Abstract
Rationale: Lifestyle intervention programs are effective for preventing diabetes among those with impaired glucose tolerance (IGT), with little evidence of benefit in other high-risk states. The selection of those who are eligible for lifestyle intervention is often made via diabetes prediction tools which identity those at high risk of diabetes. Unless such scores preferentially identify IGT, this will lead to poor performance of diabetes prevention programs. We developed an IGT-specific risk model and a diabetes-specific risk model, and compare their abilities to identify those with IGT.
Methods: Data from population-based surveys conducted in Mauritius in 1987, 1992 and 1998 were used to develop prediction models for IGT and DM. A cross-sectional sample in 1987 was used to develop the IGT model. A cohort sample from 1987 to 1992 was used to develop the DM model. Logistic regression was used to develop the scores. The performances were investigated by calculating area under the receiver operator curve (aROC). The Youden index was used to find a cutpoint in probability. Both scores were tested on independent samples.
Results: The aROC for IGT and DM models were 0.68 and 0.69 respectively. The sensitivity of the IGT model (derived on 1987 data) in identifying IGT individuals on independent samples from 1992 was 66%. Among those above the Youden index cutpoint, 23%, 9% and 68% had IGT, IFG and NGT, respectively. In the DM model, sensitivity for IGT was 60%. Among those above the cutpoint 24%, 8% and 67% had IGT, IFG and NGT, respectively. Increasing the cutpoints to levels at which 40% of those above the cutpoint had IGT resulted in sensitivities for IGT of 15% and 5% in the IGT and DM models, respectively.
Conclusion: Our results show that risk scores perform poorly in identifying people with IGT, and that without screening identified populations with an oral glucose tolerance test, a high proportion of people entering DM prevention programs will not have IGT-the only state shown to respond to lifestyle.
Disclosure
C.I. Sari: None. D.J. Magliano: None. P.Z. Zimmet: None. J. Tuomilehto: None. K. Sudhirsen: None. S. Söderberg: Other Relationship; Self; GlaxoSmithKline plc.. Speaker's Bureau; Self; Actelion Pharmaceuticals US, Inc.. Other Relationship; Self; Swedish Heart-Lung Foundation. Advisory Panel; Self; Merck Sharp & Dohme Corp. J.E. Shaw: Advisory Panel; Self; AstraZeneca. Research Support; Self; AstraZeneca, Boehringer Ingelheim GmbH. Advisory Panel; Self; Merck Sharp & Dohme Corp.. Speaker's Bureau; Self; Mylan. Advisory Panel; Self; Novo Nordisk Inc..