The authors' objective, with this study, was to determine the use in routine clinical practice of the cobalamin serum assay in evaluating patients suspected of having cobalamin deficiency. The design was a prospective observational study of a diagnostic test compared with a criterion gold standard. The setting was the Veterans Affairs hospital inpatient wards and outpatient clinics. The authors prospectively identified all patients who had serum cobalamin measured and enrolled all those with cobalamin < 133 pmol/L (180 pg/mL) and a random sample of patients with cobalamin greater than or equal to 133 pmol/L (N = 168). There were no interventions. To identify cobalamin deficiency, subjects underwent blinded clinical and laboratory examinations, Schilling test, and urinary methylmalonic acid assay, with a trial of cobalamin replacement, if needed. Of 1,599 patients assayed, 84 (5.2%) had low cobalamin levels. After evaluation, 16 of those with low cobalamin levels, but none of those with normal cobalamin levels, were confirmed as deficient. Positive predictive value of a low cobalamin level in this population was 22.2% (95% confidence interval, 12.6-31.8%). After correcting for sampling procedure, the calculated sensitivity rate was 100% (95% confidence interval, 20.6-100%), and specificity rate was 95.8% (95% confidence interval, 95.2-100%). The authors conclude that in routine use, the positive predictive value of the cobalamin assay is low. Low cobalamin levels should not be used alone to justify a lifetime of parenteral therapy or to rule out other conditions that mimic cobalamin deficiency.