The 40 and 42 amino-acid residue forms of amyloid beta (Abeta(1-40) and Abeta(1-42)) in cerebrospinal fluid (CSF) have been proposed as potential biomarkers of Alzheimer's disease (AD). Quantitative analyses of Abeta peptides in CSF have relied almost exclusively on the use of immunoassay-based assays such as the enzyme-linked immunosorbent assay (ELISA) procedure. However, due to the ability of the Abeta peptides to readily self-aggregate or bind to other proteins and glassware, such analyses are extremely challenging. Analyses are further complicated by the potential of the peptides to undergo post-translational modifications and the possibilities for cross-reaction in the ELISA assays with endogenous components of the CSF. An approach based on liquid chromatography/tandem mass spectrometry (LC/MS/MS) has now been developed which overcomes these methodological issues. The key steps in implementing this new approach involved immunoaffinity purification coupled with the use of [15N]-labeled Abeta peptides as internal standards, a basic LC mobile phase, negative ion electrospray ionization, and a basic solvent for dissolving the peptides and washing the injection needle to prevent carryover of analytes during multiple injections on the LC/MS system. The validated method had limits of quantitation of 44 fmol/mL (200 pg/mL) for Abeta(1-42) and 92 fmol/mL (400 pg/mL) for Abeta(1-40). An excellent correlation was found between the LC/MS/MS assay and an ELISA assay for Abeta(1-42) in human CSF (r2 = 0.915), although less correlation was observed for Abeta(1-40) (r2 = 0.644). Mean CSF Abeta(1-42) concentrations for samples collected 2 weeks apart from a limited number of AD patients provided additional confidence in the reproducibility of the LC/MS/MS assay. Concentrations for duplicate samples from AD patients were slightly higher than most previously reported values (mean 1.06 +/- 0.25 ng/mL; n = 7). Abeta(1-40) concentrations in duplicate samples obtained from AD patients were also reproducible but were found to be slightly lower than most previously reported values (mean 6.36 +/- 3.07 ng/mL; n = 7). Consistent with literature reports, mean Abeta(1-42) concentrations were found to be lower in AD patients compared with the normal subjects (mean 1.49 +/- 0.59 ng/mL; n = 7), whereas there was no difference in Abeta(1-40) concentrations between AD patients and normal subjects (mean 5.88 +/- 3.03 ng/mL; n = 7). The accuracy and precision of the LC/MS assay mean that it will be a useful complement to existing ELISA assays for monitoring therapeutic interventions designed to modulate CSF Abeta(1-42) concentrations in individual AD patients. Moreover, the introduction of stable isotope labeled internal standards offers the potential to achieve a more rigorous account of the influence of methodological effects related to sample collection and processing.