← All articles

🩺 Health

One Prevented Stroke Costs $476,046 Before Any Rebate

Investigators print a one-year NNT of 53. Twelve Sunharbo packs at $748.50 cost $8,982. Fifty-three of those years is $476,046 at list price, before any rebate.

A single round white tablet and a glass of water on a steel hospital tray, an empty bed behind them, and city lights at dusk beyond the window.

Fifty-three people. One ischemic stroke at the investigators' own one-year count costs $476,046 before any rebate.

On 9 November 2026 the Food and Drug Administration approved Sunharbo, asundexian 50 mg once daily, as the first factor XIa inhibitor in the United States, taken with antiplatelet therapy after a recent mild-to-moderate non-cardioembolic ischemic stroke (NIH Stroke Scale 15 or under) or a high-risk transient ischemic attack, and not for atrial fibrillation. Hylton Joffe, M.D., director of FDA's Office of Cardiology, Hematology, Endocrinology, and Nephrology, called the approval "a meaningful advance for patients and health care providers seeking a new option for secondary stroke prevention."

Bayer's list price is $748.50 for 30 days, a figure in an email to Managed Healthcare Executive, a spokesperson comment to BioPharma Dive, and TCTMD's account of Bayer's own line that the list price "does not reflect what most health plans or patients will pay." Twelve packs cover 360 days, and 12 times 748.50 is 8,982.

Conference slides for OCEANIC-STROKE print a one-year absolute risk reduction of 1.9 percent and a number needed to treat of 53, and one divided by 0.019 is 52.6, so 53 is that reciprocal rounded to a whole person. Fifty times 8,982 is 449,100, three times 8,982 is 26,946, and the sum is 476,046 if those 53 people each fill every pack and nobody receives a discount. A 365-day year is 365/30 packs times 748.50, or $9,106.75, and 53 of those is $482,657.75.

ClinicalTrials.gov lists 12,327 participants, started 26 January 2023, completed 23 October 2025, at 741 locations in 37 countries (the slides say 702 sites). People were randomized within 72 hours, 6,162 to asundexian 50 mg and 6,165 to placebo, on antiplatelet therapy, and atrial fibrillation requiring anticoagulation was excluded. Sharma and coauthors reported a median follow-up of 19 months on 15 April 2026. The College's scan says 567 days, and 567 divided by 30.4375 (365.25/12) is 18.6 months, so 19 is the rounding. FDA's median treatment, in the sentence with the 26 percent reduction, is 14.2 months, against a registry window of 3 to 31 months.

Ischemic stroke hit 384 people versus 518, integers that appear on the slides and in Bayer's 6 May severity note. 384 divided by 6,162 is 6.23 percent before rounding to 6.2, 518 divided by 6,165 is 8.40 percent, and the difference is 2.17 points, reciprocal 46.1 across the whole follow-up, while 8.4 minus 6.2 is the 2.2-point gap TCTMD called "through 2 years," reciprocal 45.5. Neither is the one-year 53.

Counts, percentages and cause-specific hazard ratios are the International Stroke Conference primary-results slides. Percentages are those slides' rounding. The one-year number needed to treat of 53 is a figure note on the same deck, not a row here. The major-bleeding P value on the slide is 0.46.
OutcomeAsundexian, n of 6,162Placebo, n of 6,165Hazard ratio (interval)
Ischemic stroke384 (6.2%)518 (8.4%)0.74 (0.65 to 0.84)
Any stroke404 (6.6%)545 (8.8%)0.74 (0.65 to 0.84)
Cardiovascular death, heart attack, or stroke568 (9.2%)685 (11.1%)0.83 (0.74 to 0.92)
Ischemic stroke in the first 90 days183 (3.0%)218 (3.5%)0.84 (0.69 to 1.02)
Disabling or fatal stroke128 (2.1%)185 (3.0%)0.69 (0.55 to 0.87)
Major bleeding117 (1.9%)107 (1.7%)1.10 (0.85 to 1.44)
Fatal bleeding14 (0.2%)8 (0.1%)1.77 (0.74 to 4.23)

