Mind Uncharted Explore. Discover. Learn.
UCL researchers discover five subtypes of heart failure with potential to predict future risk
Photograph: stockasso/Envato
Heart failure is a broad term when the heart fails to circulate blood throughout the body efficiently [1]. The current methods of categorizing heart failure fail to foresee how the disease will develop accurately.
The Study
Researchers examined detailed anonymized patient data of over 300,000 individuals aged 30 and above who had been diagnosed with heart failure in the UK over a 20-year period as part of the study, which was published in The Lancet Digital Health [2].
Using various machine learning techniques, researchers have discovered five subtypes of heart disease. These include early onset, late onset, metabolic (associated with obesity and low cardiovascular disease rates), cardiometabolic (linked to obesity and cardiovascular disease), and atrial fibrillation related (atrial fibrillation is a condition that causes an irregular heartbeat).
Findings
Per the research findings, the patients’ probability of dying within one year of diagnosis varied across subtypes. Here are the corresponding one-year mortality risks for all-cause: early onset (20%), late-onset (46%), atrial fibrillation related (61%), metabolic (11%), and cardiometabolic (37%).
The researchers designed an app that clinicians can use to identify individuals’ heart failure subtypes. This may enhance the accuracy of future risk predictions and provide important information for patient consultations.
Lead author, Professor Amitava Banerjee from the UCL Institute of Health Informatics explained that the objective was to enhance the classification of heart failure, to gain a better comprehension of its likely progression and communicate this effectively to patients. At present, forecasting disease progression is difficult for individual patients, with some being stable for several years while others deteriorate quickly.
“Better identification and classification of the various subtypes of heart failure can potentially result in more precise treatment strategies and enable a fresh perspective on potential therapies.”
“In the latest research, we have successfully identified five robust subtypes using various machine learning techniques and datasets.”
“The next step is to evaluate whether the classification of heart failure demonstrated has practical implications for patients in terms of risk prediction, quality of diagnosis and changes in patient treatment. Additionally, the cost-effectiveness of the proposed app should also be examined. Although further clinical trials and research are required, the designed app might benefit routine care.”
The research methodology
The researchers employed four methods to group heart failure cases to prevent bias from a single machine learning technique. They used these approaches on data obtained from two vast primary care datasets in the UK, accurately portraying the entire UK population.
Additionally, these datasets were linked with records of hospital admissions and fatalities. The datasets, namely, Clinical Practice Research Datalink (CPRD) and The Health Improvement Network (THIN), encompassed the period from 1998 to 2018.
The research team trained the machine learning tools on specific parts of the data. After identifying the most robust subtypes, the team cross-checked these classifications using a separate dataset.
The classification of the subcategories was determined by assessing 87 factors out of a potential 635, which encompassed age, indicators of other diseases, symptomatology, prescribed medications, and test findings such as blood pressure and evaluations like kidney function.
The research team examined information from the UK Biobank study of 9,573 heart failure patients. Their findings indicate that specific types of heart failure are associated with increased polygenic risk scores (overall risk scores due to the genes as a whole) for hypertension and atrial fibrillation.
Conclusion
The discovery of the five subtypes of heart disease and the designed app is a major milestone in improving the accuracy of future risk predictions, providing better consultation and treatment strategies, and enabling a fresh perspective on potential therapies for heart failure patients. However, further clinical trials and research are needed to evaluate the practical implications of these findings and examine the cost-effectiveness of the proposed app.
[1]https://www.nhlbi.nih.gov/health/heart-failure[2]https://www.thelancet.com/journals/landig/article/PIIS2589-7500(23)00065-1/fulltext
