In the reported study, they used a radio-labeled hydroxybenzothiazole, termed PIB (Pittsburgh compound B), which...  Read more

In the reported study, they used a radio-labeled hydroxybenzothiazole, termed PIB (Pittsburgh compound B), which selectively binds to aggregated fibrillar Aβ deposits. PIB was intravenously injected into AD patients and healthy controls, and positron emission tomography (PET) was then used to image PIB retention from different gross anatomical regions. As a group, AD patients were observed to have greater PIB retention measured from the frontal, parietal, and temporal cortices, with no difference observed in the cerebellum. The greatest difference between the groups was observed in the frontal cortex, while within the temporal cortex, a greater difference was observed in the lateral temporal lobe compared to the medial temporal lobe. The authors further demonstrated that PIB retention correlated with regional basal metabolism, as detected with PET measures of glucose uptake.

Taken together with a prior study by Shoghi-Jadid et al., Klunk’s group has unquestionably achieved the century-old goal of visualizing amyloid plaques in living subjects. With this conquest, we can begin to assess the ultimate utility of this approach. A range of imaging techniques have been developed attempting to enable researchers to visualize different features of AD—volumetric changes measured with MRI; metabolic changes measured with PET, SPECT, and fMRI; and now, histological changes measured with PET. In general, in-vivo imaging is needed to address three outstanding clinical questions:

1. How do we improve our ability to dissociate mild forgetfulness caused by early, pre-dementia AD from mild forgetfulness caused by normal aging? This is a question of early detection.

2. How do we improve our ability to dissociate dementia caused by AD from other dementing illnesses? This is a question of diagnostic specificity.

3. What is the best way to test for drug efficacy? This question is important both for drug development as well as for following the course of approved drugs.

In this regard, the Annals paper demonstrates that the precision and integrity with which the Klunk group perform their groundbreaking science extends also to their scientific reporting. For example, they highlight the fact that, although group differences were detected, there was significant overlap between AD and controls, suggesting that, at this point, imaging plaques might not be appropriate for early detection and early diagnostics. Nevertheless, although not explicitly assessed in their study, it seems plausible that imaging amyloid plaques will aid in enhancing diagnostic specificity when presented with an atypical demented patient. Furthermore, imaging amyloid plaques should aid in testing drugs designed as "plaque busters." The authors highlight an additional utility of their approach: By being able to quantify brain β amyloidosis (a term they borrow from George Glenner’s original studies), they can begin to cross-correlate plaque load against various factors—aging, disease onset, risk factors, etc. In so doing, they can ensure that this imaging approach will likely contribute to understanding basic mechanisms of plaque formation.

View all comments by Scott Small

The pathological aggregation of the β amyloid peptide into fibrillary senile plaques (SPs) and the hyperphosphorylation of the tau protein into neurofibrillary tangles (NFTs) play a central role in the pathogenesis of Alzheimer’s disease (AD). The extent and the pattern of distribution of both lesions are indicators for the progression of AD. The initial neuropathological processes—particularly the formation of NFTs—occur in the medial temporal lobe, expanding later to the rest of the temporal lobe, the parietal lobe, and finally engulfing the whole neocortex in the late stages of disease. It is the prospect of in-vivo visualization of these neuropathological lesions that has driven the Pittsburgh group (e.g., Klunk et al., 1994), the UCLA group (e.g., Shoghi-Jadid et al., 2002), the U. Penn group (e.g.,   Read more

