13.1 Classification and diagnosis of diabetes mellitus

Diabetes mellitus is a group of metabolic diseases of multiple aetiologies characterized by hyperglycaemia together with disturbances of carbohydrate, fat, and protein metabolism resulting from defects in insulin secretion, insulin action, or both. The chronic hyperglycaemia of diabetes is associated with microvascular damage affecting, particularly, eyes, kidneys, nerves, and heart, together with an increased risk of macrovascular disease (1).

Acknowledgement that diabetes is not a single disorder has been attributed to two Indian physicians: Charaka and Sushruta (600 bc). They recognized two forms of the disease, although most of the descriptions in the early literature probably relate to what, today, is known as type 1 (insulin-dependent) diabetes. Sushruta is also credited with the first observation that diabetes was associated with ‘honeyed urine’. Throughout history, such renowned scientists and physicians as Galen, Avicenna, Paracelsus, and Maimonides have made reference to diabetes. Maimonides (1135–1204) observed on his travels that diabetes was seldom seen in ‘cold’ Europe, but was frequently encountered in ‘warm’ Africa (2). During the 18th and 19th centuries, a less symptomatic, ‘milder’ variety of the disorder was noted. It was identified by heavy glycosuria, often detected in later life, and commonly associated with being overweight rather than the previously described wasting.

Thomas Willis (1621–75) noted that diabetic urine tasted ‘wondrous sweet’, and, in 1766, another Englishman, Matthew Dobson, also observed that diabetic serum tasted sweet. Dobson demonstrated the presence of sugar in diabetic urine through chemistry, and chemical tests, such as Fehling’s, were developed by the 1840s. Benedict’s urine test (1911) was the mainstay for assessing control of diabetes for decades, although a blood sugar test was introduced in 1919 by the work of Folin and Wu.

As a consequence of the 1870 Siege of Paris, during the Franco-Prussian War, a French physician, Apollinaire Bouchardat (1806–86), noted the beneficial effects of food shortages on patients with diabetes: glycosuria and ketonuria decreased or disappeared, as did the major symptoms and signs. Almost 20 years later, the development of the theory of pancreatic diabetes emerged when the results of an important study—‘Diabetes mellitus after extirpation of the pancreas’—was published by Joseph von Mering and Oskar Minkowski in 1889. The discovery was the result of a wager between Minkowski and von Mering that a dog could not survive without a pancreas. The dog did survive the experiment, but kept urinating on the laboratory floor. Minkowski tested the dog’s urine for glucose, as his mentor Bernhard Naunyn (1839–1925) had instructed him to do for patients with polyuria, and found a high glucose content. This discovery inspired the work relating to the isolation of insulin for use in the therapy of diabetes, for which Frederick G. Banting (1891–1941) and John Macleod (1876–1935) won the Nobel Prize in 1923; Banting was assisted by Charles Best (1899–1978), with whom he shared his portion of the prize money.

In 1936, Harold Himsworth (1905–93) proposed that there were at least two clinical types of diabetes: insulin sensitive and insulin insensitive (3). He suggested that patients with insulin-sensitive diabetes were insulin deficient and required exogenous insulin to survive, while the other group did not require insulin. This observation was based on clinical observation alone, as there were then no assays for the measurement of insulin. Confirmation came with the development of a bioassay for insulin by Bornstein and Lawrence (4). Subsequently, Yalow and Berson developed the radioimmunoassay for insulin in the 1950s (5), which was used to demonstrate the near-total or absolute lack of insulin in those with ‘juvenile-onset’ diabetes, while significant amounts were still found in those with the older-onset obesity-associated form of the disorder. Thereafter, there was widespread acceptance that there were at least two major forms of diabetes. As these appeared to be separated according to the age of onset, they were labelled ‘juvenile-onset’ and ‘maturity-onset’ diabetes.

