Pregnancy outcomes among women with type 1 diabetes mellitus using continuous subcutaneous insulin infusion versus multiple daily injections: A retrospective cohort study

Short abstract

Background

Insulin delivery options for pregnant women with type 1 diabetes mellitus are either continuous subcutaneous insulin infusion or multiple daily injections. The aim of this paper is to compare pregnancy outcomes in women with type 1 diabetes mellitus using continuous subcutaneous insulin infusion or multiple daily injections in pregnancy.

Methods

Retrospective single-centre cohort study of 298 pregnancies booked between 2006 and 2016. Descriptive analysis was performed for HbA1c values. Logistic regression models were created to compare selected maternal and neonatal outcomes.

Results

Continuous subcutaneous insulin infusion was associated with increased risk of large-for-gestational age (aOR 2.00, 95% CI 1.20–3.34) and preterm neonates (aOR 1.80, 95% CI 1.04–3.03). Continuous subcutaneous insulin infusion had no association with increased risk of adverse pregnancy outcomes. No difference in HbA1c values existed between groups.

Conclusion

Using continuous subcutaneous insulin infusion for type 1 diabetes mellitus through pregnancy is associated with increased risk of large-for-gestational age and preterm neonates, without increased risk of associated adverse maternal or neonatal outcomes.

Introduction

Women with type 1 diabetes mellitus (T1DM) face specific and significant challenges to maintaining optimal glycaemia during pregnancy. Poor glycaemic control during pregnancy can lead to many adverse maternal and neonatal outcomes, including maternal hypoglycaemia, caesarean section delivery, stillbirth, congenital abnormality, large-for-gestational age (LGA) neonates, and neonatal hypoglycaemia.¹ Maintaining tight glycaemic control in pregnancy can reduce the risk of these adverse outcomes. While tight control in the past has been associated with an increased risk of severe hypoglycaemia, this is becoming less of a concern with newer technologies.²

The two mainstay insulin delivery modalities in T1DM are multiple daily injections (MDIs) and continuous subcutaneous insulin infusion (CSII). Existing studies investigating MDI compared with CSII use in a general population of people with T1DM have demonstrated superiority of CSII in attaining better glycaemic control, quality of life, incidence of severe hypoglycaemia, and a reduced daily insulin requirement.³ Given the findings in the non-pregnant population, CSII would also presumably produce tighter glycaemic control in a pregnant population, thus leading to better pregnancy outcomes. However, to date there has been no study or high-level evidence to show that CSII is superior to MDI, with most existing studies using either a small population, have focused primarily on glycaemic control at intervals through pregnancy, or examined a limited scope of pregnancy outcomes.^4,5

The aim of the present study was to compare maternal and neonatal outcomes in women with T1DM using either MDI or CSII throughout pregnancy. To achieve this we conducted a retrospective cohort study at the Royal Women’s Hospital (RWH), in Melbourne, Australia, a large tertiary referral hospital with a specialised diabetes service.

Methods

We examined singleton pregnancies of women with T1DM, whose booking visit fell between January 2006 and December 2016. Deidentified patient data were retrieved from the RWH Diabetes Database. The database collects pregnancy data from patient records for all women with gestational, type 1, and type 2 diabetes and is subject to routine internal audit. Inclusion criteria required that both the entire antenatal care schedule and delivery occurred at RWH. Women were recorded as using either MDI or CSII therapy at their initial booking visit. Patients were excluded if their booking visit occurred after 12 weeks’ gestation or treatment crossover occurred.

At RWH, a specialised diabetes service provides a multidisciplinary team and a twice-weekly clinic delivering obstetric care for women through pregnancy. All patients see an obstetrician, endocrinologist, diabetes educator, dietician, and physiotherapist. In addition to initial routine pregnancy care, the initial booking visit comprises diabetes education advising glycaemic targets of ≤5.0 mmol/l fasting and ≤ 6.7 mmol/l 2 h after meals, achieved with insulin therapy as either basal-bolus (with at least one dose of medium/long-acting insulin each day and short/rapid-acting insulin before each main meal) or CSII. Patients are also advised to undertake 30 min of exercise (e.g. brisk walking) at least four times per week. Complications of diabetes are reviewed: nephropathy is assessed through a timed urine sample or albumin/creatinine ratio and serum creatinine; retinopathy is assessed by referral to an ophthalmologist, optometrist, or fundal photos; and neuropathy is assessed clinically and investigated as appropriate. These components are documented in our standardised clinical practice guideline for all women with pre-gestational diabetes that is used to guide care for the duration pregnancy. We reviewed superseded institutional guidelines and noted no significant change in practice across the period of study.