The pathological aggregation of the β amyloid peptide into fibrillary senile plaques (SPs) and the hyperphosphorylation of the tau protein into neurofibrillary tangles (NFTs) play a central role in the pathogenesis of Alzheimer’s disease (AD). The extent and the pattern of distribution of both lesions are indicators for the progression of AD. The initial neuropathological processes—particularly the formation of NFTs—occur in the medial temporal lobe, expanding later to the rest of the temporal lobe, the parietal lobe, and finally engulfing the whole neocortex in the late stages of disease. It is the prospect of in-vivo visualization of these neuropathological lesions that has driven the Pittsburgh group (e.g., Klunk et al., 1994), the UCLA group (e.g., Shoghi-Jadid et al., 2002), the U. Penn group (e.g., Kung et al., 2003), and other investigators to search for imaging biomarkers of these pathologies. The ideal AD imaging biomarker should be specific for the intended molecular targets (e.g., amyloid and/or tau aggregates), clear well from nonspecific binding areas (i.e., have low general lipid binding, like white matter), and yield a good signal-to-noise ratio for amyloid/tau to nonspecific sites. All this assumes that the probe binds to the aggregate site(s) in a saturable and specific manner, similar to neuroreceptor binding, although it is now apparent that amyloid and tau aggregates are complex conglomerates that contain multiple binding sites with different affinities for probes (e.g., [F-18]FDDNP binds at sites different from thioflavin probes in general). Several questions come to mind in this endeavor. First, can in-vivo imaging procedures with PET permit quantification of regional amyloid (or tau) aggregate concentrations throughout the brain? And second, can effective tracer kinetic models be established and validated to delineate transport of the labeled probe between plasma and tissue, as well as nonspecific and specific binding of the probe to amyloid and/or tau in the brain?

The early success with the use of [F-18]FDDNP (Shoghi-Jadid et al., 2002) to visualize NFTs and SPs in AD, and this work by Klunk at al. (2004) on amyloid labeling, offer an unprecedented opportunity to follow the neuropathological evolution of AD in living subjects. We should all bear in mind, however, the important challenges ahead for all amyloid/tau probes under development. The opportunity they provide is not only in early diagnosis, but also in early and repeated monitoring of both amyloid and tau anti-aggregation therapies with newly developed drugs—an active area of research and development in the pharmaceutical industry. In-vivo visualization of these brain pathologies will also help develop further understanding of how anti-aggregation drugs—like the unsuspected NSAIDs or new ones—directly interact with neurofibril aggregates (Agdeppa et al., 2003).

The article by Klunk et al. (2004) reports clinical results of the Pittsburgh Compound-B (PIB) labeled with C-11 (half-life = 20 min.) for a group of AD and control subjects. Results are encouraging; however, the authors note several methodological issues that closely relate to some of the aspects discussed above. For example, brain accumulation of PIB in AD subjects and controls is reported as SUV (standardized uptake value: tracer uptake in tissue normalized by bodyweight and injected activity) at 40-60 minutes after injection of the tracer. This approach has inherent drawbacks because it is subject to differences in fat content, bone mass, and peripheral metabolism among individuals, all of which are variable in elderly patients. Thus, apparent brain accumulation can be skewed by these factors without an independent means to verify the magnitude of these effects. Significant variability in the data can be expected because of this approach.

The authors acknowledge that they resorted to the use of SUV due to the inapplicability of the Logan graphical method (with the cerebellum as the reference region). The Logan method is only applicable when dynamic equilibrium of the tracer is reasonably achieved. The authors point out that equilibrium is not achieved at 60 minutes after injection of PIB and, in the absence of equilibrium, interpretation of SUV measures may be arbitrary. PIB equilibrium would be likely at later times, beyond 60 minutes, but the data has not been presented as of yet. An inherent limitation for longer scanning times is the short half-life of C-11, the radiolabel for PIB. Therefore, full characterization of the in-vivo kinetics of the probe remains somewhat challenging.

In light of the above issues, interpretation of the resulting data in AD patients is difficult. One of the issues is the frontal accumulation of PIB observed in some AD patients. In the discussion section, it is stated, "Antibodies to Aβ, or thioflavin S, do identify frontal cortex as a brain area very high in amyloid deposition…" as one possible explanation for the PET-PIB signal in frontal cortex in AD patients. However, the cited neuropathological studies do not confirm the importance of frontal Aβ deposition in AD. One of those references notes that, "Temporal and occipital lobes had the highest amyloid plaque densities, limbic and frontal lobes had the lowest, and parietal lobe was intermediate." (Arnold et al., 1991). Indeed, the authors recognize in the discussion section that the frontal lobe accumulation of PIB could be an artifact. The intense accumulation of PIB in white matter areas (1.5 times higher than cerebral cortical grey matter in normal subjects) indicates the high nonspecific lipid binding of PIB and would certainly be a factor to consider in measuring cortical amyloid-specific binding of PIB. This is particularly important in early stages of AD, with presumably lower amyloid concentrations, but this is not discussed in this publication. Partial volume effects (e.g., spillover of activity from one area to the other) could be very significant in this case because of cortical atrophy present in aging and AD patients. Partial volume effects are important to consider with all imaging probes when examining brain cortex in AD, particularly at later stages of disease, but can be more of an issue with relatively high concentrations of PIB in neighboring white matter. The reported net accumulation of PIB in the cerebellum, an area known not to have amyloid plaque deposition in early AD, further indicates nonspecific (or other target) binding of PIB. In parallel, it is interesting that no correlations between cortical PIB binding with MMSE scores or ApoE status had been established with AD subjects in this work.

