Prevention of Alzheimer's Disease — What Will It Take?

Resources

The global burden of Alzheimer's disease

Carrillo MC, Thies W, Bain LJ. The global impact of Alzheimer's disease. In: Hampel H, Carrillo M, eds. Alzheimer's Disease: Modernizing Concept, Biological Diagnosis and Therapy (Adv Biol Psychiatry). Alzheimer's Association. Basel, Karger; 2012;(28):1-14.

Prince M, Bryce R, Albanese E, et al. The global prevalence of dementia: a systematic review and metaanalysis. Alzheimers Dement. 2013;9(1):63-75.

Reitz C, Brayne C, Mayeux R. Epidemiology of Alzheimer disease. Nat Rev Neurol. 2011;7(3):137-52.

Wimo A, Jonsson L, Bond J, et al. The worldwide economic impact of dementia 2010. Alzheimers Dement. 2013;9(1):1-11.

The pathogenic mechanisms involved in Alzheimer's disease

Hampel H, Lista S. Alzheimer disease: From inherited to sporadic AD — crossing the biomarker bridge. Nat Rev Neurol. 2012;8(11):598-600.

Jack CR Jr, Knopman DS, Jagust WJ, et al. Hypothetical model of dynamic biomarkers of the Alzheimer's pathological cascade. Lancet Neurol. 2010;9(1):119-128.

Jack CR Jr, Knopman DS, Jagust WJ, et al. Tracking pathophysiological processes in Alzheimer's disease: an updated hypothetical model of dynamic biomarkers. Lancet Neurol. 2013;12(2):207-16.

Efforts to prevent Alzheimer's disease

Carrillo MC, Brashear HR, Logovinsky V, et al. Can we prevent Alzheimer's disease? Secondary "prevention" trials in Alzheimer's disease. Alzheimers Dement. 2013;9(2):123-131.

Langbaum JB, Fleisher AS, Chen K, et al. Ushering in the study and treatment of preclinical Alzheimer disease. Nat Rev Neurol. 2013;9(7):271-81.

Sperling RA, Karlawish J, Johnson KA. Preclinical Alzheimer disease — the challenges ahead. Nat Rev Neurol. 2013;9(1):54-8.

Adaptive trials for cancer and other diseases

Barker AD, Sigman CC, Kelloff GJ, et al. I-SPY 2: an adaptive breast cancer trial design in the setting of neoadjuvant chemotherapy. Clin Pharmacol Ther. 2009;86(1):97-100.

Berry DA. Adaptive clinical trials in oncology. Nat Rev Clin Oncol. 2011;9(4):199-207.

Esserman LJ, Woodcock J. Accelerating identification and regulatory approval of investigational cancer drugs. JAMA. 2011;306(23):2608-2609.

Meurer WJ, Lewis RJ, Berry DA. Adaptive clinical trials. A partial remedy for therapeutic misconception? JAMA. 2012;307(22):2377-8.

Challenges in therapy development for Alzheimer's disease

Herrup K, Carrillo MC, Schenk D, et al. Beyond amyloid: Getting real about nonamyloid targets in Alzheimer's disease. Alzheimers Dement. 2013;9(4):452-458.

Sperling RA, Jack CR Jr, Aisen PS. Testing the right target and right drug at the right stage. Sci Transl Med. 2011;3(111):111-133.

Vellas B, Carrillo MC, Sampaio C, et al. Designing drug trials for Alzheimer's disease: What we have learned from the release of phase III antibody trials: A report from the EU/US/CTAD Task Force. Alzheimers Dement. 2013;9(4):438-44.

Advocacy, research, and educational organizations for Alzheimer's disease

Accelerate Cure/Treatments for Alzheimer's Disease (ACT-AD)

Alzheimer's Association

Alzheimer's Disease International

Alzheimer's Foundation of America

Alzheimer's Society (United Kingdom)

Critical Path Institute: Coalition Against Major Diseases (CAMD)

Cure Alzheimer's FUND

USAgainstAlzheimer's

Sponsors

Alzheimer's disease (AD) and related dementias affect more than 30 million people worldwide. This number is expected to triple by 2050 unless new treatments can prevent or delay the onset of the disease. Despite enormous expenditures by the pharmaceutical industry, clinical trials for disease-modifying drugs have been disappointing. There is a need for more efficient and cost-effective drug development. Moreover, the complexity and heterogeneity of AD demands not only innovative treatments but also novel trial designs. AD trials must address two particular challenges: the disease progresses over many years, and multiple biological pathways affect neurodegeneration and dementia. Thus, researchers must determine when to intervene and which mechanism to target. This conference focused on an adaptive trial design that approaches many of these questions simultaneously.