The CROWN database was referenced to locate an appropriate core outcome set (COS). We identified one COS for reporting in pre-gestational diabetes; however, as the focus was on effectiveness of pre-pregnancy care, it was not suitable for use in our present study.⁶ Miscarriages were reported for descriptive purposes only and not included in subsequent analysis. Primary maternal outcomes of interest included hypertensive disorders of pregnancy (composite of gestational hypertension, pre-eclampsia, and HELLP syndrome) and caesarean section. We defined gestational hypertension as new onset of raised blood pressure >140 mmHg systolic or >90 mmHg diastolic after 20 weeks’ gestation; pre-eclampsia as new onset of raised blood pressure >140 mmHg systolic or >90 mmHg diastolic after 20 weeks’ gestation with one or more of renal involvement, haematological involvement, hepatic involvement, neurological involvement, pulmonary oedema, or fetal growth restriction; and HELLP syndrome as haemolysis, elevated liver enzymes, low platelets, and ankle/foot clonus. Secondary maternal outcomes of interest included induction of labour (IOL), third or fourth degree perineal tears, and hypoglycaemia requiring admission. Primary neonatal outcomes of interest included LGA above the 90th centile, LGA above the 95th centile (LGA >95), shoulder dystocia, birth injury (injury to neurological or musculoskeletal systems), and perinatal mortality. Secondary neonatal outcomes of interest included small-for-gestational age (SGA), preterm birth (birth before 37 weeks’ gestation), very preterm birth (birth before 34 weeks), jaundice, severe infection, clinical hypoglycaemia, congenital malformations, APGAR <7 at 5 min, respiratory distress syndrome (RDS), requirement of neonatal special care nursery (NSCN) admission, or requirement of neonatal intensive care unit (NICU). Maternal HbA1c values at pre-conception (baseline), first trimester, second trimester, and third trimester were chosen to represent glycaemic control. LGA, LGA >95, and SGA values were all obtained from the Australian national birthweight percentiles by sex and gestational age study.⁷

Ethics

The study was approved by the RWH Research Committee and RWH Human Research Ethics Committee (reference, Project AQA17/19, 2017).

Statistical analysis

Maternal and neonatal characteristics were analysed using descriptive statistics. Discrete variables are reported as n (%), normally distributed continuous variables are reported as mean ± standard deviation, and non-normally distributed continuous variables as median (interquartile range (IQR)). Discrete variables were analysed using Fisher’s exact test or Pearson’s chi-square test. Continuous variables were analysed using either Student’s t-test, Wilcoxon rank-sum test, or Kruskal–Wallis test where appropriate.

Obstetric and neonatal outcomes were analysed using logistic regression, which produced an odds ratio (OR) and 95% confidence intervals (CIs) for each outcome. Women treated with CSII were compared to women treated with MDI (reference group). Univariable logistic regression was initially used to produce a crude (unadjusted) OR. Outcomes with a p < 0.1 were selected for multivariable logistic regression to produce an adjusted OR. All outcomes were adjusted for age and BMI. Further outcome-specific adjustment was applied with significant confounders for their respective outcomes. These included parity, smoking, gestation, and hypertensive disorders of pregnancy. Statistical significance was set at a P-value <0.05 (two-sided). All analyses were performed using Stata Version 14.1 (StataCorp, College Station, TX, USA).

