Enrichment through biomarkers in clinical trials of Alzheimer’s drugs in patients with mild
M. Lorenzia, M. Donohued, D. Paternicòa, C. Scarpazzaa, S. Ostrowitzkig, O. Blinh, E. Irvingi, G.B.
Frisonia,c *, and the Alzheimer's Disease Neuroimaging Initiative
a LENITEM Laboratory of Epidemiology, Neuroimaging, and Telemedicine, IRCCS San Giovanni di
Dio-FBF, Brescia, Italy
c AFaR Associazione Fatebenefratelli per la Ricerca, Rome, Italy
d Division of Biostatistics and Bioinformatics, University of California, San Diego, USA
g Clinical Pharmacology, Roche Palo Alto LLC, 3431 Hillview Avenue, Palo Alto, CA 94304, USA.
h Clinical Investigation Centre (CIC-UPCET) and Department of Clinical Pharmacology, UMR-CNRS
6193 Institute of Cognitive Neurosciences, CHU Timone, Marseille, France
i Neurosciences CEDD, GlaxoSmithKline, New Frontiers Science Park, Third Avenue, Harlow, Essex,
CM19 5AW, United Kingdom
This version of the manuscript is previous to the review procedure. The work has been published
on Aug 2010 by the journal Neurobiology of Aging.
Clinical trials of disease modifying drugs for Alzheimer’s disease (AD) in patients with mild
cognitive impairment (MCI) might benefit from enrichment with true AD cases. 405 MCI patients
(143 converters and 262 non converters to AD within 2 years) of the ADNI were used. Markers for
enrichment were hippocampal atrophy on MR, temporoparietal hypometabolism on FDG PET, CSF
biomarkers (Abeta42, tau, and phospho-tau), and cortical amyloid deposition (11C-PIB PET). Two
separate enrichment strategy were tested aimed to A) maximize the proportion of MCI converters
screened in, and B) minimize the proportion of MCI converters screened out. Based on strategy A,
when compared to no enrichment and ADAS-Cog as an outcome measure (sample size of 834),
enrichment with 18F-FDG PET and hippocampal volume lowered samples size to 260 and 277 cases
per arm, but at the cost of screening out 1,597 and 434 cases per arm. When compared to no
enrichment and CDR-SOB as an outcome measure (sample size of 674), enrichment with
hippocampal volume and Abeta42 lowered samples size to 191 and 291 cases per arm, with 639
and 157 screened out cases. Strategy B reduced the number of screened out cases (740 for [11C]-
PIB PET, 101 hippocampal volume, 82 ADAS-COG and 330 for [18F]-FDG PET) but at the expense of
decreased power and a relative increase size (740 for [11C]-PIB PET, 676 for hippocampal volume,
744 for ADAS-Cog and 517 for [18F]-FDG PET). Enrichment comes to the price of an often relevant
proportion of screened out cases, and in clinical trial settings the balance between enrichment of
screened in and loss of screened out patients should be critically discussed.
Drugs aimed to modify the course of Alzheimer’s disease (AD) are under active development.
These drugs might be maximally effective when prescribed early in the course of the disease.
Amnestic mild cognitive impairment (MCI) is currently the earliest stage when patients with AD can
be captured for clinical trial purposes, but the diagnostic category of MCI is contaminated by a
sizable proportion (up to 50%) of patients who do not have AD. Indeed, all clinical trials with anti-
dementia drugs that have been carried out in the MCI populations have failed to demonstrate a
significant treatment effect (Feldman et al., 2007; Loy and Schneider, 2006; Petersen et al., 2005;
Raschetti et al., 2007; Salloway et al., 2004), and one of the proposed reasons is contamination by
non Alzheimer’s cases (Visser et al., 2005).
