**5. Neuroimaging (Table 4)**

#### **5.1. Structural imaging**

Neuroimaging is now one of the most common tools used to aid the diagnosis of AD. It is a huge and burgeoning field and only select modalities and important studies on longitudinal imaging are discussed here.

Predicting Cognitive Decline in Alzheimer's Disease (AD): The Role of Clinical, Cognitive Characteristics and Biomarkers http://dx.doi.org/10.5772/54289 389

**4.3. Multiple marker arrays**

388 Understanding Alzheimer's Disease

**4.4. Summary**

and/or inconclusive results.

peripheral nature and ease of blood sampling.

**5. Neuroimaging (Table 4)**

**5.1. Structural imaging**

imaging are discussed here.

Given the disappointing results achieved by single markers despite tremendous efforts, the field has now moved towards multiple markers that are obtained through high throughput technologies, sophisticated statistical analysis and bioinformatics. Ray [94] published a blood plasma-based proteomic screening tool to identify patients with AD and also to identify those likely to progress from MCI to AD. Biological analysis of the 18 proteins points to systemic dysregulation of hematopoiesis, immune responses, apoptosis and neuronal support. How‐

Based on current literature, no single marker has been found to be significant in all the multiple marker arrays. Moreover one can expect that utilizing high throughput array technology, more multiple marker arrays will appear and dominate the blood biomarker landscape. To sound a note of caution, however, some panels may be derived from 'over-fitting' the dataset and may not survive generalization and independent validation. To date, multiple marker arrays have not been employed to study the conversion of MCI to AD and to differentiate between

Plasma Aβ is an appealing biomarker since many AD interventions under investigation are directed against Aβ. Thus an Aβ-based biomarker is attractive for those who will benefit from such treatments. However, many studies involving various blood biomarkers have conflicting

APOEε4 influence may be more stage-dependent, with its effect on disease trajectory most evident in the earliest stages of disease and less so in moderate to severe stages. Hence it should be included as a covariate in various clinical progression and therapeutic trials. A major challenge is that the literature thus far has focused on the use of blood biomarkers for diagnosis (requiring the identification of dichotomous - disease versus normal- states), which may not be applicable to the use of such biomarkers for tracking disease progression (for which an effective biomarker must show continuous change rather than merely being present or absent). Nevertheless blood biomarkers should be employed in combination with clinical assessment and neuroimaging to improve diagnostic and prognostic accuracy, especially given the

Neuroimaging is now one of the most common tools used to aid the diagnosis of AD. It is a huge and burgeoning field and only select modalities and important studies on longitudinal

ever efforts at independent validation of Ray's findings have been discouraging [95].

fast and slow progressors. This would be a logical next step for investigation.


Small et al. [108] 22 HC and 21 Increases in [18F]FDDNP PET scanning may be useful in

memory decline yielded highest diagnostic accuracy. ROC

Doraiswamy et al. [109] 51 MCI, 69 HC, MCI Aβ+ and Florbetapir PET, which detects Aβ pathology, may

followed up over associated with for progression to AD.

MCI Aβ+ associated With greater decline in memory, DSS and MMSE (p < 0.05). Ossenkoppele et al. [110] 11 HC, 12 MCI, Global cortical [11C]PIB and [18F] FDG track molecular changes and 8 AD [11C]PIB BPND in different stages of AD.

MCI followed up frontal, posterior identifying people at risk for future cognitive over 2 years cingulate, and decline. Higher [18F]FDDNP binding at

correlated with P = 0.05 to 0.002).

and 31 AD HC Aβ+ be helpful in identifying individuals at increased risk

18 months. greater clinical Higher SUVr in MCI associated with greater decline

ADAS-Cog and MMSE (all p<0.05).

followed up over is significantly MCI subjects were found to have an increased 2.5 years. increased in amyloid load while AD subjects had increased MCI subjects, progressive metabolic impairment. but no changes [18F]FDDNP is less useful for examining disease

13 research pooled estimates: techniques for prediction of AD progression in MCI

14 HC 3 years sequence (temporal- frontal- sensorimotor)

subjects and 18 3 years than slow decliners, especially in the medial HC followed up occipitoparietal areas (specifically precuneus,

presenting dementia in regional brain metabolism was a sensitive indicator symptoms of 59% of AD. A negative PET scan indicated that dementia pathologic progression of cognitive impairment

as the disease progressed. Mirroring the sequence of neurofibrillary tangle accumulation observed in cross sections

single negative PET scan.

Left hemisphere degenerates faster (5.3 ± 2.3% per year in AD v.s. 0.9 ± 0.9% per year in controls; p<0.029) than the right

Lingual gyrus and cuneus which was not yet detected by clinical and neuropsychological

during the mean 3-year follow-up was unlikely to occur. Sn 93%, Sp 76%. –LR 0.1 (95% CI 0.06- 0.16) experiencing progressive course after a

was observed in progression.

global [18F] p=0.35).

FDDNP.

studies (7 FDG- 78.7% Sn (95% subjects.

Thompson et al. [100] 12 AD subjects, Followed up Cortical atrophy occurred in a well defined

at autopsy.

Kinkingnéhun et al.[103] 23 mild AD Followed up Fast decliners had a more extensive cortical atrophy

assessment.

Silverman et al. [107] 284 patients Progressive In patients presenting symptoms of dementia,

74% Sp (95%CI 67.0-80.3%) PIB-PET pooled estimates: 93.5% Sn (95% CI 71.3-99.9%) 56.2% Sp (95% CI 47.2-64.8%)

Zhang et al. [111] Meta-analysis of FDG-PET Both FDG-PET and PIB-PET are valuable

PET) CI 68.7-86.6%)

*Predicting rapid AD progression/ decline* 

Functional Imaging

global binding at baseline is associated with future decline in follow-up most cognitive domains (r = -0.31 to -0.56,

progression of Frontal and parietal [18F]FDDNP binding

(r = -0.32 to -0.37, 0.88 (95% CI 0.72-1.00) compared with 0.68 P = 0.03 to 0.01). (95% CI 0.45-0.91) for medial temporal binding.

worsening on on ADAS-Cog, CDR-SB, memory measure (DSS)

and CDR-SB. MCI Aβ+ had higher risk of developing AD.

AD subjects or Reduction in [18F]FDG uptake at follow-up HC. Increase observed in AD subjects only (esp frontal, parietal, most prominent temporal lobes (all p<0.01). Changes in global in lateral [11C]PIB binding (p=-0.42, p<0.05) and cingulate temporal lobe [18F]FDG uptake (p=0.43, p<0.01) correlated (p < 0.05). with changes in MMSE score over time across No changes in groups but not for [18F] FDDNP binding (p=-0.18,

Greater hippocampal atrophy in MCI converters at baseline.

months.