AD pathology offers a new and unique environment for imaging with PET with its own set of unique challenges, some of which were discussed above. The authors should be commended for their efforts and congratulated for their successes. We are well aware of the difficulties and the magnitude of the task at hand. We should all be greatly encouraged by the significant progress made on amyloid imaging in humans in the last few years. This work adds to that. It reflects the great opportunities ahead for the use of molecular imaging techniques to aid in the early differential diagnosis of the various forms of dementia, and to help guide the development of therapeutic interventions by providing direct biological assessments of the brain in living patients throughout the course of disease.—Jorge R. Barrio, Professor of Molecular and Medical Pharmacology; Gary W. Small, Professor of Psychiatry; Henry Huang, Professor of Molecular and Medical Pharmacology, Professor of Biomathemathics; and Michael E. Phelps, Professor and Chairman, Molecular and Medical Pharmacology, UCLA School of Medicine.

View all comments by Jorge Barrio
View all comments by Sung Cheng Huang
View all comments by Gary Small

Dr. Scott Small eloquently puts our work into historical perspective and into perspective with current neuroimaging technologies. Implicit in his remarks, and worthy of further emphasis, is the fact that amyloid-imaging is a technique that is complementary to existing structural (MRI) and functional (FDG-PET, blood flow, fMRI) imaging techniques. No one imaging technique will serve all purposes, but will need to be used in conjunction to answer the important questions of early diagnosis and diagnostic specificity, as well as for assessing and following the effects of new drugs. For example, while it may be obvious why one would use amyloid-imaging to assess the effectiveness of an anti-amyloid therapy such as a β-secretase inhibitor, it may make little sense to use amyloid-imaging to evaluate a more general neuroprotective drug. We appreciate Dr. Small's emphasis of our point that amyloid-imaging is first a measure of β-amyloidosis. As he suggests, the job of relating β-amyloidosis to the early diagnosis and natural history of AD remains to be accomplished.

We appreciate the congratulations and encouragement of the UCLA PET group. While there are many difficulties involved in the development of amyloid-imaging radiotracers, we share the common goals of improving early diagnosis and facilitating drug development for the benefit of those who suffer from AD and those who care for them. We regard their comments as a constructive challenge to further understand the strengths and limitations of all amyloid-imaging technologies. Towards this goal, we are in the process of performing new studies at the University of Pittsburgh and elsewhere to extend the studies presented in the Annals of Neurology paper. We are specifically addressing the methodological issues raised, and are in the process of identifying and validating a simple pharmacokinetic method for routine assessment of amyloid deposition while expanding our human studies. In the end, the field in general will decide these issues as the technologies disseminate beyond their point of origin.

View all comments by William Klunk
View all comments by Chester Mathis
View all comments by Julie Price

Fair enough.

Then, to the caveats. It is no secret that the human brain may be burdened with a huge plaque load, seen by autopsy, in the absence of cognitive deficits prior to death. PIB-PET may just as well come to prove the irrelevance of amyloid burden.

In Finland, to my knowledge, there are two PET scanners, both located in Turku. Even if we sent 100,000 people with memory impairment to Turku, the two scanners would not be enough to scan them all, let alone the baby boomers who will soon start to reach the age where they start to develop dementia.

