Aims/hypothesis. The yield of screening programmes for Type 2 diabetes in the existing healthcare setting might be lower than anticipated from tests of screening algorithms in data from epidemiological surveys. Our aims were to evaluate the reliability of the algorithms and the effectiveness of a proposed stepwise screening programme for Type 2 diabetes in general practice. Methods. The screening programme had four steps: (i) mail-distributed self-administered risk-chart; (ii) screening tests: random blood glucose (RBG) and HbA 1c; (iii) diagnostic procedure 1 for fasting blood glucose (FBG) (if RBG ≥5.5 mmol/l or HbA1c ≥6.1%); and (iv) OGTT as diagnostic procedure 2 (if 5.6≤FBG<6.1 mmol/l or HbA1c ≥6.1%). Abnormalities of glucose metabolism were classified according to the WHO 1999 criteria, based on capillary whole blood. The subjects were all patients between 40 and 69 years of age (n=60,926) who were registered in 88 general practices and had not been previously diagnosed with diabetes. Results. A total of 11,263 individuals had a high-risk risk-score and attended the screening consultation (step 1 test-positive). Of these, 30.1% needed diagnostic tests (step 2 test-positive) and 27.2% of these needed an OGTT (step 3 test-positive). The test-positive proportions were equal to the proportions obtained in data from a population-based survey from Step 2 onwards, and the algorithms were thus reliable. The identification rate was only 19% of all prevalent undiagnosed diabetes according to a recently published prevalence estimate. This was due to a large dropout rate among high-risk individuals prior to entry into the programme. Conclusions/interpretation. Population-based mail-distributed stepwise screening for Type 2 diabetes in general practice is ineffective, despite reliable screening algorithms, primarily because many high-risk individuals fail to participate.