We assessed the prevalence of previously unrecognized hemochromatosis among patients in whom diabetes mellitus was diagnosed after the age of 30 yr, and we evaluated the positive predictive value of biochemical screening tests for hemochromatosis in diabetic subjects. Thirty-eight of 572 patients screened (6.6%) had a serum ferritin level greater than 324 micrograms/L; 16 patients had normal levels on repeat testing. Four patients' serum ferritin levels fell to less than 400 micrograms/L. Seven of 18 patients with a persistently elevated serum ferritin level did not undergo a liver biopsy because of a recognized cause of hyperferritenemia (carcinoma, alcoholism, or systemic lupus erythematosus). The diagnosis of hemochromatosis seemed certain in 1 of 3 patients who were not biopsied for technical reasons. Of 8 patients biopsied, 2 had hemochromatosis, 4 had fatty liver, 1 had hemosiderosis, and 1 had a chronic inflammatory cell infiltrate with no iron deposition. Of 4 patients with a raised transferrin saturation level, 2 had raised serum ferritin levels and hemochromatosis, 1 had raised serum ferritin and hemosiderosis on liver biopsy, and 1 had a normal transferrin saturation level on repeat testing. Two of 3 cases of hemochromatosis had other clinical markers of the condition. Therefore, routine screening of diabetic patients for hemochromatosis is not necessary, because patients with hemochromatosis will often have other clinical features of the disease. When screening diabetic patients for hemochromatosis, it should be remembered that a persistently raised serum ferritin level has a low positive predictive value (16.6%) and that a normal transferrin saturation level does not exclude the diagnosis.