Blood glucose 2 h after an oral glucose load (2hBG) and glycohaemoglobin (GHb) (Corning agar-gel electrophoresis) levels were used as screening tests in a general practice diabetic screening programme. The diagnosis of diabetes (DM) was based on a separate oral glucose tolerance test (OGTT) in 223 of 1040 screened subjects, selected as a stratified sample biased towards higher levels of 2hBG and GHb. The GHb assay was also repeated at the recall examination and urine was tested for glycosuria before and after glucose administration. At a cut-off level of 8.1%, the screening GHb assay correctly identified 90% of all probable diabetics with a specificity of 85.3% (95% Cl 83.3-87.3%) and a positive predictive value of 14.0% (9.0-19.0%). The specificity of the screening GHb assay as a screening test for true DM was 45.8% (39.0-52.4%) at 90% sensitivity, and that of the recall GHb assay was 64.5% (57.9-71.1%). The screening 2hBG was 93.3% (88.9-97.7%) specific at 90% sensitivity as a screening test for true DM diagnosed by OGTT at recall. The test characteristics for fasting glycosuria were: sensitivity 16.7% (0-37.8%) and specificity 98.0% (96.0-100.0%). Equivalent values for the post-glucose test for glycosuria were: 72.7% (46.4-99.0%) and 77.4% (70.1-84.7%), respectively. While GHb assay is a poorer screening test for DM than the 2hBG at the single cut-off level quoted, comparison of the accuracy of the two tests shows that the GHb assay is only marginally less accurate. It is superior to testing for glycosuria as a screening test for DM and can be performed on any random blood sample, facilitating its use in population screening.