In the 1960s, diabetes was still a relatively rare disease, predominantly occurring in developed countries. Concerns were, however, being expressed that the prevalence was increasing, and also that there were a confusing set of different terms being employed. WHO therefore convened its first Expert Committee on Diabetes Mellitus in an attempt to bring order to the field (6). This marked the beginning of the modern era.

Although different forms of diabetes have been recognized for more than 2,000 years, the apparent diversity in the aetiology of the two major types made the development of a definitive classification difficult. The first real attempt to classify diabetes in a uniform way came with the 1964 WHO Expert Committee, which recognized that, without clear classification, it is not easy to take a systematic epidemiological approach to clinical research and develop evidence-based recommendations for the treatment and prevention of diabetes.

At that time, little was known about the aetiology of the different types of diabetes. The Committee resorted to a symptomatic classification on the one hand, and age of onset on the other. Thus, it described ‘asymptomatic’ diabetes and ‘clinical’ diabetes, with diagnostic criteria based on the oral glucose tolerance test (OGTT); it also had ‘potential’ diabetes and ‘latent’ diabetes. Its age categorization of diabetes comprised:

In addition, it confusingly retained the classification ‘juvenile’ diabetes for anyone requiring insulin and showing ketosis. It also had ‘brittle’ diabetes, ‘insulin-resistant’ diabetes, ‘gestational’ diabetes, ‘pancreatic’ diabetes and ‘endocrine’ diabetes. This was a start, but further work was required. In succeeding years, much more information became available, e.g. it became clear that ‘juvenile’-onset, ketosis-prone diabetes was an autoimmune disorder.

The big breakthrough came in 1979 and 1980: the National Diabetes Data Group (NDDG) in the USA and the second report of the WHO Expert Committee on Diabetes Mellitus offered a classification that was accepted internationally (7, 8). Two main classes of diabetes were suggested: insulin-dependent diabetes mellitus (IDDM) (type 1) and non-insulin-dependent diabetes mellitus (NIDDM) (type 2). ‘Other types’ and gestational diabetes mellitus completed the list. Two risk classes —previous abnormality of glucose tolerance and potential abnormality of glucose tolerance —were also introduced and replaced categories such as pre-diabetes and potential diabetes. The condition of impaired glucose tolerance (IGT) also appeared, replacing ‘borderline’ diabetes. Further changes in the classification took place in 1985, based on both clinical and aetiological characteristics, and the terms ‘type 1’ and ‘type 2’ diabetes were dropped (9). The 1985 WHO Study Group classification also added malnutrition-related diabetes mellitus.

Over the next decade, data from genetic, epidemiological, and aetiological studies accumulated, and understanding of the aetiology and pathogenesis of the diabetes syndromes improved. In 1995, an American Diabetes Association (ADA) Expert Committee met to decide whether changes to the classification were necessary (10), and, shortly afterwards, WHO convened a consultation to consider the issues and examine the available data. Its provisional report was published in 1998, with the document finalized a year later (11, 12). The proposed new classification was intended to include both aetiology and clinical stages of the disease, and was based on the suggestion of Kuzuya and Matsuda (13); it is depicted graphically in Fig. 13.1.1. In it, someone could, for example, have normal glucose tolerance but already have the type 1 diabetes ‘process’ occurring. Alternatively, someone with the type 2 process could move from insulin requiring to a diet-controlled state without pharmaceutical therapy. Hyperglycaemia could be subcategorized, regardless of the underlying cause, by staging into:

The 1999 classification retained the main groups of type 1, type 2, other specific types, and gestational diabetes, but dropped the terms ‘insulin dependent’ and ‘non-insulin dependent’, as these were potentially confusing. It was also expected that, as causes became known for subgroups of people, they would move into ‘other specific types’, and that the numbers of people categorized as ‘type 2’, which in many ways is a diagnosis of exclusion, would fall.

WHO re-examined the classification in 2006, when no changes were recommended (14). A further WHO Expert Committee looked at the classification again in 2009, with results to be published in 2011. A certain amount of updating occurred, but there were no fundamental changes (1) (Box 13.1.1).

