Reading the PROTECT H2H Result

01 The category's problem, and who inherits it

The data is clean. The room it walks into is not.

PROTECT H2H did what it had to do. 522 patients across 20 sites in the United States, Germany and Brazil, randomized 1:1 against Sentinel. All primary and secondary endpoints met, non-inferior on 30-day MACCE in the 485-patient modified intention-to-treat cohort (4.5% vs 5.0%) and on 30-day stroke (2.0% vs 2.1%), statistically superior on debris capture, and a technical success gap of nearly nine points. As a regulatory package it is clean, and as a first-in-category randomized result it is genuinely unprecedented. On 9 September 2026 it did its job: FDA granted 510(k) clearance, indicated to capture and remove embolic material during TAVR, with a phased US launch to follow.

Which settles the first question and sharpens the second. The regulator is satisfied. The skeptic is not, and the skeptic is now the audience.

The difficulty is the room.

Embolic protection has been tested against no protection twice, at scale. The story is more specific than it is usually told, and the specifics matter.

PROTECTED TAVR, 2022, 3,000 patients across 51 centers, missed its primary endpoint: 2.3% stroke with protection against 2.9% without, P=0.30. But a secondary endpoint the trial was not powered for showed significantly fewer disabling strokes — 0.5% against 1.3%, P=0.02, a number needed to treat of 125. Kapadia, the global principal investigator, argued that the missed primary endpoint was not the end of the story, and his own institution has continued to use cerebral protection in all TAVR patients on the strength of the disabling-stroke finding. That signal has anchored the case for the category ever since.

BHF PROTECT-TAVI, 2025, tested it at more than twice the size: 7,635 patients across 33 UK centers. The primary endpoint again showed no difference, 2.1% against 2.2%. This time the secondary endpoints showed no difference either, including disabling stroke. A cognitive substudy within the same trial, covering 3,535 of those participants, found no effect on cognition.

So the accurate reading is not that the category has been shown not to work twice. It is that the one hopeful signal from the first trial did not replicate in a trial five times its size. Kharbanda's own summary was that in his population a routine strategy did not reduce stroke, and that the study provides convincing evidence of no value in routine use.

The word doing the work in that sentence is routine, and Kharbanda said so himself. In the same session he noted that next-generation devices could potentially reduce stroke risk for certain groups, and that the field needs to understand which patients are at higher risk during TAVR and whether a more stratified approach would benefit them.

That is the opening. It was not found by reading against the trial. It was stated by the investigator who ran it.

That is the sentence the field is carrying into 2026. The assessment published on ACC.org after ACC.26 reached the same place. Writing for the College's Interventional Member Section, Robert Jay Widmer, MD, FACC, called PROTECT H2H an acceptable alternative device while noting that it continues to demonstrate a disconnect between mechanistic protection and clinical outcomes, since despite markedly increased debris capture and improved device performance, no reduction in clinical stroke was observed. His closing judgement is harder still: that although Emboliner represents a technically successful evolution in device design, its clinical advantage over existing devices — and over no protection at all — remains unproven.

That is a signed commentary rather than a position of the College, and it should be read as one man's view. But he is an interventional cardiologist writing for his own society's membership, describing Emboline's own result, three weeks after it was presented. That is the readership the launch has to persuade, and this is what it published first.

A qualification — from the literature, not the company

BHF PROTECT-TAVI stopped early for futility at 7,635 patients. Its published design had targeted 7,730, so enrollment was nearly complete — but that target assumed a 3% stroke rate in the control arm, and the rate observed was 2.2%. Against the event rate the trial actually saw, retaining 80% power to detect a one-third reduction would have required 9,712. It was therefore underpowered for a moderate effect of the order PROTECTED TAVR had hinted at.

This does not rescue routine use, and nobody should pretend it does. It does mean the trial is more precisely described as having excluded a large benefit than as having excluded any benefit at all. The point is made in the European Heart Journal, not by anyone with an interest in the answer.

