Premium
Prediction models for incident Type 2 diabetes mellitus
in the older population: KORA S4/F4 cohort study
Author(s) -
Rathmann W.,
Kowall B.,
Heier M.,
Herder C.,
Holle R.,
Thorand B.,
Strassburger K.,
Peters A.,
Wichmann H.E.,
Giani G.,
Meisinger C.
Publication year - 2010
Publication title -
diabetic medicine
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.474
H-Index - 145
eISSN - 1464-5491
pISSN - 0742-3071
DOI - 10.1111/j.1464-5491.2010.03065.x
Subject(s) - medicine , cohort , diabetes mellitus , type 2 diabetes mellitus , cohort study , population , type 2 diabetes , gerontology , endocrinology , environmental health
Diabet. Med. 27, 1116–1123 (2010) Abstract Background The aim was to derive Type 2 diabetes prediction models for the older population and to check to what degree addition of 2‐h glucose measurements (oral glucose tolerance test) and biomarkers improves the predictive power of risk scores which are based on non‐biochemical as well as conventional clinical parameters. Methods Oral glucose tolerance tests were carried out in a population‐based sample of 1353 subjects, aged 55–74 years (62% response) in Augsburg (Southern Germany) from 1999 to 2001. The cohort was reinvestigated in 2006–2008. Of those individuals without diabetes at baseline, 887 (74%) participated in the follow‐up. Ninety‐three (10.5%) validated diabetes cases occurred during the follow‐up. In logistic regression analyses for model 1, variables were selected from personal characteristics and additional variables were selected from routinely measurable blood parameters (model 2) and from 2‐h glucose, adiponectin, insulin and homeostasis model assessment of insulin resistance (HOMA‐IR) (model 3). Results Age, sex, BMI, parental diabetes, smoking and hypertension were selected for model 1. Model 2 additionally included fasting glucose, HbA 1c and uric acid. The same variables plus 2‐h glucose were selected for model 3. The area under the receiver operating characteristic curve significantly increased from 0.763 (model 1) to 0.844 (model 2) and 0.886 (model 3) ( P < 0.01). Biomarkers such as adiponectin and insulin did not improve the predictive abilities of models 2 and 3. Cross‐validation and bootstrap‐corrected model performance indicated high internal validity. Conclusions This longitudinal study in an older population provides models to predict the future risk of Type 2 diabetes. The OGTT, but not biomarkers, improved discrimination of incident diabetes.