This was previously known as juvenile-onset diabetes or insulin-dependent diabetes. Most frequently, it occurs as a result of autoimmune destruction of pancreatic β cells (type 1A diabetes) with evidence of immunological activity directed against the β cell, e.g. antibodies directed against islet cells; or constituents of the β cell, such as glutamate decarboxylase (glutamic acid decarboxylase (GAD)), or IA-2 (insulinoma-associated protein 2), a tyrosine kinase constituent of the membrane of the insulin secretory granule, or the zinc transporter 8 (ZnT8), or insulin. Evidence of such autoimmune activation can precede hyperglycaemia by several years, but the presence of the autoantibodies is diagnostic of the type 1 process. Later, subtle loss of β cell mass can be detected by failure of the rapid initial response of insulin to an intravenous glucose bolus, until so much β cell function is lost that spontaneous hyperglycaemia occurs. In a significant minority of people with insulin-deficient, ketosis-prone, type 1 diabetes, there is no evidence of autoimmune activation. This apparently non-autoimmune type 1B diabetes is more common in black people.

Type 1 diabetes is generally of rapid onset. The peak incidence is in childhood and adolescence, but it can occur at any age. There is also a slower onset form of type 1 diabetes, which has been referred to as latent autoimmune diabetes (LADA) (15). This occurs in adults, and, initially, such people may be diagnosed as having type 2 diabetes. There is a genetic predisposal to type 1 diabetes, largely accounted for by human leucocyte antigens (HLAs) DQ8 and DQ2, and type 1 diabetes is rare in the absence of predisposing HLA haplotypes.

Type 2 diabetes is characterized by relative insulin deficiency and insulin resistance. The normal response to insulin resistance is hyperinsulinaemia, and a degree of insulin deficiency is implicit in the presence of hyperglycaemia. However, the residual insulin secretion is enough to prevent lipolysis and ketogenesis. Many may progress to require insulin to achieve optimal glycaemic control, however, as the disease is progressive. In Europids, insulin resistance predominates, generally associated with obesity, whilst in some Asian populations insulin hyposecretion is more marked (16). In the great majority of people with type 2 diabetes, insulin is not required for survival, although it may be necessary to achieve good glycaemic control. The precise molecular mechanisms underlying the type 2 process are not known. By definition, there is no evidence of autoimmunity and no specific cause. It occurs largely in adults and older people, but, as the population prevalence rises, so it is being found in more younger people, e.g. in Japan, there are more adolescents with type 2 than with type 1 diabetes (17). The risk of type 2 diabetes increases with obesity, physical inactivity, family history, hypertension, dyslipidaemia, and the presence of macrovascular disease. It is also strongly associated with several ethnic groups, such as South Asians, Polynesians, First Nation Americans, and people of Arab origin. Several genetic associations have been described, the strongest being with TCF7L2 (transcription factor 7-like 2), which occurs over several ethnic groups (18), but as yet there is no characteristic genetic pattern.

These include a range of genetic defects of insulin secretion and action, specific genetic syndromes, diseases of the exocrine pancreas, endocrinopathies, infections, and diabetes caused by a range of drugs and toxins (Box 13.1.2).

These are rare and most, such as type A insulin resistance, are associated with mutations of the insulin receptor. Donohue syndrome (leprechaunism) and pineal hyperplasia, insulin-resistant diabetes mellitus, and somatic anomalies (Rabson–Mendenhall syndrome) also fall into this category.

(See also Chapter 13.3.4). Over the past decade, there have been major advances in knowledge of the molecular genetics underlying insulin secretion. This has led to the discovery of several subtypes of diabetes, and also aided in their clinical management. The best known of the defects of insulin secretion are the maturity-onset diabetes of the young (MODY) family (19), which were first described clinically more than 30 years ago (Chapter 13.3.4). The commonest subtypes are GCK (glucokinase) MODY; HNF1A (hepatocyte nuclear factor-1, alpha) MODY; and HNF4A (hepatocyte nuclear factor-4, alpha) MODY.

