Ethnic differences in HbA1c in adults in New Zealand: a cross-sectional study

Aim: To determine whether there were ethnic differences in HbA1c concentrations in adults with normal and abnormal glucose tolerance. Methods: Cross-sectional data were from 3,559 participants not previously diagnosed with diabetes aged 35-74 years. Participants underwent a 75 gm oral glucose tolerance test. Glucose and fructosamine levels were determined enzymatically, and HbA1c by cation exchange high performance liquid chromatography. Results: Compared with Europeans, and after adjusting for age and gender, HbA1c levels were on average 5.5 (se=0.25) mmol/mol higher for Pacific, 2.5 (0.24) mmol/ mol higher for Māori, and 2.3 (0.40) mmol/mol higher for Asians. These ethnic differences were attenuated in a multivariable regression model for HbA1c, mainly due to the inclusion of current smoking habit and BMI, but still retained statistical significance. HbA1c levels were higher in Māori, Pacific and Asian participants with normal glycaemia, Māori and Pacific people with impaired fasting glycaemia, and impaired glucose tolerance, and Pacific people with newly diagnosed diabetes compared to Europeans after adjusting for age, gender, fasting and 2 hour glucose. Ethnic differences of HbA1c concentrations increased with increasing glycaemia compared to Europeans. After adjusting for age, gender and body mass index, the squared semi-partial correlations for HbA1c were 50.5% for fasting glucose and 3.6% for 2 hour glucose. Conclusions: HbA1c concentrations in this population mainly reflected fasting glucose levels. The higher HbA1c level for the same degree of glycaemia in nonEuropean ethnic groups has clinical relevance for both diagnosis and management. has management implications, for example on rates of adverse events, as suggested by a recent post-hoc analysis of the Action to Control Cardiovascular Risk in Diabetes (ACCORD) trial [22]. These interindividual and inter-ethnic differences therefore have significant clinical, as well as resource and health-care cost, implications. The purpose of this study was to examine whether there were ethnic differences in HbA1c levels in a New Zealand population. Materials and Methods


Introduction
Māori and Pacific people living in New Zealand have a higher prevalence of Type 2 diabetes mellitus (DM) compared to Europeans [1,2]. They are more likely to develop renal disease and renal failure requiring dialysis [3][4][5] and the prevalence of diabetic retinopathy is higher compared to Europeans [3,5,6].
Ethnic differences in mean glycated haemoglobin (HbA 1c ) levels have been previously described in people without diabetes in the U.S. [7][8][9][10] and in people with impaired glucose tolerance or diabetes [11,12]. Ethnic differences have been reported that account for variation in HbA 1c of up to 4 mmol/mol [11] for similar levels of glycaemic control. While intra-individual biological variation of HbA 1c is small and changes little over time, inter-individual variation between patients, for the same degree of glycaemic control, is much larger [13,14]. Possible explanations include subtle differences in entry of glucose into red blood cells [15] and varying red cell lifespan in the circulation [16,17]. Genome-wide association analysis shows that only a minority of loci related to raised HbA 1c clearly relate to glucose metabolism, so a number of other factors, both physiological and potentially pathological, are likely to be involved [18].
Since HbA 1c is now regarded as either the preferred [19] or acceptable alternative [20] means of diagnosing diabetes, its use to diagnose diabetes can affect population prevalence and which individuals are diagnosed with diabetes in different ethnic groups [21]. Inter-individual variation in the rate of HbA 1c formation also Ethnic differences in HbA 1c in adults in New Zealand: a cross-sectional study has management implications, for example on rates of adverse events, as suggested by a recent post-hoc analysis of the Action to Control Cardiovascular Risk in Diabetes (ACCORD) trial [22]. These interindividual and inter-ethnic differences therefore have significant clinical, as well as resource and health-care cost, implications. The purpose of this study was to examine whether there were ethnic differences in HbA 1c levels in a New Zealand population.