Molecular Imaging

**Study variable Population Results Key findings** 

Jack et al. [96] 55 NC, 41 MCI, Atrophy rates of Rates of change from serial MRI studies together

years follow-up entorhinal progression in AD.

ventricle) Jack et al. [97] 133 MCI subjects 52 subjects MRI brain atrophy rate measures can be used as

carriers. Jack et al. [98] 72 aMCI 13 HC Larger ventricular APC (HR for a 1-SD increase

 MCI-converters. Apostolova et al. [99] 20 MCI subjects 6 subjects Smaller hippocampi and specifically CA1 and

Risascher et al. [101] 339 MCI (277 62 MCI Degree of neurodegeneration of MTL structures is

Querbes et al. [103] 122 aMCI (50 72 aMCI Normalised cortical thickness can predict AD

Lo et al. [105] 229 normal Rates of change Amyloid deposition is an early event before

Okello et al. [106] 31 aMCI 17 out of 31 MCI PIB-positive MCI subjects are more likely to

developed AD. Half (47%) converted to AD

Koivunen et al. [107] 29 MCI, 13 HC 17 MCI Hippocampal atrophy increases and amyloid

followed up over increased

(82%)

within 1 year.

progressive MCI), 130 HC, 130 AD followed up over 24

64 AD four structures with standard clinical/psychometric measures can subjects; 1-5 (hippocampus, be used as surrogate markers of disease

(45 were APOEε4 therapeutic MCI setting.

subjects, 91 HC; developed MCI 1.4, p=0.007) increased risk of AD conversion. 1-2 years follow- or AD; Both ventricular APC (HR for a 1-SD increase 1.59, up. 39 MCI subjects p=0.001) and whole brain APC (HR for 1-SD increase developed AD 1.32, p=0.009) provided additional predictive

followed up over developed AD subicular subfields are associated with increased 3 years (MCI-c), 7 risk for conversion from MCI to AD.

MCI-stable, 62 developed AD the best antecedent MRI marker of imminent MCI-converters) conversion, with decreased hippocampal volume subjects, 206 (left > right) being the most robust structural MRI HC, 148 AD feature. Effect sizes of hippocampus (0.6) and MTL subjects structures (0.53) comparing MCI-stable and converters.

stable MCI, 72 developed AD. conversion with 76% cross-validated accuracy.

,397 MCI in CSF Aβ42 hypometabolism or hippocampal atrophy, and 193 AD , glucose meta- suggesting that biomarker prediction for cognitive

subjects, 26 HC (55%) had develop AD than PIB-negative subjects.

followed up hippocampal Positive APOE4 status accelerated hippocampal

3 years [11C]PIB Fast converters have higher PIB retention levels at

followed up over developed AD deposition changes modestly during conversion to

PIB-positive MCI with faster conversion rates (p=0.035)

retention at baseline than slower converters in anterior cingulate, baseline (PIB- (p=0.027) and frontal cortex (p=0.031). positive). Only 1 out of 14 PIB-negative subjects develop AD. 14 of these 17 7 of 17 PIB-positive MCI, APOEε4 carriers associated

during evolution of MCI.

single negative PET scan.

AD converters had greater [11C]PIB retention at baseline in posterior cingulate (p=0.022), putamen (p=0.041), caudate nucleus (p=0.025). Greater hippocampal atrophy in MCI converters at

> HC = Healthy Controls SD = Standard deviation MTL = Medial Temporal Lobe

aMCI = amnestic MCI

Sn = Sensitivity Sp= Specific

PIB = Pittsburgh Compound B


in cross-study comparisons.

PET = Positron Emission Tomography MMSE =Mini Mental State Examination

FDDNP = Fluoroethyl)methylamino]-2-napthyl}ethylidene) malononitrile

**Table 4.** Neuroimaging methods in predicting AD conversion in MCI patients and rapid AD progression/ decline

With technological advances over the past three decades, MRI is now readily available and relatively economical. Currently it is widely used as a diagnostic tool, to complement clinical assessment and neuropsychological testing. Moreover, MRI has also been considered for longitudinal tracking of the disease progression and to predict whether a MCI patient may go on to develop AD, or whether an AD patient will have an indolent or rapid course. Advances in technology have led to automated data-driven methods, such as automated measurement of whole brain volume over time, voxel-based morphometry (VBM), deformation-based morphometry (DBM) and analysis of cortical thickness. These technologies ameliorate the previous problems associated with manual measurement, inter-rater reliability and difficulties

Predicting Cognitive Decline in Alzheimer's Disease (AD): The Role of Clinical, Cognitive Characteristics and Biomarkers

http://dx.doi.org/10.5772/54289

391

In a seminal paper, Jack [96] studied annualized changes in volume of four structures in serial MRI studies: hippocampus, entorhinal cortex, whole brain and ventricles of normal, MCI and AD subjects. All four atrophy rates were greater among MCI-converters compared to nonconverters and fast-progressors versus slow progressors. Although the differences in atrophy rates have been replicated consistently in several follow-up studies [97,98], given the overlap among those who did and did not convert, the authors cautioned that these measures were

Using hippocampal volumetry, a prospective longitudinal cohort study found that greater atrophy in the CA1 hippocampal and subicular subfields predicted MCI conversion, whereas

Employing a 3-dimensional cortical mapping approach, Thompson [100], demonstrated a temporal-frontal-sensorimotor sequence of cortical atrophy with AD progression in a longi‐ tudinal series of 12 AD subjects, where left brain was found to degenerate faster than right.

Employing VBM technique, Risacher [101] found that AD and MCI converters demonstrated high atrophy across regions as compared to HC in global and hippocampal grey matter (GM)

larger hippocampal volumes predicted cognitive stability and/or improvement [99].

unlikely to provide absolute prognostic information for individual patients.

subjects bolism and change is stage dependent.

3 years volume atrophy changes in MCI and AD.

remained stable Larger hippocampal volumes and relative (MCI-nc), and 7 preservation of both the subiculum and CA1 are improved (MCI- i). associated with cognitive stability or improvement.