Glucokinase acts as the glucose sensor in the β cell. In GCK MODY, higher concentrations of glucose are required to obtain normal insulin concentrations. The disorder is mild and not associated with microvascular complications.

HNF1A MODY is the commonest form. It is associated with severe hyperglycaemia and complications and is extremely sensitive to sulphonylureas.

Diabetes diagnosed before the age of six months is almost always monogenic neonatal diabetes, rather than classical type 1. In about half of cases, the diabetes resolves. The majority of this latter group have abnormalities in the chromosome 6q24 region. Abnormalities in conversion of proinsulin to insulin have also been described and inherited in an autosomal dominant fashion. The resultant hyperglycaemia tends to be mild.

Any major disease of the exocrine pancreas can be associated with the development of diabetes. Fibrocalculous pancreatopathy is a not uncommon cause of diabetes, particularly in the Indian subcontinent. It was originally considered to be a consequence of malnutrition. It is probably related to tropical pancreatitis, with the same end result as chronic pancreatitis in the developed world. Pancreatic carcinoma, infections, and trauma are all also associated with diabetes. Interestingly, adenocarcinomas, involving only a small part of the pancreatic mass, may be associated with diabetes, suggesting that a mechanism other than simple reduction of β cell mass and insulin secretion are involved.

Diseases associated with increased secretion of several hormones, e.g. growth hormone, cortisol, glucagon, and adrenaline, can cause diabetes. This occurs in acromegaly, Cushing’s syndrome, glucagonoma, and phaeochromocytoma, respectively. The diabetes generally disappears when the disease is successfully treated. Somatostatinomas may also be associated with diabetes, presumably due to inhibition of insulin secretion.

Many drugs and poisons can inhibit insulin action or secretion, and thereby cause diabetes. This may occur in subjects who already have compromised insulin secretion or action, e.g. some of the earlier thiazide diuretics. Other toxins, such as the rat poison Vacor (pyriminil; N-3-pyridylmethyl-N′-p-nitrophenyl urea), permanently destroy β cells.

Certain viral infections may induce type 1 diabetes. There remains some uncertainty, but diabetes has been associated with congenital rubella, type B Coxsackie virus, cytomegalovirus, infectious parotitis (mumps), and adenovirus infections.

There are several uncommon forms of immune-mediated diabetes. These include patients with insulin autoantibodies, the ‘stiff man’ syndrome, and people with high titres of insulin receptor antibodies.

There are a range of genetic syndromes sometimes associated with diabetes. These include the severe obesity-associated Prader–Willi syndrome (Prader–Willi–Labhart syndrome), Alström syndrome, and Laurence–Moon–Biedl syndrome (Bardet–Biedl syndrome). There are a group associated with chromosomal abnormalities such as Down’s syndrome, Klinefelter’s syndrome, and Turner’s syndrome. Diabetes is also associated with several neurological disorders including Friedreich’s ataxia, Huntington’s disease (Huntington’s chorea), and myotonic dystrophy (See also Chapter 13.4.5).

This is glucose intolerance or diabetes first diagnosed in pregnancy. It has major implications for both the fetus and the mother and is also an indicator of high risk of later type 2 disease (see Chapter 13.4.10.6).

Some phenotypes of diabetes do not fit conveniently into the above-considered classification. As prevalence has increased, the age of onset has decreased, such that many reports have appeared of type 2 diabetes occurring in adolescents and, indeed, in children. These cases are generally associated with obesity. However, type 1 diabetes can occur in an overweight person. The same is true in young adults, where it can be difficult to distinguish between slow-onset type 1 diabetes and type 2 diabetes. There are also many reports now of people who turn out to have type 2 diabetes presenting in ketoacidosis. WHO, in its most recent report, has suggested the category of ‘unclassified’ should be used, particularly at diagnosis, where there is phenotypic uncertainty (1). Formal categorization can occur at a later stage.