Fig. 1 — Stroke, every randomized comparison to date Percentage of patients

No color appears in this chart, because nothing separates in it. Timepoints differ — the first two report stroke at 72 hours against no protection; PROTECT H2H reports 30-day stroke against an active comparator — so these bars are not directly comparable to one another. They are shown together for one reason: on all stroke, whatever the design, the two arms land in the same place every time. PROTECTED TAVR did show a significant reduction in disabling stroke as an underpowered secondary endpoint (0.5% vs 1.3%, P=0.02); BHF PROTECT-TAVI, five times larger, did not replicate it.

Source — PROTECTED TAVR, Kapadia et al., NEJM 2022. BHF PROTECT-TAVI, Kharbanda et al., NEJM 2025 (n=7,635; 33 UK centers; difference −0.02 pp, 95% CI −0.68 to 0.63, P=0.94). PROTECT H2H, Emboline press release and ACC.org trial summary, March 2026.

Three consequences worth naming plainly

The comparator is doing less work than it appears to

Running against Sentinel was the right regulatory choice and the right competitive one. But it means the evidence base now shows Emboliner performing at least as well as a device that the two largest trials in the category could not distinguish from no device at all. An informed reader completes that syllogism without help.

The superiority endpoint is the mechanism those trials failed to validate

Debris capture at 150 microns and above is a real and defensible engineering result. It is also the specific causal step that two well-powered trials could not show translating into outcomes. Sentinel captures debris in nearly every case and did not move the stroke number. Leading with capture volume invites the response that capture volume was never the problem.

Technical success is the finding that does not depend on the debate

Roughly one deployment in eight falls short with the incumbent in PROTECT H2H. That is a fact about the room the operator is standing in, and it holds regardless of how the stroke question resolves. PROTECTED TAVR reported 94.4%, but BHF PROTECT-TAVI — asking whether both filters stayed correctly deployed for the duration of the procedure — reported 81.2%. The three figures do not disagree so much as measure different things, and the definition is the argument.

Fig. 2 — Where the two devices actually separate PROTECT H2H, n=522

Each dot is a captured particle. This is the only place in the dataset where the two devices meaningfully diverge — and one of the two measures is a surrogate for the outcome that Fig. 1 shows has never moved. The other is a workflow fact that survives the debate entirely. Which of those leads is the decision on page two.

Source — Emboline press release, 30 March 2026 (NCT05684146); ACC.org, ACC.26 structural intervention trials summary. Operator experience per the same release, which reports that all study operators had prior experience with the control device only and used Emboliner for the first time; independent conference reporting describes investigators as trained but not experienced with the study device.

None of this is a problem with the data. It is a problem with the story the data will be asked to tell, and it belongs to the category rather than to Emboline.

Emboline is the only company in embolic protection holding fresh, positive, randomized evidence. Boston Scientific has the installed base, the sales force, and two trials it would rather not discuss.

That asymmetry means the job in front of Emboline is larger than taking share. It is deciding what the category is now for, and being the one to say it first. Companies rarely get that opportunity. It usually arrives disguised as a communications problem.

02 Four doors

Four positions are available. One of them has to lead.

Each is defensible, each costs something, and each is aimed at a different audience. They are not mutually exclusive. But a launch that argues all four argues none.

Door one · Patient selection

The trials asked whether everyone needs protection

That was never the interesting question.

Both PROTECTED TAVR and BHF PROTECT-TAVI tested routine use in unselected populations. That is the honest opening, and Emboline already holds the literature to walk through it, most directly their own SafePass 2 first-in-human study, which found that atrial fibrillation on admission was associated with 4.1 times more particles above 150µm and 8.1 times more above 500µm, and that self-expanding valves produced roughly twice the particle count. The same study found the TASK clinical stroke-risk score correlated with particle counts above 500µm, which matters more than it appears: a clinical score is available where imaging is not. It is a 31-patient single-arm study, so these are signals rather than settled findings — but they are published, peer-reviewed, and Emboline's own, and the paper's stated conclusion is this position in the authors' own words, that larger studies are needed to identify high-risk patients for selective embolic protection. The position becomes: identify the high-burden patient, protect that patient.