Alzheimer's disease occurs over many years, with neurodegeneration leading to a progressive decline in cognition that begins long before dementia is diagnosed. (Image courtesy of Reisa A. Sperling)

There is an emerging consensus that AD treatment should involve early intervention before significant neurodegeneration has occurred. However, detecting clinical changes at this stage is difficult. Biomarkers may offer a solution, and three preventive trials currently underway are relying heavily on biomarkers as indicators of treatment efficacy. The Alzheimer's Prevention Initiative (API) comprises two complementary treatment and biomarker trials in cognitively normal individuals at high genetic risk of developing AD. The first trial will enroll individuals with a deterministic autosomal dominant mutation, which ensures development of the disease; the second trial will enroll individuals with two copies of the ε4 variant of the Apolipoprotein E gene (APOEε4), which increases the risk of AD by about 12-fold compared to those with no copies of the ε4 variant. The Dominantly Inherited Alzheimer's Network (DIAN) trial will test three different anti-amyloid drugs in individuals with an autosomal dominant mutation, primarily using biomarker endpoints. The Anti-Amyloid Treatment in Asymptomatic AD (A4) trial will test treatment strategies in cognitively normal adults who have early biomarker evidence of the pathogenic process that leads to AD.

These trials are predicated on a hypothetical model of biomarker variation over the course of the disease and should provide additional data to confirm or refute this model. However, completed studies of putative disease-modifying therapies have failed to show a correlation between biomarkers and clinical outcomes and suggest that markers of target engagement may fail to predict an effect on brain function. Thus, we need better biomarkers that are optimized in clinical trials, rather than in observational studies designed to clarify the natural history of the disease, as well as more sensitive cognitive measures that track decline throughout the disease.

Adaptive trials could help to determine whether and when (at which stages of disease) new treatments are effective and to identify theragnostic biomarkers, which track the biochemical effects of a drug. Speakers at the workshop aimed to build a plan for an adaptive trial for AD, following the model of the I-SPY 2 Trial, which is simultaneously testing several neoadjuvant treatments for breast cancer. Co-principal investigators for I-SPY 2, Laura J. Esserman from the University of California, San Francisco, and Donald A. Berry from the University of Texas MD Anderson Cancer Center, described its development and outcomes, which could inform a trial for AD. Reisa A. Sperling from Harvard Medical School then outlined challenges that are specific to AD. After this overview, participants broke into small working groups to discuss how to develop an adaptive trial for AD, how such a trial could test combination therapies, and how to support the study through partnerships. The workshop ended with the creation of a preliminary plan and the establishment of a steering committee to guide research.

Speakers:
Donald A. Berry, University of Texas MD Anderson Cancer Center
Laura J. Esserman, University of California, San Francisco

Highlights

• Adaptive designs allow clinical trials to be tailored to individual patients, increasing the likelihood of demonstrating a treatment effect and reducing patient exposure to ineffective treatments.
• Adaptive trials are especially useful for heterogeneous diseases like Alzheimer's disease.
• As many as eight drugs can be tested simultaneously in one trial with a single control group, increasing the efficiency and lowering the cost of drug development.

The I-SPY 2 model and its application to other diseases

The I-SPY 2 Trial (investigation of serial studies to predict your therapeutic response with imaging and molecular analysis) is a collaborative study that uses an adaptive design to test novel neoadjuvant treatments for metastatic breast cancer. It allows researchers to tailor the investigative treatment according to patient and tumor characteristics, relying heavily on biomarkers to measure therapeutic responses. Laura J. Esserman from the University of California, San Francisco, described the trial as efficient and effective, with an independent, patient-centered network that integrates the clinical trial process with clinical care. Its adaptive approach allows investigators to learn from experience and modify the trial as it progresses.