Mean time APCs greater in APOE ε4 non-carriers.

cortex, whole Atrophy rates greater among MCI converters. brain, and Atrophy rates greater among AD fast progressors

developed AD indicators of disease progression in a multi-site

carriers). APC was greater in converters than non-converters.

information to covariate-adjusted sectional HC volume at baseline about risk of AD conversion. However, overlap present among those converters and non-converters indicate that these measures are unlikely to provide absolute prognosis for

to conversion APCs and changes in cognitive test performance 556 day in APOE uniformly correlated in expected direction (p<0.000)

*Predicting AD conversion in MCI subjects*

Structural Imaging


NC = Normal Controls

MRI = Magnetic Resonance Imaging

APC = Annual percent change


**Study variable Population Results Key findings** 

Jack et al. [96] 55 NC, 41 MCI, Atrophy rates of Rates of change from serial MRI studies together

years follow-up entorhinal progression in AD.

ventricle) Jack et al. [97] 133 MCI subjects 52 subjects MRI brain atrophy rate measures can be used as

carriers. Jack et al. [98] 72 aMCI 13 HC Larger ventricular APC (HR for a 1-SD increase

 MCI-converters. Apostolova et al. [99] 20 MCI subjects 6 subjects Smaller hippocampi and specifically CA1 and

Risascher et al. [101] 339 MCI (277 62 MCI Degree of neurodegeneration of MTL structures is

Querbes et al. [103] 122 aMCI (50 72 aMCI Normalised cortical thickness can predict AD

Lo et al. [105] 229 normal Rates of change Amyloid deposition is an early event before

Okello et al. [106] 31 aMCI 17 out of 31 MCI PIB-positive MCI subjects are more likely to

developed AD. Half (47%) converted to AD

Koivunen et al. [107] 29 MCI, 13 HC 17 MCI Hippocampal atrophy increases and amyloid

Small et al. [108] 22 HC and 21 Increases in [18F]FDDNP PET scanning may be useful in

memory decline yielded highest diagnostic accuracy. ROC

Doraiswamy et al. [109] 51 MCI, 69 HC, MCI Aβ+ and Florbetapir PET, which detects Aβ pathology, may

followed up over associated with for progression to AD.

MCI Aβ+ associated With greater decline in memory, DSS and MMSE (p < 0.05). Ossenkoppele et al. [110] 11 HC, 12 MCI, Global cortical [11C]PIB and [18F] FDG track molecular changes and 8 AD [11C]PIB BPND in different stages of AD.

followed up over increased

(82%)

within 1 year.

progressive MCI), 130 HC, 130 AD followed up over 24

months.

Molecular Imaging

390 Understanding Alzheimer's Disease

64 AD four structures with standard clinical/psychometric measures can subjects; 1-5 (hippocampus, be used as surrogate markers of disease

(45 were APOEε4 therapeutic MCI setting.

subjects, 91 HC; developed MCI 1.4, p=0.007) increased risk of AD conversion. 1-2 years follow- or AD; Both ventricular APC (HR for a 1-SD increase 1.59, up. 39 MCI subjects p=0.001) and whole brain APC (HR for 1-SD increase developed AD 1.32, p=0.009) provided additional predictive

followed up over developed AD subicular subfields are associated with increased 3 years (MCI-c), 7 risk for conversion from MCI to AD.

MCI-stable, 62 developed AD the best antecedent MRI marker of imminent MCI-converters) conversion, with decreased hippocampal volume subjects, 206 (left > right) being the most robust structural MRI HC, 148 AD feature. Effect sizes of hippocampus (0.6) and MTL subjects structures (0.53) comparing MCI-stable and converters.

stable MCI, 72 developed AD. conversion with 76% cross-validated accuracy.

,397 MCI in CSF Aβ42 hypometabolism or hippocampal atrophy, and 193 AD , glucose meta- suggesting that biomarker prediction for cognitive

subjects, 26 HC (55%) had develop AD than PIB-negative subjects.

followed up hippocampal Positive APOE4 status accelerated hippocampal

3 years [11C]PIB Fast converters have higher PIB retention levels at

followed up over developed AD deposition changes modestly during conversion to 2 years AD, suggesting dissociation between the two

baseline.

MCI followed up frontal, posterior identifying people at risk for future cognitive over 2 years cingulate, and decline. Higher [18F]FDDNP binding at

correlated with P = 0.05 to 0.002).

and 31 AD HC Aβ+ be helpful in identifying individuals at increased risk

18 months. greater clinical Higher SUVr in MCI associated with greater decline

ADAS-Cog and MMSE (all p<0.05).

followed up over is significantly MCI subjects were found to have an increased 2.5 years. increased in amyloid load while AD subjects had increased MCI subjects, progressive metabolic impairment. but no changes [18F]FDDNP is less useful for examining disease

13 research pooled estimates: techniques for prediction of AD progression in MCI

14 HC 3 years sequence (temporal- frontal- sensorimotor)

subjects and 18 3 years than slow decliners, especially in the medial HC followed up occipitoparietal areas (specifically precuneus,

presenting dementia in regional brain metabolism was a sensitive indicator symptoms of 59% of AD. A negative PET scan indicated that dementia pathologic progression of cognitive impairment

as the disease progressed. Mirroring the sequence of neurofibrillary tangle accumulation observed in cross sections

single negative PET scan.

Left hemisphere degenerates faster (5.3 ± 2.3% per year in AD v.s. 0.9 ± 0.9% per year in controls; p<0.029) than the right

Lingual gyrus and cuneus which was not yet detected by clinical and neuropsychological

during the mean 3-year follow-up was unlikely to occur. Sn 93%, Sp 76%. –LR 0.1 (95% CI 0.06- 0.16) experiencing progressive course after a

was observed in progression.

global [18F] p=0.35).

FDDNP.

studies (7 FDG- 78.7% Sn (95% subjects.

Thompson et al. [100] 12 AD subjects, Followed up Cortical atrophy occurred in a well defined

at autopsy.

Kinkingnéhun et al.[103] 23 mild AD Followed up Fast decliners had a more extensive cortical atrophy

assessment.

Silverman et al. [107] 284 patients Progressive In patients presenting symptoms of dementia,

74% Sp (95%CI 67.0-80.3%) PIB-PET pooled estimates: 93.5% Sn (95% CI 71.3-99.9%) 56.2% Sp (95% CI 47.2-64.8%)

Zhang et al. [111] Meta-analysis of FDG-PET Both FDG-PET and PIB-PET are valuable

PET) CI 68.7-86.6%)

*Predicting rapid AD progression/ decline* 

Functional Imaging

NC = Normal Controls

MRI = Magnetic Resonance Imaging APC = Annual percent change

PIB-positive MCI with faster conversion rates (p=0.035)

retention at baseline than slower converters in anterior cingulate, baseline (PIB- (p=0.027) and frontal cortex (p=0.031). positive). Only 1 out of 14 PIB-negative subjects develop AD. 14 of these 17 7 of 17 PIB-positive MCI, APOEε4 carriers associated

during evolution of MCI.

global binding at baseline is associated with future decline in follow-up most cognitive domains (r = -0.31 to -0.56,

progression of Frontal and parietal [18F]FDDNP binding

(r = -0.32 to -0.37, 0.88 (95% CI 0.72-1.00) compared with 0.68 P = 0.03 to 0.01). (95% CI 0.45-0.91) for medial temporal binding.

worsening on on ADAS-Cog, CDR-SB, memory measure (DSS)

and CDR-SB. MCI Aβ+ had higher risk of developing AD.