One ‘uncertain’ type has been termed ‘ketosis-prone type 2 diabetes’, or ‘ketosis-prone atypical diabetes’. This has been described in sub-Saharan Africa and in people of African origin living elsewhere (20). Patients present with ketoacidosis and are definitely insulin requiring, but, after months, may come off all therapeutic agents. The cycle then repeats itself. A viral infection has been implicated. This has similarities with ‘Flatbush diabetes’ described in African Americans and may also be the same entity as ‘periodic insulin-dependent diabetes’, which was described in East Africa many years ago (21). It is described in detail in Chapter 13.4.3.4).

The term ‘intermediate hyperglycaemia’ encompasses levels of glucose that are above normal, but below those used to diagnose diabetes. It incorporates both impaired glucose tolerance (IGT) and impaired fasting glycaemia (IFG). Collectively, these are known as ‘prediabetes’, but the term intermediate hyperglycaemia is now preferred by WHO.

IGT was previously listed as a class of diabetes by WHO, but now, more properly, IGT and IFG are included as risk states. Both can be viewed as stages in the natural history of disordered carbohydrate metabolism. The two terms are not synonymous and the two conditions may have different aetiologies. Thus, IGT reflects more handling of glucose by peripheral tissues, while IFG relates more to gluconeogenesis and hepatic metabolism.

The IGT category includes people whose OGTT result is beyond the boundaries of normality by WHO criteria (Box 13.1.3). IGT may represent a stage in the natural history of diabetes, as people with it are at higher risk for diabetes than is the general population (22). People with IGT have a heightened risk of macrovascular disease, and IGT is associated with other known cardiovascular disease risk factors, including hypertension, dyslipidaemia, and central obesity. The diagnosis of IGT, therefore, may have important prognostic implications, particularly in otherwise healthy and ambulatory individuals.

IFG was introduced as a new risk category in 1997 and 1998 by ADA and WHO, respectively (10, 11). It is certainly a risk state for diabetes, but its relationship to cardiovascular disease is more doubtful. It does, however, form a component part of the metabolic syndrome (see Chapter 13.3.6), which itself indicates increased risk of both diabetes and cardiovascular disease.

There has been a tendency to measure only fasting concentrations of glucose as a screening test for diabetes. This can be misleading as significant numbers of those with IFG will have either diabetes or IGT on oral glucose tolerance testing. Someone with IFG and IGT has a much higher risk of developing diabetes than those with either IFG or IGT alone.

The diagnosis of diabetes has major implications for an individual thus diagnosed. Accordingly, the person must have confidence that the diagnosis is accurate. In people with obvious symptoms and clearly elevated blood glucose levels, this is not a problem. However, in an asymptomatic person with moderate hyperglycaemia, this can be more difficult. Whatever test is used, a second confirmatory test is essential in people without symptoms. This is of increasing importance as more screening programmes are introduced and more asymptomatic people with previously undiagnosed diabetes are discovered.

Clinical diagnosis of diabetes generally is prompted by the presence of classical symptoms. These include thirst, polyuria, weight loss, recurrent infections and, in more severe cases, drowsiness and coma (see Chapter 13.2.1). Glycosuria is almost always present. A casual venous plasma glucose level of 11.1 mmol/l (200 mg/dl) or more is then sufficient to make the definite diagnosis. If the plasma glucose is between 5.0 mmol/l (90 mg/dl) and 11.1 mmol/l (200 mg/dl), WHO suggests that a more definitive test should be performed (see below).

In the 19th century, glycosuria was used as the major diagnostic test. Measurement of blood glucose came into use as the main diagnostic test in the 20th century and has remained the diagnostic cornerstone now for more than a hundred years. The OGTT was first devised in the early years of the 20th century and slowly came into use for diagnosis. Initially, measurement of glucose was a laborious process and not very specific. It was only from the 1950s that glucose measurement became relatively easy and rapid to perform. Specificity came in shortly afterwards with the introduction of enzymatic assays. There was also no universal agreement on precisely which tests should be performed and what concentrations of glucose were diagnostic for diabetes. Glycosuria continued to be used for diagnosis until the 1960s.