It is widely believed that MCI patients with abnormal brain structure volume or metabolism, or
biochemical marker profile are more likely to develop AD than the parent MCI population. A
proposal for new diagnostic criteria has been developed that could allow diagnosis of AD at the
MCI stage based on atrophy of medial temporal lobe structures (among which the hippocampus)
on structural magnetic resonance imaging (MRI), hypometabolism in the temporoparietal cortex
on 18F-FDG PET, low Abeta42 or high tau or phospho-tau in the CSF, and positivity on amyloid
imaging with tracers such as 11C-PIB (Dubois et al., 2007) . A corollary of this is that AD markers
might be employed in clinical trials of MCI patients to screen out non-AD MCI cases and select a
population of MCI enriched with truly AD cases to be randomized.
Of course, the ideal marker is one with 100% sensitivity and specificity, which would support
screening out of all non AD and screening in all AD cases. However, this is hardly a realistic scenario
in that markers will in all likelihood merely enrich screened out with true negatives and screened in
with true positives. In a clinical trial scenario, a good marker will be one with high sensitivity: the
ratio between the proportion of AD cases which are screened positive and included, i.e. true
positive rate, and the proportion of AD cases which are screened negative and excluded, i.e. false
negative rate. Data that allow estimation of this sensitivity and specifity are available from the
Alzheimer’s Disease Neuroimaging Initative (ADNI) (Mueller et al., 2005). The ADNI has studied 399
MCI patients with structural MRI, 18F-FDG PET, CSF studies, and 11C-PIB and followed them to
detect conversion to dementia. The aim of the present study was to assess the benefit of the
enrichment of MCI patients with true AD cases by means of hippocampal atrophy on MRI,
temporoparietal hypometabolism on 18F-FDG PET, CSF biomarkers (Abeta42, tau, and phospho-
tau), and cortical amyloid deposition
n on 11C-PIB. All markers were measured on continuous scales and the optimal threshold for
screening has been defined empirically based on the distribution of the marker in the 229 healthy
elders of the ADNI database in whom the same markers have been collected.
The subjects of this study were taken from the Alzheimer’s Disease Neuroimaging Initiative (ADNI)
database (http://www.loni.ucla.edu/ADNI/Data) as of September 29th, 2009. The ADNI was launched
in 2003 by the National Institute on Aging (NIA), the National Institute of Biomedical Imaging and
Bioengineering (NIBIB), the Food and Drug Administration (FDA), private pharmaceutical
companies and non-profit organizations, as a $60 million, 5-year public-private partnership. The
primary goal of ADNI has been to test whether serial magnetic resonance (MR) imaging, positron
emission tomography (PET), other biological markers, and clinical and neuropsychological
assessment can be combined to measure the progression of mild cognitive impairment (MCI) and
early Alzheimer's disease (AD). Determination of sensitive and specific markers of very early AD
progression is intended to aid researchers and clinicians to develop new treatments and monitor
their effectiveness, as well as lessen the time and cost of clinical trials. The Principal Investigator of
this initiative is Michael W. Weiner, M.D., VA Medical Center and University of California - San
Francisco. ADNI is the result of efforts of many co-investigators from a broad range of academic
institutions and private corporations, and subjects have been recruited from over 50 sites across
the U.S. and Canada. The initial goal of ADNI was to recruit 800 adults, ages 55 to 90, to participate
in the research – approximately 200 cognitively normal older individuals to be followed for 3 years,
400 people with MCI to be followed for 3 years, and 200 people with early AD to be followed for 2
years. For up-to-date information see www.adni-info.org. Table 1 shows the salient features of the
MCI converters and non converters and healthy elderly controls used for the present study. Mean
participant age, gender and education do not differ across groups.