Disabling or fatal stroke (modified Rankin of 3 or worse at 90 days, or a one-point worsening if already there) was 128 versus 185. 185/6,165 minus 128/6,162 is 0.009236, reciprocal 108.3, not 53. Major bleeding was 117 versus 107: ten extra bleeds against 134 fewer ischemic strokes, a crude 13.4 to 1, and not a test, since the interval is 0.85 to 1.44 and the slide P value is 0.46. Fatal bleeding went the other way on the point estimate, 14 versus 8, hazard ratio 1.77, interval 0.74 to 4.23. Hemorrhagic stroke was 13 versus 20.

Bayer's approval release, quoting the October 2026 label, gives major bleeding per 100 patient-years as 1.6 versus 1.5 overall, 4.5 versus 3.1 in the first 90 days, 2.6 versus 1.7 past age 75, and 2.9 versus 1.0 after a high-risk TIA. The overall gap is 0.1, so those three gaps (1.4, 0.9, 1.9) are fourteen, nine, and nineteen times the overall point estimate, and the overall hazard interval still includes no increase. Gastrointestinal bleeding is 2.7 versus 2.1, hazard ratio 1.3, interval 1.0 to 1.6. Diarrhea was 3.4 percent versus 2.8.

Ashkan Shoamanesh, co-principal investigator, told Bayer's release that after a stroke "the risk of another potentially devastating event is highest in the days and weeks after the initial event, and that risk persists over time," which can be true of background risk even when the drug-versus-placebo contrast in that window misses its bar. Prespecified ischemic-stroke counts in the first 90 days were 183 versus 218, hazard ratio 0.84, interval 0.69 to 1.02, P of 0.08. Of the 134-event crude gap, 35 (26.1 percent) sit in those 90 days and 99 (73.9 percent) sit after, a count split rather than a landmark analysis.

OCEANIC-AF asked a different question and stopped early. Piccini and coauthors, on 1 September 2024, randomized 14,810 people with atrial fibrillation to the same 50 mg dose or to apixaban, and stroke or systemic embolism hit 98 patients (1.3 percent) versus 26 (0.4 percent), hazard ratio 3.79, interval 2.46 to 5.83, while major bleeding was lower at 17 versus 53, hazard ratio 0.32, interval 0.18 to 0.55. Median follow-up was 155 days. Bayer said on 19 November 2023 that the committee wanted the AF study stopped and the stroke study continued, which is how the registry records it.

Price inputs are Bayer's stated $748.50 for 30 days. The one-year NNT is the stroke-conference slide. Whole-trial gaps use the integer counts above. 553,320 multiplies CDC's "more than 795,000," its "about 87 percent" ischemic share, and the 80 percent non-cardioembolic share a Bayer spokesman gave Managed Healthcare Executive.
WorksheetArithmeticResult
Twelve packs12 times 748.50$8,982
One prevented ischemic stroke, year one, list price53 times 8,982$476,046
Same bill if a year is 365 days53 times 9,106.75$482,657.75
Whole-trial count NNT1 divided by 0.02170546.1 people
Rounded 8.4 minus 6.21 divided by 0.02245.5 people
Disabling or fatal, whole trial1 divided by 0.009236108.3 people
Spokesman times CDC inputs795,000 times 0.87 times 0.80553,320, not 520,000

A Bayer spokesman told Managed Healthcare Executive that about 80 percent of ischemic strokes are non-cardioembolic, and the magazine then wrote approximately 520,000 a year. CDC, citing the American Heart Association's 2023 update, says more than 795,000 people have a stroke each year and about 87 percent are ischemic. Multiply 795,000 by 0.87 to get 691,650, then by 0.80 to get 553,320, which is 33,320 above 520,000, and 520,000 divided by 691,650 is 75.2 percent, not 80.

