To Pump or Not to Pump in Pregnancy?

Serious adverse pregnancy outcomes (major congenital anomaly, stillbirth, and neonatal death) continue to affect 1 in 13 type 1 diabetes pregnancies, with devastating and costly implications for women, their families, and healthcare funders. ¹ Nationwide studies confirm substantial morbidity with more than five times the background maternity population rates of preterm delivery, birth weight more than the 90th percentile, and neonatal intensive care unit (NICU) admission among infants of mothers with type 1 diabetes. ^{1 –3} These extremely costly complications (£5000 per preterm NICU admission, £3000 per term admission) are potentially preventable by improving maternal glucose control. Outside of pregnancy, there is a large body of evidence suggesting that insulin pump therapy is associated with improved glucose control. ⁴ During pregnancy, insulin requirements fluctuate with gestational changes in insulin sensitivity, leading to reduced doses during the late first and early second trimester followed by increasing insulin resistance requiring higher insulin doses, particularly before meals, as pregnancy advances. ⁵ Insulin pump therapy allows frequent fine tuning of insulin doses and, as suggested by Abell et al., is theoretically an ideal insulin delivery method to improve glucose control during pregnancy.

Abell et al. ⁶ here present results from a retrospective cohort study involving 37 women using insulin pump therapy (with 40 pump pregnancies) and 102 women using multiple daily injections (MDI) (127 MDI pregnancies). The study was performed during 2010–2015 in a National Centre of Diabetes Excellence in Melbourne, Australia, and linked patient level data from clinical and laboratory records with a validated maternal and neonatal outcome data set. Sixty percent of women attended preconception care and all except for one woman in the insulin pump group were experienced pump users. Women in the insulin pump group were seemingly older, had lower body mass index, fewer smokers, and had considerably longer duration of diabetes [median (IQR) 20 (7–22) years vs. 12 (8–20) years] than MDI users, but none of these differences were statistically significant. The prepregnancy maternal HbA1c levels were comparable (7.8% vs. 7.9% or 62 mmol/mol vs. 63 mmol/mol) with a nadir of 6.8% (51 mol/mol) in the second trimester among insulin pumpers and of 6.6% (49 mmol/mol) in the third trimester (between 34 and 37 weeks gestation) in the MDI group. The rates of severe hypoglycemia and diabetic ketoacidosis were low in both groups (4%–5%) with no severe hypoglycemia and one diabetic ketoacidosis (DKA) episode in the insulin pump group. There were no differences in infant outcomes (preterm delivery, birthweight, large for gestational age) between groups.

These data are in contrast to a larger multicenter Canadian study that described the glycemic control and pregnancy outcomes of 113 pump pregnancies compared with 218 on MDI. ⁷ Kallas-Koeman et al. reported a significant difference in first trimester HbA1c levels among pump versus MDI users (6.9% vs. 7.6% or 52 mmol/mol vs. 60 mmol/mol; p < 0.001), which persisted until the third trimester (6.5% vs. 6.8% or 47 mmol/mol vs. 51 mmol/mol; p = 0.002). Rates of DKA were low (2%–3%) with no increased episodes of DKA or of severe hypoglycemia among insulin pumpers, despite their lower HbA1c levels. Approximately 20% more women using pumps achieved HbA1c levels <7% (53 mmol/mol) in each trimester (trimester 1: 62% vs. 38%, trimester 2: 90% vs. 69%, trimester 3: 85% vs. 67%). ⁶ However, there were no differences in infant health outcomes and apparently higher rates of large for gestational age infants (55% vs. 39%; p = 0.007) in the pump group.

Previous systematic reviews and meta-analysis on very small number of women using early generation pumps and MDI really do not help us understand whether or not there are definitive clinical benefits of modern pumps over modern MDI regimens during pregnancy. ^8,9 Abell et al. ⁶ suggest that “selected compliant, motivated and technically sophisticated patients generally manage best,” highlighting the challenges of all retrospective cohort studies, which are inherently limited by selection bias, patient, and physician preferences.

This study, like most previous studies, is limited by lack of statistical power. The sample size for a 0.5% between group differences in maternal HbA1c would require at least 64 women in each arm. This is assuming 80% power, HbA1c SD of 1%, and no gestational changes in HbA1c during pregnancy! Clearly, there are gestational changes in HbA1c, which may not be the most accurate marker of average glucose control during late pregnancy. ¹⁰ To determine whether pump therapy impacts on day-to-day glucose levels would require more detailed continuous glucose monitoring, with validated hypoglycemic and hyperglycemic outcomes as well as markers of glycemic variability.

In terms of infant outcomes, the minimum sample size for a very large (40%) reduction in LGA (large-for-gestational age) from 50% to 30% is 266 women (133 per arm). A smaller reduction would require an even larger sample size. It is, therefore, inappropriate to conclude a lack of improvement in HbA1c and pregnancy outcomes from any of the published data sets. Registry data such as those from the large National Pregnancy in Diabetes that now includes ∼700 pump pregnancies will be of interest to clinicians. ¹¹ However, to advise women “to pump or not to pump” during pregnancy, well-designed randomized clinical trials with appropriate glycemic control outcomes (using continuous glucose monitoring in addition to HbA1c) and adequate power to examine obstetric and neonatal outcomes are needed. As pump therapy becomes increasingly prevalent among women of reproductive years, recruitment to such trials will become more challenging and require large collaborative efforts. The Diabetes Pregnancy community must also decide whether to invest in an evidence base for current pump technology or await the next generation of smarter pumps and automated insulin delivery. ¹²