View all comments by Mikko Laakso

Members of this same group, most notably Gary Small, have also made significant contributions to the use of [18F]FDG PET imaging of the regional cerebral metabolic rate for glucose rCMRglu for early diagnosis of AD. While both techniques use PET imaging and compounds with similar abbreviations, it is important to point out that these are very different technologies at very different stages of development. The UCLA group has recently published a widely publicized study (  Read more

Members of this same group, most notably Gary Small, have also made significant contributions to the use of [18F]FDG PET imaging of the regional cerebral metabolic rate for glucose rCMRglu for early diagnosis of AD. While both techniques use PET imaging and compounds with similar abbreviations, it is important to point out that these are very different technologies at very different stages of development. The UCLA group has recently published a widely publicized study (Silverman et al. 2001), in which they suggest that the well-established [18F]FDG imaging technique can identify abnormalities in glucose metabolism two-three years prior to the onset of AD. In the present study, Shoghi-Jadid et al. use the new amyloid-binding agent [18F]FDDNP, developed by corresponding author Jorge Barrio, in an attempt to specifically label deposits of amyloid plaques and NFT in living AD patients. [18F]FDDNP is an 18F derivative of a very lipohilic, viscosity- and solvent-sensitive compound called DDNP. Shoghi-Jadid et al. cross-sectionally study patients already diagnosed with AD, with mild to moderate severity, and make no claims of pre-clinical diagnosis—although they allude to that potential.

The UCLA group first presented similar data in preliminary form at scientific meetings over two years ago. These preliminary reports generated considerable controversy in the PET imaging community due to difficulty in interpreting the data, a difficulty that has not been entirely overcome in the current publication. Problems in interpreting the data arise in a number of areas and stem largely from the fact that human amyloid imaging data were presented prior to any significant in-vitro characterization of [18F]FDDNP. Critical characteristics such as percent specific binding to synthetic Aβ and to homogenates of post-mortem AD brain have never been addressed and represent the major problem in the human study, as well.

Other important characteristics such as: 1) quantitative binding affinity (Kd) and binding stoichiometry (Bmax) for synthetic Aβ fibrils; 2) reversibility of binding; 3) specificity for staining Aβ and NFT deposits in post-mortem AD brain; 4) quantitative differentiation of [18F]FDDNP binding to homogenates of AD, control and non-AD dementia brains; 5) peripheral and brain metabolism and pharmacokinetics of [18F]FDDNP in animals; 6) quantitative screening for binding to other CNS receptor sites; 7) validation using ex vivo and microPET studies in transgenic mice which deposit Aβ in the brain; and 8) toxicity (determined by standard toxicological studies normally required for FDA approval) were not reported before human studies were presented, and most of these issues remained poorly defined.

A recent online publication in the Journal of Neuroscience (Agdeppa et al., 2001) has addressed some of these issues. I discuss these Agdeppa et al. findings here because they have strong bearing on the major weakness of the human study, i.e., demonstrated binding specificity in vitro and in vivo to Aβ. In Agdeppa, binding affinity to synthetic Aβ was studied by a fluorescent technique that relies on the capacity of FDDNP to change its fluorescence properties in different environments (being a solvent- and viscosity-sensitive fluorophore). The fluorescence binding technique cannot measure specific vs. non-specific binding (a critical issue in the human studies discussed below). This is because of differences in fluorescence in the two environments (fluorescence was shown to increase upon binding to Aβ). Agdeppa et al. do briefly mention attempts at 18F binding determinations that show a 50-fold lower affinity, but still report no data on specific binding.

Specificity of fluorescent labeling of plaques and tangles in port-mortem AD brain is also difficult to interpret from the Agdeppa study, since the fields of view are cropped to contain only a single plaque or single NFT. Furthermore, NFTs are reported to stain faintly, leading one to conclude that most binding may represent Aβ deposits in plaques and cerebrovascular amyloid. Autoradiographic studies performed in AD and control brain with [18F]FDDNP were more impressive, but are somewhat tarnished by the need to use 90 percent ethanol to differentiate the binding from background. Even so, some of the [18F]FDDNP binding still shows a poor correlation with plaques and NFT deposits, most notably demonstrated by an intense area of [18F]FDDNP binding located over a hole in the tissue. Thus, many questions regarding the basic pharmacological and pharmacokinetic characteristics of [18F]FDDNP remain, as does the difficulty in interpreting the current study by Shoghi-Jadid et al.