AD subjects or Reduction in [18F]FDG uptake at follow-up HC. Increase observed in AD subjects only (esp frontal, parietal, most prominent temporal lobes (all p<0.01). Changes in global in lateral [11C]PIB binding (p=-0.42, p<0.05) and cingulate temporal lobe [18F]FDG uptake (p=0.43, p<0.01) correlated (p < 0.05). with changes in MMSE score over time across No changes in groups but not for [18F] FDDNP binding (p=-0.18,

AD converters had greater [11C]PIB retention at baseline in posterior cingulate (p=0.022), putamen (p=0.041), caudate nucleus (p=0.025). Greater hippocampal atrophy in MCI converters at

subjects bolism and change is stage dependent.

3 years volume atrophy changes in MCI and AD.

remained stable Larger hippocampal volumes and relative (MCI-nc), and 7 preservation of both the subiculum and CA1 are improved (MCI- i). associated with cognitive stability or improvement.

Mean time APCs greater in APOE ε4 non-carriers.

cortex, whole Atrophy rates greater among MCI converters. brain, and Atrophy rates greater among AD fast progressors

developed AD indicators of disease progression in a multi-site

carriers). APC was greater in converters than non-converters.

information to covariate-adjusted sectional HC volume at baseline about risk of AD conversion. However, overlap present among those converters and non-converters indicate that these measures are unlikely to provide absolute prognosis for

to conversion APCs and changes in cognitive test performance 556 day in APOE uniformly correlated in expected direction (p<0.000)

*Predicting AD conversion in MCI subjects*

Structural Imaging

**Table 4.** Neuroimaging methods in predicting AD conversion in MCI patients and rapid AD progression/ decline

With technological advances over the past three decades, MRI is now readily available and relatively economical. Currently it is widely used as a diagnostic tool, to complement clinical assessment and neuropsychological testing. Moreover, MRI has also been considered for longitudinal tracking of the disease progression and to predict whether a MCI patient may go on to develop AD, or whether an AD patient will have an indolent or rapid course. Advances in technology have led to automated data-driven methods, such as automated measurement of whole brain volume over time, voxel-based morphometry (VBM), deformation-based morphometry (DBM) and analysis of cortical thickness. These technologies ameliorate the previous problems associated with manual measurement, inter-rater reliability and difficulties in cross-study comparisons.

In a seminal paper, Jack [96] studied annualized changes in volume of four structures in serial MRI studies: hippocampus, entorhinal cortex, whole brain and ventricles of normal, MCI and AD subjects. All four atrophy rates were greater among MCI-converters compared to nonconverters and fast-progressors versus slow progressors. Although the differences in atrophy rates have been replicated consistently in several follow-up studies [97,98], given the overlap among those who did and did not convert, the authors cautioned that these measures were unlikely to provide absolute prognostic information for individual patients.

Using hippocampal volumetry, a prospective longitudinal cohort study found that greater atrophy in the CA1 hippocampal and subicular subfields predicted MCI conversion, whereas larger hippocampal volumes predicted cognitive stability and/or improvement [99].

Employing a 3-dimensional cortical mapping approach, Thompson [100], demonstrated a temporal-frontal-sensorimotor sequence of cortical atrophy with AD progression in a longi‐ tudinal series of 12 AD subjects, where left brain was found to degenerate faster than right.

Employing VBM technique, Risacher [101] found that AD and MCI converters demonstrated high atrophy across regions as compared to HC in global and hippocampal grey matter (GM) density, hippocampal and amygdalar volumes, and cortical thickness values from entorhinal cortex and other temporal and parietal lobe regions. MCI-stable showed intermediate atrophy. Degree of atrophy of medial temporal structures, especially the hippocampi, was found to be the best antecedent MRI marker of imminent conversion.

clinical worsening on the AD Assessment Scale-Cognitive subscale (ADAS-Cog) and Clinical Dementia Rating-sum of boxes (CDR-SB). In MCI, Aβ + scans were also associated with greater decline in memory, Digit Symbol Substitution (DSS) and MMSE. Aβ + MCI subjects again

Predicting Cognitive Decline in Alzheimer's Disease (AD): The Role of Clinical, Cognitive Characteristics and Biomarkers

http://dx.doi.org/10.5772/54289

393

In a seminal comparison study of three modalities [110], using [(11)C]PIB, [(18)F]FDDNP and [18F]FDG, there was a significant increase in global cortical [(11)C]PIB binding (most promi‐ nent in the lateral temporal lobe) in MCI patients, but no changes in AD patients or controls. Interestingly, [(18)F]FDDNP did not show any changes in global binding potential. Moreover, changes in global [(11)C]PIB binding and posterior cingulate [(18)F]FDG uptake were corre‐ lated with changes in MMSE score over time across groups, but not with [(18)F]FDDNP binding. Hence it was postulated that [(11)C]PIB and [(18)F]FDDNP track molecular changes in different stages of AD. There was an increased amyloid load in MCI patients and progressive metabolic impairment in AD patients. The authors opined that [(18)F]FDDNP was less useful

To estimate the diagnostic accuracy of FDG-PET and PIB-PET for prediction of short-term conversion to AD in patients with MCI, Zhang [111] and co-workers performed a meta-analysis undertaken with a random-effects model. Overall diagnostic accuracy determined for both FDG-PET and PIB-PET suggests that they are potentially valuable techniques for prediction of progression in patients with MCI. Both have their advantages and their combined use is a

Villain et al recently published a longitudinal PIB study (testing conducted 18 months apart), showing a significant increase in amyloid-β accumulation in both PIB-positive and negative subjects (significantly higher in PIB-positive individuals) with a bimodal distribution of

MRI volumetry and brain atrophy rates have fairly good diagnostic and predictive value in MCI subjects. Longitudinal data on brain atrophy rates with disease progression are available and hence, can be used for monitoring disease progression in clinical trials. The limitations of structural neuroimaging as a biomarker include problems with the accurate delineation of regions of interest and lack of standardization of imaging and measurement techniques, making it difficult to compare data across the different institutions out of Europe, North America and Australia (all of which have their unified imaging consortiums). The advent of automated data-driven innovations for structural imaging holds promise. FDG-PET appears to be the leading candidate among the functional neuroimaging modalities, with available evidence for MCI diagnosis, prediction of MCI-converters and longitudinal data in monitoring serial progression. To date, [(11]C] PIB is the most extensively studied PET amyloid tracer, although 18F florbetapir proves to be an attractive alternative given the longer half-life. There is emerging evidence for amyloid imaging in the diagnosis of preclinical AD. From the standpoint of clinical trials of anti-amyloid therapy, in-vivo amyloid imaging pre-treatment allows selection of patients with demonstrable cerebral Aβ loads; repeated imaging during ongoing treatment allows detection of decrease in insoluble Aβ load in response to amyloid-

tended to convert to AD at a higher rate than Aβ- subjects [109].

individual rates of neocortical amyloid- β accumulation [112].

for examining disease progression.

promising option.