The overall goal is to identify and market effective treatments more quickly by establishing a standing network of clinical trial sites that can graduate investigative treatments to small phase III studies, drop ineffective treatments, and add new ones in response to patient outcomes. In addition to identifying the most effective treatments for subgroups of patients, the I-SPY 2 design will test and validate biomarkers and provide information about how these markers are affected by each drug.

The I-SPY 2 team developed a model in which clinical care and clinical trial processes are integrated with treatment evaluation. This core work is supported by a structural branch providing operational infrastructure and real-time data, with oversight and guidance provided by regulatory agencies and a precompetitive consortium.

The I-SPY 2 Trial adopted a process model to accelerate knowledge acquisition over the course of the trial. Patients take part in core processes (yellow) related to the specific disease and trial. Operational infrastructure (green) supports these processes. A precompetitive consortium facilitates collaboration among multiple companies and regulatory agencies. (Image courtesy of Laura J. Esserman)

There are 20 sites and approximately 400 patients enrolled in the trial, which is preparing to test multiple investigational agents. According to Esserman, developing the infrastructure for the trial was an arduous task, but she believes the investment of time in the early stages will sustain the trial over the long term and provide essential support to keep operating sites that are capable of enrolling large numbers of subjects. Biomarkers will be used to stratify patients, who will be adaptively randomized to different treatment arms based on the results of previous studies. Medical imaging will be used to assess pathological complete response (pCR), defined as no evidence of breast cancer, which the U.S. Food and Drug Administration (FDA) has accepted as an endpoint.

Applying lessons from I-SPY 2 to Alzheimer's clinical trials

Similarities between breast cancer and AD suggest that an adaptive I-SPY 2-like design might be useful in AD trials, according to Donald A. Berry from the University of Texas MD Anderson Cancer Center. Both diseases are extremely heterogeneous, with multiple subtypes. This heterogeneity has led to disproportionate outcomes in breast cancer research: while many effective therapies have been developed for non-metastatic breast cancer, few trials are underway for the more complex and poorly understood metastatic forms of the disease, resulting in a sense of desperation about the lack of effective treatments. The traditional approach to clinical trials has proved very costly for research into both AD and cancer, requiring too many patients and exposing many to drugs from which they are unlikely to benefit. Adaptive trials, by contrast, explicitly consider the heterogeneity of the disease by randomizing patients adaptively, using biomarkers to increase the probability of success.

In I-SPY 2, preliminary simulations identify characteristics of the trial such as error rate, power, and sample size. These simulations are computationally intensive and require data from longitudinal biomarker studies, but they help institutional review boards, investigators, and regulatory agencies to understand the trial design. In addition, as many as eight arms can be tested simultaneously in small phase II trials, with one arm representing a shared control group. Patients are classified into subtypes and placed in designated arms according to biomarker signatures. Successful drug/biomarker pairs graduate to focused phase III studies based on Bayesian predictive probabilities.

The I-SPY2 Trial can start with as many as 8 arms, one representing the control (standard therapy). Depending on whether interim results indicate a benefit for patients, an arm may graduate to a small, focused phase III trial or be dropped for futility. Additional arms can be added, including arms combining drugs in other arms. (Image courtesy of Donald A. Berry)

The I-SPY 2 approach offers numerous advantages. It reduces costs, with a common control and common infrastructure across multiple arms, smaller focused studies, and faster determination of whether a drug will graduate or be dropped. Patient-centered trials match patients with drugs they are likely to benefit from and reduce patient exposure to drugs from which they are unlikely to benefit. Researchers can learn from experience as the trial progresses, maximizing the likelihood of a positive outcome, and are able to answer multiple questions simultaneously.

Chairs:
Reisa A. Sperling, Harvard Medical School
Michael Krams, Janssen Pharmaceuticals
 
Speakers:
Michael Poole, AstraZeneca
Simon Lovestone, King's College London, UK
Robert A. Stern, Boston University School of Medicine
Michael T. Ropacki, Janssen Alzheimer Immunotherapy

Highlights

• There are many challenges to building an adaptive trial for Alzheimer's disease, such as the lack of defined biomarkers for AD and the need for blinded trials.
• New biomarkers and other outcome measures are needed to provide readouts in a short timeframe.
• Large numbers of individuals will be needed to conduct the trials efficiently.