Several points in the Shoghi-Jadid et al. study require further clarification. This study can be considered analogous to the many neuroreceptor studies commonly accomplished with PET ligands—with binding sites located on amyloid plaques and NFT being the "receptors" of interest. Results of PET neuroreceptor studies are typically quantified in terms of distribution volume (DV), which is proportional to the density of binding sites (Bmax in units of moles/1000 mL tissue).The radioactivity associated with a good neuroreceptor ligand will homogeneously distribute throughout the brain within minutes after bolus injection, and then clear more quickly from brain areas that lack receptor sites, leaving specifically bound ligand in areas rich in receptor sites. After allowing 60-90 minutes for clearance of free and non-specifically bound ligand, the differential in radioactivity concentration between receptor-rich and receptor-poor brain areas often exceeds 3-fold for good neuroreceptor ligands. In the Shoghi-Jadid study, [18F]FDDNP initially accumulates most in pons (an area without significant numbers of plaques or tangles) and least in hippocampus (an area of severe pathology). Five to 10 min after injection, the radioactivity in the pons exceeds that in the hippocampus by 1.6-fold. Differential clearance then results in a reversal and, at best, a 1.4-fold increase in residual radioactivity in hippocampus over the pons. Rather than presentation in standard DV terms, the authors use an unconventional parameter, the "relative residence time," to present their results. This parameter is affected both by the greater initial uptake as well as the more rapid clearance of [18F]FDDNP from the pons. The reasons given for not applying standard DV techniques of analysis are not at all convincing at the molecular level. At some point, one must set the data analysis techniques aside and simply look at the pictures. Visual examination of the [18F]FDDNP images from human brain presented by Shoghi-Jadid et al. is remarkable for the lack of obvious specific localization of [18F]FDDNP to amyloid-rich brain areas and the intense localization in amyloid-poor regions such as the pons. This is the major weakness of this study.

Absence of imaging data in the cerebellum, a commonly used reference region for PET neuroreceptor studies, is a major oversight in the Shoghi-Jadi publication. This brain region is known to have very low densities of plaques and NFT's in AD subjects. Analysis of [18F]FDDNP binding kinetics in this region would be very helpful in understanding the nature of non-specific binding.

Another potential problem may arise from the very high lipophilicity of [18F]FDDNP (logPoct = 3.92), which is a full log unit above the optimum value for good brain imaging agents. A very high permeability across the blood-brain barrier (high PS value) may cause brain uptake and clearance to be dependent on regional cerebral blood flow—a characteristic of typical cerebral blood-flow imaging agents. This may be a partial explanation of the rapid accumulation and clearance in pons and occipital cortex and the relatively slow accumulation and clearance in hippocampus. Therefore, the potential blood flow dependence of the [18F]FDDNP results need to be carefully assessed.

Despite these difficulties, Shoghi-Jadid et al. manage to discern some remarkably significant differences of "relative residence times" between AD and control and remarkably significant correlations of "relative residence time" to cognitive performance, the most intriguing and strongest aspect of this study.

Given that the most significant weakness of this first human amyloid imaging agent, [18F]FDDNP, appears to be lack of demonstration of specificity, one can conclude the following. Either [18F]FDDNP is binding to a multitude of sites in addition to plaques and tangles that change in the same direction in AD vs. control, or the amyloid signal in AD brain is so intense that even a non-specific or weakly specific probe can demonstrate it. Keep in mind that the concentration of fibrillar amyloid deposits in AD brain is 100- to 1,000-fold higher than that of typical neurotransmitter receptors. Despite this, if the signal in AD brain is dominated by [18F]FDDNP binding to sites other than plaques and tangles, then any interpretation of the data remains open to question.

In spite of these problems, Shoghi-Jadid and colleagues at UCLA achieve an important milestone in being the first to attempt to image brain amyloid in human studies. Their early findings in this relatively small group of patients may or may not stand the test of time, but they certainly herald an era of a new technology for the study of amyloid deposition in living AD patients. Once fully developed by this or other groups of investigators, this technology could provide the first direct window into the characteristic neuropathology of AD in living patients - a pathology now seen only after death. Such technology could provide confirmatory diagnostic evidence in AD and perhaps even pre-clinical diagnosis, since amyloid deposition is thought to begin a decade or more before the first clinical symptoms of AD.

Perhaps the most important short-term use of this emerging imaging technology will be speeding the development of the eagerly awaited anti-amyloid therapies. By what better means can the effectiveness of these therapies at preventing or reversing amyloid deposition be determined in living subjects? How else can effective dose levels in a given patient be determined? Although this study must be regarded with the questions typical of many first attempts, it should be welcomed as breaking new ground in our fight against this terrible illness.

