Heart failure (HF) with preserved ejection fraction (HFpEF) accounts for ~50% of all HF cases globally1, and the incidence and prevalence of HFpEF are increasing. Patients with HFpEF face a high risk for mortality, recurrent hospitalizations, and reduced quality of life and functional capacity. To date, no therapies have convincingly reduced mortality in patients with HFpEF, and few trials have shown a reduction in risk of hospitalization2.

HFpEF is therefore a strategic priority for the US National Heart, Lung, and Blood Institute (NHLBI)1. In collaboration with the Foundation for the National Institutes of Health (FNIH), the NHLBI, the US Food and Drug Administration (FDA) and multiple industry partners have developed the Accelerating Medicines Partnership in Heart Failure (AMP HF) project, with a focus on HFpEF. With $37 million in funding over 5 years, and six US-based academic medical centres and seven private-sector partners involved so far, this partnership is uniquely poised to deconstruct the heterogeneous syndrome of HFpEF in a large, diverse patient population. Here, we overview the rationale, design and deliverables for the AMP HF project, and highlight how similar initiatives could accelerate therapeutic advances for other common, heterogeneous diseases.

AMP HF rationale and design

The goal of the overall AMP programme is to bring together the resources of the US government and the private sector to improve understanding of disease pathways and facilitate better selection of therapeutic targets3,4. Managed by the FNIH, examples of AMP projects include efforts to accelerate discovery, prioritization and validation of therapeutic targets in Alzheimer’s disease, common metabolic diseases, autoimmune diseases, Parkinson’s disease and schizophrenia, and to establish bespoke gene therapies for rare diseases4. In addition to solving challenges unique to each disease, AMP projects generate valuable resources, including large repositories of data, images and biospecimens. Made broadly available to researchers, these resources are valuable for discovery of novel targets and biomarkers; validation of existing targets; classification of heterogeneous diseases into subtypes; and development of new analytical tools.

For HFpEF, disease heterogeneity is a key factor underlying the challenges in finding effective therapeutics1, and more granular classification is hypothesized to lead to improved targeted therapeutic development and disease outcomes. Prior studies have used unsupervised machine learning analyses for novel classification of HFpEF5, but are limited by previously collected, relatively low-dimensional data and lack of inclusion of raw images and multi-omics data.

The AMP HF project (via the NHLBI HeartShare programme) will use imaging, multi-omics analysis in blood and other tissues, deep clinical phenotyping, digital measurements, exercise testing and artificial intelligence methods to identify disease subtypes and novel biological pathways underlying the development and progression of HFpEF, with the goal of advancing targeted therapeutics. HeartShare is centred around a data translation centre at Northwestern University, which serves as a hub for integrating existing clinical data, images and multi-omics data with prospectively collected data and images, and for depositing results in BioData Catalyst (NHLBI’s cloud-based computing resource, which facilitates secure access to data and images). The clinical centres are developing repositories of electronic health record (EHR) data and images on patients with HF and comparator patients without HF, and will recruit individuals into a prospective registry and deep phenotyping studies.

The database developed for the first component of the AMP HF project (Supplementary Fig. 1) will include existing clinical data, multi-omics data (such as whole genome sequencing, proteomics and metabolomics data) and imaging data (such as echocardiography, cardiac magnetic resonance imaging (MRI) and computed tomography (CT) data). The second component of the AMP HF project is a prospective study in three progressive tiers (Supplementary Fig. 1). Tier 1 is a large-scale passive collection of EHR and raw image data on all patients with HF and age- and sex-matched comparator patients without HF cared for at each clinical centre since 2016. Tier 2 is a prospective, ‘low touch’ registry consisting of a subset of Tier 1 patients with HF and comparator patients without HF. Tier 2 will include eConsent and remote interaction via the Eureka platform, a direct-to-participant digital platform. It offers a wide variety of surveys for symptoms, quality of life and health status; geolocation to enhance documentation of hospitalizations; connectivity to wearables; and remote 6-minute walk test functionality. Tier 2 will also result in an engaged participant cohort, which can serve as a resource for pilot clinical trials (for example, testing wearables or other diagnostics). Tier 3 is the prospective deep phenotyping study, which will recruit up to 2,000 participants (75% with HFpEF, 25% age- and sex-matched comparators without HF).

Participants in the deep phenotyping study will undergo a comprehensive baseline assessment that includes, among other tests and surveys, cardiopulmonary exercise testing with concomitant echocardiography; cardiac MRI with assessment of myocardial fibrosis; CT of the chest, abdomen, pelvis and upper thighs for assessment of lung parenchyma and adipose and skeletal muscle composition; pulmonary function testing; arterial tonometry; laboratory testing; adipose and skeletal muscle biopsies (in a subset); and multi-omics analysis. Following the baseline examination, participants will be studied longitudinally for up to 5 years via in-person visits and remote visits through the Eureka platform. Events such as cardiovascular hospitalizations and mortality will be adjudicated centrally, while dedicated core laboratories will analyse all phenotypic tests independently.