Demonstrating clinical efficacy—biomarkers and clinical/cognitive markers

Reisa A. Sperling from Harvard Medical School outlined the challenges that may make it more difficult to implement an adaptive trial for AD than for breast cancer. Unlike AD, breast cancer can be identified through tissue diagnosis, validated biomarkers, and known subgroups. I-SPY 2 is run as an unblinded trial, making it more attractive for patients. However, blinding will almost certainly be required in an AD trial since there are no clearly defined biomarker signatures to characterize subgroups and huge placebo effects have been observed in previous studies.

Despite these difficulties, Michael Poole from AstraZeneca expressed optimism about the adaptive trial approach, noting that only ten years ago it was virtually unknown in the pharmaceutical industry but now almost all trials have some sort of adaptation. A series of working groups examined the challenges specific to an AD trial—demonstrating clinical efficacy and selecting the appropriate patient population—as well as the potential for using this structure to test combination therapies.

An adaptive design requires readouts in a short timeframe so that treatments can be modified throughout the trial; however, neither available biomarkers nor cognitive measures for AD are sufficiently sensitive to provide early evidence of a treatment effect. In addition, the trajectory of AD demands different treatment approaches and different types of markers at different stages of the disease.

Different biomarkers and cognitive measures are useful at different stages of AD. For example, increased amyloid deposition (dark red line) signals initiation of the pathogenic process, whereas cognition (the purple line) does not become abnormal until the late stages of disease. The other biomarkers depicted in this figure are synaptic function (yellow), neurodegeneration (turquoise), brain structure (blue), and clinical function (brown). (Image courtesy of Reisa A. Sperling)

Simon Lovestone from King's College London outlined the different types of biomarkers that are needed for an adaptive trial: stratification markers indicate the stage of the disease; progression markers show progression over a short timeframe and enable researchers to modify treatments.

Existing biomarkers support a hypothetical model of AD development. This model, which has some confirmation in empirical data, posits that AD begins long before symptoms appear, starting with abnormal cleavage of the Amyloid Precursor Protein (APP), which leads to aggregation and accumulation of the β-amyloid peptide (Aβ) in the brain. This peptide triggers a cascade of poorly defined processes that cause neuronal dysfunction, neurodegeneration, and, eventually, clinical symptoms. Aβ deposition in the brain can be assessed using positron emission tomography (PET) imaging with an amyloid ligand or by measuring levels of the Aβ protein in cerebrospinal fluid (CSF). Neurodegeneration biomarkers include CSF levels of the proteins tau and phosphorylated tau (phospho-tau), structural magnetic resonance imaging (MRI) detecting brain atrophy, and fluorodeoxyglucose PET (FDG-PET) imaging of metabolic activity.

Amyloid PET imaging and CSF Aβ measurement provide highly specific stratification biomarkers, but the invasiveness and high cost of PET scans and lumbar punctures limits the feasibility of using these tests on large numbers of individuals. Thus, a highly sensitive, low-intervention marker—such as a blood-based biomarker—is needed to begin to differentiate individuals. A smaller number of people could later be tested with intermediate or high-intervention markers to provide more specific stratification.

Sufficiently sensitive progression biomarkers are not yet available, but many are under investigation. These include electrophysiological markers such as electroencephalography (EEG), event-related potentials (ERP), and magnetoencephalography (MEG), as well as imaging markers such as FDG-PET, which assess regional metabolism, and several MRI measures.

Cognitive measures of disease progression are the most clinically relevant for AD patients, yet tests of cognition, including the widely used ADAS-cog (Alzheimer's Disease Assessment Scale-cognitive subscale), lack sufficient sensitivity to distinguish normal aging from the early stages of AD. In addition, rates of decline on the ADAS-cog vary at all stages of the disease, and there is no agreement on what amount of change is clinically meaningful. Thus the test is a poor progression marker.

According to Robert A. Stern from Boston University School of Medicine, we need new neuropsychological measures (which test cognition as well as behavior, motor skills, linguistic ability, and executive function), particularly to assess episodic memory, as well as new self and informant measures of cognitive and functional concerns. Self and informant measures are thought to be particularly useful in the early stages of the disease, when patients maintain an adequate level of awareness to provide meaningful data; however, these measures are subjective and thus are prone to high levels of variability. Developing better neuropsychological measures may involve applying newer psychometric approaches to existing data; using more sensitive tests of episodic memory, executive function, and processing speed; and creating composite and/or computerized measures. For example, the A4 study has developed a preclinical AD cognitive composite (ADCS-PACC), which combines multiple cognitive tests into one composite measure, and will be using rate of decline on this composite as a primary outcome measure.