The measure of hippocampal atrophy was the mean left and right baseline hippocampal volume
reported in the ADNI dataset, collected through manual tracing on high resolution 3D MR scans
following the protocol of Jack and colleagues (Jack et al., 1995). The measure of temporoparietal
hypometabolism was the t-sum developed by Herholz et al. (2002) on 18F-FDG PET images. This is
an adimensional number ranging between 0 and infinity indicative of hypometabolism in the
cortical regions found specific to AD including temporoparietal cortex, posterior cingulate and
precuneus, frontal association cortex bilaterally. A value of 11,090 was empirically found to be the
optimal threshold to distinguish AD patients from healthy elders (Herholz et al., 2002). The values
of baseline CSF biomarkers (Abeta42, tau, and phospho-tau) were those reported in the ADNI
database, measured through the Luminex xMAP platform. The measure of cortical amyloid
deposition was defined as the mean value of the 11C-PIB PET images in the gray matter. This
measure was obtained after a number of image processing steps. The [11C]-PIB PET images were
first co-registered to the respective MR images and spatial normalized using the parameters
determined from the normalization of MR images through the DARTEL procedure (Ashburner,
2007). 11C-PIB PET normalized images were then scaled to the cerebellum and the mean uptake
value was finally computed on the regions defined by the grey matter DARTEL template.
2.3 Data treatment and statistical analysis.
Data treatment is summarized in figure 1. The distribution of the markers was modelled using the
fast Fourier transform to convolve the approximation of the empirical distribution with a gaussian
kernel and using linear approximation to evaluate the density at the specific point. The analysis
was conducted in the R statistical computing environment (R Development Core Team R ISBN URL
2008) (http://www.R-project.org). Two different enrichment strategies were tested.
Enrichment strategy A: maximization of the proportion of screened-in MCI converters
Increasingly restrictive thresholds were defined based on the 70th, 85th, 95th, and 99th percentile of
the distribution of marker values in healthy elders. The number of MCI converters and non-
converters among screened out and screened in on the biomarker (with 95% confidence interval)
(Wilson, 1927) was computed for each threshold. The thresholds associated with the highest
proportion of MCI converters among screened in were then chosen (thresholds shown in
supplementary table). For temporoparietal hypometabolism, the threshold of 11,090 was also
tested following the original results from Herholz et al. (2002), and for CSF biomarkers the
threshold of 192 pg/ml was tested following Shaw et al.’s (2009).
Enrichment strategy B: minimization of the proportion of screened-out MCI converters
At each percentile distribution of the control population, the ratio between the number of non-
converters and converters among screened out was computed. The thresholds were chosen
according to an efficiency criterium in order to minimize the proportion of excluded converters
among the screened out population. The thresholds optimizing the previous criteria (shown in
supplementary table) were then employed to carry out the ensuing power computations.
For each marker, the ratio between the proportion of converters and non converters among the
screened in was computed using a classical bayesian approach (Albert, 2009). The distributions of
the proportions pconverters were inferred using a binomial likelihood and a beta(1,1) as non
informative conjugate prior, resulting in a beta distribution for p of parameters a=nconverters+1 and
b=nnon converters+1. The subsequent statistical analyses were then perfomed on the ratios between
pconverters and pnonconverters =1-pconverters obtained from drawing 10,000 samples from the posterior
Finally, the screened groups enriched with the different markers were used to compute the sample
size required to detect a hypothetical 25% difference in the rate of decline in a two-year placebo
controlled randomized clinical trial with six-month visit intervals. The group resulting from [11C]-
PIB enrichment was dropped from further consideration due to inadequate size. Longitudinal
ADAS-COG and CDR-SB scores available from the ADNI dataset were used to fit random effects
models to estimate the annual rate of change for each enrichment scenario. The models included
random intercept and slope and, based on the parameter estimates with 95% confidence intervals,
the sample size required with associated confidence intervals was then computed using the
formula of Li and Liang (1997). We note here that the resulting estimates depend on the
composition of the different groups from which the model is fitted. Sample size estimates are
inflated to account for a 30% dropout rate over 2 years.
