3,000 in-home echocardiograms, with two uninterpretable

The situation

Every previous estimate of how common valvular heart disease is in the United States had a sampling problem. The studies drew on people who already had a clinical reason for an echocardiogram, or on people well enough to travel to a medical facility for one. Both routes oversample the mobile and the already-engaged, and undersample women, older adults and minority populations.

The Cardiovascular Research Foundation designed PREVUE-VALVE to remove that problem by removing the facility. If the echocardiogram happens at the participant’s kitchen table, the sample is no longer filtered by who can get to a hospital. That only works if the imaging done at the kitchen table is good enough to read.

What made it difficult

• A comprehensive two-dimensional transthoracic echocardiogram is not a screening scan. It had to be acquired to the core laboratory’s full protocol, in living rooms, on portable systems.
• Participants were drawn from all 50 states rather than from a set of study sites, so there was no fixed footprint to staff against.
• Enrollment ran for three years. Acquisition quality had to hold steady across that whole window, not just at the start.
• Participants were 65 to 85 years old, and the study deliberately sought the people other studies miss, including those who do not travel easily.

What we did

• Contacted individuals who cleared the study’s eligibility screener, obtained informed consent, and scheduled the visit around the participant.
• Sent certified cardiac sonographers to the home, working to the acquisition protocol issued by the core laboratory rather than to a protocol of our own.
• Ran the whole visit in one appointment: the echocardiogram, an electrocardiogram, questionnaires, and optional phlebotomy, so a participant gave up one morning rather than several.
• Uploaded raw imaging data straight to the core laboratory for independent reading. We did not grade our own work.

Result

3,002 participants completed the in-home visit, 88.4% of those who consented. Two of their echocardiograms were excluded as uninterpretable, leaving 3,000 analyzable studies. The resulting cohort matched the U.S. population aged 65 to 85 on age, sex, race, ethnicity and geography.

The study found a weighted prevalence of moderate or greater valvular heart disease of 8.2%, rising to 18.4% when clinically significant regurgitant lesions are included, and estimates that 4.7 million Americans aged 65 to 85 are affected today. Those are the study team’s findings rather than ours. The finding that belongs to the delivery model is the one the authors state in their conclusion: that the work establishes the feasibility of large-scale decentralized cardiovascular imaging.

For a sponsor weighing in-home imaging, the number that matters is two. Two studies out of 3,002 could not be read, judged by an independent core laboratory rather than by us.