BMI data in 39 women were missing. To minimise bias inherent to complete case analysis, we applied a multiple imputation (MI) method to the multivariable analyses, widely regarded as best practice in managing missing data.^8,9 We found no systematic explanation for absent values, hence the missing at random assumption was plausible and applied. We included the other covariates used in the regression analyses to create 25 MI sets of the missing data, produced through MI by chained equations technique given the non-normal distribution. The multiple logistic regression models were subsequently run on each filled-in data set and using the MI estimate function in Stata to combine the parameter estimates and standard errors into a single set of results.

Results

Participants

There were 298 pregnancies in total: 287 (96.31%) live births, five (1.68%) miscarriages, and six (2.01%) stillbirths. One hundred and ninety-six (66.89%) women were treated with MDI and 97 (33.11%) with CSII. There was no difference between groups for rate of miscarriage or stillbirths.

Patient characteristics

Baseline patient characteristics are presented in Table 1. Women across the two groups were of similar age (CSII 29.8 versus MDI 29.7) and parity. Median BMI was 26 (IQR; MDI 23,29 versus CSII 23,30) for both groups. Women treated with CSII had a median duration of diabetes 3 years longer (16.0 versus 13.0, p = 0.036), and no difference in the incidence of diabetes complications. Patients using CSII had a median pregnancy gestation that was five days shorter than their MDI counterparts (37.0 weeks versus 37.7 weeks, p = 0.002). Both groups had similar attendance rates at pre-pregnancy clinic within one year prior to conception (MDI 17.86% versus CSII 23.71%). The number of smokers prior to pregnancy and during pregnancy was similar, and both groups saw a reduction of smokers during pregnancy. Use of folate supplementation was similar prior to conception and throughout pregnancy. Maternal inpatient length of stay was similar across both groups.

Table 1.

Maternal and neonatal characteristics.

Variable	MDI (n = 198)	CSII (n = 100)	P-value
Maternal age (years)	29.7 ± 5.10	29.8 ± 4.94	0.899
Maternal BMI (kg/m2)	26 [23,29]	26 [23,30]	0.651
Maternal diabetes duration (years)	13 [7,20]	16 [9,22]	0.036
Diabetic retinopathy	39 (19.90)	18 (18.56)	0.785
Diabetic nephropathy	25 (12.76)	12 (12.37)	0.926
Diabetic neuropathy	7 (3.57)	4 (4.12)	0.815
Primiparous	90 (45.92)	54 (55.67)	0.116
Gestation (weeks)	37.7 [36.4,38.1]	37.0 [35.6,37.9]	0.002
Attended pre-pregnancy clinic	35 (17.86)	23 (23.71)	0.237
Smoking pre-pregnancy	40 (20.41)	14 (14.44)	0.346
Smoking during pregnancy	25 (12.76)	7 (7.22)	0.346
Folate pre-pregnancy	71 (36.22)	49 (50.52)	0.064
Folate during pregnancy	154 (78.57)	82 (84.54)	0.075
Aspirin	22 (11.11)
Maternal post-partum LOS	4 [3,6]	5 [3,7]	0.346
Neonate male gender	100 (51.02)	42 (43.30)	0.416
Neonate birthweight (g)	3496 [3030,3880]	3438 [3058,3972]	0.572

Attended pre-pregnancy clinic: attended a pregnancy planning clinic at RWH within one year prior to conceiving; CSII: continuous subcutaneous insulin infusion; maternal age: maternal age at conception; maternal BMI: maternal body mass index at booking; maternal diabetes duration: maternal duration of diabetes in years; maternal post-partum LOS: maternal length of post-partum inpatient stay; MDI: multiple daily injection; N: number.

Values expressed as n (%), mean ± SD, or median [IQR].

Obstetric and neonatal outcomes

Obstetric outcomes are presented in Table 2. CSII treatment did not reduce the risk of a maternal hypoglycaemia episode requiring hospital admission (OR 0.79, 95% CI 0.42–1.51, p = 0.483). There was no difference between groups for developing hypertensive disorders of pregnancy. Risk of IOL was comparable between groups. CSII use did not convey any difference in third or fourth degree perineal tears. CSII use also did not increase the risk of caesarean section (OR 1.51, 95% CI 0.86–2.66, p = 0.155), and importantly there was no difference in the risk of requiring an emergency caesarean section.

Table 2.