**5.3. Summary**

A separate study also showed that occipitoparietal (specifically precuneus, lingual gyrus and cuneus) atrophy at baseline better anticipated the rate of progression (fast decliners from slow decliners) over 3 years compared to clinical and neuropsychological assessment [102].

Cortical thickness is another measure of interest in structural neuroimaging where a normal‐ ized thickness index was computed using a subset of these regions, namely the right medial temporal, left lateral temporal and right posterior cingulate. Normalized thickness index at baseline differed significantly among all the four diagnosis groups (HC, stable MCI, progres‐ sive MCI and AD). Furthermore, normalized thickness index also correctly predicted evolution to AD for 76% of aMCI subjects after cross-validation [103].

### **5.2. Functional and molecular imaging**

There are many functional imaging studies for AD though only a few specifically investigate longitudinal progression of MCI and AD using Fluorodeoxyglucose (18F) (FDG)-Positron Emission Tomography (PET) [104].

Lo [105] found that the rate of change of glucose metabolism and hippocampal volume accelerated as cognitive function deteriorated. Moreover, glucose metabolic decline and hippocampal atrophy were significantly slower in subjects with normal cognition compared to those with MCI or AD. Positive APOE4 status was also associated with accelerated hippo‐ campal atrophy.

Molecular imaging utilizes small molecule ligands that bind with nanomolar affinity to amyloid and enters the brain for imaging with PET. It is a measure to detect and quantify cerebral beta-amyloidosis. It should be noted that besides AD, there are other disease condi‐ tions that may have cerebral Aβ.The most commonly used ligand is the carbon-11(11C)-based Pittsburgh compound B (PIB), which binds specifically to fibrillar Aβ but exhibits no demon‐ strable binding to neurofibrillary tangles. However, fluorine-18 (18F)-based tracers, e.g. 2-(1- {6-[(2-fluorine 18-labeled fluoroethyl)methylamino]-2-napthyl}ethylidene) malononitrile ([(18)F]FDDNP) have a considerably longer half-life compared to [11(C)]PIB and some types have been shown to also bind to neurofibrillary tangles.

Okello [106] showed that PIB-positive subjects with MCI are significantly more likely to convert to AD than PIB-negative ones. A separate longitudinal study showed that hippocam‐ pal atrophy and amyloid deposition (in posterior cingulate, lateral frontal cortex, temporal cortex, putamen and caudate nucleus) seem to dissociate during the evolution of MCI, the atrophy increasing clearly and [(11)C] PIB retention changing modestly when conversion to AD occurs [107]. Using [(18)F]FDDNP PET, higher baseline binding was associated with future decline in most cognitive domains. Specifically, frontal and parietal [(18)F]FDDNP binding yielded the greatest diagnostic accuracy in identifying MCI-converters versus non-converters [108]. With 18F florbetapir (18F-AV-45) tracer, baseline Aβ + scans were associated with greater clinical worsening on the AD Assessment Scale-Cognitive subscale (ADAS-Cog) and Clinical Dementia Rating-sum of boxes (CDR-SB). In MCI, Aβ + scans were also associated with greater decline in memory, Digit Symbol Substitution (DSS) and MMSE. Aβ + MCI subjects again tended to convert to AD at a higher rate than Aβ- subjects [109].

In a seminal comparison study of three modalities [110], using [(11)C]PIB, [(18)F]FDDNP and [18F]FDG, there was a significant increase in global cortical [(11)C]PIB binding (most promi‐ nent in the lateral temporal lobe) in MCI patients, but no changes in AD patients or controls. Interestingly, [(18)F]FDDNP did not show any changes in global binding potential. Moreover, changes in global [(11)C]PIB binding and posterior cingulate [(18)F]FDG uptake were corre‐ lated with changes in MMSE score over time across groups, but not with [(18)F]FDDNP binding. Hence it was postulated that [(11)C]PIB and [(18)F]FDDNP track molecular changes in different stages of AD. There was an increased amyloid load in MCI patients and progressive metabolic impairment in AD patients. The authors opined that [(18)F]FDDNP was less useful for examining disease progression.

To estimate the diagnostic accuracy of FDG-PET and PIB-PET for prediction of short-term conversion to AD in patients with MCI, Zhang [111] and co-workers performed a meta-analysis undertaken with a random-effects model. Overall diagnostic accuracy determined for both FDG-PET and PIB-PET suggests that they are potentially valuable techniques for prediction of progression in patients with MCI. Both have their advantages and their combined use is a promising option.

Villain et al recently published a longitudinal PIB study (testing conducted 18 months apart), showing a significant increase in amyloid-β accumulation in both PIB-positive and negative subjects (significantly higher in PIB-positive individuals) with a bimodal distribution of individual rates of neocortical amyloid- β accumulation [112].

### **5.3. Summary**

density, hippocampal and amygdalar volumes, and cortical thickness values from entorhinal cortex and other temporal and parietal lobe regions. MCI-stable showed intermediate atrophy. Degree of atrophy of medial temporal structures, especially the hippocampi, was found to be

A separate study also showed that occipitoparietal (specifically precuneus, lingual gyrus and cuneus) atrophy at baseline better anticipated the rate of progression (fast decliners from slow decliners) over 3 years compared to clinical and neuropsychological assessment [102].

