Can a screening programme for diabetes be applied to another population?

Abstract

To compare the performance of a Danish diabetes screening protocol in populations from Denmark and Australia. The populations used were the Inter99 population from Denmark and the AusDiab population from Australia. The Inter99 study included 6270 individuals aged 30-60 years, randomly selected from Copenhagen County. The AusDiab study included 7079 individuals also aged 30-60 years, randomly selected from throughout Australia. In both studies, all individuals without known diabetes underwent an oral glucose tolerance test (OGTT). Screening using a Danish risk score as an initial step was followed by measurement of fasting plasma glucose (FPG). The characteristics of the study populations showed some differences in risk profile, with more individuals in the AusDiab study being obese, whilst in the Danish cohort, mean blood pressure was higher, less people took anti-hypertensive medication and there was a higher prevalence of screen-detected diabetes. Comparing the Australian and Danish populations, overall performance of the screening protocol was similar-area under area receiver operator characteristic (ROC) curve 0.75 vs. 0.77, sensitivity 71 vs. 76% and positive predictive value (PPV) 6 vs. 9%. Small but statistically significant differences were observed in specificity (70 vs. 66%; P < 0.001) and percentage of the population requiring further testing (31 vs. 36%; P < 0.001). Combining the risk score with FPG > or = 6.1 mmol/l showed a similar pattern. Again, specificity was slightly but significantly higher in the Australian population (95 vs. 93%; P < 0.001) and percentage of the population requiring further testing was lower (5 vs. 8%; P < 0.001). The Danish risk score performed well when applied to a geographically different Caucasian population and is a suitable tool for detecting people at high risk of undiagnosed diabetes.