Characterizing maternal glycemic control: a more informative approach using semiparametric regression

Abstract

Objective: To characterize glucose concentrations measured throughout pregnancy in women with type 1 diabetes using semiparametric regression analysis to examine the gestational time-specific association with fetal outcome. Methods: We conducted a secondary analysis of data from an interdisciplinary program of diabetes in pregnancy of women with type 1 diabetes. Semiparametric regression was used to characterize glucose concentrations measured using reflectance meters throughout pregnancy by examining the time-specific association of maternal glucose with delivery of a large for gestational age (LGA) baby. Results: The optimal model demonstrated that time-specific differences in glycemic profiles of mothers who had LGA versus AGA babies changed at various rates across gestation (p = 0.0007). AGA glucose profiles exceeded LGA profiles in the first trimester and mid pregnancy; conversely LGA glucose profiles exceeded AGA profiles initially during the third trimester. Differences were based on examination of 95% simultaneous confidence bands. Conclusions: Semiparametric regression techniques enabled synchronous inclusion of all glucose concentrations using multi-step modeling. We identified specific periods of gestation where maternal glucose concentrations differ for the LGA and AGA developing fetus, with greatest distinctions appearing in first and third trimesters. Novel statistical approaches that examine time-specific behavior garner insight into longitudinal assessment of maternal glycemic control.

Keywords::

• Functional data analysis
• mean evolution
• simultaneous confidence bands
• glucose curves
• maternal hyperglycemia
• large for gestational age

Acknowledgements

We are grateful for the support to conduct this research. The research was supported in part by (1) an Institutional Clinical and Translational Science Award, NIH/NCRR Grant Number 5UL1RR026314-03, and (2) Diabetes in Pregnancy Program Project Grant (PPG), HD11725, funded from 1978 to 1995 by the National Institutes of Health and U.S.P.H.S. The contents are solely the responsibility of the authors and do not necessarily represent the official views of the NIH.

Declaration of interest: The authors have no conflicts of interest to report for this work.