Diagnosis of diabetes using the oral glucose tolerance test

The prevalence of type 2 diabetes varies greatly by ethnic group within and across countries. The most reliable data on the prevalence of diabetes are based on two hour plasma glucose values after an oral glucose tolerance test,¹ which is currently the gold standard epidemiological and clinical diagnostic test for diabetes and impaired glucose tolerance. In Newcastle, England, on the basis of clinical evidence and oral glucose tolerance test results, about 20% of British South Asians had diabetes, compared with only 4% of white Europeans, after age adjustment in a sample of 25-74 year olds.² Might such observed differences in prevalence, at least in part, be artefacts of the diagnostic method?

In 1965, the World Health Organization expert committee drew attention to the "lack of suitable epidemiological information about glucose tolerance in various populations of various races and cultures in different countries"³ and highlighted the need for research in different populations. The call was repeated in 1980,¹ with special reference to the oral glucose tolerance test and the dose of glucose, with 75 g being recommended pending further investigations. The International Diabetes Federation in consultation with WHO and the American Diabetes Association (ADA) have raised similar concerns, particularly about the oral glucose tolerance test.^{4 5} With a 75 g dose, a venous plasma glucose value of 11.1 mmol/l or more is indicative of diabetes, as indicated by its association with complications such as retinopathy. A value of 7.8-11.0 mmol/l is indicative of impaired glucose tolerance. Yet these concerns have not been dealt with.
The prevalence of diabetes is increasing worldwide, and accurate testing is more important than ever. The best way to make a diagnosis has been debated for decades,^{1 3 4 5} and more guidance is imminent. In some ethnic groups, comparatively low fasting plasma glucose concentrations are seen in people who have two hour postload glucose values that are diagnostic for diabetes.⁶
In the light of these concerns it is vital to know whether the 75 g carbohydrate load is appropriate for all adults, regardless of ethnicity. Glucose tolerance is influenced by several factors—from genetics, to body build (height and weight), to diet and lifestyle. Differences in body composition and skeletal muscle mass are important determinants of postprandial glucose metabolism, and height measurement partly reflects such differences.
An independent inverse association with two hour plasma glucose after the oral glucose tolerance test has been repeatedly shown for height in diverse populations.^{7 8} In a study of the prevalence of type 2 diabetes in white Europeans, African-Caribbeans, and Pakistanis, height almost completely accounted for ethnic differences in two hour plasma glucose in multiple regression models. Pakistanis, in whom the prevalence was the greatest, were markedly shorter (by 2-5 cm) than people in other ethnic groups.⁷ The implications of these findings are that a uniform oral glucose load may not accurately assess glucose tolerance across populations,¹ and a high two hour plasma glucose after the oral glucose tolerance test may overdiagnose impaired glucose tolerance in some ethnic groups compared with white populations.
Other factors related to body composition that vary by ethnicity may also be important. Varying the glucose load, as is done in children, or adjusting the results according to ethnicity or height (or both), may improve measures of glucose tolerance. These general observations could have wider implications in explaining inequalities. Impaired fasting glucose is more prevalent in men, whereas impaired glucose tolerance is more prevalent in women.⁹ Women are generally shorter than men, so this difference could simply reflect height differences by sex.
Whereas height has been shown to have a marked association with two hour plasma glucose after the oral glucose tolerance test, fasting plasma glucose and glycated haemoglobin measurements vary very little with height or sex.^{8 10} We should consider whether the oral glucose tolerance test can be replaced with other measures, such as glycated haemoglobin, in everyday clinical practice. This was a topic of debate at this year’s ADA annual conference in New Orleans, and work is already under way to standardise the measurement of glycated haemoglobin. However, as with the oral glucose tolerance test, the validity of glycated haemoglobin needs to be shown across ethnic groups before it is accepted and implemented.
WHO’s warnings in 1965 about the validity of the oral glucose tolerance test across various populations were prescient and deserve continuing attention. The uniform size of the oral glucose load used in this test, even though body size and composition vary, may account for some of the variation in the prevalence of diabetes between men and women and different ethnic groups. Nonetheless, the excess of diabetes in South Asians is marked using other criteria, such as those based on fasting glucose used by the ADA.¹¹ The complications of diabetes, such as retinopathy and nephropathy, are also greater in South Asians.¹²
Clinicians must be confident that the key tests for diabetes or impaired glucose tolerance are accurate, because the consequences of these diagnoses are considerable and lifelong. Although a false positive result might lead to good advice about diet and exercise, it could also provoke anxiety and adoption of the sick role. A false negative result is potentially dangerous in view of the high levels of cardiovascular diseases and renal dysfunction in South Asians. We must always establish the validity of diagnostic tests across sexes, age groups, and ethnic groups. This still applies to the oral glucose tolerance test and its likely successor, the measurement of glycated haemoglobin.
Cite this as: BMJ 2009;339:b4354