Buys: Survives the neutral-trial objection instead of arguing with it. Compatible with future guideline language in a way routine-use claims are not.
Costs: Narrows addressable procedure volume on paper. The predictor evidence is a 31-patient single-arm study — hypothesis-generating, not established. Preprocedural CT angiography did not predict debris count, so a selection tool cannot be built on imaging. And BHF PROTECT-TAVI already ran subgroups on age, sex, bicuspid anatomy, clinical risk score, valve type and aortic calcification, finding treatment benefit in none of them. Any selection hypothesis has to sit outside ground already tested, and it will need prospective validation rather than retrospective analysis.
Aimed at: KOLs, guideline committees, trial investigators

Door two · Total embolic burden

The category measured strokes

The debris does not only go to the brain.

Kidney injury, mesenteric and renal events, everything downstream. This is a different endpoint conversation than the one the category lost, and Emboliner is built for it. Emboliner is not alone here, and this is the door where that matters most. EuroIntervention classifies Emblok and AorticLab's FLOWer alongside it as full-capture devices, with Filterlex's CAPTIS deflecting at the arch and capturing in the descending aorta. Emboliner is the only full-capture device holding completed pivotal randomized data — Protembis has completed a randomized pivotal for the cerebral-only ProtEmbo, so the distinction sits within full-body capture rather than across the whole category. But Emblok's US pivotal is enrolling up to 532 patients against a commercially available comparator, and Innovative Cardiovascular Solutions already markets Emblok as whole-body protection for the brain, heart, kidneys and intestines. The organ-by-organ argument is not open ground. It is ground a competitor is already standing on, without the data to hold it. That argues for taking the position early and evidencing it, not for assuming it will still be there later.

Buys: The neutral results become irrelevant rather than inconvenient — they measured something else. Defensible on device architecture rather than on statistics.
Costs: Requires evidence that does not fully exist yet. PROTECT H2H did report stage III acute kidney injury at 0.4% against 1.2%, the largest proportional gap among the individual endpoints — but it did not reach significance and the trial was not powered for it, so it is a signal to build on rather than a claim. Asks the field to care about an outcome it has not been trained to prioritize. And reimbursement does not reward it: CPT add-on code +33370, effective 1 January 2022, pays for placement and removal of cerebral embolic protection during TAVR. A device may protect the kidneys, but no separate code pays for having done so. Any systemic claim has to earn its way through clinical preference rather than economics — slower, and it needs the evidence to be unambiguous.
Aimed at: KOLs first, then society and guideline audiences, on a multi-year arc

Door three · Procedural performance

One in eight, and every operator has felt it

95.9 against 87.1, p = 0.0005.

Roughly one deployment in eight falls short with the incumbent in this trial. It is not an efficacy argument, which is precisely why it cannot be dragged into the efficacy debate — and it is currently buried beneath the capture figure. PROTECTED TAVR reported 94.4% successful Sentinel deployment, which will be raised as a counter — but BHF PROTECT-TAVI, asking whether both filters stayed correctly deployed for the duration, found 81.2%, and Kapadia himself publicly questioned the British trial's device success rate. The counter has to be met on what is being counted, not on which number is larger.

Buys: Immediate credibility with the person holding the catheter. Independent of the efficacy debate. Fastest to activate — needs no new evidence.
Costs: A workflow argument does not justify the category's existence or its price. Wins the operator, does not win the value analysis committee.
Aimed at: Interventionalists, cath lab staff, proctors

Door four · Beyond TAVR

The debate is about aortic stenosis

The embolic burden is worse elsewhere.

Mitral, tricuspid, valve-in-valve, complex structural work. Higher debris load, no entrenched incumbent, and no accumulated neutral evidence to argue against. Emboline already describes itself as developing embolic protection for structural heart procedures generally, not TAVR alone. This is not a hypothetical extension: Greenbaum, presenting PROTECT H2H, framed the data as supporting broader use in structural heart and transcatheter procedures carrying greater embolic burden than routine TAVR, and a 2026 report in JACC: Cardiovascular Interventions already describes Emboliner used in structural heart procedures at extreme embolic risk. Its authors include Greenbaum, who presented PROTECT H2H, and Nazif, its co-principal investigator — so the people who ran the TAVR trial are already publishing beyond it.

The constraint is now explicit. The September clearance is indicated for TAVR. Everything in this door therefore sits outside the cleared indication until a new one exists, which makes it a clinical and regulatory programme rather than a message. Third-party literature can be distributed under the rules governing scientific information; company communications cannot follow it.