A semblance of order came with the first WHO Expert Committee in 1964 (6). One of its most important actions was to consign the use of glycosuria as a diagnostic test to the scrap heap. It also commented on the wide range of glucose tolerance tests in use, and suggested that only the 50 g and the 100 g OGTTs should be used. It set the fasting glucose criterion at 130 mg/dl (7.2 mmol/l)—but this was venous whole blood, so, for plasma, it would be about 150 mg/dl (8.3 mmol/l). At that time, however, non-specific tests for glucose were used so that a true plasma glucose would have been between 130 mg/dl (7.2 mmol/l) and 140 mg/dl (7.8 mmol/l). The use of steroid-modified tests, widespread at the time, was deemed unnecessary. The WHO Committee was also the first to suggest that the 2-h value alone after an oral glucose load was sufficient, and that the intervening values were unhelpful. It suggested 130 mg/dl for venous whole blood—independent of the size of the glucose load! It also introduced the category of ‘borderline diabetes’—the forerunner of IGT. Finally, it suggested that people should be carbohydrate loaded for at least 3 days before the OGTT—advice that still holds today, but to which adherence is rare.

Despite the efforts of the first WHO Committee, there was little standardization of testing. Thus, when 20 diabetologists were asked by the late Kelly West (a pioneer in diabetes epidemiology) what diagnostic 2-h cut-point for blood glucose they used following a glucose load, 15 different answers were received. The real breakthrough came with the reports of the NDDG in the USA in 1979 and the second WHO Expert Committee in 1980, respectively. The diagnostic fasting venous plasma glucose threshold was set at 140 mg/dl (7.8 mmol/l) by NDDG and 8.0 mmol/l (144 mg/dl) by WHO (7, 8).

It was agreed that only a 75 g oral glucose (anhydrous) should be used for the OGTT—except in children, when a weight-related dose was recommended (1.75 g/kg). The 2-h post-load thresholds were very similar: venous plasma glucose of 11.0 mmol/l (198 mg/dl) for WHO and 200 mg/dl (11.1 mmol/l) for NDDG. The rationale for these values was bimodality in certain high prevalence populations, such as the Pima Indians and the Nauruans. It was also affirmed that, in an asymptomatic individual, two separate abnormal tests were required. A casual venous fasting glucose value of greater than 11.0 mmol/l (200 mg/dl) in the presence of symptoms was, however, sufficient to make the diagnosis. The category of IGT was introduced, replacing ‘borderline diabetes’. For this, fasting glucose had to be below the diagnostic value for diabetes and, for the 2-h value, between 7.8 mmol/l and the threshold for diabetes.

Subsequently, the WHO group met again in 1985 and adjusted the fasting and 2-h values to be consistent with the NDDG criteria (9). Thus, the fasting threshold was now universally 7.8 mmol/l (140 mg/dl) and the 2-h value 11.1 mmol/l (200 mg/dl).

The next major change came in 1997 and 1999 when an ADA Expert Committee (10) and a WHO Consultation (11) both agreed that the previous fasting threshold was too high and so dropped the value to 7 mmol/l (126 mg/dl). This was based, at least in part, on the cross-sectional relationship between fasting plasma glucose (FPG) and retinopathy. The ‘2 h’ value remained unchanged. A new category of IFG was also introduced. This was to cover levels of glucose that were not diagnostic for diabetes, but were clearly above normal and carried an increased risk of subsequently developing diabetes. One driver to moving to a lower value for FPG was the ADA view that few people bothered with an OGTT. ADA recommended that the FPG alone could be used to diagnose diabetes. WHO did not support this view. Many studies have shown that significant numbers of people will have diabetic values 2 h after a glucose load, even if the FPG is normal. Furthermore, WHO recommended that, if IFG is detected, a full OGTT should be applied. The only exception was for epidemiological studies where an OGTT was not practicable.

