Alzheimer's Disease Neuroimaging Initiative (ADNI) is a multisite study that aims to improve clinical trials for the prevention and treatment of Alzheimer's disease (AD). This cooperative study combines expertise and funding from the private and public sector to study subjects with AD, as well as those who may develop AD and controls with no signs of cognitive impairment. Researchers at 63 sites in the US and Canada track the progression of AD in the human brain with neuroimaging, biochemical, and genetic biological markers. This knowledge helps to find better clinical trials for the prevention and treatment of AD. ADNI has made a global impact, firstly by developing a set of standardized protocols to allow the comparison of results from multiple centers, and secondly by its data-sharing policy which makes available all at the data without embargo to qualified researchers worldwide. To date, over 1000 scientific publications have used ADNI data. A number of other initiatives related to AD and other diseases have been designed and implemented using ADNI as a model. ADNI has been running since 2004 and is currently funded until 2021.

Primary goals Detect the earliest signs of AD and to track the disease using biomarkers. validate, standardize, and optimize biomarkers for clinical AD trials. to make all data and samples available for sharing with clinical trial designers and scientists worldwide.

History and funding The idea of a collaboration between public institutions and private pharmaceutical companies to fund a large biomarker project to study AD and to speed up progress toward effective treatments for the disease was conceived at the beginning of the millennium by Neil S. Buckholz at the National Institute on Aging (NIA) and Dr. William Potter, at Eli Lilly and Company. The Alzheimer's Disease Neuroimaging Initiative (ADNI) began in 2004 under the leadership of Dr. Michael W. Weiner, funded as a private – public partnership with $27 million contributed by 20 companies and two foundations through the Foundation for the National Institutes of Health and $40 million from the NIA. The initial five-year study (ADNI-1) was extended by two years in 2009 by a Grant Opportunities grant, and in 2011 and 2016 by further competitive renewals of the ADNI-1 grant (ADNI-2 and ADNI-3, respectively) (Table 1).

Enrollment of participants ADNI enrolls participants between the ages of 55 and 90 who are recruited at 57 sites in the US and Canada. One group has dementia due to AD, another group has mild memory problems known as mild cognitive impairment (MCI), and the final control group consists of healthy elderly participants. ADNI-1 initially enrolled 200 healthy elderly, 400 participants with MCI, and 200 participants with AD. ADNI-GO, ADNI-2 and ADNI -3 added additional participants to augment the cohort, for final cohort size of over 1000 participants (Table 1).

Table 1: Study characteristics ADNI-1ADNI-GOADNI-2ADNI-3Primary goal Developed biomarkers as outcome measures for clinical trials Examine biomarkers in earlier stages of disease Develop biomarkers as predictors of cognitive decline, and as outcome measures Study the use of tau PET and functional imaging techniques and clinical trials Funding $40 million federal (NIA), $27 million industry and foundation $24 million American Recovery Act funds $40 million federal (NIA), $27 million industry and foundation $40 million federal (NIA), $20 million industry and foundation Duration/start date 5 years/October 2004 2 years/September 2009 5 years/September 2011 5 years/September 2016 Cohort 200 elderly controls 400 MCI

200 AD Existing ADNI-1 + 200 early MCIExisting ADNI-GO+ 150 elderly controls

100 early MCI

150 late MCI

150 late mild cognitive impairment

150 ADExisting ADNI-2 + 133 elderly controls

Testing of participants

ADNI uses a variety of techniques to study its participants. After obtaining informed consent, participants undergo a series of initial tests that are repeated at intervals over subsequent years (Table 2): a clinical evaluation to assess overall health and relevant history such as education neuropsychological tests to assess aspects of brain function affected by AD such as memory, executive function, and the ability to perform activities of daily living. genetic testing for the major AD genetic risk factor, APOE ε4, the gene for a form of apolipoprotein E, and other studies lumbar puncture to collect cerebrospinal fluid (CSF) which is tested for the AD biomarkers β-amyloid, the main component of amyloid plaques, and tau protein, which forms Alzheimer's brain tau tangles. Magnetic resonance imaging (MRI) scans to assess brain structure, connectivity, and the extent of white matter disease Positron emission tomography (PET) scans to assess how well the brain uses glucose, and the extent of amyloid plaques and tau tangles. Neuropathological examination if a participant dies and has consented to autopsy. This definitively diagnoses AD by identifying amyloid plaques and tau tangles, and also notes other common comorbidities. Table 2 Study techniques ADNI-1ADNI-GOADNI-2ADNI-3Imaging MRIStructural XXXXPerfusion XXXResting state XXXDiffusion XXXConnectomicsXHigh resolution XPET Glucose metabolism XXXβ-amyloid [11C] Pittsburgh compound [18F] florbetapir [18F] florbetapir[18F] florbetapir/Florbetaben Tau [18F] T807BiosamplesCSF β-amyloid, tauXXXXGenetic analysis APOEXXXXGenome wide association studiesXXXXWhole genome sequencing XXSystems biology approaches XNeuropsychological tests XXXX AutopsyXXXX