Figure 2 shows that the distribution of markers was roughly bell-shaped for all markers in the three
groups of healthy elderly controls, MCI converters, and MCI non converters. A hint of a bimodal
distribution could be appreciated in healthy elders for CSF Abeta42, consistent with the notion that
some healthy elders might host pre-symptomatic forms of the disease. A clear bimodal distribution
was present in MCI non converters for PIB, consistently with the notion that some of the MCI non
converters might convert in the near future, as also supported by the observation of a large share
of CSF Abeta42 values in MCI non converters lying in the conversion area. Interestingly, a small
bell-shaped tail can be appreciated for [11C]-PIB PET in MCI converters lying in the healthy elders
area, suggesting that some MCI converters to Alzheimer’s disease might on the contrary have
other forms of dementia.
The figure also shows that with enrichment strategy A, increasingly restrictive thresholds (from
none to the 99th percentile of the distribution of healthy elderly controls) generally lead to select a
monotonously increasingly enriched proportion of future converters among those screened in
except in the case of CSF markers, where the correspondence of the distribution curves of MCI
converters and non converters led to a monotonous increase. The highest proportion of future
converters was achieved by hippocampal volume thresholded at the 1st percentile of the healthy
elders distribution, and [11C]-PIB PET thresholded at the 95th percentile, increasing from 38% with
no threshold to 59 and 60%, respectively. However, this enrichment was obtained at the expense
of a marked increase of screened out rate, up to 77% and 84% of those MCI enrolled.
Lastly, the figure shows the thresholds found with enrichment strategy B. The lowest proportion of
screened out converters was achieved by ADAS-cog (7.5%) and [11C]-PIB PET (9%) at the 58th and
85th percentile respectively. For CSF biomarkers, the proportion of screened-out MCI converters
was monotonously decreasing with decreasing marker values, thus preventing to identify an
Figure 3 shows that with enrichment strategy A the most favourable ratio between MCI converters
and non converters is achieved by [11C]-PIB PET (ratio of 1.5), but due to the small group size the
confidence interval is very large (3.69 to 0.62) and the point estimate is poorly reliable. A slightly
less favourable ratio (1.46) is achieved by hippocampal volume, but with a much more accurate
point estimate. A lower favourable ratio (1.14) is achieved by FDG PET. However, in all of these
cases the proportion of screened out is remarkably high (84%, 77%, and 86%). All other markers
yield ratios below 1, ranging between 0.98 (ADAS-Cog) and 0.87 (CSF Abeta42 and p-tau). It should
be noted that for 5 markers ([11C]-PIB PET, hippocampal volume, ADAS-Cog, CSF tau, and CSF
Abeta42) the decreasingly favourable ratio of MCI converters to non converters was associated
with an expected decreased proportion of screened out (from 84% down to 77%, 56%, 38%, and
35%), and for two markers (CSF tau/Abeta42, CSF p-tau) the proportion of screened out was
relatively high (46% and 55%) despite an unfavourable ratio of MCI converters to non converters. It
should be noted that, although all markers led to a significant enrichment (ratios always
significantly greater than the reference condition) the ratio of hippocampal volume was
significantly greater than all other ratios except PIB PET due to its low group size and wide
confidence interval, and [18F]-FDG PET.
With enrichment strategy B, [11C]-PIB PET at the 85th percentile leads again to the most favourable
ratio (ratio of 1.00) between MCI screened in converters and non converters. This ratio is lower
than that obtained with strategy A (ratio of 1.50) as well as for the other markers, associated with
ratios ranging between 0.56 and 0.64.
Table 2 shows that enrichment strategy A leads to identify 18F-FDG PET as the marker associated
with the lowest sample size per arm for a hypothetical 24-month trial in MCI patients of a disease
modifying drug with 25% efficacy and 90% power and ADAS-COG as an outcome measure (260
cases per arm, estimated from the screened group of 28 patients), but at the cost of screening out
1,597 cases per arm. When CDR sum of boxes was used as an outcome measure, the lowest sample
size was achieved by hippocampal volume with 191 cases per arm and 639 screened out cases. CSF
Abeta42 is associated with the lowest screened out group size using both ADAS-cog (269 and
sample size of 500 per arm) and CDR sum of boxes as outcome measure (157 cases and sample size
of 291 per arm).
