Background: Alzheimer's disease (AD) pathology can be quantified in vivo using cerebrospinal fluid (CSF) levels of amyloid-β1-42 (Aβ1-42), total-tau (t-tau), and phosphorylated tau (p-tau181p), as well as with positron emission tomography (PET) using [11C]Pittsburgh compound-B ([11C]PIB). Studies assessing concordance between these measures, however, have provided conflicting results. Moreover, it has been proposed that [11C]PIB PET may be of great clinical utility in terms of identifying patients with mild cognitive impairment (MCI) who will progress to the dementia phase of AD. Objective: To determine concordance and classification accuracy of CSF biomarkers and [11C]PIB PET in a cohort of patients with MCI and AD. Methods: 68 patients (MCI, n = 33; AD, n = 35) underwent [11C]PIB PET and CSF sampling. Cutoffs of >1.41 ([11C]PIB), <450 pg/mL-and a more lenient cutoff of 550 pg/mL-(Aβ1-42), <6.5 (Aβ1-42/p-tau181p), and 1.14 (Aβ1-42/t-tau), were used to determine concordance. Logistic regression was used to determine classification accuracy with respect to stable MCI (sMCI) versus MCI who progressed to AD (pMCI). Results: Concordance between [11C]PIB and Aβ1-42 was highest for sMCI (67%), followed by AD (60%) and pMCI (33%). Agreement was increased across groups using Aβ1-42 <550 pg/mL, or Aβ1-42 to tau ratios. Logistic regression showed that classification accuracy of [11C]PIB, between sMCI and pMCI, was superior to Aβ1-42 (73% versus 58%), Aβ1-42/t-tau (63%), and Aβ1-42/p-tau181p (65%). Conclusion: In the present study, [11C]PIB proved a better predictor of progression to AD in patients with MCI, relative to CSF measures of Aβ1-42 or Aβ1-42/tau. Discordance between PET and CSF markers for Aβ1-42 suggests they cannot be used interchangeably, as is currently the case.