Objective. To determine the optimal means of identifying patients with undiagnosed haemochromatosis.Design. Case-control study where cases are defined by the presence of specific clinical diagnoses or symptoms.Setting. Primary care patients were recruited from three Oxfordshire practices and secondary care patients were recruited from those patients attending specialist clinics in Amiens University Hospital.Subjects. A total of 569 patients recruited via hospital clinics and 60 primary care patients (recruited from 4022 consultations) presenting with the following haemochromatosis associated conditions, diabetes, arthralgia/chronic fatigue, osteoporosis or arthropathy were studied. The control group, a total of 991 healthy volunteers, were recruited through a Health Appraisal Centre. Patients and controls were included in the study if they or their family members had not previously been diagnosed with hereditary haemochromatosis.Main outcome measures. Serum ferritin concentration, transferrin saturation (Tsat) and presence of HFE mutations, C282Y and H63D. The check-up in controls consisted of a questionnaire, clinical examination, biochemical tests and screening for the presence of the C282Y and H63D mutations.Results. Patient groups presenting with unstable diabetes or chronic fatigue and arthralgia together with a raised serum ferritin concentration showed an enrichment in the haemochromatosis-associated genotype HH/YY, odds ratio (OR) = 40.1, confidence interval (CI) = 8.0-202.1 and OR = 103, CI = 22.9-469.7, respectively.Conclusion. Patients presenting to hospital clinics with haemochromatosis associated conditions should be screened biochemically for iron overload. Only those with a serum ferritin >300 mug L-1 or Tsat >40% should subsequently go on to be genotyped for HFE mutations. The patients at greatest risk of having undiagnosed haemochromatosis are those presenting with unstable diabetes, or fatigue and/or arthralgia in the absence of any other explanation.