Diagnosis and Classification of diabetes mellitus
Diabetes mellitus is a group of diseases characterised by chronic elevation of blood glucose and defined by measurements that reflect this level. Since glycaemia is the only way of defining diabetes, there has been considerable debate as to how glucose exposure should be estimated, and selection of the cut-off for diabetes. Both the measures and the cut-offs have changed over time. By general agreement, diabetes is defined in terms of the risk of small vessel disease (more specifically, retinopathy). The diagnostic threshold for diabetes is defined in terms of the 2 hr value following the oral glucose tolerance test, or fasting plasma glucose, or HbA1c. Below this threshold, increasing glucose exposure is however associated with increasing cardiovascular risk, an association that extends within the non-diabetic glucose range and rises linearly with increasing glucose exposure. There is therefore a need for a second diagnostic threshold to indicate increased cardiovascular, but not microvascular risk. This intermediate range is designated as impaired glucose tolerance (IGT), impaired fasting glucose (IFG), or by various terms when HbA1c is used. The diagnostic threshold and value of treatment for intermediate hyperglycaemia, often known as prediabetes, are still subject to controversy.
How do you define a disease?
Diabetes is defined as a state of chronic hyperglycaemia. This having been said, the question remains as to how blood glucose should be measured, and the cut-off point used to define diabetes as a disease (i.e. something that causes harm). These simple questions have perplexed diabetologists for more than a century, and have yet to be answered to everyone's satisfaction.
One hundred years ago, the answer seemed easy enough. Diabetes was a condition with characteristic symptoms and glucose in the urine. Furthermore, those affected were obviously unwell and would deteriorate or die in the absence of effective therapy. Physicians such as Elliot Joslin (1870-1962) were well aware that some individuals had a higher than normal blood glucose without symptoms, and even used the term "prediabetes" to describe them, but did not consider the diagnosis useful to make.
The 1930s showed increasing awareness of the problem of late complications of diabetes. Physicians became aware that insulin was not a panacea, and that the life expectancy of insulin-treated patients was reduced by conditions such as diabetic nephropathy and heart disease. Fifty years were to pass before it was demonstrated that treatment directed towards careful control of blood glucose levels could prevent or delay the small vessel complications of diabetes.
Before the 1950s there were only two treatments for diabetes: diet and insulin. Diet was the mainstay of those deemed not to need insulin, and little effort was made to screen for diabetes. Urine tests were used for routine screening, and undoubtedly failed to detect many people with a raised blood glucose that did not exceed the renal threshold for glucose.
The basis for treating hyperglycaemia was provided by large prospective population studies, the most famous of which was the Framingham Study in the USA. By following the health of several thousand individuals prospectively, the investigators identified simple clinical measures which predicted future health problems. Prominent among these were "risk factors" (a term coined in the course of the study) such as hypertension, hypercholesterolaemia and hyperglycaemia. A disease was no longer something that made you ill, but something that - although asymptomatic - could easily turn into something that made you ill.
This was the prelude to the first large controlled intervention trials, made possible by the availability of the thiazides to treat hypertension, and of introduction of tolbutamide, the first sulfonylurea agent, in 1957. The first trials of antihypertensive therapy were dramatically successful, but the first major trial of antidiabetic therapy, known as the University Group Diabetes Program (UGDP) appeared to show that treatment with sulfonylureas or phenformin actually increased the cardiovascular mortality associated with diabetes.
Clinicians, meanwhile, still argued as to the glucose level that defined diabetes. It was generally agreed that the oral glucose tolerance test (OGTT) was the best means of diagnosis, but there was wide discrepancy as to its interpretation. Kelly West asked groups of experts to score fasting and 2-hour plasma glucose values as “clearly normal”, or “clearly abnormal”. As compared with a reference US population, the proportion of values judged “clearly abnormal” ranged from 2-62%!.
The situation was resolved by prospective population studies which examined the risks associated with hyperglycaemia, including studies of Pima Indians in the US and the Whitehall Study in the UK. These indicated that diabetic retinopathy was much more commonly seen in those with 2-hour glucose values > 200mg/dl (11.1 mmol/l) following an OGTT, and this went on to become the agreed cut-off for the definition of diabetes. Prospective study also revealed the existence of an intermediate group with 2-hour values of 140-200 mg/dl (7.2-11.1 mmol/l). These people had an increased cardiovascular mortality (as seen in those with frank diabetes), but a low risk of small vessel complications. This was termed impaired glucose tolerance (IGT), and was considered not to require specific glucose-lowering therapy.
TheDiabetes Control and Complications Trial (DCCT) and the UK Prospective Diabetes Study (UKPDS) proved beyond doubt that intensified glucose control was effective in preventing the microvascular complications of diabetes, but the benefits for arterial disease appeared more marginal. This has been confirmed by a series of large clinical trials which examined cardiovascular risk in relation to glucose control. The results were generally disappointing, and mainly served to emphasise the importance of other risk factors (smoking, blood pressure, lipids) in the management of diabetes.
What makes diabetes mellitus a disease is the (vascular) complications that are the result from (chronically) high glucose values. Therefore, one would expect that those who have glucose measurements that meet the criteria for diabetes would be at risk for these complications, and those who have normal glucose levels would not. Epidemiologically, however, the relation between measures of glycaemia and the risk for complications does not always have a clear threshold. For instance, even in healthy people, those with the higher average glucose levels have a higher risk of mortality.
Figure 1. Cut-offs used to define diabetes, IGT and IFG. Please note that these are arbitrary: for many complications there is a continuous relation between the glucose level and complication risk, even in the 'normal' glucose range. Thus, it is important to note that the cut-off levels that separate those with diabetes from those without are arbitrarily defined (see figure 1). There are arguments in favour of or against most of the criteria used, but the criteria for the diagnosis of diabetes have been (and probably will be) revised about once a decade.
The most recent change has been the introduction of the glycosylated hemoglobin (HbA1c) level as a criterion. The rationale for this is that the HbA1c reflects average glycaemia over a longer period (of about 6 weeks) and predicts both risk and – when decreasing during therapy – risk reductions for the development of complications. However, opponents argue that as a predictor for risk fasting glucose serves just as well, and that HbA11c is a difficult and expensive measurement not easily performed in less affluent countries.
Regardless of which test is chosen, it is important to remember that different tests will identify different people as (not) having diabetes.
Figure 2. Results of screening tests for diabetes in an American population. In total 5.4% of people were found to have undiagnosed diabetes by any of the three criteria. The 2-hour glucose value after an oral glucose tolerance test identified most people.For instance, when studying a population from the United States 86.9% of people did not have diabetes and 7.8% were already known to have diabetes. Using all three possible criteria, i.e. FPG >= 7.0 mmol/l, 2-hour glucose after OGTT>= 11.1 mmol/l and/or HbA1c >= 6.5% (48 mmol/mol) about 5.4% of the population had undiagnosed diabetes, but not all met all three criteria (see figure 2).
As can be seen, it is unexpectedly complicated to equate diabetes as a disease (something that makes you ill) with a given measure of glucose exposure. Many elderly people, for example, have diabetes by glucose criteria, but may not necessarily enjoy a better quality or duration of life as a result of efforts to lower their blood glucose. Better ways of estimating individual risk in relation to glucose levels (and better estimates of the utility of treating these) may in the future lead to ways of diagnosing and treating hyperglycaemia according to individual needs and likely benefit.
^ West KM. Substantial differences in the diagnostic criteria used by diabetes experts. Diabetes 1975;24: 641-644