Hemodialysis (HD) is conducted through a hemodialysis vascular access, such that the
patient's blood can be removed from the patient, circulated through the dialysis machine,
and then returned to the patient. HD vascular accesses may take the form of an
arteriovenous fistula (AVF), an arteriovenous graft (AVG), or a temporary or permanent
dialysis catheter. Because of its low complication rate and good longevity (durability),
the AVF is the preferred access; current Kidney Disease Outcomes Quality Improvement
"Fistula First" recommendations are that any patient with an estimated glomerular
filtration rate of 25 mL/min should undergo AVF creation surgery in preparation for
eventual HD.
Stage IV is chosen as the optimal time for AVF creation, even though the patient does not
yet need HD. This is because (1) it usually takes ~8 weeks to schedule a patient for AVF
creation surgery, (2) it takes another 12-16 weeks after creation for the AVF to "mature"
and become ready for use, and (3) as many as 60% of surgically created AVFs fail to
mature at all. Considering the time lag to maturation and the possible need to intervene
on an access that is failing to mature, a significant lead time is required to avoid
encountering a patient in stage V CKD who needs HD immediately and then has to resort to
a tunneled dialysis catheter with all its attendant risks of infection, thrombosis and
central vein stenosis.
Being able to actively monitor an AVF from the time of surgical creation with a device
that can predict whether that AVF will fail vs. mature, will empower patients and their
physicians to decide much earlier to intervene and put that AVF "back on track" to
maturation. This means that many more patients will be able to start HD when needed,
using their AVF as a dialysis access, and therefore reduce the use of tunneled dialysis
catheters.
This type of monitoring, however, requires frequent measurements in order to track
progression toward maturity. By training an AI/ML algorithm to analyze sound signals
recorded from the patient's nascent AVF, the algorithm can detect when an AVF is likely
or not to mature. If trained properly, it should be able to detect an "unlikely to
mature" condition very early after surgery, perhaps even within the first two weeks. This
will prompt the patient's physician to consider intervention on the AVF that would
increase the likelihood of maturation and increase the overall maturation rate. A
digital, AI/ML-enabled stethoscope, perhaps even used in the patient's home in the
future, could easily facilitate that monitoring requirement as well as transmit the
relevant information to the patient's physician so that timely intervention could be
arranged.
The Eko CORE is an FDA-cleared and CE-marked electronic stethoscope. The CORE allows the
audio recording of lung, heart, and other body sounds. The CORE features 40x audio
amplification, ambient noise reduction, a 4000Hz sample rate, and 3 audio filters. The
de-identified auscultatory CORE sound recordings transmit wirelessly via Bluetooth to the
secure, HIPAA-compliant Eko application on a smartphone or tablet, which allows the user
to playback sound recordings, annotate notes on recorded audio, and save recordings. This
data is synced in real-time to a secure, HIPAA-compliant, cloud-based Amazon Web Services
(AWS) database server managed by Eko Devices.
This prospective, observational, longitudinal, proof of concept study will be enrolling
patients at the University of North Carolina (UNC) Kidney Center. This study will enroll
a minimum of 25, maximum of 50, patients with end-stage renal disease (ESRD) patients in
the hemodialysis unit (HDU) with maturing AVFs who are being dialysed through a tunneled
dialysis catheter (TDC). Recordings will be taken once per week during the patient's HD
appointment. Duration of the study will extend from the time of enrollment until one of
the following occurs first: (a) abandonment of the AVF, or (b) 2 weeks of measurement
after documentation of successful cannulation defined by 2-needle cannulation for 3
successive dialysis treatments, or (c) 6 months of study participation. These datasets
will be used to derive an AI/ML model to predict the likelihood of a new AVF to mature or
fail.