Deep and Frequent Phenotyping; combinatorial biomarkers for dementia experimental medicine

After years of failure the first promising signs of progress in disease modification of Alzheimer's disease (AD), arguably the disorder with the greatest unmet need relative to prevalence and cost in the developed world, are emerging. Already though it is clear that there are lessons for drug development. First, that biomarkers are essential - it is only with their use that it became apparent that 20-30% of trial participants do not have the targeted pathology. Second, that efficacy is most likely in early phases of disease including prodromal stage. Third, and most important, that the proof of concept phase for disease modification is significantly deficient using currently available approaches. It is hard, if not impossible, to conduct short term, modest-size proof of concept in prodromal disease whilst outcome measures remain long term and clinical.

The Deep and Frequent Phenotyping study is designed to rectify this by generating a biomarker set for proof of concept in prodromal AD. The objective is to build on previous studies, focussing largely on established biomarkers in annual assessment, with very deep phenotyping using established, developing and novel biomarker modalities applied frequently over a year. A pilot study demonstrated that such complex protocols can be effectively established across multiple sites and that trials participants are sufficiently committed to research for this devastating disease that they remain engaged despite taxing study demands.

Specifically, we aim to implement PET imaging and CSF biochemistry for amyloid and tau, functional and structural MRI, electrophysiology for synaptic function including EEG and MEG, measures of gait and use of remote monitoring for ecologically valid assessment of a range of phenotypes, measures of retinal pathology and a collection of bio-samples unparalleled in potential utility for molecular biomarkers and to establish stem cells for in vitro studies. Some measures such as PET will be applied at baseline and follow up, others up to every 2 months and some, such as the peripheral devices, continuously.

The study builds upon the Dementias Platform UK; we will recruit from constituent cohorts using the information workstream, will utilise the imaging workstream and will bank materials through the cells workstream. Furthermore, the programme will be nested within the IMI-European Prevention of AD combined registry, cohort and adaptive trial programme (www.ep-ad.org), the largest and the leading initiative within the rapidly growing Global Alzheimer's Platform Prevention of Alzheimer's Disease (GAP-PAD) initiative. These international public-private partnerships have committed over £100m to a linked proof of concept phase trials initiative; an initiative dependent on identifying biomarkers to speed the trials process. This proposal is designed to provide the data for such markers.

Taking the lead from ADNI and its partner studies, we will make summary data very widely available to the scientific community and work to enable access to the immense volumes of underlying primary data. Within group we will use modelling and machine learning approaches to analyse these data for markers of change in prodromal AD, combining different modalities for markers that track or improve upon advanced measures of cognition and PET measures of pathology.

The outcome will be definitive for biomarkers in this phase of disease; if such markers are achievable then this study will identify them. The deliverable will be both the data to the scientific community for wide further analysis as well as a combinatorial biomarker for use in phase II, proof of concept trials. Such a marker would speed decision making, reduce expense of clinical trials, increase the number of compounds tested at this phase. The absence of such a marker is one of the most grievous single obstacles to progress in the search for a disease modification or secondary prevention therapy for AD.

The D&FPhen study follows from a pilot phase that has established technical feasibility and participant acceptability. The study is for a repeated measures observational design in prodromal disease defined as no dementia but presence of episodic memory impairment >1.5SD from age adjusted norms in the context of AD pathology assessed using PET and CSF. Key design points include:

- Participant recruitment through pre-existing parent cohorts. Ethical approval for this process, agreement from cohorts, technical platforms and proof of numbers meeting recruitment criteria are all in place
- Utilisation of an algorithm established on ADNI data, tested in other datasets and proposed for US GAP-PAD studies to reduce screen failure based on age, cognition and APOE genotype.
- Screen by PET amyloid with entry to study on a 4:1 ratio of screen positive and negative to avoid inadvertent disclosure of genotype or biomarker data
- Repeated measures of both outcome comparator modalities (cognition and pathology) and assessment modalities between 2 months and 6 months in frequency based on projected one year change and acceptability criteria (see main section for full details)
- Protocols harmonised with EPAD where shared and established in pilot phase for structural and functional MRI, electrophysiology by EEG and MEG, optical measures including ultra wide field studies and OCT and connected devices for gait, activity and cognition (see main section for full details)
- Participant engagement including a start-up with a track record of increasing participation, a study within a study for participant engagement and burden in precision medicine studies and a process led by the Alzheimer's Society for Patient and Public Engagement (see main section for full details)
- Sharing of data with minimal delay and facilitated data-use including building analytical communities through the leading global not-for-profit organisation for scientific data sharing.

The purpose of the Deep and Frequent Phenotyping study is to provide the means to assess outcomes in early phase, proof of concept for clinical trials in AD more effectively than currently possible. Success in doing so will speed trials, reduce costs, increase productivity and ultimately contribute to more effective pipelines for early clinical drug development. This will benefit all those prosecuting such trials, including academia and the pharmaceutical industry, and those benefiting from drug development in dementia including future patients and their families.

This claim to impact is based on three pillars. The first substantial and immediate need for the trials that will utilise our outputs. The Global Alzheimer's Platform Prevention of Alzheimer's Disease (GAP-PAD) effort for trials ready cohorts and adaptive proof of concept trials, is underway in Europe with responses to invited bids for funding submitted in Canada and USA and in process in Japan. Adaptive designs mandate assessment that is swifter than clinical outcomes and currently only available in AD for amyloid based therapies (ie PET amyloid).

The second pillar is that our proposal - for combinatorial biomarkers for short term outcomes - is based on evidence. Multiple studies support the notion that combinatorial approaches add value (technically, they increases AUC of a ROC). Furthermore, this proposal will be the first to combine established biomarkers used in the context of repeated measures for short-term change with novel and innovative markers. The impact of predecessor studies such as ADNI, AIBL and AddNeuroMed has been enormous as evidenced by citations; the impact of D&FPhen with increased data from both established and novel biomarkers is likely to scale such impact.

The third pillar is the opportunity to increase trials productivity. Too many trials progress to phase III without robust proof of concept due to the inability to conduct short-term, rapid decision making studies. The cost of phase III in AD is huge and their serial failure represents the single largest threat to the development of disease modifying therapies. An intermediate, combinatorial biomarker would enable rapid decision-making at phase II, with better selection of compounds through the clinical development pipeline. Such an increase in productivity stands to have a very major impact in this therapeutic area.

The impact to patient, academic and industrial beneficiaries will be achieved firstly through the GAP-PAD consortium and secondly through open access and facilitated usage of the data. The first study within the GAP-PAD consortium to be funded is the European Prevention of AD (EPAD) with 58m Euros commitment to a trials ready cohort and with a standing, proof of concept, adaptive trial funded by Pharma partners. Critically however, the EPAD PoC trial is unable to proceed as envisaged with therapeutics other than those targeting amyloid in the absence of an intermediate phenotype of change. The D&FPhen study will read out in time for uptake by EPAD and enable transition to the standing adaptive trial to which multiple Pharma have committed. Beyond EPAD, applications are in process to NIH and CIHR in North America and these three GAP-PAD studies, all partnerships with Pharma will immediately utilise D&FPhen outputs if we are successful in our objectives.

We will not rely only on our close partners in GAP-PAD, despite the global nature of this partnership. We will make all D&FPhen data Open Access and go beyond passive sharing to active facilitation through partnership with SAGE BioNetworks. Such active sharing of the huge data volumes utilises both the technical infrastructure established by SAGE and their experience in building analytical communities. Evidence strongly suggests that the scale of impact is increased very substantially through such facilitated data sharing.