OUP user menu

Two-year follow-up of amyloid deposition in patients with Alzheimer's disease

William E. Klunk, Chester A. Mathis, Julie C. Price, Brian J. Lopresti, Steven T. DeKosky
DOI: http://dx.doi.org/10.1093/brain/awl281 2805-2807 First published online: 27 October 2006

Understanding the natural history of amyloid deposition is of great importance for advancing our knowledge of the pathophysiology of Alzheimer's disease—both during the clinically apparent phase as well as during the antecedent pre-clinical phase (Goldman et al., 2001). Awareness of the natural history of amyloid deposition also will be necessary to interpret experimental anti-amyloid drug studies that might extend over a year or more. Post-mortem studies cannot, of course, directly assess the progression of amyloid deposition in an individual over time and attempts to deduce the natural history by comparing a series of post-mortem cases with different clinical severities and amyloid loads have significant limitations (Hyman and Gomez-Isla, 1997). Nevertheless, several very extensive post-mortem studies have provided useful information (Braak and Braak, 1997). For example, although some degree of correlation has been reported between plaque load (Cummings et al., 1996) or Aβ levels (Naslund et al., 2000) and measures of cognition, it is generally believed that neither the number nor the total area of neocortical plaques correlate well with cognitive deficits before death (Terry et al., 1991; Braak and Braak, 1998). It is with this backdrop that we must interpret the manuscript by Engler et al. (2006) in which they report the first longitudinal study of Aβ amyloid deposition in living subjects. The original cohort was scanned in 2002–2003 and was the subject of the initial report on amyloid imaging using the positron emission tomography (PET) tracer, PIB (Pittsburgh Compound-B) (Klunk et al., 2004). Remarkably, Engler and co-workers succeeded in re-scanning all 16 of the original subjects who had a clinical diagnosis of Alzheimer's disease 1.5–2.5 years after the baseline PIB scan. They also re-scanned the one control who showed evidence of high PIB retention in the baseline study. As has been shown in previous studies (Lopresti et al., 2005; Price et al., 2005), Engler et al. found PIB retention to vary only by ∼5% between two scans performed within hours to days of each other. This small interscan variability, coupled to the 60–95% higher PIB retention in Alzheimer's disease patients (versus controls) reported in this and the original study (Klunk et al., 2004) suggests that, if applied at regular intervals over the entire period of active amyloid accumulation, PIB–PET could accurately track the increase in amyloid deposition that must occur between the time a subject begins to deposit amyloid and when that subject attains an amyloid load typical of Alzheimer's disease.

However, Engler and co-workers do not find increases in PIB retention (and presumably amyloid deposition) over the 2 year follow-up. Instead, their two primary findings were (i) PIB retention showed no significant change over the 2 ± 0.5 years of follow-up and (ii) regional cerebral metabolism rate for glucose (rCMRglc; indexed by FDG–PET) fell an additional 20% from baseline. Since cognition typically worsens steadily over time in Alzheimer's disease and since post-mortem amyloid deposits do not correlate well with cognition (Terry et al., 1991; Braak and Braak, 1998), the lack of progression in PIB retention is not unexpected. Still, several considerations make these findings less straightforward. First, it is well established that brain atrophy proceeds throughout the course of Alzheimer's disease at a rate of 2–4% per year (Fox and Freeborough, 1997; Jack et al., 1998). Since PIB is a positive change and rCMRglc is a negative change, progressive atrophy would tend to blunt increases in PIB retention and exaggerate decreases in rCMRglc. The Engler et al. study was done without MRI co-registration and atrophy correction, so the contribution of brain atrophy to the reported findings cannot be determined. Second, in contrast to the baseline study, this study normalized cortical PIB retention to that in the cerebellum. The use of cerebellar data to correct for non-specific uptake is an important quantification step that should be performed consistently across studies and the small test–retest variability of cerebellar reference methods makes them attractive for longitudinal studies such as this (Lopresti et al., 2005; Price et al., 2005). However, there are known caveats related to the use of cerebellum as a reference region and these have been discussed (Lopresti et al., 2005). The typical cerebellar region of interest (ROI) lies between and in close proximity to non-specific PIB retention in the cerebellar peduncles and specific PIB retention in the occipital lobe in Alzheimer's disease patients, so the use of MRI co-registration to guide accurate placement of the cerebellar reference ROI is highly advisable. The effect of not employing MRI co-registration to place the reference ROI in this study is not clear. A third consideration is the difference between changes in group averages and trajectories of individual subjects. The individual ‘heterogeneity’ to which Engler and co-workers allude in their discussion is evidenced in the difference in PIB changes between the cognitively stable Alzheimer's disease (Alzheimer's disease-S) subjects and the progressively declining AD patients (Alzheimer's disease-P). Comparing areas of highest PIB retention (frontal and posterior cingulate/precuneus) we see a slight increase in the Alzheimer's disease-S group and small decrements in the Alzheimer's disease-P group. Further, if one overlaps the graphs in Fig. 2A and B and draws a vector between baseline and follow-up values for each of the 13 subjects with two data points, the following patterns emerge: (i) eight subjects showed both increased PIB retention and decreased rCMRglc; (ii) subjects 13 and 16 showed essentially no change in either PIB retention or rCMRglc; (iii) the two subjects (3 and 7) who had the highest PIB retention and lowest rCMRglc at baseline (i.e. the two most pathologically advanced) both showed a further large decrease in rCMRglc and a decrease in PIB at follow-up; and (iv) subject 15 was anomalous—showing both decreased PIB retention and increased rCMRglc at follow-up. Thus, over 60% of the subjects showed some increase in PIB retention, but this was offset in the group average by larger decreases in PIB retention in subjects 3, 7 and 15. Although subject 15 appears to be an anomaly, subjects 3 and 7 may be examples of the dynamic nature of amyloid deposits emphasized in the discussion by Engler and co-workers. At least one prior study indicated that Aβ deposits declined late in the course of Alzheimer's disease in Down syndrome (Wegiel et al., 1999), and another suggested that the same phenomenon occurred in sporadic Alzheimer's disease (Hyman et al., 1993). Thus, it is intriguing that the subjects who had the most advanced disease at baseline (and who also showed significant cognitive worsening over the 2 year follow-up) were the only subjects who had both decreased PIB retention and rCMRglc.