Table 2 also shows that with enrichment strategy B and ADAS-cog as outcome measure the lowest
sample size per arm is associated with 18F-FDG PET (517 cases estimated from 127 patients and
330 screened out cases). On the other hand, considering the CDR sum of boxes as outcome
measure, 11C-PIB PET leads to the lowest sample size with 351 cases per arms and an equal
number of screened out cases. ADAS-cog achieves the lowest number of screened out using both
ADAS-cog (82 cases and a sample size of 744 cases per arm) and CDR sum of boxes (56 cases and
sample size of 509) as an outcome measure.
We have shown that the screening procedure with imaging and biological markers can lead to a
significant enrichment of groups of MCI patients enrolled in clinical trials of AD drugs with “true AD
cases”, i.e. patients who will convert in the following months. In this study, two different strategies
enabled respectively to (A) enrich the screened in group with true AD patients (MCI converters)
and (B) control the number of screened out cases, reducing the loss of MCI converters. The un-
enriched ratio between converters and non converters of 0.56, i.e. almost 1:2, can be reversed
with strategy A to 1.46, i.e. about 3:2. This enrichment comes to the price of a sometimes relevant
proportion of screened out MCI patients falling below threshold, that increase with increasing
enrichment and can amount to as much as 84% of all MCIs. This percentage is reduced with
strategy B, varying between 10% and 50%, and comes with the advantage of a reduced number of
true converters lost (between 7.5% and 17% of the whole MCI population), although the screened-
in populations are characterized by a lower ratio of converters to non-converters, albeit
significantly higher than the un-enrichment scenario. Interestingly, CSF biomarkers did not exhibit
a consistent threshold minimizing the number of excluded converters, reflecting high specificity.
Thresholds resulting from strategy A often led to an unrealistic high percentage of screened out
patients (up to 86%) as well as markers values lying in the pathological range, such as for the 11C-
PIB whole brain uptake close to the value of 2 or ADAS-Cog score of 19.4. This inconvenient is
mitigated by the adoption of strategy B, where marker values are closer to the thresholds
employed for diagnostic purposes (supp Table). This result is achieved at the expense of a less
favourable proportion of MCI converters in the screened in population.
A key point emerging from the current study is the role of the markers thresholds chosen for the
screening procedure, and the impact of their use in the resulting clinical practice.
In a hypothetical clinical trial, the balance between enrichment of screened in and loss of screened
out patients should be viewed in the light of the gain of power and the relative decreased costs
brought about by enrichment and the increased costs brought about by the exclusion of screened
A large number of studies have recently shown that MCI patients positive to one or more AD
biological and imaging markers have greater chance to convert to AD (de Leon et al., 2007; Hampel
et al., 2008). Some (Risacher et al., 2009; Ferris, 2002) have suggested that markers may help
identify MCI individuals at increased risk of conversion to AD, thus assisting researchers striving to
enrich clinical trial populations with people with latent AD, but to the best of our knowledge no
study has so far estimated the extent of enrichment as well as the inevitable costs in terms of
screened out. Back in 2002, Ferris argued that “one approach to reducing the cost would be to
recruit ‘enriched’ samples of subjects who are at greater risk of developing AD during the trial” and
underlined that the major effort required to screen and recruit large numbers of subjects for such
trials would contribute to the cost. While acknowledging that research to develop more efficient
assessment methods is needed, he suggested that data acquisition over the Internet might be an
efficient and practical tool. Thanks to the recent availability of the public ADNI dataset, we can
now conclude that hippocampal volumetry might be an efficient strategy for enrichment. The
opportunity to resort to alternative strategies with lower enrichment power such as FDG PET, PIB
PET or CSF markers should be judged in the context of the lower costs for screening and the
biological mechanism of the drug under trial.
Most disease modifiers presently in phases II and III clinical trials are targeting beta amyloid and an
enrichment strategy aiming to select MCI patients with brain amyloidosis might be appropriate.
