Persistent fasting hyperglycaemia is more predictive of Type 2 diabetes than transient fasting hyperglycaemia
Article first published online: 19 JUN 2012
© 2011 The Authors. Diabetic Medicine © 2011 Diabetes UK
Volume 29, Issue 7, pages e75–e81, July 2012
How to Cite
Inoue, K., Inoue, M., Matsumoto, M. and Akimoto, K. (2012), Persistent fasting hyperglycaemia is more predictive of Type 2 diabetes than transient fasting hyperglycaemia. Diabetic Medicine, 29: e75–e81. doi: 10.1111/j.1464-5491.2011.03536.x
- Issue published online: 19 JUN 2012
- Article first published online: 19 JUN 2012
- Accepted manuscript online: 12 DEC 2011 02:01AM EST
- Accepted 6 December 2011
- cohort study;
- diabetes mellitus;
- fasting plasma glucose;
- persistent/transient fasting hyperglycaemia
Diabet. Med. 29, e75–e81 (2012)
Aims We investigated the value of persistent fasting hyperglycaemia as assessed by repeated elevated fasting plasma glucose in predicting the progression to diabetes.
Methods A retrospective cohort study was conducted from 1998 to 2006 inclusive among 7929 persons (37 742 person-years), with a mean age of 53.0 years at baseline. The cumulative incidence of diabetes was measured. A baseline and follow-up fasting plasma glucose were categorized as normal fasting glucose (< 5.56 mmol/l), or impaired fasting glucose (5.56–6.94 mmol/l).
Results The cumulative incidence and incidence density of diabetes were 3.5% (275 cases) and 7.3 per 1000 person-years over a mean follow-up period of 4.8 years. The cumulative incidence of diabetes among subjects with impaired fasting glucose at both previous examinations (persistent impaired fasting glucose) was 30.4% (222/1518) compared with 0.6% (15/5063) of those with normal fasting glucose at both baseline and initial follow-up. The hazard ratios to develop diabetes, adjusted for possible confounders, was 37.10 (95% CI, 21.6–63.7) for persistent impaired fasting glucose versus persistent normal fasting glucose. Persistent impaired fasting glucose predicted diabetes at 80.7% (222/275) sensitivity and 83.1% (6358/7654) specificity, whereas first baseline impaired fasting glucose only predicted diabetes at 86.9% (239/275) sensitivity and 74.9% (5730/7654) specificity. The model using both previous fasting plasma glucose levels had a greater AUROC (area under receiver operating characteristic) than that using first baseline fasting plasma glucose only (0.92 vs. 0.88; P < 0.001).
Conclusions Repeated measurements of fasting plasma glucose better predicts incidence of diabetes than a single test. In particular, persistent fasting hyperglycaemia adds more substantial precision to the prediction of future diabetes than transient impaired fasting glucose. This combination is cost efficient and may be practical for early detection of high-risk individuals.