Diabetes mellitus has been recognized as one of the greatest health threats of this century.1 By the year 2030, 7.7% of the world's population, some 439 million people, are expected to have the disease.2 This epidemic, which is primarily of type 2 diabetes mellitus, is a significant burden on the international community. It has been estimated that US$376 billion was spent on diabetes in 2010, accounting for 12% of global health expenditure.3 Dramatic increases in the prevalence of diabetes are being seen worldwide in almost all populations.
The Gulf region is one of the epicentres of the diabetes epidemic. The United Arab Emirates (UAE) is especially affected, with an estimated prevalence of 21%.4 Figure 1 crudely illustrates the high prevalence of diabetes in Gulf nations compared with a mostly Europid population from Australia.4–7
Given the enormity of the global diabetes burden, the means by which diabetes is diagnosed is a critical issue. Moving away from the traditional approach of quantifying glucose levels in order to diagnose diabetes, in early 2011, the World Health Organization (WHO) included haemoglobin A1c (HbA1c) as an option in its recommended diagnostic criteria.8 Although HbA1c has played a fundamental role in the management of diabetes for decades now, this modification to the way in which diabetes is defined has prompted considerable debate over the relative benefits and possible disadvantages of using HbA1c for diagnosis.
A historical perspective on diabetes diagnosis
Over the centuries, diabetes has been diagnosed by various means, all of which, until recently, have been based on measuring glucose concentrations in either blood or urine. In ancient times, Indians used to diagnose diabetes on the basis of whether or not ants were attracted to a patient's urine.9 Later, scholars, such as Ibn Sīnā of the Islamic Golden Age, recommended tasting urine to see if it was sweet.10 It was not until the twentieth century that blood glucose levels were first used in diagnosis.
Unlike type 1 diabetes, which usually has an acute presentation with characteristic signs and symptoms and marked hyperglycaemia, type 2 diabetes mellitus generally has a slow onset, with glycaemia gradually worsening over time. In many cases, type 2 diabetes mellitus is diagnosed in asymptomatic individuals, who are unlikely to experience overt complications in the immediate future. Furthermore, the relationship between cardiovascular risk and glycaemia is approximately linear. Thus, the diagnosis of type 2 diabetes mellitus is somewhat arbitrary, being reliant on diagnostic thresholds that have largely been determined through epidemiological studies.
Until recently, the primary means for diagnosing diabetes was the measurement of blood glucose levels either in the fasting state or following an oral glucose load. Initially, diagnostic thresholds were derived from examination of the distribution of glucose levels in populations. In the 1970s, it was observed that some populations with a high prevalence of diabetes, such as the Pima Indians and the Micronesian population of Nauru, demonstrate a bimodal distribution of glucose levels.11,12 The intersection of the two curves in such a distribution can be used to classify the population into two groups, those with disease and those without. These data formed the basis of the 1979 criteria recommended by the National Diabetes Data Group (NDDG) in the United States, and subsequently the equivalent criteria published by the WHO Expert Committee on Diabetes Mellitus.13,14
In 1997, these diagnostic criteria were updated based on knowledge of the relationship between plasma glucose levels and prevalent retinopathy. Underpinning this decision were three studies, which were conducted in Pima Indians, Egyptians and participants from the National Health and Nutrition Examination Survey (NHANES) in the USA.15–17 These fasting and post-load glucose criteria remain part of WHO's recommended diagnostic criteria today (Table 1).18
Although the use of HbA1c in the diagnosis of diabetes was postulated as early as the mid-1980s,19 it was not until 2009 that an international expert committee recommended its incorporation into diagnostic criteria.20 The committee's recommendation was adopted by the American Diabetes Association the following year and subsequently by WHO in 2011 (Table 1).8,21
Haemoglobin (Hb), a metalloprotein found within the cytoplasm of erythrocytes, is responsible for the transport of oxygen around the circulatory system. It is composed of four subunits, each comprising a protein chain and an iron-containing haem group. In the case of normal adult haemoglobin (HbA), there are two α-chains and two β-chains. The N-terminal valine, an available amine group, is located at one end of the paired β-chains. A molecule of HbA1c is formed when glucose binds at this site.
Glycation is the non-enzymatic process by which glucose binds to proteins. The rate at which this occurs is largely proportional to the concentration of glucose in the blood. Hence, HbA1c levels provide an indication of glycaemia over the preceding few months.