Selecting the right patient population

Identifying and enrolling patients for an adaptive trial will require a completely different approach from that taken in previous studies, according to Michael T. Ropacki from Janssen Alzheimer Immunotherapy. In traditional trials, potential subjects are selected based on a clinical diagnosis of AD, assessed through a variety of measures. This produces a heterogeneous sample in which each person is treated equally despite their differences. As a result, test results are highly variable for reasons unrelated to the trait they intend to test, mean scores are not informative, and the trial fails to meet predetermined endpoints. Moreover, in an attempt to select enriched populations for trials, those who do not meet certain predetermined criteria are lost as "screen failures," resulting in lengthy enrollment periods at dozens of sites and thus prolonged, expensive, and inefficient clinical trials.

In contrast, future trials—especially adaptive trials—will require objective evidence-based diagnoses of AD using psychometrically validated tests, rather than relying on the more subjective clinical diagnosis. Performance on memory tests will be stratified based on premorbid intellectual functioning, estimated using demographic, historical, and performance data, to enable more accurate determination of disease stage. Screen failures, which Ropacki called lost opportunities, will not occur.

Ropacki proposed a Registry Recycling Model. In this model, a registry of hundreds of thousands of unselected volunteers is established, with only minimal data collected on each person. From that registry, an observational cohort study is created using selection criteria determined by the needs of the study. For example, the observational cohort might be based on age or a family history of dementia. Participants are then followed until disease progression begins, at which time they are considered ready to enroll in various clinical studies. Readiness groups could also be created based on characteristics such as genetic markers and other risk factors that have been discovered in the observational study. For instance, a readiness group could be composed of individuals with Down syndrome, about 70% of whom will develop AD by age 50. The observational cohort study will also provide opportunities to test and validate biomarkers, neuropsychological measures, composites, and other measures to identify precise progression predictors. Data collected may also be used in a subsequent clinical trial as "run-in" data to provide information about disease progression before treatment begins. Importantly, there are no screen failures in this model: individuals who have not progressed enough to be enrolled in a trial are simply "recycled" to the observational cohort.

In the Registry Recycling Model, individuals may be enrolled in an observational cohort study, where they are followed with biomarkers, neuropsychological tests, and other measures to track the natural history of the disease. Clinical studies can then draw from this cohort to create a population of well-characterized subjects, shortening the enrollment period and increasing both the efficiency and efficacy of trials. (Image courtesy of Michael T. Ropacki)

Participants at the workshop agreed that it would be valuable to leverage registries that have already been established, such as the Alzheimer's Prevention Registry (APR) or the United Kingdom BioBank. To begin an adaptive trial, Esserman suggested identifying a group that is likely to progress rapidly, while continuing to follow other people in the registry. A genetic signature could identify this group, or it might be possible to use run-in data collected in the observational cohort.

Chairs:
Samantha L. Budd, AstraZeneca
Jeffrey Sevigny, Biogen Idec

Highlights

• Combinations of drugs may be required to achieve disease modification in Alzheimer's disease.
• Combination therapy could target multiple pathways or multiple points within a single pathway.
• Adaptive designs are ideal for testing combination therapies.

Rationale for combination therapy for AD

One advantage of the I-SPY 2 design is that multiple targets and molecules can be tested simultaneously, either as monotherapies or in combination. In fact, given the complex pathogenesis of AD, there are many who believe that combination therapies will be required to achieve disease modification. Combination therapy was itself born of the need to find effective therapies for other complex diseases such as cancer and HIV/AIDS.

Samantha L. Budd from AstraZeneca provided an overview of the mechanisms of AD pathogenesis that have been proposed, as well as the targets and classes of drugs that emerge as potential therapeutics. Amyloid and tau, and their role in the plaques and tangles in the AD brain, are the most widely investigated components of AD. In addition, or possibly as a consequence of amyloid and tau deposition, other states such as neuroinflammation and synaptic dysfunction contribute to pathology, resulting in neurodegeneration and cognitive deficits.

