Letters

Glycated haemoglobin, diabetes, and mortality in men

Maybe disturbance in physiological mechanisms regulating blood glucose is risk factor for cardiovascular death

Analytical information is required for generalisation of data

Authors' reply

Medicine is now using diagnostic criteria rather than reference ranges

Maybe disturbance in physiological mechanisms regulating blood glucose is risk factor for cardiovascular death

EDITORKhaw et al contribute to the substantial literature showing an association between measures of glycaemia and subsequent morbidity and mortality.¹ They suggest that "preventive efforts need to consider not just those with established diabetes but whether it is possible to reduce the population distribution of HbA_1c [glycated haemoglobin]." This implies that the determinant is the degree of hyperglycaemia.

In the non-diabetic population, glycated haemoglobin principally reflects the fasting blood glucose concentration, which has been shown by several groups to predict morbidity and mortality. One of these groups, the DECODE (diabetes epidemiology: collaborative analysis of diagnostic criteria in Europe) Study Group, also measured the blood glucose concentration two hours after a load and found that fasting concentrations were not additionally predictive within two hour blood glucose categories.²

In a cross sectional study with carotid intima media thickness as a marker of atherosclerosis, blood glucose concentrations after a load (in particular the two hour value) and incremental values (that is, above fasting) were more strongly related to intima media thickness than were fasting glucose or glycated haemoglobin concentrations.³ Furthermore, in a population based study in Italy instability of fasting blood glucose concentrations over the years of observation was an independent predictor of cardiovascular mortality.⁴

Clearly there are several possible interpretations of these findings. One is that it is not glycaemia itself that is the risk factor but the disturbance(s) in the physiological mechanisms that regulate the blood glucose concentration. This explanation would agree with the disappointing results of hypoglycaemic treatment in type 2 diabetes, commented on by Barrett-Connor and Wingard.⁵

R J Jarrett, emeritus professor of clinical epidemiology, University of London. 
45 Bishopsthorpe Road, London SE26 4PA

Analytical information is required for generalisation of data

EDITORAnalytical information is required before data can be generalised. Khaw et al's data suggesting that the relation between cardiovascular disease and glycaemia is a continuum extending throughout the non-diabetic population are fascinating.¹ Unfortunately, as the paper stands, the results are not generalisable since neither the methodology for assessing glycated haemoglobin nor calibration data have been included.

Professional organisations throughout the United Kingdom agree that percentage glycated haemoglobin concentration should be harmonised in relation to a common standard² and that, in the absence of a primary calibrant, this standard should be that used in the diabetes control and complications trial.³ Without this information, the results of the present study cannot be compared against others; it is inappropriate to discuss "a threshold commonly accepted for diagnosis of diabetes."

With an ion exchange method (HA-8140, Menarini Diagnostics) widely used in Europe the upper limit of the reference range observed in non-diabetic subjects has been reported as 5.1%⁴equivalent to 5.9% after standardisation as in the diabetes control and complications trial. If either of these cut off values was applicable to the present study a large (although vastly different) number of subjects in the study could not be considered to have normal glycaemic control. This could be explicable in terms of the age range of the cohort (45-79 years) and the association of increased age and decreased glucose tolerance.⁵

The suggestion that a large percentage of the cohort might have abnormal glycaemic control would not be surprising: Harris et al, for example, showed that 27.1% and 42.9% of the male population of the United States aged 55-64 and 65-74 respectively had either impaired glucose tolerance or diabetes.⁵ The present data could therefore be explained on the basis that many patients with impaired glucose tolerance were included in the upper two thirds of the non-diabetic population and that these patients have an increased rate of cardiovascular disease.

The authors should be encouraged to provide at least a non-diabetic reference range for their assay and preferably some data relating their results to a method aligned to the diabetes control and complications trial. This study can then be compared with other work in this area.

Finally, the authors state that "HbA_1c [glycated haemoglobin] may provide a practical screening tool for diabetes or impaired glucose tolerance." Although this statement could be true, the authors do not provide any evidence to support it. The World Health Organization has categorically stated that glycated haemoglobin concentration should not be used to establish the diagnosis of diabetes, and this has been reiterated by Diabetes UK.

Edmund Lamb, consultant clinical biochemist. 
East Kent Hospitals NHS Trust, Kent and Canterbury Hospitals, Canterbury CT1 3NG edmund.lamb{at}kch-tr.sthames.nhs.uk

1.	Khaw KT, Wareham N, Luben R, Bingham S, Oakes S, Welch A, et al. Glycated haemoglobin, diabetes, and mortality in men in Norfolk cohort of European prospective investigation of cancer and nutrition (EPIC-Norfolk). BMJ 2001; 322: 15-18. (6 January.)
2.	Marshall SM, Barth JH. Standardization of HbA1c measurementsa consensus statement. Diabetic Med 2000; 17: 5-6[CrossRef][Medline].
3.	Diabetes Control and Complications Trial Research Group. The effect of intensive treatment of diabetes on the development and progression of long-term complications in insulin-dependent diabetes mellitus. N Engl J Med 1993; 329: 977-986[Abstract/Free Full Text].
4.	John WG, Braconnier F, Miedema K, Aulesa C, Piras G. Evaluation of the Menarini-Arkray HA 8140 hemoglobin A1c analyzer. Clin Chem 1997; 43: 968-975[Abstract/Free Full Text].
5.	Harris MI, Hadden WC, Knowler WC, Bennett PH. Prevalence of diabetes and impaired glucose tolerance and plasma glucose levels in US population aged 20-74 yr. Diabetes 1987; 36: 523-534[Abstract].