Haemoglobin A1c was discovered in the late 1950s,22 but it was not until a decade later that Rahbar first made the connection with diabetes.23,24 Subsequently, several studies were conducted which established a relationship between blood glucose concentrations and HbA1c.25,26 Knowledge of this association underpinned the introduction of HbA1c into clinical practice as a measure of glycaemic control in patients with diabetes. Since that time, it has become a cornerstone of diabetes management.
Haemoglobin A1c and glycaemia
More recent studies have sought to define the relationship between HbA1c and chronic glycaemia with greater precision. The A1c-Derived Average Glucose (ADAG) study examined this relationship using a combination of continuous glucose monitoring and frequent capillary glucose testing as a measure of chronic glycaemia. HbA1c was found to correlate with average blood glucose levels over the preceding three months.27
The relationship between HbA1c and glycaemia is somewhat complicated in that it is liable to interference from factors affecting erythrocytes. The longer an erythrocyte is in circulation, the greater the probability that it will be glycated. Obviously, the reverse situation is also true. Generally, there is a preponderance of young erythrocytes in circulation at any one time. The average half-life of an erythrocyte is approximately 30 days.28 Accordingly, it has been demonstrated that average blood glucose levels in the month before HbA1c testing contribute about 50% to the final result, whereas levels 3–4 months prior contribute only around 10%.29
Advantages of using haemoglobin A1c for diagnosis
Haemoglobin A1c possesses several attributes that make it compelling for routine diagnostic use. As discussed, it provides an indication of blood glucose levels over the preceding few months rather than at a single point in time. Diabetes is a disease of chronic hyperglycaemia, and thus diagnosing it with a marker of chronic glycaemia seems very logical. Furthermore, many management decisions for patients with diabetes are based on their HbA1c. Consistency between the measures used for diagnosis and those used for monitoring also makes sense.
The oral glucose tolerance test (OGTT) is often considered the gold standard tool for diabetes diagnosis. This test is time-consuming, requires that patients fast beforehand and uses multiple blood samples. Conversely, HbA1c testing does not require patients to fast and only requires a single blood sample, making it a more convenient process. Considering the large proportion of diabetes cases that go undiagnosed,7 it has been hypothesized that a simpler diagnostic test, such as HbA1c, may improve detection rates.30
Another argument in favour of HbA1c for diagnosis is the wealth of data that connect HbA1c with clinical outcomes. This includes many trial data that demonstrate the importance of reducing HbA1c in order to lower complication risk. Two key randomized control trials that provided a foundation for the clinical use of HbA1c were the Diabetes Control and Complications Trial (DCCT) and the United Kingdom Prospective Diabetes Study (UKPDS).31,32 These studies showed that improving glycaemic control, defined as a reduction in HbA1c, is associated with a reduced risk of microvascular complications in both type 1 and type 2 diabetes mellitus . There is also considerable evidence of an association between HbA1c and cardiovascular disease, the primary cause of mortality among people with diabetes. A recent meta-analysis demonstrated that HbA1c predicts cardiovascular events, even in people without diabetes.33 Furthermore, the DCCT also highlighted the benefits, in regard to cardiovascular complications, of reducing HbA1c in people with type 1 diabetes.34 The results from trials in type 2 diabetes mellitus have been somewhat confusing. Follow-up of UKPDS participants at 10 years found a lower risk of cardiovascular events among those who had undergone intensive glucose control during the intervention.35 However, more recent studies, including ADVANCE and ACCORD, have had less conclusive results for patients with type 2 diabetes mellitus.36,37
Evidence from observational studies has shown that HbA1c has at least as strong a relationship with diabetes-specific complications, such as retinopathy, as either fasting or 2-hour plasma glucose.38 Some of the most compelling evidence in this regard has come from the DETECT-2 (evaluation of screening and early detection stategies for type 2 diabetes and impaired glucose tolerance) study, which pooled data on prevalent retinopathy from over 44 000 individuals across five countries.39 In this analysis, HbA1c was shown to possess a narrow threshold range in which the prevalence of retinopathy begins to increase substantially. This study provides further evidence in favour of using HbA1c at a threshold level of ≥6.5% to diagnose diabetes.
Finally, from an analytical point of view, HbA1c also has some advantages. It is known to have better preanalytical stability than glucose assays and has less intraindividual variability day to day.40,41
Concerns regarding the use of haemoglobin A1c in diagnosis
Despite its apparent strengths, there has been much debate about the role of HbA1c in diagnosis. Much of the concern relates to the fact that HbA1c is not a direct measure of glycaemia and, consequently, is susceptible to variation as a result of factors unrelated to diabetes risk. Numerous physiological and disease states have been observed to alter HbA1c. There are essentially three mechanisms by which HbA1c can be altered: first, by affecting the amount of glucose entering erythrocytes; secondly, through changing the average age of erythrocytes in circulation; and, finally, by varying the rate of glycation.