Maternal outcomes of women with T1DM using CSII compared to MDI.

Outcomes	MDI (n = 198)	CSII (n = 100)	P-value	Crude OR	P-value	Adjusted OR	P-value
Hypo admission	16 (8.16)	5 (5.15)	0.472	0.61 [0.22,1.72]	0.351	–	–
Hypertension	38 (19.39)	22 (22.68)	0.511	1.22 [0.67,2.21]	0.510	–	–
IOLa	48 (76.19)	20 (83.33)	0.471	1.50 [0.44,5.14]	0.519	–	–
Dystociaa	2 (3.33)	0 (0.00)	1.00	–	–	–	–
Perineal tears (third/fourth)	0 (0.00)	1 (4.55)	0.268	–	–	–	–
Caesarean	133 (67.86)	73 (75.26)	0.192	1.51 [0.86,2.66]	0.155	–	–
Electiveb	68 (34.69)	40 (41.24)	0.271	1.84 [0.93,3.65]	0.082	1.83 [0.92,3.64]	0.087
Emergency	65 (33.16)	33 (34.04)	0.271	1.61 [0.80,3.25]	0.183	–	–

CSII: continuous subcutaneous insulin infusion; hypertension: hypertensive disorders of pregnancy composite; hypo admission: hypoglycaemic episode requiring admission to hospital; hypo unawareness: episode of hypoglycaemic unawareness during pregnancy; IOL: induction of labour; MDI: multiple daily injection; N: number; third/fourth degree tears: third or fourth degree perineal tears; T1DM: type 1 diabetes mellitus.

Values expressed as n (%), crude and adjusted odds ratio (OR), and 95% confidence interval (CI).

All outcomes adjusted for age and BMI.

^aReported only for vaginal deliveries.

^bAdjusted for smoking and hypertensive disorders of pregnancy.

Table 4.

Maternal HbA1c values through pregnancy.

Variable	MDI (n = 198)	CSII (n = 100)	P-value
Conception HbA1c	7.3 [6.6,8.5]	7.3 [6.7,8]	0.860
First trimester HbA1c	6.8 [6.2,7.6]	6.8 [6.25,7.5]	0.867
Second trimester HbA1c	6.5 [5.8,7.2]	6.5 [6,7.1]	0.980
Third trimester HbA1c	6.5 [5.9,7.4]	6.8 [6.2,7.4]	0.201

CSII: continuous subcutaneous insulin infusion; MDI: multiple daily injection

Values expressed as median [IQR].

Neonatal outcomes are presented in Table 3. Women using CSII had both a higher incidence (58.95% versus 41.71%, p = 0.010) and higher risk of having an LGA neonate (aOR 2.00, 95% CI 1.20–3.34, p = 0.008). CSII treatment also increased the risk of LGA >95 neonates (aOR 2.05, 95% CI 1.22–3.45, p = 0.007). CSII did not increase the risk of birth injury, neonatal hypoglycaemia, jaundice, RDS, or significant infection. There was no difference between therapies for SGA neonates. There was no difference in Apgar score <7 at 5 min between treatment groups. CSII therapy increased the incidence (44.21% versus 32.29%, p = 0.048) and risk (aOR 1.80, 95% CI 1.04–3.03, p = 0.034) of preterm birth. Notably, there was no increased risk of very preterm birth with CSII. Finally, despite earlier gestation at delivery and increase in LGA risk, CSII therapy did not convey an increased risk of requiring admission for intensive neonatal care, in either the NSCN or NICU. HbA1c values through pregnancy are presented in Table 4. We observed no difference in HbA1c values between groups at conception and all trimesters.

Table 3.

Neonatal outcomes of women with T1DM using CSII compared to MDI.

