Is it possible to develop a ‘one-size fits all’ prediction model for undiagnosed Type 2 diabetes?
Article first published online: 12 DEC 2013
© 2013 The Authors. Diabetic Medicine © 2013 Diabetes UK
Volume 31, Issue 1, pages 116–117, January 2014
How to Cite
Diabet. Med. 31, 116–117 (2014)
- Issue published online: 12 DEC 2013
- Article first published online: 12 DEC 2013
- Accepted manuscript online: 3 AUG 2013 01:51AM EST
- Manuscript Accepted: 30 JUL 2013
- Novo Nordisk Foundation
- Novo Nordisk A/S
We read with interest the study by Xu et al. . In their observational study of 7567 Chinese participants, they compare the ability of different indices of obesity (BMI, waist circumference, waist–hip ratio and waist-to-height ratio) to discriminate between individuals with and without previously undiagnosed Type 2 diabetes. The authors conclude that the waist-to-height ratio seems to be the best indicator for undiagnosed Type 2 diabetes. We have three comments on this study—the first is related to the diagnosis of diabetes, the second to potential sex differences and the third to the general assumption of a ‘one-size fits all’ indicator for Type 2 diabetes.
In the study by Xu et al. , diagnosis of Type 2 diabetes was based on fasting plasma glucose only and not by an oral glucose tolerance test or HbA1c. However, the pathogenesis of Type 2 diabetes is different between individuals developing diabetes by elevated fasting glucose, 2-h glucose or both , and the question arises whether the waist-to-height ratio is also the best obesity predictor in those with elevated 2-h plasma glucose levels. The authors state that the agreement between diabetes diagnosed by fasting and 2-h plasma glucose concentration is 91% in a subset of participants who have ingested a meal consisting of 75 g of glucose. The diagnostic criterion for classifying diabetes by a meal test is not provided, and the overlap is surprisingly high given that other studies have found that approximately half of the cases of Type 2 diabetes identified by screening have diabetes by elevated 2-h plasma glucose only (i.e. non-diabetic fasting glucose) [2, 3]. Thus, the results provided by Xu et al.  only seem to apply to half of the population developing Type 2 diabetes.
Another issue of interest is related to potential sex differences. Xu et al. nicely stratify their statistical analysis by sex and show that waist-to-height ratio is the best obesity predictor for both Type 2 diabetes and impaired fasting glycaemia in men, whereas waist circumference is an equally good predictor in women. These potential sex differences are more or less ignored by the authors in the discussion and conclusion. However, the findings are in accordance with previous studies showing that different measures of obesity predict diabetes and changes in glucose metabolism differently in men and women [4, 5]. Also, the fact that men develop diabetes at a lower BMI than women , and that men in general have higher fasting glucose levels than women before the onset of Type 2 diabetes , suggest that the underlying pathogenesis of Type 2 diabetes may differ between sexes. Thus, it is likely that Type 2 diabetes prediction models should be developed separately for men and women, in line with prediction models for cardiovascular disease .
A more general question arising from this study (and others) is whether it is relevant to continue searching for ‘one-size fits all’ indicators for Type 2 diabetes. Several studies have examined predictors and risk factors for Type 2 diabetes and cardiovascular disease, and corresponding prediction models have been developed [8-10]. However, such models have proved to have limited predictive ability, resulting in a large proportion of the population being misclassified. As a consequence, researchers continue to search for new risk factors and biomarkers to improve prediction and identification of diabetes and its complications. In our opinion, it is unlikely that single ‘indicators’ or global prediction models can adequately discriminate between individuals who will and will not develop Type 2 diabetes and cardiovascular disease. It is more likely that some risk factors (e.g. genetic information ) will only add value to risk prediction in subgroups of individuals. Thus, future prediction models may benefit from taking into account the heterogeneity in disease development between individuals or groups of individuals. Instead of using global arbitrary cut points for risk factors (e.g. waist-to-height ratio > 0.5 and/or BMI ≥ 24 kg/m2 ), we hope to face a future where risk prediction is tailored to individuals or groups of individuals with similar expected patterns of disease development.
Steno Diabetes Center A/S receives part of its core funding from unrestricted grants from the Novo Nordisk Foundation and Novo Nordisk A/S.
K.F. and D.V. are employed by Steno Diabetes Center A/S, a research hospital working in the Danish National Health Service and owned by Novo Nordisk A/S. K.F and D.V. own shares in Novo Nordisk A/S.
- 2Trajectories of cardiometabolic risk factors before diagnosis of three subtypes of type 2 diabetes: a post-hoc analysis of the longitudinal Whitehall II cohort study. Lancet Diabetes Endocrinol 2013; 1: 43–51., , , , , et al.
- 9The value of genetic information for diabetes risk prediction – differences according to sex, age, family history and obesity. PLoS One 2013; 8: e64307., , , , .