Cortical thickness is another measure of interest in structural neuroimaging where a normal‐ ized thickness index was computed using a subset of these regions, namely the right medial temporal, left lateral temporal and right posterior cingulate. Normalized thickness index at baseline differed significantly among all the four diagnosis groups (HC, stable MCI, progres‐ sive MCI and AD). Furthermore, normalized thickness index also correctly predicted evolution

There are many functional imaging studies for AD though only a few specifically investigate longitudinal progression of MCI and AD using Fluorodeoxyglucose (18F) (FDG)-Positron

Lo [105] found that the rate of change of glucose metabolism and hippocampal volume accelerated as cognitive function deteriorated. Moreover, glucose metabolic decline and hippocampal atrophy were significantly slower in subjects with normal cognition compared to those with MCI or AD. Positive APOE4 status was also associated with accelerated hippo‐

Molecular imaging utilizes small molecule ligands that bind with nanomolar affinity to amyloid and enters the brain for imaging with PET. It is a measure to detect and quantify cerebral beta-amyloidosis. It should be noted that besides AD, there are other disease condi‐ tions that may have cerebral Aβ.The most commonly used ligand is the carbon-11(11C)-based Pittsburgh compound B (PIB), which binds specifically to fibrillar Aβ but exhibits no demon‐ strable binding to neurofibrillary tangles. However, fluorine-18 (18F)-based tracers, e.g. 2-(1- {6-[(2-fluorine 18-labeled fluoroethyl)methylamino]-2-napthyl}ethylidene) malononitrile ([(18)F]FDDNP) have a considerably longer half-life compared to [11(C)]PIB and some types

Okello [106] showed that PIB-positive subjects with MCI are significantly more likely to convert to AD than PIB-negative ones. A separate longitudinal study showed that hippocam‐ pal atrophy and amyloid deposition (in posterior cingulate, lateral frontal cortex, temporal cortex, putamen and caudate nucleus) seem to dissociate during the evolution of MCI, the atrophy increasing clearly and [(11)C] PIB retention changing modestly when conversion to AD occurs [107]. Using [(18)F]FDDNP PET, higher baseline binding was associated with future decline in most cognitive domains. Specifically, frontal and parietal [(18)F]FDDNP binding yielded the greatest diagnostic accuracy in identifying MCI-converters versus non-converters [108]. With 18F florbetapir (18F-AV-45) tracer, baseline Aβ + scans were associated with greater

the best antecedent MRI marker of imminent conversion.

to AD for 76% of aMCI subjects after cross-validation [103].

have been shown to also bind to neurofibrillary tangles.

**5.2. Functional and molecular imaging**

Emission Tomography (PET) [104].

392 Understanding Alzheimer's Disease

campal atrophy.

MRI volumetry and brain atrophy rates have fairly good diagnostic and predictive value in MCI subjects. Longitudinal data on brain atrophy rates with disease progression are available and hence, can be used for monitoring disease progression in clinical trials. The limitations of structural neuroimaging as a biomarker include problems with the accurate delineation of regions of interest and lack of standardization of imaging and measurement techniques, making it difficult to compare data across the different institutions out of Europe, North America and Australia (all of which have their unified imaging consortiums). The advent of automated data-driven innovations for structural imaging holds promise. FDG-PET appears to be the leading candidate among the functional neuroimaging modalities, with available evidence for MCI diagnosis, prediction of MCI-converters and longitudinal data in monitoring serial progression. To date, [(11]C] PIB is the most extensively studied PET amyloid tracer, although 18F florbetapir proves to be an attractive alternative given the longer half-life. There is emerging evidence for amyloid imaging in the diagnosis of preclinical AD. From the standpoint of clinical trials of anti-amyloid therapy, in-vivo amyloid imaging pre-treatment allows selection of patients with demonstrable cerebral Aβ loads; repeated imaging during ongoing treatment allows detection of decrease in insoluble Aβ load in response to amyloidclearing drugs such as immunotherapy. Amyloid imaging needs to be more practically accessible and affordable before it can be transferable to the clinical diagnostic routine.

**Subjects Follow-up (years) Biomarker Results** 

[72] No change in Aβ42 or p-tau 181.

MCI (n=8) 1 CSF p-tau231 MCI: 5.0; NC: 3.0 \* NC (n=10) CSF Aβ40 MCI: 4.0; NC: 8.0 [70] CSF Aβ42 MCI: 4.0; NC: 2.0 MCI (n=7) 2 CSF p-tau231 MCI: 2.0; NC: 20.0 \* NC (n=9) CSF Aβ40 MCI: 0.5; NC: 3.5 [71] CSFAβ42 MCI: 0.35; NC: 1.5

MCI (n=62) 2 CSF isoprostane NC:-1.9; MCI:-0.4; AD: 5.0 \*\* AD (n-68) CSF neurofilaments light NC:-0.18; MCI:-0.79; AD: -0.96 NC (n=24) CSF Aβ40 NC: 0.61; MCI:0.28; AD:0.43

MCI (n=41) AD fast -15.4

AD fast -22.7

AD fast -3.6

AD fast 1.9

MCI (n=49) (neocortical PiB rate (non-acc) -0.01

MCI (n=72) 1-2 Hippocampus \* -3.3 (2.7) [104] Entorhinal cortex -7.0 (4.3) Whole brain -0.7 (1.0) Ventricle 3.3 (2.3)

NC = Normal Controls AD = Alzheimer's Disease CSF = Cerebrospinal fluid PIB = Pittsburgh Compound B

\*\* expressed as annual change β MCI = Mild Cognitive Impairment

MMSE = Mini Mental State Examination CDR-SB = Clinical Dementia Rating – Sum of Boxes RAVLT = Rey Auditory Verbal Learning Tes

**Table 5.** Longitudinal biomarker studies

\* expressed as % change per year compared to baseline values

FDG-PET = Fluorodeoxyglucose (18F)-Positron Emission Tomography

MCI (n=57) 3 CSF Aβ42 MCI(stable): 3.42, MCI (converters):0.78, AD:-11.9\*\* AD (n=56) [65] CSF tau MCI(stable):19.7, MCI(converters):17.4, AD: 0.55 NC (n=8) CSFp-tau MCI(stable):1.24, MCI(converters):-0.21, AD: -2.2 CSFAβ42/tau MCI(stable):-0.54.MCI(converters):-0.4,AD: -0.008 CSFAβ42/ptau MCI(stable):-0.19, MCI(converters):-0.07,AD:0.18 NC (n=55) 1.2-2.4 Hippocampus\* MCI (stable):-4.4,MCI(converters):-7.8, AD slow -9.4,

Predicting Cognitive Decline in Alzheimer's Disease (AD): The Role of Clinical, Cognitive Characteristics and Biomarkers

http://dx.doi.org/10.5772/54289

395

AD (n=64) [99] Entorhinal cortex MCI(stable):-15.9, MCI(converters):-16.0, AD slow-20.5,

