Efficacy of automated insulin delivery in pregnant women with type 1 diabetes: a meta-analysis and trial sequential analysis of randomized controlled trials.

Abstract

Automated insulin delivery (AID) devices have shown to be a promising treatment to improve glycemic control in patients with type 1 diabetes mellitus (T1DM). However, its efficacy in pregnant women with T1DM remains uncertain. PubMed, Scopus, Cochrane Central and ClinicalTrials.gov were systematically searched for randomized controlled trials (RCTs) comparing AID to standard care (SC), defined as use of sensor-augmented pump and multiple daily insulin injections. Outcomes included time in range (TIR), nocturnal TIR, time in hypoglycemic and hyperglycemic ranges, among others. Sensitivity and trial sequential analyses (TSA) were performed. CRD42023474398. We included five RCTs with a total of 236 pregnant women, of whom 117 (50.6%) received AID. There was a significant increase in nocturnal TIR (mean difference [MD] 12.69%; 95% CI 8.74-16.64; p < 0.01; I2 = 0%) and a decrease in glucose variability (standard deviation of glucose; MD -2.91; 95% CI -5.13 to -0.69; p = 0.01; I2 = 0%). No significant differences were observed for TIR, HBGI, LGBI, mean glucose and time spent in hyperglycemia and hypoglycemia. Regarding TSA, the statistical significance obtained in nocturnal TIR was conclusive and with minimal risk of a type I error. Our findings suggest that AID systems can significantly improve nocturnal glycemic control and potentially reduce glycemic variability in pregnant women with T1DM, with no effect in the risk of hypoglycemia and hyperglycemia compared with current insulin treatments.