View all comments by William Klunk

The authors should be commended for their very careful step-by-step approach to the problem. Although considerably more work needs to be done in this area, this work is extremely promising for the development of a tool which could be used for diagnosis, monitoring the effects of treatment, and even possibly early detection of Alzheimer's disease.

View all comments by Michael Weiner

Alzheimer's disease is diagnosed definitively by the presence of amyloid and tau pathology in the context of dementia. Because plaques and tangles (the canonical AD lesions) can be detected with certainty only by histological analysis, a method of visualizing either or both of these lesions non-invasively in the living brain would be a boon to patients, physicians, and researchers. In vivo imaging would permit a longitudinal analysis of the amplification and spread of the lesions, as well as the relationship of the lesions to specific behavioral impairments; Furthermore, the ability to detect Aβ plaques early in the course of the illness could help to rule out other causes of cognitive decline, and the amyloid signal would be invaluable as a biomarker for assessing the effects of disease-modifying treatments for AD. At present, the most promising imaging method for AD is a PET method using radiolabeled Pittsburgh compound-B. The drawbacks of this method are the exposure of the patient to radiation, short half-life of the compound's radioactivity, expense, low...  Read more

Alzheimer's disease is diagnosed definitively by the presence of amyloid and tau pathology in the context of dementia. Because plaques and tangles (the canonical AD lesions) can be detected with certainty only by histological analysis, a method of visualizing either or both of these lesions non-invasively in the living brain would be a boon to patients, physicians, and researchers. In vivo imaging would permit a longitudinal analysis of the amplification and spread of the lesions, as well as the relationship of the lesions to specific behavioral impairments; Furthermore, the ability to detect Aβ plaques early in the course of the illness could help to rule out other causes of cognitive decline, and the amyloid signal would be invaluable as a biomarker for assessing the effects of disease-modifying treatments for AD. At present, the most promising imaging method for AD is a PET method using radiolabeled Pittsburgh compound-B. The drawbacks of this method are the exposure of the patient to radiation, short half-life of the compound's radioactivity, expense, low resolution, background noise and (from a research standpoint) the relatively poor binding of the compound to Aβ deposits in experimental transgenic mice.

The ideal imaging technique, then, would be safe, sensitive, specific, uncomplicated and inexpensive; comparable binding in humans and animal models also would be a plus. While we are still far from the ultimate imaging method, Higuchi et al. make important advances in several of these domains. Specifically, they have developed a non-radioactive, 19F-containing compound (FSB) that crosses the blood-brain barrier, binds selectively to amyloid deposits in transgenic mouse brain, and can be visualized by MRI both in the 19F and 1H modes. The compound is comparatively safe, selective, long-lived, and works well in a transgenic murine model of cerebral β-amyloidosis. The brain structures affected can be visualized with high resolution MRI. As the authors note, there is still need for improvement before this method becomes practical for use in humans. Relatively long imaging times in a powerful (9.4T) magnet are employed to achieve the reported results in mice. The differential sensitivity of T1-weighted 1HMRI for detecting plaques in different brain areas requires further analysis, and the sensitivity of the protocol for diffuse vs. dense deposits as well as for vascular β-amyloid in vivo is unclear. It would also be useful to know if the compound, or similar compounds, binds to pre-fibrillar Aß structures. Higuchi and colleagues note that experiments are in progress to assess the binding of the compound to tangle-like lesions in a mouse model of tauopathy. As a bridge between mice and humans, MR imaging of FSB could be profitably undertaken in larger animals such as aged nonhuman primates, which naturally develop senile plaques and cerebral amyloid angiopathy (and, in some instances, tauopathy). It is heartening to note the acceleration of progress in imaging AD pathology in vivo; it appears likely that clinicians and scientists soon may have multiple options for visualizing proteopathic lesions in the living brain.

View all comments by Lary Walker

In summary, the paper by Small et al. is welcome news for those seeking a test to improve the accuracy and early diagnosis of AD. Improvements in tracer design should allow this technology to become a useful and widely available clinical tool.

View all comments by Christopher Rowe

Since some of the MCI patients were on cholinesterase inhibitor treatment, it might be possible that due to treatment effect they are not correctly classified as AD based on the cognitive tests.