Buys: Clean ground with no history to overcome. Positions Emboline as a structural heart company rather than a TAVR accessory. Strategically the largest of the four.
Costs: Slowest, and off-label until the indication moves. Evidence and indication work that follows the current clearance rather than accompanying it.
Aimed at: Investors, partners, the long-horizon clinical community

Fig. 3 — The decision is sequencing, not selection Indicative horizon

Doors three and one can run together in year one. Door two is the multi-year build that determines whether the category survives. Door four is where the company ends up if the first three work.

Source — Author's construction. Horizons are indicative, not drawn from company guidance.

References

Emboline, Inc. Emboline Announces Positive Pivotal PROTECT H2H Trial Results for Emboliner, Presented as a Late-Breaking Clinical Trial at ACC 2026. Press release, 30 March 2026. ClinicalTrials.gov NCT05684146. Primary source for all PROTECT H2H endpoint figures.
Widmer RJ. ACC.26 Structural Intervention Trials Offer Viable Options. Expert Analysis published on ACC.org, 23 April 2026, for the ACC Interventional Member Section. Source of the disconnect assessment quoted on page one, of the judgement that clinical advantage over existing devices and over no protection at all remains unproven, and of the modified intention-to-treat cohort of 485 in which MACCE was 4.5% vs 5.0%. This is a signed expert commentary hosted by the College, not an institutional position of the College.
American College of Cardiology. PROTECT H2H: Emboliner vs. Sentinel Embolic Protection Devices in TAVR. ACC.org trial summary, March 2026. Trial geography: 20 sites, United States, Germany, Brazil; mean age 79; 33% women.
Kapadia SR, Makkar R, Leon M, et al. Cerebral embolic protection during transcatheter aortic-valve replacement (PROTECTED TAVR). New England Journal of Medicine, 2022;387:1253–1263. doi:10.1056/NEJMoa2204961. 3,000 patients, 51 centers, North America, Europe and Australia. Primary endpoint not met: stroke 2.3% (34/1,501) vs 2.9% (43/1,499); difference −0.6 pp, 95% CI −1.7 to 0.5, P=0.30. Disabling stroke, a secondary endpoint for which the trial was not powered: 0.5% (8) vs 1.3% (20), 95% CI −1.5 to −0.1, P=0.02; NNT 125. Sentinel deployed successfully in 94.4% of patients and captured debris in 99%.
Kharbanda RK, Kennedy J, Jamal Z, Dodd M, et al. Routine cerebral embolic protection during transcatheter aortic-valve implantation (BHF PROTECT-TAVI). New England Journal of Medicine, 2025 Jun 26;392(24):2403–2412; published online 30 March 2025. doi:10.1056/NEJMoa2415120. PMID 40162661. ISRCTN16665769. 7,635 participants (mean age 81; 39% women), 33 UK centers, October 2020 to October 2024; 3,815 assigned to CEP and 3,820 to control; enrollment stopped early for futility. Stroke within 72 hours 2.1% (81/3,795) vs 2.2% (82/3,799); difference −0.02 pp, 95% CI −0.68 to 0.63, P=0.94. Disabling stroke 1.2% vs 1.4%. Both filters were fully and correctly deployed for the duration of the procedure in 3,058 of 3,768 patients (81.2%); deployment departed from protocol in 710 of 3,768 (18.8%). No prespecified subgroup showed treatment benefit. Funded by the British Heart Foundation and Boston Scientific.
BHF PROTECT-TAVI Investigators. Impact of cerebral embolic protection on cognitive function after transcatheter aortic valve implantation: data from the BHF PROTECT-TAVI randomized trial. Circulation, 2025. doi:10.1161/CIRCULATIONAHA.125.076761. PMID 40884786. Secondary analysis of 3,535 participants (1,763 CEP, 1,772 control; mean age 81.0; 37.7% women). Adjusted mean change in telephone MoCA score 0.83 (95% CI 0.70 to 0.96) with CEP vs 0.91 (0.79 to 1.04) without; between-group difference −0.07 (95% CI −0.22 to 0.09), P=0.42.