Interfering with the deposition of amyloid in the brain could be accomplished by targeting the production, removal, aggregation, distribution, and/or degradation of the Aβ peptide; for each of these mechanisms a number of molecules have been identified that interfere with the target. For example, Aβ is produced through the action of enzymes called secretases, and inhibitors of alpha, beta, and gamma secretases have been investigated as possible AD treatments. The tau pathway is somewhat less well understood but no less complicated. Tau accumulates as neurofibrillary tangles in the brain that are thought to be the primary driver of neurodegeneration. Tau pathology can be caused by microtubule destabilization, aggregation, phosphorylation, or abnormal spreading. Targeting phosphorylated tau alone is complicated, as there are 85 potential phosphorylation sites on the human tau protein and 29 abnormally phosphorylated sites on tau have been identified in the AD brain.

Most clinical studies in AD target Aβ, and the agents tested in these studies are often aimed at several different goals, such as treating symptoms, modifying disease pathophysiology, preventing disease by addressing risk factors, and promoting synaptic plasticity or neurogenesis.

Treating and/or preventing AD will probably involve targeting multiple aspects of the disease. Therapies that are currently available only treat the symptoms; however, emerging drug-development programs are focusing on disease modification and prevention. (Image courtesy of Samantha L. Budd)

Combination therapies could target multiple proteins, such as amyloid and tau, or multiple stages along a single pathway, such as the production or removal of amyloid. Adaptive trials could enroll patients across many disease stages simultaneously, which could help determine when different treatments and drug combinations are effective. Indeed, combinations may be particularly useful in later stage disease when multiple pathogenic mechanisms are evident. While there is significant theoretical interest in tackling multiple targets and disease stages together, a more practical approach might be to combine drugs that are available now or are in the late stages of development. These available drugs, for the most part, target Aβ; thus, several participants suggested that a first test of combination therapy could combine a beta-secretase (BACE) inhibitor (which interferes with production of Aβ) with a monoclonal antibody against Aβ (which facilitates removal of Aβ). Combinations can be accommodated by an I-SPY 2-like design, with a factorial design embedded into the trial.

FDA framework for combination trials

In 2010 the FDA issued draft guidelines for the co-development of two or more new (unmarketed) investigational drugs for use in combination trials. Jeffrey Sevigny from Biogen Idec summarized the FDA guidance, which provides a framework for selecting drugs based on the stage of development for each drug.

The criteria established by I-SPY 2 for drug selection could easily be adapted for a similar trial in AD. These criteria, adapted for AD, specify that selected drugs should have completed phase I testing for safety. Each drug should have already been shown to be efficacious in the treatment of AD, or a rationale for using it as a treatment for AD should be established. Each drug should target key pathways or proteins that contribute to AD pathogenesis and should fit into a strategic model for optimizing combinations of drugs targeting single or multiple molecular pathways. And there should be sufficient quantities of each drug available to complete the trial.

Summary of the FDA Guidance on the co-development of two or more unmarketed investigational drugs for use in combinations. (Image courtesy of Jeffrey Sevigny)

Chair:
Chas Bountra, University of Oxford, UK
 
Speakers:
Celia Merzbacher, Semiconductor Research Corporation
Laura J. Esserman, University of California, San Francisco

Highlights

• Consortia such as the one built for I-SPY 2 enable a group of companies to share both the risks and the benefits of drug discovery.
• Consortia allow companies to share data and assets while protecting intellectual property.
• New models are needed in medicine to combine clinical trials with clinical care.

Creating a consortium

I-SPY 2 was built as a public–private partnership using a consortium model, which allowed the parties involved to share risks and benefits, create a continuous standing trial, and develop novel funding sources. The Foundation for the NIH (fNIH), an independent nonprofit organization established to support the mission of the NIH, brought together investigators from academia, industry, the FDA, and the National Cancer Institute to build the I-SPY 2 infrastructure. Five companies collaborate in I-SPY 2, contributing assets and sharing data, with the caveat that each company has six months' exclusive use of the data it contributes. All the companies gain from new knowledge and no single one bears the entire risk (cost) of drug development.