Outcomes	MDI (n = 198)	CSII (n = 100)	P-value	Crude OR	P-value	Adjusted OR	P-value
LGAa	82 (42.71)	56 (58.95)	0.010	1.93 [1.17,3.17]	0.010	2.00 [1.20,3.34]	0.008
LGA>95a	59 (30.73)	44 (46.32)	0.010	1.94 [1.17,3.23]	0.010	2.05 [1.22,3.45]	0.007
SGA	3 (1.56)	1 (1.05)	0.729	0.67 [0.07,6.53]	0.730	–	–
Apgar<7b	15 (7.81)	6 (6.38)	0.663	0.80 [0.30,2.15]	0.664	–	–
Pretermc	62 (32.29)	42 (44.21)	0.048	1.66 [1.00,2.75]	0.049	1.80 [1.04,3.03]	0.034
Very pretermc	16 (8.33)	12 (12.63)	0.248	1.59 [0.72,3.51]	0.251	–	–
Neonatal hypo	102 (53.13)	53 (55.79)	0.670	1.11 [0.68,1.83]	0.670	–	–
Jaundiced	82 (42.19)	54 (56.84)	0.019	1.80 [1.10,2.97]	0.020	1.68 [0.99,2.87]	0.056
Phototherapy	52 (62.65)	37 (68.52)	0.482	1.30 [0.63,2.68]	0.482	–	–
Malformation	18 (9.38)	7 (7.37)	0.571	0.77 [0.31,1.91]	0.571	–	–
Infection	34 (17.71)	21 (22.11)	0.373	1.32 [0.72,2.43]	0.374	–	–
Birth injury	4 (2.08)	0 (0.00)	0.157	–	–	–	–
Resp distress	15 (7.81)	6 (6.32)	0.647	0.80 [0.30,2.12]	0.647	–	–
NSCN	97 (50.52)	49 (51.58)	0.866	1.04 [0.64,1.71]	0.866	–	–
NICU	29 (15.10)	14 (14.74)	0.935	0.97 [0.49,1.94]	0.935	–	–
Miscarriagee	2 (0.98)	3 (3.00)	0.193	–	–	–	–
Stillbirthe	4 (1.98)	2 (2.06)	0.962	–	–	–	–

Apgar <7: Apgar <7 at 5 min; birth injury: damage to musculoskeletal or neurological systems attributed to delivery; CSII: continuous subcutaneous insulin infusion; infection: significant infection requiring active intervention; LGA: large-for-gestational age; LGA >95: large-for-gestational age greater than 95th centile; malformation: congenital malformation; MDI: multiple daily injection; miscarriage: loss of pregnancy <20 weeks’ gestation; N: number; neonatal hypo: neonatal hypoglycaemia; NICU: admission to the neonatal intensive care unit; NSCN: admission to the neonatal special care nursery; phototherapy: jaundice requiring treatment with phototherapy; preterm: preterm birth <37 weeks’ gestation; resp distress: respiratory distress syndrome; SGA: small-for-gestational age <10th centile; stillbirth: loss of pregnancy >20 weeks’ gestation; T1DM: type 1 diabetes mellitus; very preterm: preterm birth <34 weeks’ gestation.

Values expressed as n (%), crude and adjusted odds ratio (OR), and (95% confidence interval (CI)).

All outcomes adjusted for age and BMI.

^aAdjusted for parity and smoking.

^bReported only for vaginal deliveries.

^cAdjusted for smoking and hypertensive disorders of pregnancy.

^dAdjusted for parity and gestation.

^eFor descriptive purposes only.

Comment

Main findings

Surprisingly we found that CSII use in T1DM during pregnancy significantly increased the risk of having both an LGA and an LGA > 95 neonate. CSII use also increased the risk of preterm birth, however did not increase the risk of very preterm birth. Most importantly, we observed no increased risk of associated adverse obstetric or neonatal outcomes among the CSII cohort. Women using CSII also tended to have a longer duration of diabetes. We observed no difference in HbA1c between groups at conception, nor during each trimester.

Strengths and limitations

To our knowledge, this is the largest single-centre cohort study examining the impact of insulin delivery systems on obstetric and neonatal outcomes.^10,11 Producing data from a single centre was an important aspect of our study, as significant between-centre practice variation observed in pregnancy outcomes and glycaemic control can introduce treatment bias.¹² Maternal characteristics were similar between treatment groups, except that CSII women tended to have a longer duration of diabetes, yet without any difference in macrovascular or microvascular complications of diabetes. We established a cohort of patients whose pregnancy care was uniform, to represent the impact of each treatment as accurately as possible. When dealing with missing data, we applied MI which ultimately underestimates ORs for outcomes, rather than the complete case method, which would add a selection bias to the results. We analysed standardised outcomes that increased generalisability of our results. Our regression models used evidence-based covariates to adjust for confounding.

