Then, in October 2024, it stopped visiting them: 6,082 meals in September, 184 in October, and 53 in all of 2025.
In short
In October 2024, Xarelto stopped paying for meals with physicians. In September it had reported 6,082; in October, 184, and in December, 13. In all of 2025 it reported 53. Eliquis, that same year, 115,819.
Before that cut it was already losing, and not for lack of effort: in 2023 it reported more than twice the dollars in payments to physicians that its rival did, and in Medicare, the public insurance for people over 65 and people with disabilities, it now sells a quarter as much (6.5 against 26.4 million prescriptions in 2025). What it lost was the physician's choice with a new patient: since 2015 Eliquis adds more patients than it does every year, counted net, because the public file does not separate a new start from a brand switch.
My hypothesis, with the dates in view, is that it was weighed down by the reputation of causing more bleeding than Eliquis: advertising and the lawsuits built it, and from 2019 Medicare studies and the geriatrics criteria reinforced it. There is evidence against, and it is below.
Bayer, which collects royalties in the United States, sued Johnson & Johnson for abandoning in-person detailing, and the payments line up with what it alleges. My recommendation is not to rebuild detailing, and to measure with the 2025 physician-level prescriptions what the cut cost, because both products can go generic in 19 months and, with my cost assumptions, a sales force does not pay for itself in that window. The other decision comes first, defending access in 2027, which I cover in the other case.
Xarelto lost the new patient to Eliquis in the years when advertising, the lawsuits and Medicare evidence gave it the reputation of causing more bleeding, and when its reported meals with physicians fell from 6,082 to 184 in October 2024, it left behind, without meaning to, the only public way to measure what that detailing was worth.
Xarelto reached the market first and in 2013 held 92 of every 100 prescriptions of the two in Medicare, but Eliquis passed it in 2016.
Xarelto did not lose patients until 2024. What it lost was most of the ones being added each year, in a market moving from warfarin to direct anticoagulants. In 2014 it still added more patients than Eliquis; from 2015 Eliquis adds more every year, and between 2020 and 2024 it took 93 of every 100 patients the two of them added.
With the patients it already has, it does fine. Measured in days of therapy per patient per year, which comes from standardizing fills to 30 days, Xarelto delivered 270 days in 2024 and Eliquis 263, and it comes out ahead every year since 2016. In raw prescriptions it looks like the opposite, 5.04 against 5.44, only because Xarelto dispenses more 90-day fills. Nor should this be read as retaining better: a patient who starts mid-year lowers the average, and Eliquis adds patients every year, so the bias works against Eliquis. What this does say is where the problem is not. It is not in the ones who stay, it is entirely in the ones who start.
Plan restrictions do not explain it: on the 10, 15 and 20 mg tablets, the ones a physician picks instead of Eliquis, no Medicare plan required step therapy, which means making the patient try another drug first, in 2023 or in 2026. That is covered in the access case.
The second suspicion is that Eliquis was putting in more. Manufacturers report every payment to a physician in a public registry, Open Payments. In 2023, payments that name Xarelto add up to $6.73 million; those that name Eliquis, $3.14 million. Almost half of Xarelto's are fees to 340 physicians for speaking, more than ten times what Eliquis paid.
The same holds for visits, counted through reported meals: of every 100 meals that name one of the two products, 44 are Xarelto's, and of every 100 Medicare prescriptions for the two, 23. That is share of voice against share of market, and the first nearly doubles the second.
57% of Eliquis meals also name other Pfizer products and here they count in full for Eliquis; split up, Xarelto's lead would be larger.
If it was short of neither money nor visits, the question is where they went in 2023, the last full year with visits. Each dot on the chart is a county: to the right, where Xarelto sells above its 22.6% national share of the two; above, where it visits more than it sells.
Almost half of the visits by the two products (49.1%) fall in the upper left, in the 439 counties where Xarelto sells little and visits more than it sells.
The top 10% who prescribe the two the most write 49.4% of their prescriptions, the concentration consulting calls Pareto, and receives 25.8% of Xarelto's visits. In reach, how many of those physicians each brand sees, Eliquis got to 37.3% and Xarelto to 28.3%. And 29 of every 100 Xarelto visits go to professionals with fewer than 11 prescriptions of each product.
