Controlling the diabetes epidemic: how should we screen for undiagnosed diabetes and dysglycaemia?

Aims To compare the detection of undiagnosed diabetes and dysglycaemia (impaired glucose tolerance, impaired fasting glucose, diabetes) using risk factors and laboratory measures of glycaemia. Methods Casual blood glucose samples were taken from 1899 (69.4% of 2737 invited) European, Maori and Pacific Islands subjects aged 40-79 years from randomly selected households in South Auckland, New Zealand. Of these, 534 attended for a 75-g oral glucose tolerance test (OGTT) if an elevated result was identified [327/478 (68.4%)] or if randomly selected with a 'normal' screening result [207/308 (67.2%)]. Results Several Europeans with undiagnosed diabetes (25.0%) and dysglycaemia (31.4%) had no diabetes risk factors. Most Maori and Pacific Islanders had at least one risk factor. The area under the receiver operating curve (ROC) for the detection of undiagnosed diabetes was 0.92 (0.89-0.95) using fasting glucose, 0.86 (0.82-0.90) using HbA(1c), 0.75 (0.69-0.80) using random glucose, but 0.60 (0.55-0.66) using risk factor screening. The ROC for detecting any dysglycaemia was 0.88 (0.85-0.90), 0.68 (0.64-0.71), 0.72 (0.69-0.75), 0.61 (0.58-0.65), respectively. Screening using fasting glucose (the best test) detected 90.4% of new diabetes and 78.4% of dysglycaemia; risk factor screening followed by fasting glucose detected significantly less cases [88 (82-93)% and 86 (82-89)%, respectively] with 9.2% less OGTTs. Conclusions Using risk factors for the identification of who should receive a blood test for dysglycaemia adds little to direct screening with the risk of missing some with significant hyperglycaemia. Screening for dysglycaemia may best be undertaken using blood tests without initial risk factor symptom screening.