Ventricle MCI (stable):0.8, MCI(converters):1.8, AD slow -6.5,

MCI (n=131) 3 Hippocampus \* MCI (converters) -6.78; MCI (non-converters) -3.86 [101] Entorhinal cortex MCI (converters) -15.08 ; MCI (non-converters) -8.32 Whole brain MCI (converters) -0.88 ; MCI (non-converters) -0.36 Ventricle MCI (converters) 5.66 ; MCI (non-converters) 3.33

AD (n=32) 1.5 PiB-PET AD: PiB-(acc) +0.06; PiB+(acc) +0.05; PiB

NC (n=103) of change) MCI: PIB-(acc) +0.04; PiB-(non-acc) -0.001; [116] (SuVRpons/year) \* PiB+ (acc) +0.04; PiB (non-acc) -0.01 HC: PiB-(acc)+0.03; PiB- (non-acc) -0.01; PiB+(acc) +0.04; PiB+ (non-acc) -0.004 NC (n=210) 2 CSF Aβ42 NC: -0.94; MCI: -1.4; AD; -0.1 \* MCI (n=357) CSF tau NC: 3.45; MCI: 2.34; AD: 1.24 AD (n=162) PIB NC: 0.098; MCI: -0.008; AD: -0.004 [135] FDG-PET NC: -177; MCI: 752; AD: 2993 Hippocampus NC: -40; MCI: -80; AD: -116 Ventricles NC: 848; MCI: 1551; AD: 2540 ADAS-Cog total NC: -0.54; MCI: 1.05; AD: 4.37 MMSE NC: 0.0095; MCI:-0.64; AD: -2.4 CDR-SB NC: 0.07; MCI 0.63; AD: 1.62 RAVLT (5 trial total) NC: 0.29; MCI: -1.37; AD: -3.62

Whole brain MCI (stable):-0.8, MCI (converters):-2.5, AD slow -2.4,

## **6. Combinational biomarkers**

Many of the aforementioned biomarker modalities are not separate discrete entities but have an effect on each other. For example, the association of hypertension with CSF tau and ptau-181, was found to be modified by APOEε4 phenotype, where hypertension is directly related to tau pathology (and not Aβ42) in APOEε4 homozygous carriers [113]. Elevated CSF t-tau and p-tau in presence of APOEε4/ε4 genotype has also been shown to influence faster AD progression in MCI subjects [114].

For the identification of MCI-converters, various literature showing combination biomarkers have been published. They include looking at clinical measures (such as cognitive or neuro‐ psychological tests) in combination with CSF biomarkers [115], neuroimaging measures [116, 117], or in combination with both CSF and neuroimaging measures [118-119].

A combination of CSF and neuroimaging biomarkers [120-4] has found improved predictive accuracy of MCI-converters, supported by slope analyses of annual cognitive decline [120]. Okamura showed that a high ratio between cerebrospinal fluid (CSF) tau and posterior cingulate perfusion on SPECT is useful in identifying MCI converters [125]. Using a machinelearning approach (support vector machines), Furney et al examined the utility of adding cytokine and neuroimaging biomarkers to conventional measures, and found that the combi‐ nation of cytokine and neuroimaging with clinical and APOEε4 genotype improved accuracy [126]. Recent studies have also looked at multimodal neuroimaging techniques to predict MCI progression [127-129].

Other recent studies have used endophenotype-based approach and found single nucleotide polymorphism (SNP) such as rs1868402 to have strong, replicable association with CSFptau181 association with rate of AD progression [130].

### **7. Conclusion and future directions**

Clinical criteria alone, often subjective and dependent on clinical judgment, are insufficient to identify the pre-clinical stages of AD accurately. This has prompted the past decade-long intensive research into the use of more objective neuroimaging and biochemical markers to either replace, or complement, clinical approaches to facilitate an early and accurate diagnosis of the illness [131,132]. The chapter thus far details the rationale (most evident from Table 1) for the combined approach of clinical measures with other biomarkers in predicting AD progression; but in the earlier stages (prodromal and especially preclinical AD stages), biomarkers would play an increasingly important role. Combination biomarker approaches appear to be superior to a single biomarker approach, with the recent focus of researchers being Predicting Cognitive Decline in Alzheimer's Disease (AD): The Role of Clinical, Cognitive Characteristics and Biomarkers http://dx.doi.org/10.5772/54289 395


\* expressed as % change per year compared to baseline values

\*\* expressed as annual change β

MCI = Mild Cognitive Impairment

NC = Normal Controls

clearing drugs such as immunotherapy. Amyloid imaging needs to be more practically accessible and affordable before it can be transferable to the clinical diagnostic routine.

Many of the aforementioned biomarker modalities are not separate discrete entities but have an effect on each other. For example, the association of hypertension with CSF tau and ptau-181, was found to be modified by APOEε4 phenotype, where hypertension is directly related to tau pathology (and not Aβ42) in APOEε4 homozygous carriers [113]. Elevated CSF t-tau and p-tau in presence of APOEε4/ε4 genotype has also been shown to influence faster

For the identification of MCI-converters, various literature showing combination biomarkers have been published. They include looking at clinical measures (such as cognitive or neuro‐ psychological tests) in combination with CSF biomarkers [115], neuroimaging measures [116,

A combination of CSF and neuroimaging biomarkers [120-4] has found improved predictive accuracy of MCI-converters, supported by slope analyses of annual cognitive decline [120]. Okamura showed that a high ratio between cerebrospinal fluid (CSF) tau and posterior cingulate perfusion on SPECT is useful in identifying MCI converters [125]. Using a machinelearning approach (support vector machines), Furney et al examined the utility of adding cytokine and neuroimaging biomarkers to conventional measures, and found that the combi‐ nation of cytokine and neuroimaging with clinical and APOEε4 genotype improved accuracy [126]. Recent studies have also looked at multimodal neuroimaging techniques to predict MCI

Other recent studies have used endophenotype-based approach and found single nucleotide polymorphism (SNP) such as rs1868402 to have strong, replicable association with

Clinical criteria alone, often subjective and dependent on clinical judgment, are insufficient to identify the pre-clinical stages of AD accurately. This has prompted the past decade-long intensive research into the use of more objective neuroimaging and biochemical markers to either replace, or complement, clinical approaches to facilitate an early and accurate diagnosis of the illness [131,132]. The chapter thus far details the rationale (most evident from Table 1) for the combined approach of clinical measures with other biomarkers in predicting AD progression; but in the earlier stages (prodromal and especially preclinical AD stages), biomarkers would play an increasingly important role. Combination biomarker approaches appear to be superior to a single biomarker approach, with the recent focus of researchers being

117], or in combination with both CSF and neuroimaging measures [118-119].

CSFptau181 association with rate of AD progression [130].

