Cardiology trial recruitment: finding patients by ejection fraction, Lp(a) and events

How does eligibility work in cardiology trials?

Cardiology protocols define a patient in layers: ejection fraction sets the type of heart failure, NYHA class sets severity, an NT-proBNP or Lp(a) threshold narrows the group, qualifying events must fall inside a time window, and kidney function, background medicines and safety exclusions remove many of the rest.

Ejection fraction is the first split. US and European guidelines call heart failure reduced at an LVEF of 40% or lower, mildly reduced at 41% to 49%, and preserved at 50% or higher. The US guideline adds an improved category for patients whose LVEF has risen above 40%.^⁠[3] Trials draw their own lines inside these bands, and some check every past measurement, not only the latest.^⁠[1]

A patient can pass on a diagnosis code and fail on a line in an old echo report. The therapeutic areas pages cover the same pattern in other specialties.

What drives screen failure in cardiology trials?

Published screen failure data in cardiology is thin. PARAGON-HF, a trial in heart failure with preserved ejection fraction, randomized fewer than half of the patients it screened. Patients went through drug run-in periods during screening, so that gap reflects side effects during run-in as well as strict eligibility criteria.^⁠[1] Most other evidence comes from applying trial criteria to real patient groups.

In the Swedish cohort of 681 patients with reduced ejection fraction, 39% met EMPEROR-Reduced criteria, and the main reasons for ineligibility were low NT-proBNP or low eGFR.^⁠[4] Both are lab values that need context. In DAPA-HF, the NT-proBNP cutoff changed with heart rhythm and recent admissions, so a result had to be read next to the ECG and hospital history.^⁠[2] For site-level fixes, see how to reduce screen failure.

Why do structured fields miss cardiology eligibility?

The facts that settle cardiology eligibility are often written in text. For COPILOT-HF, a heart failure trial at Mass General Brigham, structured EHR data could settle only 5 of 6 inclusion criteria and 5 of 17 exclusion criteria, and study staff reviewed charts by hand for the rest.^⁠[7]

Skipping the notes costs accuracy. At Brigham and Women's Hospital, a rule-based system searched the EHR for heart failure patients eligible for guideline-directed therapy. Staff manually screened 5,460 of the patients it flagged and found 1,754 truly eligible, an accuracy of 32.1%. Over 38% of the false positives came from misjudging symptomatic heart failure and medication history, and the authors named clinical notes as the way to improve it.^⁠[8]

What does Bond read in a cardiology chart?

Bond's Identify stage reads structured and unstructured records against each inclusion and exclusion criterion, ranks candidates, and shows the evidence behind every criterion-level decision.^⁠[15] It uses imaging data, clinical notes and other unstructured documents, including pathology, radiology and molecular reports.^⁠[16] It connects to all the major EHRs, including Epic, Oracle Health (Cerner), MEDITECH, athenahealth, eClinicalWorks, NextGen, Veradigm and OncoEMR, through FHIR or HL7 interfaces or through an aggregator.^⁠[15],[16]

Each decision draws on Bond's terminology graph of 18 sources, including SNOMED CT, LOINC, RxNorm and ICD-10-CM, so a coded myocardial infarction and "STEMI" in a discharge summary can be matched to the same concept.^⁠[17] In a cardiology chart, Bond reads these sources where the site's interface exposes them:

• Echo, cardiac MRI and catheterization reports, older ones as well as the latest, so an exclusion based on any prior LVEF can be checked.
• Cardiology notes for NYHA class, symptoms and why a medicine was stopped.
• Discharge summaries and problem lists for the type and date of a heart attack, stroke or heart failure admission.
• ECG reports for atrial fibrillation or flutter, which changes some NT-proBNP cutoffs.
• Labs for NT-proBNP, Lp(a) and eGFR, each dated against the protocol's window.
• Medication records for doses and start dates of background therapy.

Criteria the chart cannot settle, such as current NYHA class, are left for the pre-screening call by Bond's Engage agents or site staff.

How would Bond screen for an Lp(a)HORIZON-style trial?

Lp(a)HORIZON tested pelacarsen in 8,323 patients with established cardiovascular disease. It required Lp(a) of 70 mg/dL or more at a central lab, plus a myocardial infarction or ischemic stroke 3 months to 10 years before screening, or symptomatic peripheral artery disease. Its key exclusions included NYHA class IV heart failure, a history of cancer, and a history of hemorrhagic stroke or other major bleeding.^⁠[9] On September 4, 2026, Novartis announced that the trial did not meet its primary endpoint.^⁠[10] Its criteria still show the search problem other Lp(a) trials face.

Few charts hold an Lp(a) result. In Epic Cosmos, a data set of more than 300 million patient records, 0.2% of the US population had Lp(a) tested from 2015 to 2024.^⁠[11] Across the University of California health system from 2012 to 2021, 0.3% of 5.55 million adults had been tested, and so had only 2.9% of those with ischemic heart disease.^⁠[12] Novartis estimates that about 20% of people worldwide have elevated Lp(a), so most patients who might qualify have no result in the chart.^⁠[10]

1. 1

   Find the qualifying events

   Bond looks for myocardial infarction and ischemic stroke in codes, problem lists, discharge summaries and notes, and for symptomatic peripheral artery disease in vascular notes.

2. 2

   Check the window

   Each event needs a date. Bond takes it from the admission or the note that describes it and keeps patients whose event falls between 3 months and 10 years before screening.^⁠[9]

3. 3

   Look for any Lp(a) result

   A prior result at or above the threshold moves a patient up the list. Untested patients stay on it, since the trial measured Lp(a) at a central lab at screening. Units need care, because labs report Lp(a) in mg/dL or nmol/L and the protocol sets its threshold in one of them.

4. 4

   Apply the exclusions

   NYHA class IV heart failure, past cancer, hemorrhagic stroke and major bleeding are checked in notes and imaging reports as well as codes.

5. 5

   Hand off what the chart cannot settle

   Current symptoms and ability to attend visits go to the pre-screening call. The investigator confirms eligibility at screening.

How many cardiology trials are recruiting in the US?

As of September 21, 2026, 2,845 studies with at least one US site were recruiting under a cardiology keyword search, about 13% of the 22,041 recruiting studies with a US site.^⁠[14] The search matches heart failure, coronary, atrial fibrillation, hypertension or cardiovascular in the condition field. It therefore counts hypertension and risk-factor studies, misses studies listed only under terms such as stroke, lipoprotein(a) or peripheral artery disease, and lets a study count in more than one area.

About 6.7 million Americans aged 20 and older have heart failure, and the number is expected to reach 8.5 million by 2030.^⁠[13] Yet of 19,191 cardiovascular trials registered on ClinicalTrials.gov from 2000 to 2025, 11.5% stopped early, and low recruitment was the most common reason given.^⁠[6] The hard part is finding the patients whose echo, labs and history fit a given protocol.