There is a follow-up of 12 controls and four MCI subjects (two converted to AD) both with PET and cognitive testing. Since only two out of 28 MCI subjects are reported to convert to AD during 24...  Read more

Since some of the MCI patients were on cholinesterase inhibitor treatment, it might be possible that due to treatment effect they are not correctly classified as AD based on the cognitive tests.

There is a follow-up of 12 controls and four MCI subjects (two converted to AD) both with PET and cognitive testing. Since only two out of 28 MCI subjects are reported to convert to AD during 24 months, the conversion rate is very low.

The differences between the PET ligand FDDNP and PIB have been widely discussed since the initial papers on FDDNP (Shoghi-Jadid et al., 2002; Klunk et al., 2004) and concerns have been raised about the sensitivity of FDDNP. This is also illustrated in the paper of Small et al. where the authors report significant differences between control/MCI/AD, but the differences between changes in control and AD/MCI 5-10 percent should be compared to the robust differences between controls/AD/MCI with PIB of 40-70 percent (Klunk et al., 2004; Engler et al., 2006; Nordberg et al., 2006 IDAC Madrid 2006).

The authors report a significant difference between FDDNP binding in MCI versus AD patients. As illustrated in Figure 2 there is a broad variation in FDDNP binding in the MCI group overlapping both the AD and control group as well (see medial temporal cortex). The authors do not discuss whether there might be two groups of MCI patients—those who show FDDNP binding comparable with MCI, and those MCI patients with FDDNP binding comparable with AD. This is a phenomenon discussed for PIB binding in MCI (Price et al., 2005; Nordberg et al., ICAD Madrid 2006). The histopathological studies performed in one autopsy case with earlier FDDNP PET investigation showed abundance of especially immunoreactive tangles in the medial temporal regions. These observations may suggest that FDDNP might be a more sensitive marker for tangles than for amyloid plaques. If this is the case, we presently may have two PET ligands, namely PIB (amyloid plaques) and FDDNP (tangles), that could be quite useful for both the understanding of the time course of disease progression as well as for evaluation of new treatment strategies. The multi-tracer PET studies, which could also include tracers for inflammation processes and neurotransmitter function, would then provide a deeper understanding of the pathophysiological disease processes that we now, 100 years after Alois Alzheimer, are starting to understand in living AD patients.

I do not think we should discuss choosing between these two PET ligands. They probably are a good complement to each other and further evaluation is needed, especially now in subjects with hereditary forms of AD as well as in those with other forms of dementia. The PET ligands should also now be applied to the different types of anti-amyloid therapy that are being evaluated presently.

References:
Shoghi-Jadid K, Small GW, Agdeppa ED, Kepe V, Ercoli LM, Siddarth P, Read S, Satyamurthy N, Petric A, Huang SC, Barrio JR. Localization of neurofibrillary tangles and beta-amyloid plaques in the brains of living patients with Alzheimer disease. Am J Geriatr Psychiatry. 2002 Jan-Feb;10(1):24-35. Abstract

Klunk WE, Engler H, Nordberg A, Wang Y, Blomqvist G, Holt DP, Bergstrom M, Savitcheva I, Huang GF, Estrada S, Ausen B, Debnath ML, Barletta J, Price JC, Sandell J, Lopresti BJ, Wall A, Koivisto P, Antoni G, Mathis CA, Langstrom B. Imaging brain amyloid in Alzheimer's disease with Pittsburgh Compound-B. Ann Neurol. 2004 Mar;55(3):306-19. Abstract

Engler H, Forsberg A, Almkvist O, Blomquist G, Larsson E, Savitcheva I, Wall A, Ringheim A, Langstrom B, Nordberg A. Two-year follow-up of amyloid deposition in patients with Alzheimer's disease. Brain. 2006 Nov;129(Pt 11):2856-66. Epub 2006 Jul 19. Abstract

Price JC, Klunk WE, Lopresti BJ, Lu X, Hoge JA, Ziolko SK, Holt DP, Meltzer CC, DeKosky ST, Mathis CA. Kinetic modeling of amyloid binding in humans using PET imaging and Pittsburgh Compound-B. J Cereb Blood Flow Metab. 2005 Nov;25(11):1528-47. Abstract

View all comments by Agneta Nordberg

However, considering the large variation in the distribution and severity of both neuropathological changes and neurochemical abnormalities among AD cases, it is unlikely that any single biomarker will fulfill the requirements of high enough sensitivity and specificity. Instead, combinations of...  Read more

