The four ICBs still writing brand where the guidelines say generic.

The national picture is clean: branded apixaban prescribing has fallen from 84% of items in 2022 to under 6% in May 2026. Four ICBs sit stubbornly above the national line, at 11–19%.

The fall had three phases, and the four outliers only make sense against them. Phase one was the switch year: between early 2022 and late 2023 the branded share dropped from 84% to around 40%, as medicines optimisation teams ran programmed switches practice by practice. Phase two was the long middle of 2024, when the line slid from 40% to the mid-teens as repeat prescriptions came up for their annual reviews and quietly flipped. Phase three is the grind we are in now: fractions of a point a month, because everything easy has already been done. Thirty-eight of the 42 ICBs now sit between 3% and 8% branded. The remaining four sit at roughly 19%, 16%, 13% and 11%. The shape itself is familiar from every big generic switch of the past twenty years, and atorvastatin traced much the same curve after its 2012 expiry. The interesting information is never in the curve. It is in who lags it, and by how much, and why.

One measurement note before we name anyone. We count an item as branded where the prescriber wrote a brand name, an AMP in dm+d terms, rather than the plain generic. Brand-written is not the same as brand-dispensed, because reimbursement follows what is written and a generically written script can be dispensed from any supplier's stock. Everything below is rebuilt from the open EPD releases, the same series that powers the rest of DoseTrend, so every number in this piece can be checked by anyone with a laptop and patience.

Why apixaban as the case study? Because it is the cleanest natural experiment the data offers: the most prescribed anticoagulant in England at around 17.1m items a year, written in every ICB, with a generic switch recent enough that the habits formed before it are still visible in the series.

Who they are

Cheshire & Merseyside. Hampshire & Isle of Wight. Bath, Swindon & Wiltshire. Devon. Three of the four share a historical formulary quirk, one doesn't.

A caution before we profile anyone. Prescribing data shows what happened. It never shows why. What follows is the pattern in the monthly series, plus the most plausible reading of public formulary documents and committee papers. Where we are inferring rather than observing, we say so plainly. That distinction matters more here than in most pieces we publish, because this one names names.

Cheshire and Merseyside, at 19%

The largest of the four, serving about 2.7 million people, and assembled in 2022 from nine former CCGs. That inheritance matters, because the nine legacy formularies agreed with each other on very little, and anticoagulation was one of the sharper disagreements. Several Merseyside localities recommended writing DOACs by brand in the late 2010s, on a patient-safety argument: people moving off warfarin were juggling unfamiliar tablets, and a consistent box was thought to reduce confusion. When the national commercial arrangement on branded apixaban arrived in early 2022 and briefly made the brand the cost-effective choice, brand-first templates stopped being a quirk and became policy. Repeat templates across the patch were built accordingly.

The arrangement lapsed. The templates did not. The monthly series shows branded share sitting on a plateau between 21% and 22% for the whole of 2024 and most of 2025, which is what committee inertia looks like when you plot it. Then, in March 2026, the area prescribing committee adopted a generic-first line with a template amendment programme attached. April printed 19.4%, the sharpest monthly fall any of the four has recorded. Two data points are not a trend. But the first months of a switch programme are usually its fastest, and this one has started at pace.

Hampshire and the Isle of Wight, at 16%

Just under two million people, and the odd geography of the four: a stretch of the south coast plus an island whose medicines arrive by ferry. During the supply disruptions of 2022 the formulary retained branded apixaban with a 'continuity of supply' rationale, which was a sensible hedge at the time. The interesting part is what happened next, which is nothing. The wording has not been revisited, and anticoagulation has not appeared on the published committee agenda since.

The series is the giveaway. Branded share has sat in a band between 15.5% and 16.5% for fourteen consecutive months. No slope, no steps, no seasonal shape. A flat line in prescribing data almost always means the same thing: nobody has been asked to change, so nobody has. Committee time is the scarcest resource in medicines optimisation, and apixaban has simply never won the argument for a slot.

Bath, Swindon and Wiltshire, at 13%

The smallest of the four at roughly 950,000 people, formed from three CCGs in 2020. One of the three legacy areas ran an anticoagulation safety initiative in 2019 that recommended brand-name prescribing while multiple generic suppliers were entering the market, so that patients would keep seeing one consistent box. The legacy still shows in the geography: practices in that locality run close to 20% branded, while the rest of the ICB sits near the national line. It is a useful reminder that an ICB average is often three different stories wearing one name.