Several issues would need to be resolved in order to implement an I-SPY 2 trial model for AD, including: defining the ownership and protection of intellectual property (IP), establishing data standards that allow data to be pooled from multiple studies and companies, crafting rules for data sharing that all participants agree to follow, and determining the characteristics of the placebo control group (that is, dosing, drug delivery method, etc.) so that a single group could serve as a control for all experimental arms. In addition, participants would need to establish rules of governance, determine the distribution of responsibilities for implementing the project, and identify sources of funding to guarantee the completion of the early precompetitive stages of the trial.

Open-access research

Other partnerships have demonstrated similar benefits through precompetitive collaboration. Chas Bountra from the University of Oxford described the Structural Genomics Consortium (SGC), a public–private partnership established to generate the structures of novel human proteins and to make them freely available to researchers. SGC was built on the concept of open-access research, requiring participants to share molecules generated through the partnership with no claims of intellectual property and to immediately publish both positive and negative research results. Eight companies and 200 academic labs have joined the consortium, speeding dissemination of knowledge and resulting in the discovery of many new novel human targets (about two are published per week) and at least one spin-off company. The public have unrestricted access to protein structures that are established, but companies are free to initiate proprietary efforts afterwards.

Bountra and his colleagues are building another public–private partnership with a similar philosophy. It will advance new candidates to phase II trials in order to identify molecules that fail—and will publish results immediately to reduce duplication and to prevent needless patient exposure. Through this precompetitive process, the partnership hopes to identify and reduce the risk (expense) to individual companies of developing compounds that show clinical potential. Patient groups have expressed interest in helping with participant recruitment and regulators have offered to provide input on study design and biomarker validation.

The SGC model is designed to share information from both successful and unsuccessful POCM (proof of concept/mechanism) studies in order to prevent duplication of negative studies and to advance the development of targets and molecules that show promise. (Image courtesy of Chas Bountra)

Lessons from the semiconductor industry

In the early 1980s the semiconductor industry was confronted with dwindling federal investment in research at a time when it badly needed innovation to sustain advances. The industry responded by establishing the Semiconductor Research Corporation (SRC), an independent, not-for-profit industry-driven organization that supports precompetitive academic research to drive innovation. Since 1982, SRC participants have invested more than $1.8 billion to support more than 10 000 students, 2000 faculty, and 250 universities in 27 countries.

SRC adopted an intellectual property (IP) model in which all member companies have rights to all the IP that results from the research. Celia Merzbacher from the SRC explained that this is feasible in the semiconductor industry because every device has so many components, each with multiple patents, that companies are willing to share IP in some areas, while they compete in others. SRC pays to prosecute a patent if the members decide they want to protect an invention after it is disclosed, but ownership remains with the university that developed it and SRC maintains a non-exclusive royalty-free license for its members.

SRC sets the direction of research, with input from members and multi-level strategic advisory boards that aim to align projects with industry needs. Members make significant investments in the consortium and benefit through access to information about and results from projects, relationships with university researchers, and connections with other members of the consortium.

The semiconductor industry has benefited from an industry-wide roadmap, the International Technology Roadmap for Semiconductors, which has identified technology barriers to advancement in semiconductor design. The industry has evaluated the general criteria to warrant and support such a roadmap, which are shown below.

Building a successful roadmap to align research projects with industry needs requires not only identification of and consensus on the overall goals of the program but also agreement among participants on the scope of the project and rules of engagement. (Image courtesy of Celia Merzbacher)

The right model for an I-SPY for AD project

Laura J. Esserman commented on how a successful business model could be built to support adaptive trials. She said that the venture capital sector has become risk averse and unwilling to invest in infrastructure development; therefore, new models of social venture philanthropy and investing are needed for medical research. In particular, clinical care and clinical research should be merged so that they can share infrastructure and data. Institutional review boards and patient consent forms should be revamped so that collecting data to advance knowledge in the field becomes a routine part of care. Common standards and common tools are needed, along with new structures to support standing trials and new business models that enable efficient data acquisition in real time.

Chairs:
Simon Lovestone, King's College London, UK
William C. Mobley, University of California, San Diego
 
Speakers:
Roger Bullock, Kingshill Research Centre, UK
Reisa A. Sperling, Harvard Medical School
Donald A. Berry, University of Texas MD Anderson Cancer Center

Highlights

• Participants at the workshop agreed to build a roadmap toward the development of an adaptive trial for Alzheimer's disease.
• A Steering Committee was established to guide development of the roadmap, to engage partners across all sectors, and to raise funds for the planning and early-stage development of an adaptive trial network.