Visiting where it sells little could be building share. To find out I looked at 2024, with the bar set before opening the data: one share point or more against Eliquis.
By county. Where it sells little, heavily visited counties lost 1.86 points and comparable lightly visited ones, 1.91; where it sells a lot, 2.58 against 3.23. The differences, 0.05 and 0.65 points, stay below the bar.
By physician. Between those who received 4 to 9 visits and comparable ones with none, the difference runs from −0.15 to +2.94 points depending on what is assumed about those who in 2024 dropped out of the file for prescribing fewer than 11 times, more common among the unvisited.
If it was not the restrictions or the money, what remains is why the physician chose Eliquis. Public data does not show what each one thought, so what follows is a hypothesis: that the literature, the news and the advertising gave Xarelto the reputation of causing more bleeding, and that this reputation weighed on the choice.
Until 2016 there was nothing to compare them with. The two trials that approved them came out the same month of 2011 and measured against warfarin, not against each other. ARISTOTLE, paid for by BMS and Pfizer, found apixaban superior at preventing stroke and with less major bleeding; ROCKET AF, paid for by J&J and Bayer, found rivaroxaban non-inferior, without superiority. Read side by side they seem to favor Eliquis, but the 2023 cardiology guideline points out that warfarin was better dosed in ARISTOTLE, so the comparison is uneven.
What did grow in those years was the noise. In December 2014 the bleeding lawsuits against Xarelto were consolidated, and J&J counted 5,000 of them in January 2016 and 16,900 a year later; Bristol Myers Squibb and Pfizer put $712 million into advertising between 2014 and 2016, and a public health report described Eliquis as "heavily marketed as a safer alternative to Xarelto".
From 2019 the evidence followed that reputation. The FDA, the agency that approves drugs in the United States, using Medicare data, associated rivaroxaban with 2.7 times the extracranial bleeding of apixaban; the Beers Criteria of the American Geriatrics Society asked for caution in people over 75 and in 2023 moved to asking that it be avoided in long-term treatment of atrial fibrillation and venous thromboembolism (VTE); the largest study in Medicare, with 581,451 patients and public funding, associated rivaroxaban with 18% more serious events. These are observational studies: they measure association, and their authors warn about unmeasured confounding. In those years Xarelto added fewer and fewer patients: 83,508 in 2019, 12,425 in 2023, and in 2024 it lost 2,831.
The first head-to-head trial arrived in 2026, and not in atrial fibrillation. COBRRA, publicly funded in Canada, compared the two in venous thromboembolism and found relevant bleeding in 3.3% with apixaban and 7.1% with rivaroxaban. For atrial fibrillation, the trials are still recruiting and would finish at the end of 2027.
What does not fit the hypothesis:
According to the lawsuit Bayer filed against Janssen in October 2025, J&J cut the people visiting physicians for Xarelto from 1,198 to 345 at the end of 2022 and in November 2024 eliminated its in-person sales force. This is an allegation: in August 2026 the judge refused to dismiss it without ruling on the facts.
The payments J&J reports confirm the dates with independent data. In Open Payments, Xarelto meals with physicians ran from 5,745 to 11,080 per month between January and September 2024, and fell to 184 in October. Speaker payments ended in November and in all of 2025 it reported 53 meals. The 2022 cut shows up too: monthly meals fell from 14,907 in September to 7,256 in October of that year.
A reported meal is not a visit and does not give the reason for the cut, but it shows that Xarelto's in-person effort stopped. Its medical information channel also acknowledges that no randomized trial compared the two in atrial fibrillation. In the first quarter of 2026 its United States sales fell 7.0% on "continued share erosion".
What J&J already chose. It cut detailing and stopped paying for talks: in a product's life cycle that looks like harvesting, protecting margin until the generic arrives instead of defending share.
Whose decision it is, and what the goal is. J&J's, which answers for the brand in the United States, and in part Bayer's, which collects royalties without controlling promotion: to know what the detailing that was cut was worth, before deciding whether to rebuild it or keep harvesting.