In existing studies examining CSII versus MDI use in pregnant women with T1DM, glycaemic control is the primary comparator for determining superiority. Contrary to this, we opted to focus on clinical outcomes rather than HbA1c as has been previously recommended.¹³ Despite wide acceptance that hyperglycaemia correlates with adverse pregnancy outcomes, particularly first trimester hyperglycaemia with spontaneous abortion and congenital anomalies,¹⁴ there is no clear HbA1c threshold under which a significant reduction in adverse pregnancy outcomes is observed.^13,15,16 Previous studies have found associations between HbA1c and adverse pregnancy outcomes throughout varied points in pregnancy.¹⁷ In lieu of evidence-based HbA1c/glycaemia targets through pregnancy, the task of defining targets becomes the burden of expert consensus recommendations.¹⁵ This can lead to an inherent bias when interpreting results as achieving HbA1c/glycaemia ‘targets’ through pregnancy will correlate with an increase in adverse outcomes.^18,19 Finally, limitations to the HbA1c test itself fail to capture important changes in glycaemia that could significantly impact pregnancy outcomes.¹⁸

We acknowledge several limitations including the retrospective nature of our study which may introduce selection bias, recall bias, and constrained to data collected a priori. We specifically recognise the lack of data on maternal weight change through pregnancy, as this has been highlighted as a confounder for several pregnancy outcomes, specifically LGA neonates.²⁰ Similarly, there is evidence that maternal ethnicity may be a confounder for adverse pregnancy outcomes; however, of the few maternal ethnicity data that were obtained, significant inaccuracy was present where most commonly nationality was reported rather than ethnicity. Furthermore, we note that our sample size may be underpowered to identify differences in rare complications such as shoulder dystocia and very preterm birth. We also acknowledge that our findings may be impacted by inherent factors of a large subspecialist tertiary hospital, experienced in pre-empting and well equipped to manage adverse pregnancy outcomes. Finally, we acknowledge that rationale for using either insulin delivery modality was lacking, specifically acknowledging an often stepwise transition from MDI to CSII where labile blood glucose persists despite strict adherence, and therefore potential unidentified confounders were controlled for by study design and statistical analysis which is suboptimal.

Interpretation

To our knowledge this is the first study demonstrating that CSII therapy significantly increases the risk of having an LGA neonate in women with T1DM. Importantly, in the CSII cohort we observed no increased risk of adverse outcomes commonly associated with LGA neonates. Previous studies have found mixed results of CSII use impacting the risk of having a LGA neonate. Some have found a comparable risk of LGA between CSII and MDI.²¹ However, others also report no difference between treatment groups, suffer from unaddressed confounding issues, selective reporting bias, or most frequently an underpowered study design.^5,11,22 Inconsistent reporting of pregnancy outcomes has made comparability of findings difficult, particularly when addressing macrosomia or LGA, as definitions and study designs vary markedly.^4,21 An interesting finding has been a persistence of LGA in women with T1DM despite tight antenatal glycaemic control.^19,23 The extended Pederson hypothesis offers a potential explanation for these findings, expanding beyond glucose to include other sugars and factors such as amino acids, lipids, and products of carbohydrate metabolism, involved in the materno-fetal metabolome that can influence fetal growth. Many components of this metabolome have shown independent associations with increased neonate size at birth.^24,25 Implementation of new technology such as continuous glucose monitoring (CGM) has shown promise in recognising maternal glycaemic profiles associated with LGA neonates.²⁶ Furthermore, with the addition of CGM values into existing diabetes therapies, both an improvement in glycaemic control and reduction in risk of macrosomia has been observed.²⁷