However, considering the large variation in the distribution and severity of both neuropathological changes and neurochemical abnormalities among AD cases, it is unlikely that any single biomarker will fulfill the requirements of high enough sensitivity and specificity. Instead, combinations of biomarkers, each reflecting different aspects of the neuropathological and neurochemical processes, will probably prove to be useful. Indeed, a number of promising diagnostic tools have to a different extent been validated for their capacity to identify AD early in the disease process, including the cerebrospinal fluid (CSF) biomarkers total tau (T-tau), phospho-tau (P-tau) and Aβ42; magnetic resonance imaging (MRI) measurements of hippocampal and entorhinal cortex atrophy; fluoro-deoxy-glucose (FDG) positron emission tomography (PET) measurements of regional abnormalities in glucose and oxygen metabolism; and visualization of β amyloid deposition using Pittsburgh Compound-B (PIB) PET [1].

In this paper, Gary Small and coworkers introduce a new PET method using the tracer 2-(1-{6-[(2-[F-18]fluoroethyl)(methyl)amino]-2-naphthyl}ethylidene)malononitrile (FDDNP), which binds to both plaques and tangles. In AD cases, FDDNP binds to cortical brain regions known to be affected by plaques and tangles.

In the study, data on FDDNP binding were presented as relative distribution volumes (DVR), which is the tracer distribution in the region of interest (ROI) divided by the distribution volume in the reference region (the cerebellum). In AD and MCI cases, approximately 10 percent higher DVRs were found in several cortical brain regions. Since the standard deviation (SD) within each diagnostic group was notably small, the estimated effect sizes (i.e., the difference between the group means divided by the pooled SD) were high, between 2.5 and 4.5.

The diagnostic accuracy of the method, evaluated by receiver operating characteristics (ROC) curves, was very high; 0.95 and 0.98 to differentiate MCI and AD from controls. Interestingly, ROC values for FDDNP were clearly higher than for either FDG-PET or MRI measurements of hippocampal volume.

Thus, the new FDDNP-PET shows great promise as a diagnostic tool for MCI and AD. FDDNP-PET may also be useful as a surrogate marker in clinical trials on the new type of drug candidates with disease-modifying potential, both in trials on drugs targeting Aβ deposition and tangle formation. Indeed, biomarker data from small clinical trials suggesting that a drug has positive effects on these pathogenic processes would be of great value to make a go/no-go decision for an expensive clinical trial with clinical improvement as the endpoint.

The research community will be waiting for future studies on the clinical usefulness of FDDNP-PET imaging. First of all, we need replication studies on prospective patients with lower dropout rates than in the present study, and also studies including cases with other dementia disorders such as frontotemporal dementia, Lewy body dementia, and Creutzfeldt-Jakob disease. It will also be interesting to see a direct comparison of the regional binding of the two ligands for plaques (PIB) and plaques and tangles (FDDNP). Further, since a recent study showed that all subjects, regardless of clinical status, with positive PIB binding had low CSF Aβ42 levels, and vice versa [2], it will also be interesting to learn how FDDNP binding correlates not only to CSF Aβ42 levels, but also to CSF T-tau and P-tau levels.

A last linguistic comment on the paper is that since the term “invasive” is defined as “puncture of the skin with entry of foreign material into the body as part of a diagnostic technique,” and FDDNP-PET is based on the injection of a radioactive substance into the body through an indwelling venous catheter, it may be considered unfortunate to claim FDDNP-PET as “a noninvasive method,” as in the conclusion of the Abstract. Nevertheless, FDDNP-PET is a promising new clinical tool for visualization of plaques and tangles directly in living patients, and is likely to provide useful diagnostic information in early AD.

References:
1. Blennow K, de Leon MJ, Zetterberg H. Alzheimer's disease. Lancet. 2006 Jul 29;368(9533):387-403. Review. Abstract

2. Fagan AM, Mintun MA, Mach RH, Lee SY, Dence CS, Shah AR, LaRossa GN, Spinner ML, Klunk WE, Mathis CA, DeKosky ST, Morris JC, Holtzman DM. Inverse relation between in vivo amyloid imaging load and cerebrospinal fluid Abeta42 in humans. Ann Neurol. 2006 Mar;59(3):512-9. Abstract

View all comments by Kaj Blennow