Effect of biological variation in HbA1c and blood glucose on the diagnosis of prediabetes

Abstract Introduction People with a low or high haemoglobin glycation index (HGI) have lower or higher HbA1c than other people with the same FPG. This study compared the prevalence of prediabetes based on FPG, 2hOGTT and HbA1c in people with low, moderate or high HGI. Methods Prediabetes was diagnosed based on ADA cutpoints in 10,488 NHANES participants without self‐reported diabetes. HGI was calculated as the difference between a participant's observed HbA1c and a predicted HbA1c where predicted HbA1c = 0.024 FPG + 3.1. Participants were divided into low (HGI < −0.15%), moderate (HGI −0.15% to +0.15%) and high (HGI > +0.15%) HGI subgroups. Results The prevalence of prediabetes was 42.4% based on FPG, 27.2% based on HbA1c and 17.2% based on 2hOGTT. FPG and HbA1c thus overdiagnosed prediabetes by 25.2% and 10.0%, respectively, compared to the OGTT gold standard. Prevalence was (1) similar in low, moderate and high HGI participants based on 2hOGTT, (2) highest in low HGI participants based on FPG, and (3) highest in high HGI participants based on HbA1c. Among participants with mismatched FPG and HbA1c, OGTT was normal in (1) 79.5% of participants with normal FPG but prediabetic HbA1c (mean HGI = +0.53%), and (2) 75.2% of participants with normal HbA1c but prediabetic FPG (mean HGI = −0.30%). Conclusions FPG overdiagnosed prediabetes in people with low HGI. HbA1c overdiagnosed prediabetes in people with high HGI. Clinical use of HGI could improve prediabetes diagnosis and help health care providers avoid inappropriate or delayed treatment of people with extremes of HGI.

of what can be expected in metabolically normal people compared to those with Type 2 diabetes. 1 Periodic prediabetes screening is now recommended for all people between 35 and 70 years of age based on FPG, 2hOGTT or HbA1c. 2 The problem with this multimetric approach is that prediabetes diagnoses based on these three common clinical tests often disagree. 3-6

| Are two metrics better than one?
In an effort to develop a more specific, timely and unambiguous approach to diabetes diagnosis, Selvin et al. 7 proposed a twodimensional, single-sample definition of confirmed undiagnosed diabetes as having both FPG and HbA1c levels above their respective ADA diabetes cutpoints. The investigators used this confirmationby-agreement approach to analyse data from the Atherosclerosis Risk in Communities study and found that 'A single-sample confirmatory definition of diabetes had a high positive predictive value for subsequent diagnosis and was strongly associated with clinical end points'. The authors defined unconfirmed undiagnosed diabetes as having either FPG or HbA1c but not both above their respective cutpoints and suggested that diagnostic ambiguity should be addressed by monitoring FPG and HbA1c over time. Figure 1A graphically depicts the linear relationship between blood glucose and HbA1c observed in a 2002 longitudinal study of paediatric of Type 1 diabetes patients. 3 Figure 1B shows results from three individuals within the population who had HbA1c levels that were consistently lower than average (green), average (blue) or higher than average (red) across a wide range of blood glucose concentrations. Many individuals in the study population had individual regression lines with 95% confidence intervals that did not overlap the population regression line or the confidence intervals of other individuals. A vertical line drawn across all three individual regression lines at any blood glucose level shows that the HbA1c range was about 5%. Similarly, a horizontal line drawn across all three personal regression lines at any HbA1c level shows that the blood glucose range was about 200 mg/dL. One interpretation is that some people have consistently lower or higher HbA1c levels compared to other people with the same blood glucose concentration. Alternatively stated, some people have consistently lower or higher blood glucose concentrations compared to other people with the same HbA1c.

| Person-to-person variation in HbA1c and blood glucose
Either way, persistent person-to-person variation in the quantitative relationship between blood glucose and HbA1c is a fact of nature that complicates the clinical diagnosis of prediabetes and diabetes.

| The Haemoglobin Glycation Index (HGI) measures person-to-person variation in HbA1c and blood glucose
HGI was introduced as a way to quantify the magnitude and direction of population variation in the quantitative relationship between blood glucose and HbA1c. 3 HGI is calculated as the difference between an individual's observed HbA1c and a predicted HbA1c obtained by inserting a date-matched blood glucose concentration into a linear regression equation describing the quantitative relationship between HbA1c and blood glucose in a reference population. This approach naturally incorporates information contained in HbA1c and FPG measurements into a single value. As reviewed in greater detail elsewhere 8,9 HGI varies between races, ranges from about −2.0% to +2.0% in populations without known diabetes and tends to be consistent within individuals over time and over the physiological range of blood glucose concentrations ( Figure 1B). HGI is simple

Moderate HGI
to calculate and has potential clinical application because high HGI is associated with both (1) greater risk for hypoglycaemia in people with diabetes, and (2) greater risk for chronic vascular disease in people with or without diabetes.

| ME THODS
This study compared single-metric and two-dimensional diagnosis of prediabetes and diabetes based on ADA cutpoints for FPG, 2hOGTT and HbA1c in people with low, moderate or high HGI.

| Study population
Data from the 1999-2016 cohorts of the National Health and Nutrition Examination Survey (NHANES) were used to identify participants who (1) were at least 20 years of age, (2) had FPG, 2hOGTT and HbA1c data and (3) had not been previously diagnosed with diabetes. NHANES is an ongoing national survey directed by the Centers for Disease Control that uses a stratified multi-stage probability sampling design to represent the non-institutionalized U.S. civilian population. 10 The National Centers for Health Statistics Ethics Review Board approved the NHANES study protocol and each participant provided written informed consent.

| Calculating HGI
HGI is the mathematical difference between an individual's observed diabetes. 9 The current study population was subdivided into three roughly equal sized groups of participants with low (HGI < −0.15%), moderate (HGI −0.15% to +0.15%) or high (HGI > +0.15%) HGI based on cutpoints derived from the reference population.

| Diagnosis of prediabetes and diabetes
Single-metric diagnosis of prediabetes and diabetes was based on the ADA cutpoints listed in

| Data analysis
Descriptive statistics were used to describe demographic and other characteristics of the study cohort. Categorical variables were sum-      Figure 3D graphs the relationship between HbA1c and postchallenge glucose increment (2hOGTT-FPG). Table 2 summarizes the number of observations and mean HGI of the three nine-compartment diagnostic models produced by the two-dimensional combinations of FPG, 2hOGTT and HbA1c as depicted in Figure 3A-C. The combination of FPG + HbA1c (    participants based on the observation that 2hOGTT was normal in 79.5% of the 940 mismatched participants diagnosed as normal by FPG but prediabetic by HbA1c (mean HGI = +0.53%). In contrast,

| Biological variation complicates two-dimensional diagnosis of prediabetes
HbA1c was a more accurate predictor of prediabetes in low HGI participants based on the observation that 2hOGTT was normal in 75.2% of the 2632 participants diagnosed as normal by HbA1c but prediabetic by FPG (mean HGI = −0.30%).
Diagnoses based on the clinically feasible two-dimensional combination of FPG + HbA1c (Table 2A)   and prediabetes reference ranges, had HbA1c <7.0% ( Figure 3C). Figure 3D shows there was little or no quantitative relationship between HbA1c and postchallenge glucose increment in the HbA1c response to increasing 2hOGTT-FPG. We conclude that there is a wide range in postchallenge glucose increment over which the magnitude of glycaemic excursion had little or no effect on HbA1c in participants with low, moderate or high HGI.

CO N FLI C T O F I NTE R E S T S TATE M E NT
The authors have no conflicts to disclose.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are openly available