Analysis of data from six previously published studies of diabetes prevalence indicates that a change in the diagnostic criteria for diabetes will have different effects on diabetes prevalence across different populations and ethnic groups, with some of the differences being marked.

An expert group recently recommended that the criteria for diagnosis of diabetes shift from the oral glucose tolerance test (OGTT) now used to a measurement of haemoglobin A1c (HbA1c) level, with a level of 6.5% or above being considered diagnostic. The authors of this paper note that HbA1c levels in people with diabetes vary by race and ethnicity, and therefore that the proposed new criteria will alter the prevalence of diabetes variably in different countries. They used data from previously published population studies to investigate this and determine the likely degree of agreement between the two criteria.

The studies used collected their data between 1999 and 2009 and were based on populations in Denmark, United Kingdom, Australia, Greenland, Kenya, and India. They ranged in size from n=296 to n=7800 (total n=23,094). For this analysis, participants were divided into four groups: diabetic or non-diabetic based on OGTT, and HbA1c <6.5% or 6.5% and above. The primary analysis was the probability of having a HbA1c 6.5% or above in those with diabetes diagnosed by OGTT (i.e. the sensitivity of the 6.5% cut-off for identifying diabetes).

According to the analysis, the prevalence of diabetes was lower than that found by OGTT in four studies and higher in two. The differences were marked - probability of a person identified as having diabetes using the OGTT having a HbA1c of 6.5% or above ranged from 17% to 78%. Overall, using HbA1c with a 6.5% cut-off would reduce the prevalence of diabetes by 18%.

In the UK study, data were available for people of Black and South Asian origin - in a subsidiary analysis, differences in diagnostic agreement between these ethnic sub-groups were of the same magnitude as in the between-country comparison.

The authors conclude that shifting to HbA1c-based diagnosis of diabetes will have a significant effect on diabetes prevalence in different populations and ethnic groups. They note that while some of the differences may be due to methodological variations, the difference found between ethnic groups in the UK study suggests that such variations do not account for the whole difference.

[Editor's comment: a recent BMJ article (BMJ 2009; 339: b4432) also raised concerns over the proposed change for other reasons.]