This study investigated the comparability of cerebrospinal fluid (CSF) cutoffs for Elecsys immunoassays for amyloid beta (Aβ)42/Aβ40 or Aβ42/phosphorylated tau (p-tau)181 and the effects of measurement variability when predicting Alzheimer's disease (AD)-related outcomes (i.e., Aβ-positron emission tomography [PET] visual read and AD neuropathology). We studied 750 participants (BioFINDER study, Alzheimer's Disease Neuroimaging Initiative [ADNI], and University of California San Francisco [UCSF]). Youden's index was used to identify cutoffs and to calculate accuracy (Aβ-PET visual read as outcome). Using longitudinal variability in Aβ-negative controls, we identified a gray zone around cut-points where the risk of an inconsistent predicted outcome was>5%. For Aβ42/Aβ40, cutoffs across cohorts were<0.059 (BioFINDER),<0.057 (ADNI), and<0.058 (UCSF). For Aβ42/p-tau181, cutoffs were<41.90 (BioFINDER),<39.20 (ADNI), and<46.02 (UCSF). Accuracy was ≈90% for both Aβ42/Aβ40 and Aβ42/p-tau181 using these cutoffs. Using Aβ-PET as an outcome, 8.7% of participants fell within a gray zone interval for Aβ42/Aβ40, compared to 4.5% for Aβ42/p-tau181. Similar findings were observed using a measure of overall AD neuropathologic change (7.7%vs. 3.3%). In a subset with CSF and plasma Aβ42/40, the number of individuals within the gray zone was ≈1.5 to 3 times greater when using plasma Aβ42/40. CSF Aβ42/p-tau181 was more robust to the effects of measurement variability, suggesting that it may be the preferred Elecsys-based measure in clinical practice and trials.