Subjects
The Auckland Diabetes, Heart and Health Survey was carried out between December 2001 and November 2003 in 4,049 randomly sampled adults aged 35 to 74 years (response rate 65%). A more detailed description of the sampling method is provided in Sundborn et al. [2]. Adults with missing HbA 1c results (n=27) and previously diagnosed diabetes (n=463) were excluded leaving 3,559. Participants comprised 47.8% males and 52.2% females, 46.3% Europeans, 24.8% Māori, 22.0% Pacific and 6.9% Asian people. Asian people comprised 20.8% Indian (South Asian) and 79.2% from China, Hong Kong, Korea, Taiwan and the Philippines.
Interviews were carried out in community venues or clinics close to participant's homes. Personnel were trained in the administration of the questionnaires and in taking blood pressure and other measurements. Ethnicity was classified according to the 2006 NZ census [23]. Around 80% of New Zealanders can claim some British ancestry and an estimated 17% are entitled to British passports [23]. The European (white Caucasian) ethnic group also included other white ethnicities such as Caucasian people from South African and the United States of America.

Approval and Consent
The New Zealand Ministry of Health Auckland Ethics Committees granted ethnical approval (NTX/12/EXP/008/AM03). Written informed consent was obtained from all subjects.

Blood collection
Participants fasted from 10pm the evening before the interview. A 75 gm oral glucose tolerance test (oGTT) was carried out in all participants and a fasting and 2 hour post 75gm glucose challenge (Glucaid drink) samples were collected for glucose measurements.

Laboratory measurements
Plasma glucose was measured using an enzymatic method (Glucose oxidase Roche (NZ)). Categorisation of glucose tolerance status was evaluated by 1998 WHO criteria (using fasting glucose ≥ 7.0 mmol/L and/or 2 h post glucose load of ≥ 11.1 mmol/l for diabetes, fasting glucose < 7.0 mmol/L and 2 h glucose between 7.8 to 11.0 mmol/l for impaired glucose tolerance (IGT) and fasting glucose of 6.1 to 6.9 mmol/L for impaired fasting glucose (IFG) [24]. Serum total cholesterol and (direct) HDL-cholesterol were measured using standard colourimetric autoanalyser methods (Roche Hitachi NZ). After phlebotomy, glucose samples were stored on ice and centrifuged within 1 hour of collection.

Other measurements
Weight and height were measured to the nearest 0.1 kg and 0.5 cm, respectively. Body mass index (BMI) was calculated as weight (kg) divided by the square of height (m). An Omron-Hem-706 oscillometric blood pressure monitor was used to measure systolic and diastolic blood pressure twice (5 minutes apart) after at least 15 minutes in the sitting position.

Data analysis
The characteristics of participants by ethnicity were compared using analysis of covariance for normally distributed variables and the χ 2 test for categorical variables. Independent correlation coefficients were compared using Fisher's Z transformation [25]. Multivariable regression was used to determine whether ethnic differences in HbA 1c could be explained by concentrations of other variables. Statistical analyses were carried out using SAS version 9.4 [26]. Table 1 shows the characteristics of participants by ethnicity. Compared to Europeans, Māori, Pacific and Asians were of lower age and had lower total-and HDL-cholesterol concentrations and higher fasting and 2 hour glucose, HbA 1c , and diastolic blood pressure levels. BMI levels were higher in Pacific and Māori participants and lower in Asians compared to Europeans, systolic blood pressure was higher in Pacific and lower in Asian participants, and current smoking levels were higher in Māori and Pacific adults. Prevalence of newly diagnosed diabetes was significantly higher in all three non-European groups. Pacific people had a higher prevalence of impaired fasting glycaemia (IFG).