A further change occurred in 2003, when another ADA Expert Committee recommended changing the lower threshold for the diagnosis of IFG—with the range now being 100 mg/dl to 125 mg/dl (5.6 mmol/l to 6.9 mmol/l) (23). WHO reviewed this in 2006, and could find no reason to change from the previous range of 6.1 mmol/l to 6.9 mmol/l (14).

Both WHO and ADA further considered the diagnostic criteria in 2010 (1,24). Plasma glucose criteria remain the same as previously set out (Box 13.1.3). The ADA continues to espouse the use of a fasting glucose alone as a definitive diagnostic test, with WHO continuing to advocate the OGTT if IFG is present, although accepting that an FPG can be used as a screening test.

The implications for an individual of a diagnosis of diabetes should not be underestimated. The diagnosis needs to be secure and the number of false positive results limited. Given the day-to-day variability of blood glucose measurements, it is paramount that diabetes be only diagnosed when two abnormal values have been found on separate days—except in the presence of appropriate symptoms and signs. It should be noted, however, that in all the studies relating blood glucose to risk of retinopathy, on which the diagnostic thresholds are heavily based, only a single OGTT was performed.

Another significant detail is that all the evidence on which diagnostic thresholds are based is for venous plasma glucose. Previous WHO reports gave equivalent values for capillary and venous whole blood glucose, but the WHO has now said that only venous plasma glucose values should be used for definitive diagnosis. Obviously, capillary values, using meters, can be used for screening purposes, and, in some situations, only capillary samples may be available. If the latter is the case, then meticulous quality assurance is required and adjustments made for the 2-h value. If capillary whole blood is used, and the measurement is not converted to a plasma equivalent, the fasting sample values are different.

The use of glycated haemoglobin (HbA_1c) as a diagnostic test for diabetes has long been discussed. It has been used as a marker for glycaemic control for 30 years, reflecting plasma glucose values over the previous 8 to 12 weeks. It has major attractions as a diagnostic tool: no dietary preparation is required, the sample can be taken at any time, it will be relatively unaffected by acute stress, and the sample is stable at room temperature for at least a week. Its use for diagnosis, however, has consistently been rejected by successive WHO and ADA committees. The reasons are manifold. Until recently, the assay has not been standardized and there has been lack of an international quality assurance system or a single international standard. This was, in part, mitigated by many laboratories calibrating against the Diabetes Control and Complications Trial assay and now the US National Glycohemoglobin Standardization Program, but this is still limited to a small number of countries. The assay is also expensive, compared with measuring glucose, and is not available at all in many parts of the world. More importantly, HbA_1c is also affected by a range of genetic, haematological, and disease-related factors (25). This is particularly true for any condition in which there is accelerated red cell turnover, such as haemolytic anaemias and malaria. With some of the commonly used assays, haemoglobinopathies are also a problem. This is a problem in many Middle Eastern and African countries.

The key question is: how well does HbA_1c predict the development of specific diabetic complications, such as retinopathy? In fact, HbA_1c was measured in the same cross-sectional studies that were used to determine the fasting and post-load glucose diagnostic thresholds: the National Health and Nutrition Examination Survey in the USA, a Pima Indian study, an Egyptian study, and a Japanese study (10, 26, 27, 28). Sensitivity and specificity were similar for FPG, 2-h plasma glucose, and HbA_1c. These were, however, cross-sectional studies; longitudinal studies would be better. Based on these studies, the threshold for HbA_1c would be around 6.5% (48 mmol/mol). In a more recent, very large, multinational study that included 48 416 participants aged 20 to 79 years old, there appeared to be thresholds for any retinopathy at 6.4 mmol/l (115 mg/dl) for FPG, 11.8 mmol/l (212 mg/dl) for 2-h plasma glucose (dropping to 9.7 mmol/l (175 mg/dl) for moderate and severe retinopathy), and 6.2% (44 mmol/mol) to 6.4% (46 mmol/mol) for HbA_1c (29). It is worth stressing that the thresholds were not as clear as in previous studies. Some of this may be due to the fact that modern methods for detecting retinopathy are more sensitive than those used previously.

