NordPerMed: PM-Heart - Precision Diagnostics and Predictions in Ischemic Heart Disease including Identification of Over-Treated Patients

Globally, the leading cause of lost years of life, is Ischemic Heart Disease (IHD). In the EU 13.2 mio. patients are diagnosed with IHD, 700,000 live in the Nordic countries. IHD causes chest pain, myocardial infarctions, reduced physical capacity and reduces life-expectance. IHD is not caused by a single mechanism. Many risk factors and disease mechanisms are known, but to drive precise sub-classifications and risk stratification of IHD we need to manage complex data. At present, patients with IHD are generally diagnosed and treated using one-size-fits-all standard regimes. This leads to inefficient, costly, potentially harmful over-management of the disease. At the same time, patients with residual high risk are not identified and treatment not optimized or compliance re-enforced. The objective of this project has been to develop and clinically implement personalized medicine (PM) with the dual purpose of avoiding futile overtreatment as well as under-treatment in IHD. In this Nordic interdisciplinary collaboration (Denmark, Norway, Island), we have established and merged large Nordic cohorts with well-described IHD genotypes and phenotypes, by combining existing data as well as recruiting new data, e.g. from electronic health records (EHR). Echocardiography data from a large cohort of patient treated at Oslo University Hospital (OUS) were extracted and anonymised, and machine learning (ML) algorithms were developed and trained on a secure off-line platform in order to automatically measure anatomical and physiological parameters on a large scale. The aim was to employ the OUS-developed ML algorithms on a large Danish echocardiography database, but technical and regulatory issues prevented to achieve this goal before the end of employment for the relevant researchers. Overall, the purpose of the project has been to differentiate between different subgroups of IHD, and from the deep phenotypic characterization, identify each patient's cause of IHD. Using a machine learning approach, we have created a clinical integrative IHD algorithm, by aggregating the available data and in each diagnostic subgroup estimate the risk for future complications in the individual patient. The data foundation will routinely be obtained with clinical data, supplemented by data from the Nordic national registries and biobanks. For studies of primary prophylaxis, muliti-omics data has been retrieved from the prospective, longitudinal, Norwegian HUNT study, such as polygenic risk scores for IHD, lipid- based metabolites, protein scans and LP(a)-analyses. These biomarkers will significantly enrich our analyses and risk predictions, to be replicated and validated based on the Danish Diet, Cancer and Health study. Not surprisingly, accessing registry data transnationally, has been associated with both technical and legal constraints, but eventually we have all the required data in place to complete the analyses for the study part on primary prophylaxis and have set aside resources for this task within the next three months. All the supplementary analyses and detailed validation of IHD as a clinical entity, have paved the way for new projects and studies, also attracted an interest from Industry. This Nordic collaboration has targeted this patient group to cross-validate and benchmark the results in a new and unprecedented manner, a major value to the patient, the healthcare system and society in general, reducing use of medication, hospital visits and healthcare expenses.

- The carefully built biomedical research infrastructure in Norway offers a number of unique features. They include long running population-based cohorts with extensive health information on a large number of participants, state-of-the-art biobank infrastructure facilitating comprehensive experiments in a short timeframe, high-quality and well-defined national and local disease registries to assess the biological impact of molecular discoveries, and expertise in epidemiology and health outcome. •We have all the required data in place to complete the analyses for the study part on primary prophylaxis and have set aside resources for this task within the next three months. - Echocardiography data from a large cohort of patient treated at Oslo University Hospital were extracted and anonymised, and a machine learning (ML) algorithms were developed and trained on a secure off-line platform in order to automatically measure anatomical and physiological parameters on a large scale. These have been important contributions to the automation of echo analysis using AI, in the early stages of this technology. - We have developed polygenic risk score for IHD and acquired additional omics data, e.g. lipid- based metabolites from 17 000 (the Nightingale platform), proteomics (Somascan) from 2200 (5700 in total), and LP(a)-analyses from approx. 25 000. These are biomarkers that will significantly enrich our analyses and the coming publications. - All the supplementary analyses and detailed validation of IHD as a clinical entity, has paved the way for new projects and studies, also attracting a clear interest from Industry.

Globally, the leading cause of years of life lost is ischemic heart disease (IHD). In the EU 13.2 mio. patients are diagnosed with IHD, 700,000 live in the Nordic countries. IHD causes chest pain, myocardial infarcts, reduced physical capacity and reduces life-expectance. IHD is not caused by a single mechanism but rather by a variety of different ones. Many risk factors and disease mechanisms are known, but we urgently need to manage complex data that can drive precise sub-classifications and risk stratification. At present, patients with IHD are generally diagnosed and treated using one-size-fits-all standard regimes. This leads to inefficient, costly, potentially harmful over-management. At the same time, patients with high risk are not identified and treatment not optimized. The objective is to develop and clinically implement personalized medicine (PM) with the dual purpose of avoiding both overtreatment as well as under-treatment in IHD. In this Nordic interdisciplinary collaboration, we intend to establish and merge large Nordic cohorts with well-described IHD genotypes and clinical expressions by combining existing and new data. The purpose is to differentiate between different subgroups of IHD, and from the better characterization identify each patient's cause of IHD. Using a machine learning approach, we will create a clinical integrative IHD algorithm, that will aggregate the available data and in each diagnostic subgroup estimate the risk for future complications in the individual patient. The data foundation will be routinely obtained clinical data, supplemented by data from the Nordic national registries and biobanks, This Nordic collaboration will target this patient group to cross-validate and benchmark the results in a new and unprecedented manner, a major value to the patient, the healthcare system and society in general, reducing use of medication, hospital visits and healthcare expenses.