**7. Conclusion and future directions**

**6. Combinational biomarkers**

394 Understanding Alzheimer's Disease

AD progression in MCI subjects [114].

progression [127-129].

AD = Alzheimer's Disease

CSF = Cerebrospinal fluid

PIB = Pittsburgh Compound B

FDG-PET = Fluorodeoxyglucose (18F)-Positron Emission Tomography

MMSE = Mini Mental State Examination

CDR-SB = Clinical Dementia Rating – Sum of Boxes

RAVLT = Rey Auditory Verbal Learning Tes

**Table 5.** Longitudinal biomarker studies

on multimodal approach using various systems biology and multivariate modeling methods. Additionally, multi-site prospective studies, such as the Alzheimer's Disease Neuroimaging Initiative (ADNI), allow for global summary of results and patterns of change observed in clinical measures and candidate biomarkers [133] (Table 5). It must also be highlighted that some of the heterogeneity of biomarker findings thus far is related to the different periods of follow-up and hence AD conversion rates in MCI subjects.

**Author details**

Mei Sian Chong1

**References**

and Tih-Shih Lee2

2 Duke University Medical School, USA

Dement 2008; 4(1): 22-29.

Neurol Neurosurg 2009; 111(4):327-30.

Dement Geriatr Cogn Disord 2008; 26(2): 109-16.

Alzheimer disease. Diabetes 2004; 53:474-481.

1 Department of Geriatric Medicine, Tan Tock Seng Hospital, Singapore

occurrence in the world. Alz Dis Assoc Disord 2003; 17: 63-67.

Alzheimer disease. Ann Neurol 1999; 45:358-68.

[1] Wimo A, Winbald B, Aguero Torres H, von Strauss E. The magnitude of dementia

Predicting Cognitive Decline in Alzheimer's Disease (AD): The Role of Clinical, Cognitive Characteristics and Biomarkers

http://dx.doi.org/10.5772/54289

397

[2] Price JL, Morris JC. Tangles and plaques in nondemented aging and preclinical

[3] Bateman RJ, Xiong C, Benzinger TL, Fagan AM, Goate A, Fox NC, Marcus DS, Cairns NJ, Xie X, Blazey TM, Holtzman DM, Santacruz A, Buckles V, Oliver A, Moulder K, Aisen PS, Ghetti B, Klunk WE, McDade E, Martins RN, Masters CL, Mayeux R, Ringman JM, Rossor MN, Schofield PR, Sperling RA, Salloway S, Morris JC; the Dominantly Inherited Alzheimer Network. Clinical and Biomarker Changes in

[4] Jack CR Jr, Albert MS, Knopman DS, McKhann GM, Sperling RA, Carrillo MC, Thies B, Phelps CH. Introduction to the recommendations from the National Institute on Aging-Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's

[5] Cores F, Nourhashemi F, Guerin O et al. Prognosis of Alzheimer's disease today : a two-year prospective study in 686 patients from the REAL-FR Study. Alzheimers

[6] Roselli F, Tartaglione B, Federico F, Lepore V, Defazio G, Livrea P. Rate of MMSE score change in Alzheimer's disease: Influence of education and vascular risk factors. Clin

[7] Soto ME, Anrieu S, Cantet C, Reynish E et al. Predictive value of rapid decline in Mini Mental State Examination in clinical practice for prognosis in Alzheimer's disease.

[8] de la Monte SM. Contributions of Brain Insulin Resistance and Deficiency in Amyloid-Related Neurodegeneration in Alzheimer's disease. Drugs 2012; 73(1):49-66.

[9] Janson J, Laedtke T, Parisi JE, O/Brien P et al. Increased risk of type 2 diabetes in

[10] Francis GJ, Martinez JA, Liu WQ, Xu K, Ayer A, Fine J, Tuor UI, Glazner G, Hanson, LR, Frey WH 2nd, Toth C. Intranasal insulin prevents cognitive decline, cerebral

Dominantly Inherited Alzheimer's Disease. N Engl J Med. 2012 Jul 11.

disease. Alzheimers Dement. 2011 May;7(3):257-62. Epub 2011 Apr 21.

The dynamic biomarker model, in the AD pathological cascade first proposed by Jack in 2010 [134], has been an area of intense interest. However, this inverse relationship between fibrillar amyloid plaque burden (on PIB imaging) and corresponding decrease in CSF Aβ42 and elevated tau, has led to the simplistic interpretation that the AD pathological cascade is purely driven by the amyloid cascade (Figure 1). This is partly due to extrapolation from crosssectional studies, where in fact, longitudinal studies are required to determine the temporal order of the appearance of various pathogenic processes involved in this complex disease. Storandt et al [135] has recently demonstrated in a community cohort that CSF Aβ42 and tau were minimally correlated, suggesting that they represent independent processes. Addition‐ ally, they accounted for only 60% of variance on PIB imaging, suggesting that a third process may be related to brain atrophy or plaque formation [136].

In addition, understanding longitudinal biomarker change allows its potential inclusion in clinical trials, with recent studies advocating the use of neuroimaging biomarkers [137,138], CSF biomarkers [139] and/or combination biomarkers [137,140] to boost the power of clinical trials and decrease sample size in MCI trials. An integrated analyses approach using patient (age) severity- and disease-related (severe baseline cognitive, global or behavioural status) factors in established AD has been shown, with the potential of symptomatic AD therapy, to decrease likelihood of faster decline [141].

Further work on biomarkers is important because of their multiple potential roles. Biomarkers have the potential to be used as a prognostic tool for the prediction of AD conversion in MCI subjects and rapid AD progression, with translation into clinical practice by using a most practical algorithm, and as a diagnostic tool in prodromal/ preclinical stages of AD. Biomarkers may also lead to a deeper understanding of the complex pathogenesis of AD disease – including stage-specific and stage-independent processes. There is also currently an unfulfil‐ led potential in biomarker-enriched clinical trials and the use of biomarkers in preclinical AD, especially in the advent of newer therapeutic targets. Finally there is also potential to extrap‐ olate biomarker findings 'backwards' into the earliest stages of disease so that we may be able to identify those at risk and consider instituting interventions. This would enable earliest therapeutic intervention for at-risk subjects most amenable to disease-modifying treatments, and exclude those for whom the possible risks from investigational treatment would be more difficult to justify. At the very least, it would identify those who might benefit most from intensive monitoring and management of clinical factors, e.g. blood pressure, diabetes and lipids, and also non-invasive interventions, e.g. cognitive training. This vital work can only been done through multi-center studies and standardized evaluation techniques using various systems biology and statistical modeling approaches.
