Aims/hypothesis. Patients with glucokinase mutations are characterised by mild, persistent fasting hyperglycaemia, a small increment in glucose in response to an oral load and a dominant family history. These patients frequently present with gestational diabetes and often require insulin treatment during pregnancy. We assessed whether the selection of gestational diabetic subjects by clinical criteria would result in a high detection rate of glucokinase mutations.¶Methods. Caucasian gestational diabetic subjects from the United Kingdom who had fasting hyperglycaemia in pregnancy but did not meet the diagnostic criteria for maturity-onset diabetes of the young (MODY) were selected for direct sequencing of the glucokinase gene if they fulfilled the following four criteria; (1) persisting fasting hyperglycaemia outside pregnancy (5.5–8 mmol/l) (2) a small increment ( < 4.6 mmol/l) during a 2-h oral glucose tolerance test (3) insulin treatment during at least one pregnancy but subsequently controlled on diet and (4) a history of Type II (non-insulin-dependent) diabetes mellitus, gestational diabetes or fasting hyperglycaemia ( > 5.5 mmol/l) in a first-degree relative.¶Results. Of the 15 subjects 12 (80 %) with all these clinical criteria had glucokinase gene mutations. These included four previously unreported mutations (N180K, R191W, Y215X and L288–1G → A).¶Conclusion/interpretation. Phenotypic selection of subjects with gestational diabetes greatly increases the likelihood of detecting a mutation in the glucokinase gene as previous studies have suggested a prevalence of 2.5 % (range 0–6 %). Our study in gestational diabetes to successfully used clinical criteria to assist in the definition of a genetic subgroup. [Diabetologia (2000) 43: 250–253]