Recently, an International Expert Committee has concluded that HbA_1c can indeed be used for the diagnosis of diabetes (30). It felt that the advantages far outweighed the disadvantages. The assay was now standardized in many countries, the coefficient of variation of the assay was low and freely available in countries such as the USA, Canada, and many European countries. The Committee selected a threshold of 6.5% (48 mmol/mol). This received further support from the ADA in 2010, where it has appeared in their practice guidelines (24). In addition, the ADA document suggested that the range of 5.7% (39 mmol/mol) to 6.4% (46 mmol/mol) should be considered as a high risk range—equivalent to IFG and IGT.

The use of HbA_1c for diagnosis has also been considered by a March 2009 WHO Consultation, which has cautiously accepted that it may be used, but with a series of caveats about quality assurance, calibration against the new International Federation of Clinical Chemistry (IFCC) standards, and its inappropriateness in countries where there are high rates of haemoglobinopathies or high red cell turnover states, such as haemolytic anaemias or malaria (1). The consultation also suggested a threshold of 6.5% (48 mmol/mol), but felt that there was insufficient evidence to suggest a high-risk range. WHO also said that glucose will continue to be the test of choice in most situations. Overall, the widespread use of HbA_1c for diagnosis and as a screening test is inevitable, although measurement of glucose will continue to be the mainstay of diagnosis in many countries for the foreseeable future for both technical and economic reasons.

It is worth summarizing the benefits and problems of both glucose and HbA_1c for diagnosis. The benefits of HbA_1c are stability, low variation of the assay, reproducibility day to day, convenience (in that no special preparation of the patient is needed and it can be performed at any time), the sample is stable at room temperature, and it is not affected by acute illness. It also reflects a period of hyperglycaemia, rather than just a single point of time. The disadvantages are cost, lack of availability on a world-wide basis, problems in countries where there is a high prevalence of haemoglobinopathies or rapid red cell turnover rates, and lack of standardization in many places.

Glucose has the advantage of long use and familiarity, good knowledge of relationship to specific complications of diabetes, cheap assay and good assay precision in most places, and availability almost everywhere. The disadvantages are that there are major pre-assay problems that are poorly recognized. Thus, there is major day-to-day variation of both fasting values and glucose tolerance; the patient needs to be fasting, but using the fasting test alone misses many cases; glucose tolerance tests are inconvenient and rarely performed outside pregnancy; results cannot be interpreted with confidence in acutely ill patients; and, unless the sample is separated rapidly and kept cool, the result will be artefactually low.

There are thus advantages and disadvantages to both tests—and, in both cases, the thresholds used require confirmation to see what level of which test is the best predictor of the long-term complications of diabetes. It should also be noted that different individuals may be detected using one test or the other. The significance of this will only become apparent with longer-term studies.

Currently, gestational diabetes is screened for at 28 to 32 weeks of pregnancy. WHO recommends use of the 75 g OGTT, and anyone who exceeds the 2-hr threshold of 7.8 mmol/l (140 mg/dl) is deemed to have gestational diabetes. New criteria based on prospective studies are being worked on at present and are likely to be agreed in the near future (see Chapter 13.4.10.6).

Overall, there have been relatively minor changes in the classification of diabetes over the past decade. More is known of the genetics of ‘other specific types’ of diabetes, but the major classification into type 1, type 2, other specific types, and gestational diabetes remains unchanged. The pragmatic introduction of ‘unclassified’ diabetes recognizes the heterogeneity of the diabetes, but has a relatively minor impact. Similarly, there has been no change in the diagnostic thresholds for glucose for diabetes and intermediate hyperglycaemia (IFG and IGT). The misleading term ‘prediabetes’ has, however, been dropped. By far the biggest change is the recommendation to use measurement of HbA_1c as a diagnostic test for diabetes—the impact of this will appear over the next 2 or 3 years.