Impact of demographics and disease progression on the relationship between glucose and HbA1c


    loading  Checking for direct PDF access through Ovid

Abstract

ContextSeveral studies have shown that the relationship between mean plasma glucose (MPG) and glycated haemoglobin (HbA1c) may vary across populations. Especially race has previously been referred to shift the regression line that links MPG to HbA1c at steady-state (Herman & Cohen, 2012).ObjectiveTo assess the influence of demographic and disease progression-related covariates on the intercept of the estimated linear MPG-HbA1c relationship in a longitudinal model.DataLongitudinal patient-level data from 16 late-phase trials in type 2 diabetes with a total of 8927 subjects was used to study covariates for the relationship between MPG and HbA1c. The analysed covariates included age group, BMI, gender, race, diabetes duration, and pre-trial treatment. Differences between trials were taken into account by estimating a trial-to-trial variability component.ParticipantsParticipants included 47% females and 20% above 65 years. 77% were Caucasian, 9% were Asian, 5% were Black and the remaining 9% were analysed together as other races.AnalysisEstimates of the change in the intercept of the MPG-HbA1c relationship due to the mentioned covariates were determined using a longitudinal model.ResultsThe analysis showed that pre-trial treatment with insulin had the most pronounced impact associated with a 0.34% higher HbA1c at a given MPG. However, race, diabetes duration and age group also had an impact on the MPG-HbA1c relationship.ConclusionOur analysis shows that the relationship between MPG and HbA1c is relatively insensitive to covariates, but shows small variations across populations, which may be relevant to take into account when predicting HbA1c response based on MPG measurements in clinical trials.Graphical abstractForest plot showing model estimates for the change in HbA1c at steady state at any given MPG for each covariate compared to a reference value (with point estimate and 95% confidence interval).

    loading  Loading Related Articles