TOPLINE

A wrist-worn photoplethysmography device detected cardiac arrest with very high sensitivity in sedated patients in the ICU after the withdrawal of life-sustaining treatment, with only one false positive alert in the test cohort.

METHODOLOGY

• A prospective, validation study evaluated whether a photoplethysmography-based algorithm could automatically detect non-induced cardiac arrest during planned withdrawal of life-sustaining treatment.
• A total of 44 patients were included (median age, 65 years; 75% men) at a Netherlands center between February 2024 and October 2025. The algorithm was iteratively refined in two training cohorts of 10 and 11 patients each before evaluation in an independent test cohort of 23 patients.
• Patients wore a photoplethysmography wristband (CardioWatch 287-2, Corsano Health, Netherlands) before the withdrawal of life support until death was confirmed, with continuous ECG and invasive arterial blood pressure serving as reference standards.
• The device alerted when the wrist pulse fell by at least 67% from the previous 2 hours and no pulses were noted for 20 seconds, and it stopped after 10 pulses returned.
• The primary endpoint was sensitivity for detecting cardiac arrest, defined as pulse pressure < 5 mm Hg. Secondary outcomes included false positive alerts, positive predictive value, and mean arterial pressure and pulse pressure at the time of algorithm detection.

TAKEAWAY

• In the test cohort, sensitivity for cardiac arrest detection was 100%. The algorithm detected cardiac arrests at a median mean arterial pressure of 30 mm Hg and pulse pressure of 13 mm Hg. 
• The positive predictive value in the test cohort was 96%, with one false positive alert occurring in 172.5 hours of photoplethysmography data. 
• The algorithm alerted a median of 2.5 minutes before cardiac arrest. 

IN PRACTICE

"[The] findings provide the first formal evidence in automatically detecting cardiac arrest events outside the induced setting, supporting further development of this potentially lifesaving technology," the authors of the study wrote.

SOURCE

The study was led by Roos Edgar, Radboud University Medical Center in Nijmegen, Netherlands. It was published online on July 24 in The Lancet Regional Health - Europe.

LIMITATIONS

Inclusion of sedated, immobile patients limited the assessment of algorithm performance during movement. The sample size was small, and tweaks in the algorithm were validated on the same data used to optimize the model. Data of invasive arterial blood pressure were unavailable at death in some patients.

DISCLOSURES

The study received support from the Dutch Heart Foundation and Radboud University Medical Center. One author reported receiving a research grant from the funding sources. Two authors reported receiving research grants from public/nonprofit organizations. One author reported receiving industry research funding and speaker fees from medical device and pharmaceutical companies, and one reported serving on an editorial board. Detailed disclosures are available in the original article.

This article was created using several editorial tools, including AI, as part of the process. Human editors reviewed this content before publication.