissions for neonatal hypoglycemia (14.3% vs 63.6%, P = .017), and numerically lower LGA (55.8% vs 67.4%). The Pregnancy Intervention with a Closed-Loop System (PICLS) study compared the earlier generation MiniMed 670G AID with SAPT in 23 women and found no benefit for TIR from AID. ⁸

The performance of C-AID in pregnancy has also been described in real-world case series.^9-13 The MiniMed 780G ⁹ and Tandem Control IQ systems^10,11 have some algorithm features potentially somewhat more suited to pregnancy glycemia (e.g. overall glucose target option of 100 mg/dL, or sleep mode with narrower glucose target) and have been associated with pregnancy TIR >70% in some real-world data. Twenty-one real-world MiniMed 780G users ⁹ had a higher TIR than that seen in the CRISTAL trial, with a mean TIR of 64%, 71%, and 76% in the first, second, and third trimesters, respectively. Fifteen Control IQ users achieved a TIR of >70% during most of gestation, specifically >70% between 6 and 21 weeks of gestation and after 30 weeks of gestation. The mean TIR values were 68.8% for weeks 22–25 and 68.7% for weeks 26–29. ¹⁰ There were similar findings in four Control IQ users who all had improved TIR and all reached >70% TIR by the third trimester. ¹¹ However, when studied in a prospective observational cohort against MDI comparator, 59 users of various C-AID systems (including MiniMed 780G and Tandem Control IQ) did not have superior TIR. ¹² The Omnipod 5 system, which has a lowest-available glucose target of 110 mg/dL, has been studied in a 17-patient retrospective cohort and was associated with median TIR of 56% to 65% throughout gestation. ¹³ Compared to these studies of C-AID, in the current study our patients had at least a comparable TIR with the mean TIR in trimester 2 of 68% and trimester 3 of 70%.

Beyond C-AID, open-source and investigational algorithms provide a potential option for AID with pregnancy-specific features. An investigational pregnancy-specific zone model predictive control algorithm which runs on open-source iAPS was studied in eleven pregnant participants for two days, with significant improvements in TIR and TBR. ²⁰ The real-world use of OSAID during pregnancy has been described in case reports and series, with findings of the studies providing TIR per trimester summarized in Supplementary Table 1. Excellent glycemic outcomes were described, with many cases having TIR >70% throughout pregnancy, and in some, >80%. Most of the described patients independently installed and managed their OSAID systems prior to pregnancy. In many cases, they would be considered “expert users,” for example, in one study not shown in Supplementary Table 1, all patients had >2 years of experience with OSAID and A1c ≤6.5% prior to conception. ²¹ Survey data from 37 respondents found that 86% of women using OSAID in pregnancy self-reported TIR >70%. ²² They also reported excellent quality of life benefits. Given that these independent-installation OSAID users are likely to represent a group with high education levels and technical skills, it is reassuring that our cohort, who were clinic-supported with OSAID installation and were new users in most cases, also achieved close to a mean TIR of 70% during pregnancy.

Strengths of the present study include this being the largest clinic-based cohort reporting the use of SOSAID in pregnancy to date, and full availability of glycemic data from throughout pregnancy. Limitations include firstly, a relatively small sample size. Second, there was variability of gestation at the commencement of SOSAID, resulting in the first trimester glycemic data capturing some AID use and some use of other insulin delivery methods. Third, we relied upon patient self-report of obstetric and neonatal outcomes, as direct records from maternity admissions were not available. Due to the frequent use of virtual care appointments, body weight measurements were also self-reported in some cases. Finally, this was a single-center study, which limited generalizability to other settings, including to T1D populations with different baseline glycemic control, diabetes knowledge, technology access, and ethnicity.

The benefits of OSAID systems should be balanced with risks. Firstly, there is a lack of a commercially-available support to address technical issues that may arise, making many OSAID users largely responsible for finding solutions through the online community. Secondly, there may be issues with lack of support or knowledge from health care professionals. Thirdly, all insulin delivery systems can fail, so appropriate education for pump failure and AID failure is required to avoid complications. The women included in this study had a high level of clinician support from a diabetes clinic with expertise in SOSAID.

As for all women with T1D, preconception optimization of glycemic management is strongly recommended in those considering OSAID. The data from this case series showed a transition to OSAID during pregnancy was safe; however, optimal preconception glycemia (A1c ≤6.5% without excess hypoglycemia) was more often found in those already on SOSAID preconception.

Conclusion

OSAID systems represent a promising tool for managing T1D in pregnancy, offering the potential for improved glycemic control and pregnancy outcomes. Prospective studies are warranted to establish OSAID use in pregnancy.