Perhaps the first diagnostic test for diabetes was described in ancient China with the observation that ants were attracted to the urine of persons with diabetes. The London physician Thomas Willis (1621-1675) noted that diabetic urine tasted 'wondrous sweet' and in 1766 another Englishman, Matthew Dobson, demonstrated the chemical presence of sugar in diabetic urine and observed that serum from diabetics was sweet-tasting. By the 1840s, such chemical tests as Fehling's were developed for sugar in urine. Benedict's urine test was described in 1911 and for many decades remained the mainstay for assessing control of diabetes.
Urine glucose measurement, though cheap and convenient, has inadequate sensitivity49, and therefore, for more accurate and reliable diagnostic tests the blood-based assessment of hyperglycaemia is relied upon. Evidence concerning the sensitivity and specificity of non-blood tests is considered in full in Chapter 8 on screening.
BLOOD-BASED ASSESSMENT OF HYPERGLYCAEMIA Random blood glucose
Few studies have examined the properties of random blood glucose (measured by reflectance meter in both studies) in relation to other means of diabetes testing (Table 2.4). In these two studies an OGTT was performed in the whole population irrespective of the random glucose value obtained. In order to achieve a sensitivity of 80-90%, the specificity of a random glucose determination seems to be significantly lower than that of fasting glucose. Furthermore, WHO 1985 criteria (in which the fasting plasma threshold was 7.8 mmol/l) were used as the gold standard in both studies, but if current criteria were applied, performance of the test is likely to be slightly worse. People who are diabetic only on the new, lower fasting value (and who have a non-diabetic 2 h value) are more likely to have normal random blood glucose values.
Blood glucose meters and whole blood glucose determination
Blood glucose meters have often been used as part of screening programmes for diabetes. However, the precision of these meters is limited, and it is
Table 2.4. The performance of a random whole blood glucose determination in comparison to an OGTT.
*The cut-off value of random whole blood glucose for a sensitivity of 90% was 4.4-6.7, depending on age and post-prandial period.
**Sensitivities and specificities were worse in women than men at all thresholds.
generally recommended that the diagnosis of diabetes is made on laboratory measurements of glucose. Even for the initial screening test, laboratory testing is preferred, and meters should only be considered if they are the only way of providing a screening service to a given population or individual. If meters are used, their imprecision, and the possibility of a false negative result, should be considered when interpreting the results. The accuracy of glucose results measured by glucose meters may be low, with only 35-83% of readings being within ±10% of the adjusted laboratory plasma glucose values52,53. The evidence on which the diagnostic thresholds are based comes from studies using laboratory plasma glucose measurements. The thresholds for other types of blood samples (e.g. whole blood in Table 2.1) equivalent to the threshold given for the plasma values have been calculated as being about 10% lower than in plasma. However, formal comparisons in large populations have not been published, and these values may not be as reliable as the plasma thresholds.
Was this article helpful?