Authors' reply

EDITORJarrett suggests that it is not glycaemia itself that is a risk factor but disturbances in the physiological mechanisms that regulate the blood glucose concentration. The point of our analysis was to examine the predictive value of glycated haemoglobin measures for the risk of death and the shape of the risk curve. In this cohort, glycated haemoglobin predicted risk of death continuously across the whole population distribution.

Most of the excess events associated with raised glycated haemoglobin concentrations occurred at values below those that would be used to define diabetes. Indeed, in the diabetes control and complications trial there was no glycaemic threshold for the development of long term complications; as the glycated haemoglobin concentration was reduced below 8% there were continuing relative reductions in the risk of complications such as microalbuminuria and retinopathy.¹ In the United Kingdom prospective diabetes study, strict control of blood glucose reduced microvascular complications significantly by 25%, but the study had inadequate power to detect a smaller difference (10%) in mortality related to diabetes.²

In response to Lamb, we would point out that full details of the assays, though omitted from our paper in the printed journal, are given in the longer version of the paper on the BMJ website (www.bmj.com/cgi/content/full/322/7277/15). As stated in that longer paper, the glycated haemoglobin measurements were made in a single laboratory using high performance liquid chromatography on a Biorad Diomat. This is a diabetes control and complications trial standardised method.

The World Health Organization and Diabetes UK may well have categorically stated that glycated haemoglobin should not be used to establish the diagnosis of diabetes. We would hope, though, that such consensus statements are based on evidence rather than opinion and may be reconsidered in the light of new evidence from studies such as ours about the nature of the relation between glycated haemoglobin concentration and risk of future events.

Of course we accept that there may be other reasons not to use measurements of glycated haemoglobin, such as the problems of standardisation between methods and the limited availability of the test in less developed parts of the world. However, both these technical issues could be rectified. There are no fundamental obstacles to using glycated haemoglobin concentration as a predictor of macrovascular complications and risk of death associated with hyperglycaemia.

Kay-Tee Khaw, professor of clinical gerontology. 
kk101{at}medschl.cam.ac.uk

Nick Wareham, Medical Research Council clinician scientist. 
Institute of Public Health, University of Cambridge, Addenbrooke's Hospital, Cambridge CB2 2QQ

Medicine is now using diagnostic criteria rather than reference ranges

EDITORKhaw et al's study showed that glycated haemoglobin concentration is associated with mortality in men without diabetes.¹ It is another illustration of the general rule for a continuous variable: subjects with characteristics outside the normal range have an increased incidence of acute and chronic conditions and higher mortality, and subjects with normal characteristics do not have equal mortalityin most cases the lowest mortality is not close to mean values.²

This rule is true for many variables, from height and weight to glycated haemoglobin concentration. It is true for population samples and for very healthy subjects.³ For most diseases the shape of the curve describing incidence against the variable is individual to that disease. For example, the curve of cancer mortality versus blood cholesterol concentration is different from that of cardiovascular mortality versus blood cholesterol concentration.⁴ Every curve can be explained if we try hard enough.

Medicine is drifting from using reference ranges to using diagnostic criteria. This means that the diagnostic threshold is based not on the distribution of values in healthy subjects but on the trade-off between getting false negative and false positive results.⁵ Current diagnostic criteria for diabetes trade off the probability of complications. Already there is a tendency for people with blood glucose concentrations in the normal range (but with impaired tolerance) to receive interventions; now comes the proposal to involve the whole population.

The situation is similar for the diagnosis of obesity: interventions are now given to people with a body weight indicating not obesity but overweight, and the population is under pressure to reduce weight. Khaw et al are wrong in stating that "it is uncertain whether the relation between blood glucose concentration and . . . diseases has a threshold or is a continuum." A threshold makes for a simplified decision rule; a blurred border between healthy subjects and people who are sick or at risk makes simple explanation hard, but it is reality.

Medicine is historically limited to more or less definite groups of people (those who are "diseased"). The alternative is a medicalisation of traits, habits, and risks. This expansion to the whole population arises because only a small proportion of subjects at risk is in the group above the diagnostic threshold; it is a simple consequence of the low effectiveness of the diagnosis. Indiscriminate interventions are usually inefficient.

Vasiliy Vlassov, professor. 
Saratov State Medical University, PO Box 1528, Saratov, 410601 Russia vvvla{at}sgu.ru