Consequently, interpretation of amyloid changes over time must take into account the status of any particular subject relative to the full spectrum of amyloid changes that may occur in Alzheimer's disease. Figure 1 shows a schematic of three hypothetical phases of Aβ amyloid deposition as follows: (i) very early initiation (ei); (ii) continuously progressive (p); and (iii) late equilibrium/symptomatic (eq). The study by Engler et al. suggests that these mild-moderate AD patients are in the late equilibrium phase because most subjects show small increases in amyloid load, some are stable and the two more advanced subjects show declining amyloid loads. The continuously progressive phase (p) must exist, and PIB–PET may be well-suited to track it, but when in the clinical course of Alzheimer's disease will we find it? A second question is how long does the progressive phase last? Will it be a decade or more (e.g. p1 and t1 in Fig. 1) or just a few years (e.g. p2 and t2). Preliminary data suggest that most subjects with mild cognitive impairment (MCI) have levels of PIB retention equal to that of Alzheimer's disease patients. Therefore, these MCI subjects may already be in the equilibrium phase (Lopresti et al., 2005; Price et al., 2005). If this is true, we must look for the progressive phase among asymptomatic subjects. Mintun et al. (2006) have recently shown that asymptomatic subjects can be identified with levels of PIB retention that are intermediate between those typically observed in normal controls and clinically diagnosed Alzheimer's disease patients. It will be very important to document the natural history of amyloid deposition in these subjects.

Fig. 1

Hypothetical schematic of the progression of amyloid deposition over time from the very early initiation (ei) phase, to the continuously progressive (p) phase and finally the late equilibrium (eq) phase. Relatively long (p1/t1) and brief (p2/t2) progressive phases are shown. Symptoms are not evident until the equilibrium (eq) phase, but the cascade of pathological events that leads to these symptoms (i.e. neurofibrillary pathology and synapse loss) is initiated during the progressive phase (p).

In addition to the findings discussed above, Engler et al. addressed two important questions left unanswered by the baseline study. The first question was whether the three subjects (4, 12 and 14) that carried a clinical diagnosis of Alzheimer's disease, but had control-like levels of PIB retention were false-negatives due to insufficient sensitivity of PIB, or were misdiagnosed (and did not show amyloid deposition because they did not have Alzheimer's disease). All three of these subjects are now classified as MCI in the follow-up study and Engler et al. state, ‘Their clinical symptoms may have a different pathological basis than Alzheimer's disease.’ Nevertheless, as Engler and co-workers note, further follow-up will be required to definitively diagnose these subjects. The second unanswered question was whether the oldest control—who showed PIB retention in the ‘low Alzheimer's disease’ range at baseline—was a false-positive or had pre-symptomatic Alzheimer's disease and would eventually progress to a diagnosis of MCI or Alzheimer's disease. Engler et al. report that this subject showed no change in cognition or rCMRglc over the follow-up period, and showed either slight (temporal and parietal) or no increases in PIB retention (frontal and striatum). This could be consistent with a false-positive result if PIB retention follows a fairly rapid course (i.e. p2 and t2 in Fig. 1) or it could be consistent with a true positive if PIB retention begins long before clinical symptoms and follows a fairly lengthy course (p1 and t1 in Fig. 1). As in the case of the three atypical AD patients, more time will be needed to answer this question with certainty.

This study is a landmark description of the natural history of amyloid deposition in living subjects. Other issues still remain for future longitudinal studies. For example, it would have been useful to see longitudinal PIB and rCMRglc data on the controls. In addition, the ongoing, more extensive evaluation of the neuropsychological testing (beyond the MMSE and Rey Auditory Verbal Learning test) alluded to by Engler et al. may produce further insights since the MMSE is not a sufficiently sensitive global cognitive measure, as the authors point out. The study also reminds us that neither the clinical course nor the pathology of Alzheimer's disease progresses in a simple, linear fashion from beginning to end, but that both are highly variable. Any interpretation of changes over time must be made from the perspective of where a given patient stands in the disease process. Further, these results imply that in order for an anti-amyloid therapy to be shown to be effective at altering the natural course of amyloid deposition, it cannot simply stabilize amyloid load in symptomatic Alzheimer's disease patients, but the treatment must reduce amyloid load. The stability in group average PIB retention and the ∼5% test–retest variability means that in order to detect effects of an anti-amyloid therapy by PIB–PET in symptomatic AD patients, therapy will probably need to induce at least a 15% decrease in amyloid load. Given the apparent dynamic equilibrium of amyloid load, this is a realistic goal. From a clinical perspective, anything less than a 15% decrease in amyloid load may not be sufficient to produce meaningful clinical effects anyway. While these estimations are speculative, one thing is certain; the Engler et al. study allows us to consider power calculations for such studies that were unimagined even 5 years ago.


View Abstract