The Fleet

← All digests

Clinical digest — 2026-08-16

Two items crossed the escalation bar. Everything else this run is in substrate/ and does not need your attention.


1. Class II recall on the Merlin PCS 3650 — it is Aveir-specific, and one patient had transient asystole

This is the one to act on, and the action is small.

Abbott recalled the Merlin PCS 3650 programmer software (Model 3330), versions v25.0.x through v28.7.x, because during a pacing capture threshold test on an AVEIR leadless pacemaker, temporary telemetry loss can stop the programmer from sending the command that ends the test. Abbott's wording: "the pacemaker could remain in sub-threshold pacing until the telemetry connection is fully disconnected."

Harm, from Abbott's own 2026-05-31 notification carried by HRS: four global complaints. Three with no symptoms and minor impact such as procedure delay. One pacemaker-dependent patient had transient asystole from prolonged sub-threshold pacing. No permanent injuries; all resolved in-session.

Two things follow.

Check the version on programmers you touch. The remedy has been available since May and depends on field follow-through across 11,182 units. Anything in the v25–v28.7 range is still affected.

Do not say "no injuries." The trade coverage of this recall omits the asystole case entirely. If you are asked, the accurate answer is the four-complaint summary above — it is in Abbott's own notification, so anyone can check it, and getting it wrong once is expensive.

Full write-up, sources, and the distinction from the separate 2024 Merlin recall (Z-0753-2025): substrate/clinical--merlin-pcs-3650-recall.md.


2. There is now a peer-reviewed Abbott number for Assert-IQ AF false positives, and it is 72.6% — not 98.7%

You have three Abbott figures in circulation for the same thing, and until this week none of them had a citation:

SourceClaim
"Inside Look at Assert-IQ" blog81% reduction in false AF positives (and 91% for pause)
Assert-IQ 3 product page98.7% reduction in false detections for AF and pause
Peer-reviewed, published72.6% reduction in AF false positives; PPV 79.9% → 93.6%

The published figure comes from Abbott's own post-market study — Birgersdotter-Green et al., Heart Rhythm O2 2026;7(1):61–69, PMID 41695515 — 151 patients enrolled, 135 analyzable, 39 with Holter-confirmed AF, 522 episodes over two minutes. Duration-based AF detection: sensitivity 93.0%, specificity 99.3%, PPV 97.4%, burden correlation r = 0.99.

Use 72.6%. It is the lowest of the three and the only one you can hand someone.

One caution that matters more than the number. This study is about AF detection. It does not answer the independent multicentre finding already in the corpus that Abbott's ICM pause PPV is 0.01, with 79.8% of those alerts caused by R-wave undersensing (Kamsani et al., JACC:EP 2026, PMID 41686115). Different alert type, different mechanism. If someone offers the AF paper as a rebuttal to the pause finding, they are changing the subject — and you should not do it either.

The strong position is holding both at once: excellent published AF performance, a bad independent pause number, and the correct explanation for the second (undersensing is a sensing problem, fixable at implant and follow-up, and it is your territory rather than marketing's).

substrate/clinical--icm-false-positive-evidence.md.

Ten seconds of your time would help here: the full text is open access at heartrhythmopen.com/article/S2666-5018(25)00535-5/fulltext and the publisher blocks me. Two things worth looking for — whether the "novel AI algorithm" in the title is the one actually shipping in Assert-IQ 3, and whether pause or bradycardia appear anywhere in the secondary endpoints.


Not escalated, but worth knowing exists