Alzheimer's disease biomarker CSF Aβ1-42 is susceptible to numerous factors that compromise measurement accuracy. We evaluated assay stability, Aβ peptide recoveries, and Aβ peptide stabilities over a 2-year period to inform CSF handling and analysis protocols for clinical trials. A UP-RPLC-MS/MS assay was developed at Pharmaceutical Product Development with the Alzheimer's Association's Global Consortium for Biomarker Standardization to quantitate CSF Aβ1-42, 1-40, 1-38, 1-37, 1-34 using stable isotope-labeled peptides. CSF was collected from young normal controls and aged patients with dementia. CSF was denatured with 6M guanidine hydrochloride (GuHCl) at different stages of the collection and aliquoting procedure (after collection (“Collection”), before sub-aliquoting (“Aliquot”), or before solid phase extraction (“Analysis”)) and then stored at -80oC. In addition, a 1 mL aliquot was prepared in LoBind tubes at the time of collection, stored at -80oC, and GuHCl was added before extraction (“LoBind”). Measurements were made at baseline, and then every 3 months for 2 years. Longitudinal stabilities were assessed according to CLSI standards, with an acceptability cutoff of +/-10% of baseline. Earlier denaturation increased baseline CSF Aβ recovery relative to later denaturation (Mean (SD) CSF Aβ1-42collection=1061(476) pg/mL, CSF Aβ1-42Aliquot=996(414) pg/mL; vs. CSF Aβ1-42Analysis=822(385) pg/mL, CSF Aβ1-42LoBind=899(376) pg/mL (p<0.0002)). The Aβ1-42/Aβ1-40 ratio reduced recovery variability due to collection condition, and improved discrimination between control and dementia groups. The assay was stable for 2 years, but CSF Aβ peptide measurements and the Aβ1-42/Aβ1-40 ratio were not longitudinally stable for the study period. CSF Aβ1-42 was stable for 15months (“Collection”,“Aliquot”) or 18months (“Analysis”). The Aβ1-42/Aβ1-40 ratio was stable for 8 months (“Collection”,“Aliquot”) and 11months (“Analysis”). When measured with a UP-RPLC-MS/MS assay, earlier denaturation improved CSF Aβ peptide recovery. CSF Aβ1-42, Aβ1-40, Aβ1-38, Aβ1-34, and the Aβ1-42/Aβ1-40 ratio were not stable for the study period when samples were stored frozen. The Aβ1-42/Aβ1-40 ratio was useful for normalizing adsorption-related artifacts at baseline and for separating Dx groups, but was more sensitive to longitudinal instability. We recommend that UP-RPLC-MS/MS analysis of CSF Aβ peptides from frozen samples occur within the cutoffs specified or close to the time of collection.