A three-part argument on payer power, what the payer should build, and the limits of a complicated solution to a complex problem — read across the EU, the UK and China
Power, commoditisation, and the limits of procurement — and what the EU, the UK and China are each teaching us about where payer power stops
Pathway not molecule, collaboration in the open, and giving the patient back a voice we already learned how to hear — then put down
The first Joint Clinical Assessment has landed — a triumph of machinery over question, and a complicated solution to a complex problem
Power, commoditisation, and the limits of procurement across the EU, the UK and China
There was a time when the prescriber held the power. A physician decided a patient should have a medicine, named it, and the system paid. Clinical judgement and financial consequence sat in the same hand, and the question of whether the price was worth it was rarely asked out loud — certainly not by the person being asked to find the money.
Then we built an apparatus to ask it. Health technology assessment, value frameworks, the QALY, the reference price, the tender. A layer of bureaucracy grew up between the prescription and the payment — and bureaucracy is not an insult here; it is institutional memory with a budget attached — and its single job was to test a claim the prescriber had once been trusted to make alone: is this worth what it costs?
That apparatus has not stopped growing, and the obvious extrapolation is that it never will: that over the next decade the people who hold the cheque — payers, procurers, the assessment bodies that stand behind them — simply become the dominant players in the industry, and pharma becomes a supplier negotiating from a weaker chair each year. It is a tidy story. It is also, I think, wrong in the way that linear extrapolations of a complex system are usually wrong. The interesting question is not whether the payer is gaining power. It plainly is. The question is what stops that power running to completion — and what each of the three big experiments currently underway is teaching us about the limit.
Notice who is mostly absent from the apparatus. We say we want quality of life. We say we want better survival. We say we want the patient at the centre. And then the patient rarely sits at the procurement table, because procurement is a negotiation between a payer and a manufacturer about a price, and the patient is the subject of that negotiation rather than a party to it.
Should they be a party to it? In many ways that is for the health system to settle with the people it serves — what the contract between them actually is. An ethics-of-care reading, after Joan Tronto, reframes the whole exchange: the live question is not "what does this molecule cost" but "what does the system owe this person, and who is answerable for meeting it?" Those are not the same question, and one of the quiet failures of the payer-dominance debate is that it keeps collapsing the second into the first. A procurement frame can buy a medicine at the right price and still leave the question of responsibility unanswered.
The procurer has one durable instinct, and it is not a villainous one: value for money, and — past a certain point — a race to the bottom line, for want of a better phrase. The instinct produces commoditisation. A class of medicine becomes a category; the category becomes a tender; the tender becomes a price; the price falls toward marginal cost. For mature, well-substituted products this is exactly what a payer should do, and it has freed up enormous sums.
But here is the trap. Push commoditisation far enough and you begin to kill the innovation you were trying to buy — not through malice, but through arithmetic. Once the standard of care is cheap, the incremental benefit a new entrant must demonstrate becomes punishingly expensive to manufacture: bigger trials, harder endpoints, a steeper hill to clear for a smaller marginal gain over a now-commoditised comparator. The payer's own success erodes the supply of the thing it is procuring.
The payer's power is real, but it is self-limiting. Run procurement pressure to completion and you commoditise the floor of the market so thoroughly that nobody can afford to build the next storey.
That is a feedback loop, not a victory — and it is the single most important reason "the payer wins" is too simple. In a complex adaptive system, no node gets to apply its own logic without limit, because the system's other parts react. The instructive thing about the present moment is that all three of the world's major experiments are now visibly hitting this limit at once, and each is responding by recasting the value equation rather than tightening it further.
Set the EU, the UK and China side by side and the pattern is hard to miss. Each is consolidating payer power and, at the same time, importing a second logic specifically to stop procurement running to its own conclusion.
Power consolidates — and industrial strategy walks back in.
On the assessment side, payer-type power is genuinely consolidating. The Joint Clinical Assessment went live in January 2025 for oncology and ATMPs, reaches orphan drugs in 2028 and all new medicines by 2030 — a single, EU-level read on relative clinical effect, with a much broader PICO scope feeding 27 national pricing decisions. That is centralisation of the evidentiary high ground.
And yet, at the very same time, the EU is recasting the value equation in the opposite direction. The Critical Medicines Act — political agreement reached in May 2026 — explicitly tells contracting authorities to stop leaning on "lowest price", rewards EU-based manufacturing, and reframes medicines as a question of sovereignty and supply resilience rather than unit cost. Industrial strategy, having been pushed out of the pricing conversation for two decades, has walked back in through the procurement door.
A decade of tightening, deliberately loosened.
The UK is the cleanest live case of the value equation being recast. After years of escalation — the VPAG rebate on newer medicines climbing to 22.9% in 2025 as branded spend outran the cap — the screw has been deliberately loosened. The 2026 headline rate falls to 14.5%, capped at 15% for three years under the UK–US trade deal; and from April 2026 NICE lifts its cost-effectiveness threshold by a quarter, to roughly £25,000–£35,000 per QALY.