A further novel finding was the increased risk of preterm birth in the CSII cohort, without an increased risk of neonatal RDS, hypoglycaemia, jaundice, or requirement for NSCN/NICU. Other studies have suggested an increase in preterm birth with CSII use, however lack statistical power to show significance.²⁸ We recognise that our institution guidelines may, however, introduce a selection bias: women with optimal glycaemic control and uncomplicated pregnancy should deliver at 38–40 weeks, and women with any complicating factor should deliver at 38–39 weeks or earlier if indicated. Complicating factors include poor glycaemic control, microvascular or macrovascular disease, hypertension, fetal macrosomia, intrauterine growth restriction, or smokers. Recent NICE guidelines advise women with no other complications to deliver between 37 + 0 weeks and 38 + 6 weeks, with consideration of delivery before 37 weeks for women with T1DM if there are metabolic or any other maternal or fetal complications.²⁹ It is possible that this CSII cohort are having earlier elective deliveries, given the higher incidence of LGA neonates. Furthermore, while not significant, the CSII cohort had an increased proportion of caesarean sections (75.26% versus 67.86%). Reassuringly, this difference may be occurring from an increased number of elective rather than emergent caesarean sections. Clinical risk aversion may warrant elective caesarean section over vaginal deliveries due to the increased risk of disproportionate growth in LGA neonates of T1DM mothers, to prevent birth trauma to both mother and neonate.³⁰

Of note was the low interaction with the pre-pregnancy clinic prior to conception for both groups. There is very clear evidence that pre-pregnancy planning and optimisation can lead to a significant reduction in adverse pregnancy outcomes.^22,31 Despite a significant difference in aspirin use between groups, no significant difference was noted for hypertensive disorders of pregnancy. We had incomplete data on both gestation commenced and indication for commencement; however, in the limited data (not presented) we noted a high degree of variance in gestation of aspirin being commenced. Furthermore, not only is there evidence that commencing prophylactic aspirin prior to 16 weeks’ gestation conveys benefit, but the literature regarding aspirin prevention for adverse pregnancy outcomes suggests even existing meta-analyses are underpowered to provide any results of clinical importance.³² While no significant difference in neonatal hypoglycaemia was observed, this finding could be a reflection of our institutional definition where we record only neonatal hypoglycaemia of clinical significance that requires an active intervention to correct. Furthermore, a non-significant difference could be a result of an underpowered sample size.

Finally, with no clear benefit to either CSII or MDI use in pregnancy, some authors leap to recommend against CSII use in pregnancy, basing this on cost incurred to both patient and health services to manage the pump system.²⁸ Existing meta-analyses have shown no difference between the two modalities, but these findings are continually qualified as inconclusive, yet often overlooked by authors citing them.^4,5,33 There is evidence that CSII may be beneficial in the subset of women with complicated diabetes or high-risk pregnancies.¹¹ Furthermore, a recent meta-analysis of cost-effectiveness concludes that CSII is cost effective when compared to MDI in several settings, particularly for high-risk women or those with difficult glycaemic control during pregnancy.³⁴

Conclusion

Women with T1DM using CSII during pregnancy are at an increased risk of having LGA, LGA > 95, and preterm neonates, with no increase in risk of associated adverse maternal or neonatal outcomes. These results provide insight into the clinical impact of each insulin delivery modality, allowing for more accurate pregnancy planning and management in women with T1DM. We can recommend CSII use in T1DM through pregnancy; however, the decision should be based on mother’s health status, requirements for achieving optimal glycaemic control, and most importantly patient preference. Further large, well-designed randomised controlled trials are necessary to provide insight into the benefit of CSII versus MDI use in pregnancy.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Ethical approval

The study was approved by the RWH Research Committee and RWH Human Research Ethics Committee (reference, Project AQA17/19, 2017). The Committee deemed individual patient consent was not required.

Guarantor

BRSD

Contributorship

BRSD and TJC conceived the study and its design. AN provided assistance with study design. BRSD conducted the analysis, wrote the manuscript, and is corresponding author for the study. SCNH made a significant contribution to data acquisition. SCNH and AN provided access to the data. BRSD, TJC, and AN contributed to the interpretation of the data. TJC, AN, and SCNH provided revisions to the manuscript. All authors approved submission of the final draft.