Results
Correlations between HbA 1c and fasting and 2 hour glucoses in the study group were highly significant (r=0.70 and 0.60, respectively). For each ethnic group they were: Europeans r=0.47 and 0.44; Māori r=0.68 and 0.59; Pacific r=0.84 and 0.73; and Asians r=0.79 and 0.75, respectively. After adjusting for age, gender and body mass index, the proportion of the HbA 1c variance explained in the study population was 50.5% by fasting glucose and 3.6% by 2 hour glucose when both were entered into the multivariable model. Table 2 shows the squared semi-partial correlation coefficients for HbA 1c in the multivariable model for fasting and 2 hour glucose levels by ethnicity and diabetes status after adjusting for age and gender. Apart from IFG in Europeans and Asians, the contribution of fasting glucose dominated the 2 hour glucose contribution. The squared semi-partial correlation coefficients for HbA 1c were highest in those with newly diagnosed diabetes ( Table 2) followed by those with IGT (with the exception of Māori in whom both fasting and 2 hour squared partial correlations were higher for IFT).
Glycaemic measures by diagnostic category and ethnicity are shown in Table 3. Fasting and 2 hour glucose levels were higher in Māori and Asian people with normal glycaemia (NG), but only 2 hour results were significantly higher in Māori and Asian with newly diagnosed diabetes compared to Europeans, respectively. No significant differences for fasting and 2 hour glucoses were seen for Māori and Asian with intermediate degrees of glucose intolerance (IFG, IGT). By contrast, Pacific adults had slightly higher fasting glucose at all degrees of glycaemia, but no significant differences for 2 hour glucose results.
With the exception of newly diagnosed diabetes, Māori with NG, IFG and IGT had higher HbA 1c levels compared to Europeans, which remained after further adjusting for fasting and 2 hour glucose results (Table 3). Pacific adults also had higher HbA 1c levels compared to Europeans, both before and after adjusting for fasting and 2 hour glucoses concentrations in all glycaemic groups. Asians had higher HbA 1c levels compared to Europeans in adults with normal glycaemia and with newly diagnosed diabetes, but that only remained significant in the normal glycaemic group after further adjusting for fasting and 2 hour glucose levels.
In the study population, HbA 1c differences (adjusted for age and gender) for different ethnicities compared with Europeans were +5.5 (se=0.25) mmol/mol for Pacific, +2.5 (0.24) mmol/mol for Māori, and +2.3 (0.40) mmol/mol in Asians. After further adjusting for fasting and 2 hour glucose in the total population, HbA 1c differences were on average +3.6 (se=0.18) mmol/mol for Pacific, +2.5 (0.17) mmol/ mol for Māori, and +0.8 (0.28) mmol/mol for Asians compared with Europeans. When these ethnic groups were included in the multivariable regression model as dummy variables (with Europeans as the reference group) adjusted for all other variables in the model these differences were attenuated (Table 4) with the exception of Asians, but still retained statistical significance. The attenuation of these ethnic differences in HbA 1c was mainly due to the inclusion of current smoking and body mass index. In the multivariate model, Pacific people showed the greatest difference compared to Europeans, followed by Māori and then Asians. The significance levels were unchanged after further adjusting for fructosamine concentrations. Substitution of waist for BMI did not change the conclusions, but the t-value for BMI was greater in both normal glycaemic people and the total study population. Total energy, carbohydrate and alcohol intakes were not related to HbA 1c , nor was physical activity (individual data not shown). The inclusion of interaction terms between each ethnic group and smoking were not statistically significant. Figure 1 shows the relationship between HbA 1c and mean plasma glucose (average of fasting and 2 hour glucose concentrations) concentrations in each ethnic group (using a quadratic regression model). The differences between the European line and Asian, Māori, and Pacific lines increased as mean plasma glucose levels increased, with the greatest difference in Pacific adults. These regression models explained the highest percentage variation of HbA 1c in Asians (70%), followed by Pacific (69%), Māori (49%) and Europeans (26%) (Figure 1).

