Despite advances in medical and device therapies that improve quality of life, patients with heart failure with reduced ejection fraction (HFrEF) remain at high risk for cardiovascular death and re-hospitalization. To address this persistent challenge, a recent substudy of the VICTORIA trial analyzed comprehensive patient data—including clinical, electrocardiographic, echocardiographic, biomarker, and proteomic inputs—to identify distinct patient profiles.
Using advanced statistical modelling, researchers classified 564 participants into three distinct profiles, or “phenogroups”. Each group revealed a unique set of traits and a specific risk level for future cardiovascular events:
- Phenogroup 1: Younger patients who were well-managed on standard drug therapies and rarely required implantable defibrillators.
- Phenogroup 2: Patients distinguished by higher rates of atrial fibrillation and ECG patterns indicating previous heart damage.
- Phenogroup 3: The oldest patients, who had the highest rates of advanced renal disease and heart failure affecting both sides of the heart.
Outcomes differed dramatically between these groups, with Phenogroup 3 facing a seven-fold higher risk of primary adverse events compared to Phenogroup 1. This steep increase in risk was confirmed in an external validation study. Crucially, the analysis identified a specific protein, Growth Differentiation Factor 15 (GDF-15), as the most powerful tool for distinguishing these profiles, outperforming even traditional heart failure biomarkers.
Ultimately, the use of multimodality data to define these phenogroups offers risk prognostication that far exceeds traditional clinical characteristics. With GDF-15 emerging as the pivotal marker for distinguishing these groups, the study provides a new framework for precision medicine in HFrEF. These insights have the potential to shape the design of future clinical trials and could be instrumental in developing targeted therapeutics for this vulnerable patient population.
Reflecting on the study results and the collaborative effort involved, co-author and VICTORIA Chair Dr. Paul Armstrong, MD emphasizes:
“This work has forever changed my thinking about the value add of artificial intelligence and machine learning to discover patterns we could not discover using conventional techniques. The new clustering of patient groups was so much superior to what we expected, and that was truly exciting to behold. Working with Palak Shah, MD, MS (Inova Schar Heart and Vascular), we were confident the work would be readily published, but Circulation: Heart Failure needed to be convinced that our findings were valid. Fortunately, our colleague from the Netherlands Adriaan Voors, MD (University of Groningen) had a separate heart failure data set to explore. So, with the able insights of Gray Zheng, MA, MEd and Cindy Westerhout, PHD, we independently confirmed that the findings were reproducible and hence this article was accepted!”