The series tells a gentler story than the other three. A steady decline of roughly a quarter of a percentage point a month, held for two years, with no step changes anywhere. That is not a programme. That is attrition: prescribers retiring, patients switching as reviews come round, templates being tidied one at a time. At the current pace BSW meets the national line around the middle of 2027 without anyone lifting a finger. Whether anyone should wait that long is a fair question for its committee.

Devon, at 11%

Devon is the one that does not fit, and we flagged it above. There is no formulary quirk to point at: the Devon-wide formulary has said generic-first since 2023, unambiguously. What Devon has instead is structure. One of the oldest age profiles in England, a high density of care homes along the coast, and a rural west where dispensing practices hand the medicine over themselves. Dispensing practices matter because the practice supplies what it stocks, and a small dispensary rationally stocks one line and stays with it. The branded residue is not spread evenly. Around a fifth of practices account for well over half of the branded items.

The series is the noisiest of the four, wandering between 11% and 13% with no trend in either direction. That is what a residual made of many small pockets looks like. There is no single committee decision that would move it, which is exactly why it may prove the most stubborn of the four.

Why habits persist

Three mechanisms explain most of what you have just read, and none of them involves anyone behaving badly. The first is formulary inheritance. ICBs formed in 2022 inherited as many as nine local formularies each, and harmonising them is a queue sorted by clinical risk and by savings. Generic apixaban is cheap, so switching the stragglers saves comparatively little, so the item sits near the bottom of the agenda, meeting after meeting. Nobody decided to keep the brand. It simply never came up.

The second is the electronic prescription system. The brand name lives inside the repeat template, and a repeat can be re-authorised for years without a human rereading the drug field. Fixing it is a per-patient edit, done in software that was not built for bulk changes, by practices that are not paid for the tidying. Nothing about it is difficult. It is merely nobody's job. Practice pharmacists have been chipping away at these templates for years regardless, which is a large part of why the national line still falls at all.

The third is the care-home loop. Medicines administration record charts are printed with the name on the box. If the box changes, the chart, the ordering sheet and the staff briefing all change with it, and any mismatch between chart and box is an incident form waiting to be filled in. So homes quite reasonably ask prescribers to keep the name stable, and prescribers quite reasonably agree. Multiply that by a coastline of care homes and you have a good share of Devon's residual on its own.

Why it matters commercially

For a marketing team, these ICBs are outliers worth understanding. For a payer team, they may be a lever.

Run the originator arithmetic first. Under 6% of a 17.1m-item market is still around a million branded items a year, and these four ICBs hold a disproportionate share of them. At list, a 56-tablet branded pack is £53.65 against a generic reimbursement price of a fraction of that, which is exactly why medicines optimisation teams will get to the stragglers eventually. The commercial question is what kind of share each ICB holds. Habit-backed share, like Devon's, erodes slowly and rewards presence on the ground. Committee-inertia share, like Cheshire and Merseyside's until March, can vanish on a single agenda item, which is what the April print suggests has begun.

For a generic or biosimilar team the same chart reads in mirror image. The remaining branded share is not won nationally. It is won template by template and home by home, and the monthly series tells you where the ground is already moving. A switch programme has visibly started in Cheshire and Merseyside, so supply reliability will be the first question every pharmacist there asks. Nothing at all is moving in Hampshire and the Isle of Wight, which means the first well-evidenced nudge to that committee could bring ten points of share with it. None of the four justifies a national campaign on its own. All four together are worth a morning of any brand manager's week.

The honest caveats

Now the paragraph our lawyers would insist on, if we had lawyers. EPD carries no indication data, so we cannot see whether an item was written for atrial fibrillation, VTE or anything else, and dose is an unreliable proxy. There is no patient-level view, so everything here counts items, with all the caveats set out in our items-versus-patients piece. Brand-written is not brand-dispensed. And at practice level the monthly numbers become small enough to be noisy, which is why we profile ICBs and not surgeries.

The larger caveat is about the stories themselves. The data supplies the what. The why is inference from public documents, and we have marked it as inference throughout. Treat each profile as a hypothesis worth taking into a meeting, not a finding you could take into court.

What we will be watching: whether Cheshire and Merseyside's fall holds its early pace, and whether anything at all lands on Hampshire's committee agenda this summer. The next data arrives on the first working day of the month. If the picture changes shape, this piece will too.

Written by the Cooply Solutions team. Corrections welcome at hello [at] dosetrend [dot] com.