What You Can Do

If the event fits the label (non-cardioembolic stroke or high-risk TIA, NIH Stroke Scale 15 or under, no atrial fibrillation, an antiplatelet already planned, liver not moderately or severely impaired), ask the pharmacy for the 30-day copay rather than $748.50, and ask whether the plan is 14.2 months of median treatment, 567 days of median follow-up, or for as long as another stroke remains a risk, the horizon a Bayer spokesperson described to Managed Healthcare Executive. Cheryl Bushnell told TCTMD that cost is "probably going to be the main limiting factor." If the copay is impossible, Bayer's approval release points to its U.S. Patient Assistance Foundation, patientassistance.bayer.us or 1-866-228-7723. Shipping starts mid-to-late November, China was approved on 26 October, and Europe and Japan are still in review.

Atrial fibrillation is the wrong shelf: hazard ratio 3.79 against apixaban, and the label excludes it. Past 75, in the first 90 days, or after a high-risk TIA, ask about the label's higher major-bleeding rates before treating the overall 1.6 versus 1.5 as the whole story. A year-one payer model should use 53 and $476,046 at list, before rebate, and should not paste the whole-trial 46.1 into a one-year cell or treat 53 as the disabling-stroke figure of 108.3. A net price, a day-365 table that rebuilds the 1.9 percent, and post-marketing fatal-bleeding counts are what would move the bill.

Limitations

This arithmetic does not prove a failed cost-effectiveness test, a clinic NNT of 53, or a rise in fatal bleeding. $476,046 assumes list price, twelve filled packs, and no rebate, the 1.9 percent is a figure note without published day-365 arm totals, and 1/0.019 only shows that 53 is the rounded reciprocal. Whole-trial NNTs are crude proportions, not the hazard model and not the Aalen-Johansen curves cut off at day 820, so 2.17 beside 1.9 mixes estimators, and 13.4 to 1 uses a bleeding gap compatible with chance. CONSORT safety sets are 6,124 and 6,130 people who took drug; the table header uses 6,162 and 6,165. Subgroup intervals are unadjusted. Black participants were 2.3 percent of the trial (282 of 12,327) and the U.S. subset was 8.3 percent (1,027, of whom 194 were Black), against CDC's note that first-stroke risk is nearly twice as high for non-Hispanic Black adults as for White adults. Bayer paid and employs authors on the paper. Patient-years are missing, so label rates cannot be reversed into counts. Marion Boulanger, via the College, wants 5 or 10 years before the net benefit can be called durable.

The Strongest Case Against This Analysis

Grant the trial its result. In 12,327 people ischemic stroke fell, hazard ratio 0.74, interval 0.65 to 0.84, P under 0.001, major bleeding did not clearly rise, disabling stroke fell, and hemorrhagic stroke was numerically lower. Earlier combination strategies, as the College quotes the paper, failed on efficacy, bleeding, or both. The atrial-fibrillation loss used apixaban, not placebo, which is why the label stays narrow. A $476,046 sticker ignores rebates, disability, and Bushnell's point that access will slow uptake, and six extra fatal bleeds inside an interval from 0.74 to 4.23 are not a demonstrated harm, any more than a non-significant bleed difference is proof of safety.

That case can be right and still leave the bill, because 384 versus 518 is not a rounding trick. The investigators' own one-year reduction, converted as they converted it, costs $476,046 per ischemic stroke at the price Bayer gave three reporters, before a discount, and a disabling stroke is a 108.3-person problem over the whole follow-up rather than 53 at one year. CDC's nearly $56.2 billion for 2019-2020 is not a per-stroke credit, 520,000 does not survive the multiplication above, and the early window called the dangerous one is the window whose drug-versus-placebo test missed 0.05.

The Bottom Line

Approval rests on 384 ischemic strokes versus 518, hazard ratio 0.74, and a major-bleeding comparison that did not separate from placebo. The slide's one-year number needed to treat is 53, twelve packs cost $8,982, and 53 of those years cost $476,046 at list price. The same 50 mg dose lost to apixaban in atrial fibrillation, the first 90 days missed their own test, and fatal bleeding's point estimate went the wrong way inside an interval that includes no increase. The sticker is not the settlement.

Related