Data sharing One defining characteristic of ADNI is the commitment by all participating research groups to share ownership of the data prior to the completion of the research and by collaborators to forgo any patent opportunities. This has been described by the head of the ADNI and data publications committee as "a radical experiment in open data access". All data generated by the ADNI study are entered into the data archive hosted at the Laboratory of NeuroImaging (LONI) at the University of Southern California. In 2013, whole genome sequencing data for the entire ADNI cohort were added to the LONI database. Qualified researchers worldwide can access image and clinical data sets that have undergone quality control procedures. To date nearly 1800 applications for data use have been received from investigators in multiple disciplines, and over 7 million brain scan images and clinical data sets have been downloaded.

ADNI contributes data to a number of consortia and big data projects which have the potential to unlock many of the mysteries of neurological diseases. It shares imaging and genetic data with the Enhancing Neuro Imaging Genetics through Meta-Analysis (ENIGMA) consortium which uses imaging genetics to study 12 major brain diseases including schizophrenia, bipolar disease and depression. The ADNI dataset was also used as the "test" dataset in the Dialogue on Reverse Engineering Assessment and Methods (DREAM) Alzheimer's disease Big Data Challenge #1 for the discovery of novel predictive AD biomarkers. One measure of the success of this open data sharing approach is the number of scientific publications arising from ADNI data: currently over 1000 and a wide variety of fields including areas outside of Alzheimer's disease.

Development of standardized protocols ADNI has developed standardized protocols that allow results from multiple centers both within the study and worldwide to be directly compared. These include methods for the acquisition and quality control of both MRI and PET scans on scanners differing in the vendor, software platform, and field strength, and also for the analysis of CSF biomarkers. The standardized methods are now used by pharmaceutical companies, and in clinical trials of preventive and disease modifying AD treatments.

Biomarker trajectories throughout disease progression An initial goal of ADNI was to understand the development of AD pathology by tracking imaging and CSF biomarkers throughout disease progression according to the amyloid hypothesis.

A model of how different AD biomarkers change during the development of the AD proposed that biomarkers become abnormal in the following order: β-amyloid (indicating deposition of amyloid in plaques outside the cell, measured in CSF and by amyloid PET) Tau (indicating the formation of tau fibrils with the neurons) Glucose metabolism (measured on PET, indicating damage to neurons) Structural MRI (indicating damage to brain structure) Cognitive impairment This model has been largely validated using longitudinal ADNI data in patients who have abnormal levels of amyloid deposition, consistent with the amyloid hypothesis.

Other significant findings Studies using ADNI cross-sectional and longitudinal MRI, PET, genetics, cognitive, biological fluid, and autopsy data have reported that: AD pathology is already present in people with no outward sign of memory loss. These cognitively normal people may already have subtle brain atrophy. AD progresses in a typical pattern, beginning in the entorhinal cortex in the preclinical AD, spreading to other regions affecting memory, then function and other cognitive abilities in mild to moderate AD and finally severe AD (Figure 1).

The spread can be seen in patterns of changes in biomarkers of β-amyloid deposition (Figure 2), glucose metabolism (Figure 3) and in structural changes in the brain measured by MRI (Figure 4).

AD is characterized by the progressive disruption of the brain connectome. As the disease progresses there are fewer connections between essential brain regions (Figure 5).

Many genes in addition to APOE ε4 underlie late onset AD. ADNI data has helped to identify or confirm 10 of the approximately 20 risk genes currently identified Cerebrovascular disease can accelerate disease progression in AD. Both the cognitively normal and MCI groups are pathologically heterogeneous. Some people show no signs of AD, some show signs of progressing to AD quickly, and others show signs of progressing to dementias other than AD.

Diagnosis of AD and the prediction of future AD ADNI data has been used to test many diagnostic and prognostic machine learning algorithms. The most successful to date have used deep learning approaches that combine longitudinal data chronicling changes in biomarkers over time from more than one imaging, genetic, or biological modality.

Diagnosis

One example of a combination of biomarkers that can accurately diagnose AD is: Changes in brain atrophy patterns over time (measured by MRI) Levels of β-amyloid and tau (measured in CSF) A second approach to diagnosis is to extract the most pertinent information from MRI scans alone. Deep learning algorithms can diagnose AD with greater than 95% accuracy, and can diagnose MCI due to AD with greater than 82% accuracy.

As imaging scans are expensive and sometimes unavailable, and the analysis of CSF requires an invasive lumbar puncture procedure, ADNI blood samples are being used to develop diagnostic blood tests for clinical use. These are currently not as accurate as other methods.

Prediction

Deep learning algorithms which extract the most pertinent information from MRI scans can also predict the progression of MCI patients to AD several years in advance with accuracies of greater than 90%.

Development of biomarkers for clinical trials