Recent evidence purports that, even in a healthy population, there is heterogeneity in both the glucose concentration gradient across the erythrocyte membrane and the average lifespan of erythrocytes.42,43 Additionally, many disease states affect erythrocyte turnover. If the average age of erythrocytes is lowered, through either increased haemolysis or increased erythropoiesis, HbA1c will be reduced accordingly. On the other hand, should the rates of haemolysis or erythropoiesis be slowed, HbA1c would rise because the average exposure of erythrocytes to glucose would be increased. Therefore, care should be taken in interpreting HbA1c results from patients with concurrent conditions that might affect erythrocyte turnover, such haemoglobinopathies, renal failure or liver disease.
In the past, a lack of international standardization of assays was an important consideration in regard to using HbA1c in diagnosis. However, over recent years much effort has been directed at unifying the calibration of assays worldwide. The International Federation of Clinical Chemistry and Laboratory Medicine (IFCC) has now developed and implemented an international reference system, which constitutes a significant step forward on the path towards universal standardization.44
On a more pragmatic level, there is also concern about the expense of HbA1c testing. It is reported to be unaffordable in many low- and middle-income countries.8 However, if HbA1c does offer the claimed benefits of greater convenience and possibly improved detection of diabetes, one must consider whether or not on the whole its use may be economically sound.
Discrepancies between the recommended diagnostic measures
Another common criticism of HbA1c as a diagnostic test is that it will detect a different population as having diabetes to that identified by traditional plasma glucose criteria. In particular, it has been claimed that HbA1c may vary, independently of glycaemia, on the basis of demographic factors such as ethnicity and age.45 These assertions do not in themselves preclude the use of HbA1c for diagnosis. The aim of diagnostic criteria is to identify people at risk of complications and who would benefit from intervention, not to detect people with particular glycaemic levels. Nevertheless, having multiple criteria for diagnosis can create confusion and uncertainty at both the individual patient and epidemiological levels.
The prevalence of diabetes has been shown to vary greatly depending on which diagnostic measure is employed. An investigation of people from the UAE without previously diagnosed diabetes found a significantly lower prevalence of diabetes (10.1%) when using HbA1c criteria than the OGTT-defined prevalence of 14.2% (P < 0.05).46
Moreover, the degree of discordance between the measures is also known to vary between different populations. A recent analysis compared the consistency of plasma glucose and HbA1c criteria across six countries. The probability of someone who meets plasma glucose criteria for diabetes also having an HbA1c ≥ 6.5% was found to range between 17% and 78% depending on the study.47 In a population from the UAE, the same statistic was 55%,46 whereas in an Arab population living in the USA it has been estimated at 19%.48 These observed discrepancies could be related to methodological differences between the studies. However, they still illustrate that, as HbA1c becomes more widely used, the global epidemiology of type 2 diabetes mellitus may be altered.
Variability of haemoglobin A1c with ethnicity
The possibility of glucose-independent ethnic differences in HbA1c has been the subject of much discussion. Numerous reports from the USA have indicated that people of African ethnicity may have higher HbA1c than people of Caucasian background, irrespective of their fasting or 2-hour plasma glucose levels.45,49 There is also evidence from Europe suggestive that the relationship between HbA1c and plasma glucose is not the same among South Asians or Inuits as it is in Caucasians.50,51 Again, this by itself does not suggest that HbA1c is incorrect in its detection of diabetes. It is possible that these differences in the relationship between HbA1c and plasma glucose are the result of HbA1c detecting differences in chronic glycaemia that are not reflected in plasma glucose. If this is found to be the case, it could be argued that HbA1c is a superior test for diagnosing diabetes among these ethnic groups. On the other hand, should future research show that the differences are caused by factors unrelated to diabetes, such as the prevalence of conditions affecting haemoglobin turnover, it would necessitate consideration of whether ethnicity-specific diagnostic thresholds of HbA1c should be developed.
Type 2 diabetes mellitus is one of the most prevalent non-communicable diseases worldwide, inflicting an immense burden on the global community. The way in which we define diabetes is a critical issue. HbA1c has many compelling attributes and is an appropriate diagnostic test when performed using an internationally standardized assay and in the absence of any clinical conditions that may impede accurate assessment of chronic glycaemia. Consideration of the various criticisms of HbA1c is warranted, with further research needed into demographic variability and its clinical significance. It is imperative that we understand fully the impact of changing the diagnostic criteria for diabetes on the epidemiology of this important disease.