BACKGROUND AND PURPOSE: In vivo detection of β-amyloid (Aβ) plaques in Alzheimer's disease (AD) is now possible with 11 C-PiB positron emission tomography (PET). Conventionally, a cortical:cerebellar PiB uptake ratio threshold of 1.4-1.5 has been used to categorize at-risk subjects as "amyloid-positive" and "amyloid-negative." It has been suggested that this threshold is too conservative and may miss early amyloid pathology. We investigated the relationship between conventional and lower baseline 11 C-PiB PET thresholds for raised amyloid load and the subsequent clinical and radiological progression of mild cognitive impairment (MCI) cases longitudinally. METHODS: We serially determined the cortical amyloid load with 11 C-PiB PET of 44 MCI subjects over 2 years and compared findings with those for 12 healthy controls (HC) and 5 AD cases. RESULTS: Twenty-four subjects were classified as normal at baseline with mean cortical PiB standard uptake value ratios (SUVR) between 1.2 and 1.5. Their cognitive status remained stable over time. Three of these cases increased their amyloid load above a threshold of 1.5 over 2 years. Twenty-seven "raised amyloid" MCI cases with baseline cortical SUVRs above 1.5, showed deteriorating cognition. Note that 50% of these cases converted clinically to AD during the follow-up period. CONCLUSION: Use of a PiB SUVR threshold of >1.5 for raised amyloid missed 14.3% of MCI cases who likely had Thal stage 1 or 2 pathology and showed a progressive amyloid increase over 2 years. Lowering the threshold for abnormality to 1.3 abolished all false negatives but resulted in 75% of HCs being falsely diagnosed as raised amyloid subjects.