What is at stake. One share point with the 4,142 high-volume prescribers only Eliquis sees is 20,415 prescriptions a year, around $7.2 million as a ceiling, at the price negotiated for 2026. With my assumption of 150 to 175 physicians per representative at $250,000 to $300,000 a year each, covering them again takes 24 to 28 representatives, between $6.0 and $8.4 million a year: the count is driven by coverage rather than by calls, because they are spread across the country.
Why it can be measured. The October 2024 cut is an experiment nobody designed: every physician who was receiving Xarelto visits lost them the same month, while the ones getting Eliquis visits kept them. The comparison works without inventing a control group.
The first step. When CMS, the agency that runs Medicare, publishes the 2025 physician-level prescriptions (the 2024 ones came out in May 2026), compare the share against Eliquis of the physicians who were receiving visits with that of comparable physicians who never received any, using the same method I used for 2023 and 2024. It answers the question that decides the rest.
The definitions, before measuring. For the 2025 comparison to hold against the 2023 one, three rules have to be set before opening the file: what counts as a visit, what counts as dropping out of the file, and what counts as a patient on therapy, which here means 30-day fills and not prescriptions. If the commercial team and the analysis use different definitions, the result compares to nothing.
What I would measure, and how far.
Physicians who lost the visit
Volume, which warns earlier than the physician-level data
What I would say to the physician with no visit. The one thing with public backing: once-daily dosing, which even the 2023 Beers Criteria accept as a reason to choose it. In days of therapy per patient Xarelto does not fall behind, although that measure cannot support better retention.
What I would not build.
The kill criterion and the trade-off (what is being risked). With those assumptions of mine, detailing would have to hold between 0.8 and 1.2 points with those physicians just to break even, and the clock runs against it because Eliquis cannot have a generic before April 1, 2028, 19 months away, while a sales force, by my estimate, takes two or three quarters to build and to show up in prescriptions. So the measurement is not there to reopen the field discussion, but to know whether the harvest came out cheap or expensive. What is being risked meanwhile is the relationship with Bayer and the lawsuit it opened.
The CMS files were downloaded between September 12 and 15, 2026. There are no market projections: the only modeled block is what rebuilding detailing would cost.
| Item | Status | Source or decision |
|---|---|---|
| Prescriptions by physician and drug | data | Part D prescribing, CMS, 2023 and 2024 |
| Payments from each manufacturer to each physician | data | Open Payments, CMS, 2023 to 2025 |
| Counting visits through reported meals | my call | they are 97.5% of Xarelto's payments and 99.8% of Eliquis's |
| Counting a shared meal in full for Eliquis | my call | 57% of its meals; splitting it favors Xarelto |
| Placing each visit where the physician prescribes | my call | the physician's ZIP code in Part D; the payment's own ZIP fails in one of every five Xarelto meals |
| Prescription and patient series, 2013 to 2026 | data | prescribers by geography and drug and quarterly spending, CMS |
| Days of therapy per patient | data | CMS-standardized 30-day fills over the patients of that year; someone starting mid-year lowers the average, and that works against Eliquis |
| What rebuilding detailing would cost | assumption | 150 to 175 physicians per representative and $250,000 to $300,000 a year fully loaded, industry ranges; J&J publishes neither |
| Studies, guidelines and clinical criteria | attributed | each one with its design and its limits; the case does not claim which blood thinner is better |
| That clinical perception explains the choice | hypothesis | dates that partly line up; there is no data on what each physician thought |
| The cut and end of Xarelto's in-person detailing | allegation, with the cut visible in payments | Bayer's lawsuit, undecided; the meals fall in October 2024 in Open Payments |
| Measuring the cut with the physicians who lost it | my call | against comparable physicians who never received visits, with the one-point bar set beforehand |
What would refute the thesis:
This is independent work, built from public sources. I have no commercial relationship, past or present, with Johnson & Johnson, Bayer, Bristol Myers Squibb or Pfizer.
If you read this far, something about the problem caught your interest. I post every new case on LinkedIn and discuss these decisions there with people who live them. To take it private, write me what you would do differently or pick a time to talk.
See all projects