Grubman D, Ahmad Y, Leipsic JA, Blanke P, Pasupati S, Webster M, Nazif TM, Parise H, Lansky AJ. Predictors of cerebral embolic debris during transcatheter aortic valve replacement: the SafePass 2 first-in-human study. American Journal of Cardiology, 2023 Nov 15;207:28–34. doi:10.1016/j.amjcard.2023.08.137. PMID 37722198. Prospective, non-randomized, single-arm study; 31 patients across three centers in New Zealand. Technical success 31/31. 30-day MACCE 6.5% (two cerebrovascular events, neither causally related to the device). Atrial fibrillation on admission associated with 4.1× more particles >150µm (P=0.0130) and 8.1× more >500µm (P=0.0086); a history of atrial fibrillation predicted more particles >500µm (P=0.0259); self-expanding valves approximately twice the particles >150µm (P=0.0281); TASK score positively correlated with particles >500µm (P=0.0337). None of the tested preprocedural CT angiography features identified higher embolic risk. The authors conclude that larger studies are needed to identify high-risk patients for selective embolic protection device use.
Berkovitch A, et al. Clinical predictors for procedural stroke and implications for embolic protection devices during TAVR: results from the multicenter TAVR in-hospital stroke (TASK) study. Journal of Personalized Medicine, 2022;12:1056.
Jimenez Diaz VA, Kapadia SR, Linke A, et al. Cerebral embolic protection during transcatheter heart interventions. EuroIntervention, 2023;19(7):549–570. Device classification: full-capture devices (Emblok, Emboliner, FLOWer); primarily deflective (TriGUARD 3, ProtEmbo, POINT-GUARD); deflection-and-capture (CAPTIS).
Vergallo R, Leone AM. Weekly Journal Scan: routine cerebral embolic protection during TAVI — a sleeping sentinel. European Heart Journal, 2025;46(43):4620–4622. doi:10.1093/eurheartj/ehaf569. Source of the power calculation: early termination for futility left 7,635 enrolled against the 9,712 required for 80% power to detect a one-third reduction in stroke.
Tat E, Camaj A, George I, Vahl TP, Hahn RT, Sethi S, Byku I, Babaliaros V, Gleason PT, Greenbaum A, Nazif TM. Embolic protection with the Emboliner for structural heart procedures with extreme embolic risk. JACC: Cardiovascular Interventions, published 11 May 2026. doi:10.1016/j.jcin.2026.03.007. Cited by DOI; volume and page range not independently confirmed.
Emboline, Inc. Emboline Receives U.S. FDA 510(k) Clearance for Emboliner Embolic Protection System. Press release, 9 September 2026. Indicated to capture and remove embolic material during TAVR; phased US commercial launch planned for later in 2026.
Innovative Cardiovascular Solutions. Evaluation of Safety and Effectiveness of the EMBLOK Embolic Protection System During TAVR. ClinicalTrials.gov NCT05295628. Prospective, multicenter, single-blind randomized controlled trial; up to 532 subjects at up to 30 US sites, randomized against a commercially available embolic protection device. Company website, emblok.com, markets Emblok 360° as whole-body embolic protection for brain, heart, kidneys and intestines.
Protembis GmbH. Cerebral Protection in Transcatheter Aortic Valve Replacement: The PROTEMBO Trial. ClinicalTrials.gov NCT05873816. Randomized against a hybrid control of no device and Sentinel; status completed, primary completion 18 October 2025.
American Medical Association. CPT add-on code +33370, transcatheter placement and subsequent removal of cerebral embolic protection device(s), effective 1 January 2022. Source for the reimbursement constraint under door two.
Emboline, Inc. Corporate website and management team page, emboline.com, accessed August 2026.

Where a journal volume or page range could not be verified from public sources, it has been omitted rather than estimated.

What cannot happen

The launch defaulting to better than Sentinel — which is the natural gravity of the situation, and the one position the evidence base will not support.

That gravity is already visible. The clearance announcement describes the highest technical success ever reported for a TAVR embolic protection device in a pivotal study. It may well be right. But it is a cross-trial comparison against results measured on different definitions of success, and it is the kind of claim that holds only with the trial and the definition attached to it. That is not a criticism of the sentence. It is an illustration of how quickly the gravity asserts itself, and of why the thing worth building now is the file that decides which version of a sentence like that can be defended.