AMP HF deliverables

The AMP HF project is designed to provide the following deliverables:

• Creation of a rich repository of clinical data, multi-omics data and images from multicentre HF trials, large-scale cardiovascular disease epidemiology and observational studies that will facilitate the investigation of HFpEF molecular mechanisms and pathophysiology

• Development of algorithms for novel, machine learning-derived HFpEF subtype identification

• Discovery and prioritization of targets in mechanistic pathways for diagnosis, risk assessment and development of new therapeutics for HFpEF

• Establishment of a large, EHR-based cohort of patients with HF who can be contacted for future remote or in-person diagnostic and therapeutic studies

• Identification and recruitment of specific patient phenotypes (either from the deep phenotyping protocol or based on automated, machine learning-based, in silico analyses of EHR data and images collected in HeartShare) to support target validation and proof-of-concept precision medicine trials for HFpEF.

In addition (via HeartShare), the project is training a cohort of new HF-proficient data scientists in machine learning, artificial intelligence and multi-omics, and will develop a curriculum that could be used by industry partners to train data scientists in their organizations.

Relevance to other heterogeneous diseases

Like HFpEF, many common diseases such as autoimmune diseases, neurodegenerative diseases and psychiatric disorders are inadequately defined, heterogeneous clinical syndromes that have limited treatment options or limited treatment efficacy. In theory, precision medicine could improve targeting of the right treatment to the right patient at the right time. In reality, such initiatives require massive data and computing resources to facilitate identification of disease subtypes and molecular or pathophysiological targets3. Furthermore, access to diseased organ tissue is often difficult, and the combination of clinical data, images, and multi-omics data from blood and urine needed to gain insights into disease subtypes, pathophysiology and molecular mechanisms is rarely pursued. The AMP HF project, with its large-scale curation of accessible datasets, remote patient engagement and prospective in-person deep phenotyping, can serve as an exemplar to accelerate drug discovery and advance precision medicine for other heterogeneous diseases.

Acknowledgements

The authors would like to thank those colleagues who dedicated 9 months to the Design Phase to shape the AMP HF, with adept facilitation by L. Nguyen of FNIH. Members comprise leadership at the NHLBI, FDA and industry (Abbott, Amgen, Inc., AstraZeneca, Bayer AG, Boehringer Ingelheim Pharmaceuticals & Eli Lilly Alliance, Bristol Myers Squibb, Cytokinetics, Inc., Eli Lilly and Company, Ionis Pharmaceuticals, Inc., Novartis AG and Novo Nordisk), as well as academic advisors. D. Goff from NHLBI provided valuable leadership support throughout the design phase. The authors would further like to recognize, in memoriam, Robb Kociol, who served in the leadership role and whose insights in helping to shape this programme will be an indelible part of his legacy. The Design Phase work was made possible with funding from the industry partners and from the FNIH Charles A. Sanders Lagacy Fund.

Disclaimer

The content of this manuscript is solely the responsibility of the authors and does not necessarily reflect the official views of the NHLBI, NIH or the United States Department of Health and Human Services.

Competing Interests

S.J.S. has received research grants from the National Institutes of Health (U54 HL160273, R01 HL107577, R01 HL127028, R01 HL140731, R01 HL149423), Actelion, AstraZeneca, Corvia, Novartis, and Pfizer; and has received consulting fees from Abbott, Actelion, AstraZeneca, Amgen, Aria CV, Axon Therapies, Bayer, Boehringer-Ingelheim, Boston Scientific, Bristol-Myers Squib, Cardiora, Coridea, CVRx, Cyclerion, Cytokinetics, Edwards Lifesciences, Eidos, Eisai, Imara, Impulse Dynamics, Intellia, Ionis, Ironwood, Lilly, Merck, MyoKardia, Novartis, Novo Nordisk, Pfizer, Prothena, Regeneron, Rivus, Sanofi, Shifamed, Tenax, Tenaya, and United Therapeutics. J.B. has received consulting fees from Abbott, Adrenomed, Amgen, Array, AstraZeneca, Bayer, Boehringer Ingelheim, Bristol Myers Squibb, CVRx, G3 Pharmaceutical, Impulse Dynamics, Innolife, Janssen, LivaNova, Luitpold, Medtronic, Merck, Novartis, Novo Nordisk, Roche, and Vifor. S.H.S. has received research funding through sponsored research agreements between AstraZeneca and Verily, and Duke University. The other authors have no relevant disclosures.