We have shown that the use of CSF Abeta42 to select MCI patients to enrol in a trial has
significantly lower effectiveness at enrichment with fast converters than hippocampal volume.
Although it might be contended that CSF Abeta42 has high sensitivity and specificity to recruit
patients with brain amyloidosis (Jagust et al., 2009), it should also be acknowledged that some of
these patients might convert significantly later than the 24 months of a clinical trial (Jack et al.,
2010). Thus, a judgement should be made over which criterion should be followed for enrichment,
i.e. efficiency or biological plausibility. Unfortunately, the low group size of MCI patients for whom
11C-PIB PET is available prevents accurate estimates of the effectiveness of this enrichment
strategy. Future studies with larger group sizes will allow to answer this question.
The present is a technical exercise that should be translated into practice with some caveats. The
enrichment strategy in a clinical trial of drug “x” that will prove effective in slowing disease
progression should be viewed in light of the intended licence. Showing the effectiveness of a drug
in a specific subpopulation positive to a biomarker (e.g. MCI patients with small hippocampi),
might exclude from the benefit of prescription the proportion of negative patients, that for some
biomarker might be much larger than the proportion of the biomarker positive. However, cases
such as the one above are not unprecedented in medicine: tamoxifen is currently used for the
treatment of estrogen receptor positive but not estrogen receptor negative breast cancer (Jordan,
This study has at least a couple of limitations. First, the proportion of converters enrolled in the
ADNI is going to change as MCI patients are followed for longer periods of time and more will
convert to Alzheimer’s dementia. Some studies of MCI patients enrolled in clinical settings are
presently available with long follow-up (Ganguli et al., 2004; Mitchell and Shiri-Feshki, 2008; Busse
and Angermeyer, 2006; Tyas et al, 2007) showing that the vast majority of conversions occur in the
first 5 years after first assessment. Since the mean follow-up of the patients of this study is 11
months , it seems likely that a sizable proportion of converters will show up in future years, and
the present estimates of the ratio between MCI converters and non converters will need to be
updated. Second, the ratio estimates for some markers are poorly accurate for the small number of
patients and healthy controls in whom the marker has been collected. Future expansion of the
ADNI dataset will allow to increase the accuracy of the estimate.
Data used in the preparation of this article were obtained from the Alzheimer's Disease
Neuroimaging Initiative (ADNI) database (www.loni.ucla.edu/ADNI). As such, the investigators
within the ADNI contributed to the design and implementation of ADNI and/or provided data but
did not participate in analysis or writing of this report. The complete listing of ADNI investigators
include is available at www.loni.ucla.edu/ADNI/Collaboration/ADNI_Manuscript_Citations.pdf.
The Foundation for the National Institutes of Health (www.fnih.org) coordinates the private sector
participation of the $60 million ADNI public-private partnership that was begun by the National
Institute on Aging (NIA) and supported by the National Institutes of Health. To date, more than $27
million has been provided to the Foundation for NIH by Abbott, AstraZeneca AB, Bayer Schering
Pharma AG, Bristol-Myers Squibb, Eisai Global Clinical Development, Elan Corporation, Genentech,
GE Healthcare, GlaxoSmithKline, Innogenetics, Johnson & Johnson, Eli Lilly and Co., Merck & Co.,
Inc., Novartis AG, Pfizer Inc., F. Hoffmann-La Roche, Schering-Plough, Synarc Inc., and Wyeth, as
well as non-profit partners the Alzheimer's Association and the Institute for the Study of Aging.
Albert, J., 2009. Bayesian Computation with R. ISBN. 978-0-387-92297-3.
Ashburner, J., 2007. A fast diffeomorphic image registration algorithm. Neuroimage. 38, 95-113.
Busse, A., Angermeyer, M.C., Riedel-Heller, S.G., 2006. Progression of mild cognitive impairment to
dementia: a challenge to current thinking. Br. J. Psychiatry. 189, 399-404.