Discussion
This study shows that HbA 1c concentrations are higher in Māori, Pacific and Asian adults compared to Europeans with both normal and abnormal glycaemia. These higher HbA 1c concentrations persisted after adjustment for fasting and 2 hour glucose concentrations, but were mainly attenuated by current smoking and BMI (Table 4). Further adjustment for fructosamine levels, another marker of glycation outside of the red cell, did not change these significant differences. Compared to Europeans, Maori and Pacific participants had higher mean body mass indices and had significantly more current smokers, whereas Asians had a lower mean body mass index and a lower prevalence of smokers compared to Europeans (Table 1).
Our study adds to the growing evidence for interethnic differences in the rate of HbA 1c formation for the same level of glycaemia. In U.S. adults without diabetes, HbA 1c levels were higher overall in African American men and women compared to non-Hispanic whites [27]. Another study in the US combined two cross-sectional studies and reported that black ethnicity in adults without known diabetes was associated with a 2 mmol/mol higher HbA 1c level in the Screening for Impaired Glucose tolerance study and 3 mmol/mol higher in NHANES III when estimated using a multivariable regression model that included gender, BMI, haemoglobin, level of education, fasting and 2 hour glucose levels [28]. This is similar to the higher levels of 3.1 mmol/mol in Pacific, 1.2 mmol/mol in Māori and 1.2 mmol/mol in Asians compared to Europeans in our own fully adjusted model in adults without known diabetes (Table 4; All) and 3.9 mmol/mol in  Adj HbA 1c and Adj Fructosamine are further adjusted for fasting and 2 hour glucose levels. * 0.01 < P < 0.05; ** 0.001 < P < 0.01; *** P < 0.0001 compared to Europeans. Pacific, 1.6 mmol/mol in Māori and 1.7 mmol/mol in Asians compared to Europeans in the fully adjusted model in adults without abnormal glycaemia (individual data not shown). Furthermore, ethnic differences of HbA 1c concentrations increased with increasing glycaemia compared to Europeans (Figure 1).
A U.K. study of adults without known diabetes, reported that HbA 1c and fructosamine levels were higher in Black than in White adults, even though fasting glucose levels were similar [29]. By contrast, in the normoglycaemic group in the current study some of the ethnic differences were explained by higher glucose concentrations in Māori, Pacific and Asians compared to Europeans, nevertheless the significant differences in HbA 1c persisted after glucose adjustment (Tables 3  and 4). Fructosamine levels were higher only in Asians with normal glycaemia and newly diagnosed diabetes that remained significant after adjusting for fasting and 2 hour glucose levels (Table 3).
Another U.S. meta-analysis that was carried out in adults with impaired glucose tolerance reported that adjusted HbA 1c levels were highest in African Americans, followed by American Indians, Asians, and Hispanics and lowest in non-Hispanic whites [12]. In the current study, HbA 1c levels in participants with impaired glucose tolerance were higher in Pacific, followed by Māori and Asians than in Europeans (Table 3).
A U.S study in adults without diabetes reported significant associations between glycated haemoglobin and older age, male sex, non-Hispanic black ethnicity, hypercholesterolaemia, higher BMI, and lower attained education [30]. HbA 1c has also been reported to be correlated with smoking and clinically overt atherosclerosis, but not caloric intake, physical activity or alcohol intake in a U.S. population [13], as was observed here (individual data not shown). These were similar to the associations with HbA 1c reported in Table 3 in normoglycaemic patients. However, educational attainment failed to reach significance in the current study, and we did not have the ability to measure clinically overt atherosclerosis.
While pathological states leading to a shortened red cell lifespan (e.g. haemolysis, venesection, chronic blood loss) are well recognised to cause misleadingly low HbA 1c levels, some evidence suggests that the inter-individual and possibly ethnic variation in HbA 1c levels may be at least be partly due to differences in the mean age of circulating red cells even in otherwise healthy patients with no haematological evidence of any pathology [31]. Inter-individual differences have also been described in the entry of glucose into red cells, and the steady state intracellular concentration of glucose compared with plasma [16].
A small clinical study has shown that only 19 to 48% of the variance in glycated haemoglobin in normal glycaemic individuals was explained by fasting or 2 hour post glucose load glucose levels [14]. In the current study, the proportion of the HbA 1c variance explained was 50.5% for fasting glucose and 3.6% for 2 hour glucose after adjusting for age and gender. In a small study (of 12 people), 85% of the total variance was due to inter-individual variation, 6% from intra-individual variation, and the remaining 9% from assay variation [29]. It has been hypothesised that ethnic differences may be due to differences in glycation, erythrocyte survival, biological differences, types of lifestyles, health care access and utilization, quality of care, or socioeconomic factors [9,32]. However, heritability studies indicate that there is a significant genetic contribution [33,34], including a majority of genetic loci that appear unrelated to glucose metabolism [34].
When comparing traditional glucose criteria with HbA 1c , Christensen [35] noted a wide variation between different countries and different studies in rates of diagnosis. While some of these differences may be related to study methodologies [35], putative biological explanations also exist. These may include differences in rates of haemoglobinopathy (e.g. higher rates of thalassemia in Africans and Southeast Asians) and nutritional iron deficiency (e.g. higher rates in East Indians). The prevalence of haemoglobinopathies in New Zealand is unknown. Differences between ethnic groups in the frequency of polymorphisms of multiple genetic loci may also contribute [34,36].
On a population basis, these findings have significant potential healthcare resource implications for rates of diagnosis of diabetes. Even in a relatively homogeneous European population in the A 1c -Derived Average Glucose (ADAG) study [37], there was significant variation in the relationship between mean glucose levels and HbA 1c and this variation was greater still when ethnic differences are considered. In view of the potential for HbA 1c results to misclassify patients [36], sole reliance on HbA 1c as a diagnostic criterion has been discouraged by expert groups especially in those in whom the possibility of an erroneous result is increased [19,20].
Inter-individual and inter-ethnic differences may also have relevance to the monitoring of patients, especially those in whom tight control is desired. Differences in glycation rates tend to be consistent for an individual patient and will bias the clinical assessment to be relatively favourable (for those with low glycation rates; i.e. low 'haemoglobin glycation index') or unfavourable (for those with high glycation rates; i.e. high 'haemoglobin glycation index'), especially if not considered carefully along with glucose results [18]. Hempe [18] showed that in a multi-ethnic population of young patients with Type 1 DM, about 30% of patients had HbA 1c results that were statistically significantly higher or lower compared with the overall relationship for that population, potentially leading to errors in management. A retrospective analysis of the ACCORD study, this time in patients with Type 2 DM [22], also showed that in tightly controlled patients hypoglycaemic events were more likely in those forming HbA 1c more quickly (i.e. high haemoglobin glycation index), perhaps by misleading caregivers into believing that mean glucose levels were higher than they truly were.
Some evidence shows that a faster rate of glycation (high glycation index) is associated with a greater prevalence of retinopathy, nephropathy and mortality [38]. The higher rate of complications in non-European ethnic groups such as Māori and Pacific Island patients is well known [3][4][5][6]. It is therefore possible that this association may reflect not only ethnic differences in glucose control, but also different rates of glucose entry and attachment to proteins in multiple tissues (for which faster HbA 1c formation in red cells may be a coincidental but clinically relevant surrogate).
One study reported that the better association between HbA 1c and fasting glucose could be related to the fact that HbA 1c correlated better with fasting glucose in the high range and with 2 hour glucose levels in the low range [39]. However, we did not observe the same association as fasting glycaemia explained a higher proportion of the variation in HbA 1c at all categories of glucose tolerance (apart from Europeans and Asians with IFG: Table 2). Black adults living in the U.S have been reported as having higher fructosamine levels compared to white persons [7]. In the current study, only Asians in the normal glycaemic and newly diagnosed diabetes groups had significantly higher mean fructosamine levels compared to Europeans. We could find no reports of higher fructosamine levels in Asians.