The road ahead: future steps for developing an AD adaptive trial

The second day of the workshop was devoted to creating a roadmap for the development of effective therapies for AD. Roger Bullock from Kingshill Research Centre started off the discussion by outlining the current status of AD therapy and the dilemmas that have hindered efforts to identify disease-modifying drugs; in particular, lack of clarity about how to select appropriate subjects, methodology, and outcome measures. Other fields have faced similar crises and responded with differing degrees of success. In the HIV/AIDS field, for example, a collective approach has been critical to recent advances. The AD field needs that kind of collaborative approach as well.

Participants highlighted the urgent need to move beyond current drug-development models in order to evaluate potential interventions more quickly and cheaply. Building on existing infrastructure will help to facilitate rapid and sustained progress. It will also be necessary to engage multiple international partners to ensure the global reach of the project. Rapid dissemination of knowledge obtained in the study should expand the reach of the project and encourage participation of partners from all sectors, especially as the benefits for various stakeholder groups become more clearly articulated.

However, participants diverged over how to achieve the greatest benefits for drug-development companies; that is, whether the field is ready to move forward with an adaptive trial or should instead concentrate on broader goals such as shared trials and shared recruitment. Many agreed that an adaptive design is ideal for AD drug development, given the large number of unanswered questions, multiple targets, and heterogeneity of the disease. Moreover, building a collaborative adaptive trial would yield many secondary benefits—especially progress on biomarkers—that would open new paths for drug development and demonstrate that the field has aligned in taking a new and different approach. Achieving consensus on outcome measures and biomarkers would also help to build the case for regulatory approval.

Although there are many thorny issues to be tackled, particularly with regard to standardization, data sharing, and protection of IP, a path forward is evident in efforts that are already underway. ADNI, API, DIAN, and A4 have solved many of the problems, reaching agreements in which participants work collaboratively and share data and companies retain control of IP. These projects have demonstrated the willingness of many companies to work together for collective gains. The I-SPY 2 investigators pointed out that complete buy-in is not necessary to begin building a foundation for an adaptive trial; as the value of the project increases, it is likely that more companies will wish to join.

Guided by models of other consortia presented at the workshop, participants agreed that a roadmap should be created to guide the adaptive trial project. The meeting established a steering committee to craft this map, beginning with a mission statement to direct research activities. The committee will develop a common assessment platform and a network of sites built on existing infrastructure.

Rules of engagement and the specifics of the roadmap will need to be defined in the coming days and weeks through a series of subcommittees. These subcommittees will: create a new registry, or build onto existing registries, and establish a mechanism to funnel participants from the registry to a readiness cohort; establish or identify a network of clinical trial sites and develop rules of engagement for these sites, including a plan to expedite institutional review board approval; identify appropriate assessment tools that will enable the design of an adaptive trial for AD, building on existing imaging, biomarker, and cognitive assessment tools; work with statisticians and modeling/simulation experts to design an adaptive trial for AD; establish a selection committee, based on models provided by I-SPY 2, DIAN, and A4, to identify assets that are most appropriate for an adaptive trial and/or combination therapy; establish a framework for negotiating agreements regarding sharing of drugs, protection of IP, and data sharing; work with regulatory agencies to ensure that the trial design is compatible with a regulatory path to approval; and secure funding for planning and early-phase work.

Who should drive the effort to build a novel, possibly adaptive trial design to test AD therapies?
 
Should this effort be driven by academia, industry, non-profit/patient advocacy organizations, or a combination of these groups?

Which group of patients should be the focus of this new drug development effort?
 
Should prevention efforts be aimed at individuals with mild cognitive impairment (MCI), or at those who are presymptomatic? Or is it possible to build a platform that can accommodate multiple populations?
 
When is the right time to intervene?
 
What are the right measures, particularly in early stages of disease?

How much of the necessary infrastructure is already available, and can we incorporate an adaptive trial design into existing infrastructure?
 
Should Contract Research Organizations (CROs) be used to implement the trials and collect data, or would it be more efficient to build a new and more nimble trial network and data system?