A critique of measurement of defective insulin secretion and insulin sensitivity as a precision approach to gestational diabetes

Aims/hypothesis

Precision medicine approaches to gestational diabetes mellitus (GDM) have categorised patients according to disease pathophysiology (insulin resistance, insulin insufficiency or both), and demonstrated associations with clinical outcomes. We aimed to assess whether using enhanced processing to determine indices of insulin secretion and sensitivity is analytically robust, reproducible in a different population, and useful diagnostically and prognostically in clinical practice.

Methods

A total of 1308 pregnant women with one or more risk factors for GDM who underwent a 75 g OGTT at one of nine hospital sites were recruited to this observational study. Specimens were collected for determination of glucose levels using standard and enhanced procedures, HbA_1c and insulin analysis. GDM diagnosis and management followed National Institute for Health and Care Excellence guidance. We categorised women into pathophysiological subtypes: insulin-resistant GDM (HOMA2-S < 25th centile of the population with normal glucose tolerance [NGT]), insulin-insufficient GDM (HOMA2-B < 25th centile), both or neither. We assessed associations with pregnancy outcomes using logistic regression.

Results

Using enhanced specimen handling, 1027/1308 (78.5%) women had NGT, with 281/1308 (21.5%) being classified as having GDM. Of this group, 135/281 (48.0%) had insulin-resistant GDM, 73/281 (26.0%) had insulin-insufficient GDM and 2/281 (0.7%) had both insulin-resistant and insulin-insufficient GDM. Unexpectedly, 71 patients (25.3%) had GDM with both HOMA2-S and HOMA2-B ≥ 25th centile (GDM-neither). This novel subgroup appeared to be relatively insulin-sensitive in the fasting state but developed marked post-load hyperglycaemia and hyperinsulinaemia, suggesting an isolated postprandial defect in insulin sensitivity that was not captured by HOMA2-B or HOMA2-S. Women within most GDM subgroups had comparable pregnancy outcomes to those of normoglycaemic women, and HOMA2-B and HOMA2-S were weak predictors of pregnancy outcomes. Maternal BMI predicted a similar number of outcomes to HOMA2-S, suggesting that there was no additional predictive value in adding HOMA2-S. Similar findings were obtained when using different indices and standard specimen handling techniques.

Conclusions/interpretation

Precision categorisation of GDM using HOMA2-S and HOMA2-B does not provide useful diagnostic or prognostic information, but did distinguish a novel subgroup of patients with GDM, characterised by an isolated postprandial defect in insulin sensitivity.

Graphical Abstract