Treating elevated pulmonary vascular resistance (PVR) alone in patients with heart failure with preserved ejection fraction (HFpEF) and combined postcapillary and precapillary pulmonary hypertension (CpcPH-HFpEF) may target the wrong mechanism and only address downstream consequences, according to a recent JACC: Heart Failure Leading Edge commentary by Milton Packer, MD, FACC; Mardi Gomberg-Maitland, MD, FACC; and Barry Borlaug, MD, FACC. Instead, based on the adipokine hypothesis, therapies such as sotatercept that target activin A may lead to improved outcomes.

In their commentary, Packer and colleagues discuss clinical trials, including DYNAMIC, PASSION, MELODY-1 and SERENADE, that used vasodilators (sildenafil, riociguat, tadalafil and macitentan) to target pulmonary arterial hypertension (PAH), which all failed to significantly improve symptoms and outcomes in patients with CpcPH-HFpEF.

In contrast, the CADENCE trial, which used sotatercept and targeted a different mechanism entirely, produced meaningful reductions in left atrial pressure and volume, lowered natriuretic peptide levels and decreased worsening heart failure events by 82% in patients with CpcPH-HFpEF, despite only having a modest effect on PVR – indicating that activin A targeting may be a potential path forward in treating this specific patient population.

"This conceptual framework provides a coherent explanation of the effects of sotatercept on the left heart (and explains the disappointing results with conventional pulmonary vasodilators)," write Packer and colleagues. "It also underscores the advantages of adipokine-directed rather than phenotype-directed targeting in the treatment of HFpEF, even when the disease is clinically advanced."

Another Leading Edge commentary by Steven A. Hamilton, MBBS, FACC, and colleagues examines how artificial intelligence (AI), particularly large language models (LLMs), could transform the use of electronic health record (EHR) data for HF research and care.

Although EHRs contain vast amounts of clinical information, many of the most meaningful details relating to HF, such as NYHA functional class, worsening congestion, genetic testing results, symptom trajectory, quality-of-life limitations, intolerance to guideline-directed medical therapy (GDMT), medication titration, frailty, caregiver support and reasons for referral, can be buried beneath unstructured clinical notes and are inadequately captured in billing codes alone. These have historically required labor-intensive and costly manual identification, synthesis and review.

However, AI has the potential to streamline workflows and support equitable patient identification while cutting down on time and costs in a variety of areas, including observational research, national registries and quality reporting, clinical trials and postmarket surveillance.

While discussing near- and long-term clinical applications, Hamilton and colleagues highlight studies that indicate AI performance varies by task complexity and requires ongoing validation against expert review. Rather than replacing humans entirely, the authors envision AI assisting researchers and clinicians by handling time-consuming and costly tasks that do not require complex clinical reasoning. Success will depend on careful implementation, governance and human oversight.

"The larger clinical opportunity is proactive identification of patients eligible for proven but under-used therapies, including [guideline-directed medical therapies], device-based therapies, advanced HF evaluation and hemodynamic monitoring," write the authors. "Realizing that opportunity will depend less on detection accuracy alone than on workflow, access, governance and multidisciplinary follow-up infrastructure. AI-enabled chart review should therefore be viewed as a scalable abstraction layer for evidence generation and implementation, with clinician oversight and equitable deployment determining whether technical gains translate into better HF outcomes."