Recently, Jack and colleagues [10
] proposed a hypothetical model that positioned established and novel biomarkers in the continuum of AD. This model elaborates on the amyloid-cascade hypothesis [11
], which states that accumulation of Aβ initiates a cascade of neuropathological events, such as the formation of neurofibrillary tangles and neuroinflammation. This may in turn lead to neurodegeneration that presumably is followed by cognitive deterioration and finally results in dementia. Accumulation of Aβ is thought to set in decades before the first cognitive complaints arise, starts to accelerate already in the preclinical stage of AD, and reaches a relative plateau at the time symptoms emerge [12
In the study by Koivunen and colleagues, the overall amyloid burden in MCI patients increased approximately 2.5% in 2 years. This increase was most prominent for those patients who did not convert to AD. However, the overall increase in amyloid deposition was only modest, whilst hippocampal volumes decreased more strongly. This is in line with the idea that Aβ starts o the cascade and may uncouple at a later time point from neurodegenerative processes. Previous studies that related the presence of amyloid plaques with the course of brain atrophy [13
] and glucose hypometabolism [14
] also found this dissociation between amyloid deposition and structural and functional changes. The findings of Koivunen and colleagues suggest that the time span between deposition of amyloid and actual neurodegeneration may perhaps be shorter than previously assumed, given the dynamic Aβ changes in MCI patients in this study.
The finding of increased amyloid burden over time in MCI patients is consistent with other recent longitudinal amyloid imaging studies. Villemagne and colleagues [15
] reported that 65 MCI patients had a mean increase in [11
C]PIB retention of 2.1% over a 20-month follow-up period (annual increase of 1.3%). A study by Jack and colleagues [16
] showed an annual increase of 1.7% in 32 MCI patients. Moreover, in an unpublished study from our group in 12 MCI patients, [11
C]PIB binding increased by 5.0% over a 30-month period (annual increase 2%). These results are in accordance with the 2.5% increase over 24 months (annual increase 1.25%) reported by Koivunen and colleagues. In all these longitudinal studies, the increased amyloid burden was primarily driven by MCI patients that displayed high [11
C]PIB retention at baseline. Taken together, amyloid deposition seems to increase in both MCI converters and non-converters, thereby challenging the theory that amyloid plaque load is stable in the prodromal stage of AD.
A limitation of the study by Koivunen and colleagues, especially given its longitudinal design, is the use of SUVr, which overestimates [11
C]PIB binding in comparison with fully quantitative kinetic models [17
]. Furthermore, SUVr does not correct for factors that are inherently associated with disease progression, such as heterogeneous flow effects in the region of interest compared to the reference region. We therefore argue that quantitative modeling, and thus dynamic PET scanning, is essential for longitudinal amyloid imaging studies.