De Leon, M.J., Mosconi, L., Blennow, K., DeSanti, S., Zinkowski, R., Mehta, P.D., Pratico, D., Tsui, W.,
Saint Louis, L.A., Sobanska, L., Brys, M., Li, Y., Rich, K., Rinne, J., Rusinek, H., 2007. Imaging and CSF
studies in the preclinical diagnosis of Alzheimer's disease. Ann. N Y Acad. Sci. 1097, 114-45.
Dubois, B., Feldman, H.H., Jacova, C., Dekosky, S.T., Barberger-Gateau, P., Cummings, J., Delacourte,
A., Galasko, D., Gauthier, S., Jicha, G., Meguro, K., O'brien, J., Pasquier, F., Robert, P., Rossor, M.,
Salloway, S., Stern, Y., Visser, P.J., Scheltens, P., 2007. Research criteria for the diagnosis of Alzheimer's
disease: revising the NINCDS-ADRDA criteria. Lancet Neurol. 6, 734-46.
Feldman, H.H., Ferris, S., Winblad, B., Sfikas, N., Mancione, L., He, Y., Tekin, S., Burns, A.,
Cummings, J., del Ser, T., Inzitari, D., Orgogozo, J.M., Sauer, H., Scheltens, P., Scarpini, E., Herrmann,
N., Farlow, M., Potkin, S., Charles, H.C., Fox, N.C., Lane, R., 2007. Effect of rivastigmine on delay to
diagnosis of Alzheimer's disease from mild cognitive impairment: the InDDEx study. Lancet Neurol. 6, 501-
Ferris, S.H., 2002. Clinical trials in AD: are current formats and outcome measures adequate? Alzheimer
Dis. Assoc. Disord. Suppl. 1,S13-7.
Ganguli, M., Dodge, H.H., Shen, C., DeKosky, S.T., 2004. Mild cognitive impairment, amnestic type: an
epidemiologic study. Neurology. 63, 115-21.
Hampel, H., Bürger, K., Teipel, S.J., Bokde, A.L., Zetterberg, H., Blennow, K., 2008. Core candidate
neurochemical and imaging biomarkers of Alzheimer's disease. Alzheimers Dement. 4, 38-48.
Herholz, K., Salmon, E., Perani, D., Baron, J.C., Holthoff, V., Frölich, L., Schönknecht, P., Ito, K.,
Mielke, R., Kalbe, E., Zündorf, G., Delbeuck, X., Pelati, O., Anchisi, D., Fazio, F., Kerrouche, N.,
Desgranges, B., Eustache, F., Beuthien-Baumann, B., Menzel, C., Schröder, J., Kato, T., Arahata, Y.,
Henze, M., Heiss, W.D., 2002. Discrimination between Alzheimer dementia and controls by automated
analysis of multicenter FDG PET. Neuroimage. 17, 302-316.
Jack, C.R., Jr, Knopman, D.S., Jagust, W.J., Shaw, L.M., Aisen, P.S., Weiner, M.W., Petersen, R.C.,
Trojanowski, J.Q., 2010. Hypothetical model of dynamic biomarkers of the Alzheimer's pathological
cascade. Lancet Neurol. 9, :119-128.
Jack, C.R.; Theodore, W.H.; Cook, M.; McCarthy, G., 1995. MRI-based hippocampal volumetrics:
data acquisition, normal ranges, and optimal protocol. Magnetic Resonance Imaging. 13, 1057-1064.
Jagust, W.J., Landau, S.M., Shaw, L.M., Trojanowski, J.Q., Koeppe, R.A., Reiman, E.M., Foster, N.L.,
Petersen, R.C., Weiner, M.W., Price, J.C., Mathis, C.A., Alzheimer's Disease Neuroimaging Initiative,
2009. Relationships between biomarkers in aging and dementia. Neurology. 73,1193-9.
Jordan, V.C., 1993. "Fourteenth Gaddum Memorial Lecture. A current view of tamoxifen for the
treatment and prevention of breast cancer". Br. J. Pharmacol. 110, 507–17.
