OBJECTIVE: To evaluate the diagnostic potential of cerebrospinal fluid (CSF) levels of tau and beta-amyloid protein ending at amino acid 42 (Abeta42) as biomarkers for Alzheimer disease (AD) in clinical practice. DESIGN: A 1-year prospective study. SETTING: Community population-based sample of all consecutive patients admitted for investigation of cognitive symptoms to the Piteå River Valley Hospital, Piteå, Sweden. PATIENTS: A total of 241 patients with probable AD (n = 105), possible AD (n = 58), vascular dementia (n = 23), mild cognitive impairment (n = 20), Lewy body dementia (n = 9), other neurological disorders (n = 3), and psychiatric disorders (n = 5) and nondemented individuals (n = 18). MAIN OUTCOME MEASURES: Cerebrospinal fluid tau and CSF-Abeta42 were assayed each week as routine clinical neurochemical analyses. Sensitivity and specificity were defined using the regression line from 100 control subjects from a multicenter study. Positive and negative predictive values were calculated for different prevalence rates of AD. RESULTS: We found increased CSF-tau and decreased CSF-Abeta42 levels in probable and possible AD. Sensitivity was 94% for probable AD, 88% for possible AD, and 75% for mild cognitive impairment, whereas specificity was 100% for psychiatric disorders and 89% for nondemented. Specificity was lower in Lewy body dementia (67%) mainly because of low CSF-Abeta42 levels and in vascular dementia (48%) mainly because of high CSF-tau levels. Sensitivity for CSF-tau and CSF-Abeta42 increased in patients with AD possessing the ApoE epsilon4 allele, approaching 100%. At a prevalence of AD of 45%, the positive predictive value was 90% and the negative predictive value was 95%. CONCLUSIONS: Cerebrospinal fluid tau and CSF-Abeta42 have so far been studied in research settings, under conditions providing data on the optimal performance. We examined a prospective patient sample, with assays run in clinical routine, giving figures closer to the true performance of CSF-tau and CSF-Abeta42. The predictive value for AD was greater than 90%. Therefore, these biomarkers may have a role in the clinical workup of patients with cognitive impairment, especially to differentiate early AD from normal aging and psychiatric disorders.