Strengths and Limitations
Strengths of the work include an important clinical research topic, and the inclusion of multiple ethnic groups studied using the same methods, as well as the random selection of participants and a relatively high response rate of 65% into the study. However, the crosssectional design of the study does not allow examination of the effect of duration. Another limitation is the relatively small sample size of Asians, which may have reduced the study's ability to detect differences between this group and Europeans. Further limitations include only one measurement of the glycaemic measures.
Compared to Europeans and after adjusting for fasting and 2 hour glucose concentrations HbA 1c levels were higher in Māori, Pacific and Asians participants without diabetes, Māori and Pacific people with IFG and IGT, and Pacific people with newly diagnosed diabetes. Although attenuated after adjusting for variables known to be associated with HbA 1c levels, the positive bias for HbA 1c formation remained significant in Māori, Pacific and Asian participants. HbA 1c levels were more highly associated with fasting glucose than 2 hour glucose. Furthermore, ethnic differences of HbA 1c concentrations increased with increasing glycaemia compared to Europeans.

Conclusion
HbA 1c concentrations in this population mainly reflected fasting glucose levels. The higher HbA 1c level for the same degree of glycaemic control in non-European ethnic groups has clinical relevance for both diagnosis and management.

Author's contributions
CK, PM and RJ conceptualised and designed the study. CK collected the data. PM analysed the data with input from CK, TK, GS and RJ. All authors were involved in data interpretation. PM and CK drafted the manuscript and all authors critically revised the manuscript. All authors approved the final submitted version and agreed to be accountable for the manuscript.

Ethical approval
All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/ or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

Informed consent
Written informed consent was obtained from all individual participants included in the study.

Funding and Acknowledgements
This work was supported by a grant from the Health Research Council (HRC) of New Zealand (Grant number: 12/223). Grant applicants were: PA Metcalf, T Keneally, C Kyle, RT Jackson. The HRC had no involvement with the study design, data collection, and analysis, interpretation or writing of the manuscript.