Li, G., Sokal, I., Quinn, J.F., Leverenz, J.B., Brodey, M., Schellenberg, G.D., Kaye, J.A., Raskind, M.A.,
Zhang, J., Peskind, E.R., Montine, T.J., 2007. CSF tau/Abeta42 ratio for increased risk of mild
cognitive impairment: a follow-up study. Neurology. 69, 631-9.
Liu, G., Liang, K. Y., 1997. Sample size calculations for studies with correlated observations.
Biometrics. 53, 937-47.
Loy, C., Schneider, L., 2006. Galantamine for Alzheimer's disease and mild cognitive impairment.
Cochrane Database Syst. Rev. 1, CD001747.
Mitchell, A.J., Shiri-Feshki, M., 2008. Temporal trends in the long term risk of progression of mild
cognitive impairment: a pooled analysis. J. Neurol. Neurosurg. Psychiatry. 79, 1386-91.
Mueller, S.G., Weiner, M.W., Thal. L.J., Petersen, R.C., Jack, C., Jagust, W., Trojanowski, J.Q., Toga,
A,W,, Beckett, L., 2005. The Alzheimer's disease neuroimaging initiative. Neuroimaging Clin. N Am. 15,
Petersen, R.C., Thomas, R.G., Grundman, M., Bennett, D., Doody, R., Ferris, S., Galasko, D., Jin, S.,
Kaye, J., Levey, A., Pfeiffer, E., Sano, M., van Dyck, C.H., Thal, L.J.; Alzheimer's Disease Cooperative
Study Group, 2005. Vitamin E and donepezil for the treatment of mild cognitive impairment. N. Engl. J.
Med. 352, 2379-88.
R Development Core Team, 2008. R: A language and environment for statistical computing. R
Foundation for Statistical Computing, Vienna, Austria ISBN 3-900051-07-0.
Raschetti, R., Albanese, E., Vanacore, Maggini N., M., 2007. Cholinesterase inhibitors in mild Download full-text
cognitive impairment: a systematic review of randomised trials. PloS Med. 4, e338.
Risacher, S.L., Saykin, A.J., West, J.D., Shen, L., Firpi, H.A., McDonald, B.C.; Alzheimer's Disease
Neuroimaging Initiative (ADNI), 2009. Baseline MRI predictors of conversion from MCI to probable AD in
the ADNI cohort.. Curr. Alzheimer Res. 6, 347-61.
Salloway, S., Ferris, S., Kluger, A., Goldman, R., Griesing, T., Kumar, D., Richardson, S.; Donepezil 401
Study Group, 2004. Efficacy of donepezil in mild cognitive impairment: a randomized placebo-controlled
trial. Neurology. 63, 651-7.
Shaw, L.M., Vanderstichele, H., Knapik-Czajka, M., Clark, C.M., Aisen, P.S., Petersen, R.C., Blennow,
K., Soares, H., Simon, A., Lewczuk, P., Dean, R., Siemers, E., Potter, W., Lee, V.M., Trojanowski, J.Q.;
Alzheimer's Disease Neuroimaging Initiative, 2009. Cerebrospinal fluid biomarker signature in
Alzheimer's disease neuroimaging initiative subjects. Ann. Neurol. 65,403-13.
Tyas, S.L., Salazar, J.C., Snowdon, D.A., 2007. Transitions to mild cognitive impairments, dementia, and
death: findings from the Nun Study. Am. J. Epidemiol. 165, 1231.
Visser, P.J., Scheltens, P., Verhey, F.R., 2005. Do MCI criteria in drug trials accurately identify subjects
with predementia Alzheimer's disease? J. Neurol. Neurosurg. Psychiatry. 76, 1348-54.
Wilson, E.B., 1927. Probable Inference, the Law of Succession, and Statistical Inference. Journal of
the American Statistical Association. 22, 209-212.