Read the motive plainly: R&D investment had visibly drained away, and the government decided that life-sciences competitiveness — industrial strategy, again — now matters as much as in-year affordability. The most disciplined payer in Europe spent ten years tightening and has just, on purpose, eased off. That is not the behaviour of an actor on an inevitable march to dominance. It is an actor discovering its own constraint.
The strongest single payer meets the same limit.
China runs the most concentrated payer power on earth: the NHSA as near-monopsony, volume-based procurement driving generic prices toward the floor, annual NRDL negotiation as the price of entry. If pure payer dominance worked anywhere, it would work here. And what is happening is the most telling signal of all — China is pulling back from pure commoditisation.
The 2025 cycle (effective January 2026) added 114 drugs, half of them genuinely novel, but declined for the first time to publish an average price cut — a deliberate de-emphasis of the discount. The 11th VBP round moved from aggressive price-only bidding to a comprehensive evaluation weighing quality, supply stability and clinical need. A stricter innovation filter is rejecting "me-too" entrants. And a brand-new commercial insurance innovative drug catalogue opens a second, higher-value reimbursement lane outside the basic scheme. The purest procurer in the world is building escape hatches to protect innovation from its own procurement.
Each jurisdiction is, in its own idiom, importing a second logic to sit alongside cost-effectiveness: resilience and sovereignty in the EU, competitiveness and investment in the UK, a protected innovation lane in China. The next ten years are not a simple story of payers winning. They are a story of the value equation being recast — and in the recasting, the payer becomes simultaneously more powerful as gatekeeper and assessor and less sovereign as decision-maker, because the system keeps adding other logics specifically to stop procurement running to its own conclusion.
This is what dominance looks like in a complex adaptive system: not a single victor, but a node whose power is continuously bounded by the reactions it provokes. The payer pushes; industrial strategy pushes back; supply security pushes back; the innovation feedback loop pushes back. Equilibrium is not the absence of power. It is power held in tension by the other forces it has woken up.
Which leads to the sharpest question of all. At present the payer's request is impossibly broad: bring us quality of life, bring us incremental benefit, bring us better mortality — across every disease at once. Should payers instead behave like sophisticated buyers and issue a genuine request for proposal? "This is the condition. This is the outcome we will pay for. Bring us that." Specify the target, rather than accepting whatever the pipeline happens to deliver.
In the complicated domain — where the problem is well-characterised, the unmet need is legible, the causal path is mapped — there is a strong case for it. An RFP for a specified outcome in a specified disease is good procurement, and it would sharpen an industry that currently optimises for what the assessment machinery rewards rather than for what the system most needs.
You cannot tender for the thing you did not know you would want.
But then: GLP-1s. No payer issued an RFP for obesity to become the defining budget event of the decade. No one specified that brief. The demand emerged — from biology, from social change, from a molecule that turned out to do far more than its original indication. That is the signature of the complex domain, where the most consequential shifts are precisely the ones nobody scoped in advance. The RFP model works where the problem is complicated and fails where the problem is complex — and the therapeutic events that reshape a health system tend to come from the second category, not the first. So the honest design is not "RFP or market" but knowing which domain you are standing in before you choose your instrument.
The payer will be more powerful, and more constrained, than the question assumes. More powerful, because the assessment apparatus is consolidating and the cheque has only grown heavier. More constrained, because every jurisdiction that has run procurement pressure hard is now discovering the same limit and reaching for the same correction — recasting value to include the industrial, the strategic, the resilient, and carving out lanes to keep the innovation feedback loop alive.
A healthy market in medicines was never going to be won by anybody. It was going to be held, by several hands at once.
No single actor gets to be sovereign in a complex adaptive system; the system will not let one logic extinguish the others without eventually breaking. The useful work for the next decade is therefore not predicting who wins. It is designing the feedback deliberately — so that the payer's necessary discipline and the innovator's necessary risk-taking hold each other in productive tension, rather than one quietly starving the other. Get that design right and the question of dominance dissolves.
Pathway not molecule, collaboration in the open, and a patient voice we built once and let lapse
The previous part ended on a deliberately unsatisfying note. No single actor wins in a complex adaptive system; the useful work is not predicting the victor but designing the feedback, so that the payer's discipline and the innovator's risk-taking hold each other in tension. Fair enough — but a design principle is not a strategy. So let me be concrete. If the payer is going to be more powerful and more constrained at once, what should it actually build?
Not a tougher tender. The payer already has price discipline in abundance; that is the one muscle every system in the last part has over-trained. What it lacks are three capabilities of a different kind — the ability to see the pathway rather than the product, to work in the open rather than across a table, and to hear the patient as a participant rather than a subject. None of the three is new. The third we built, in this country, more than twenty years ago, and quietly let lapse. The strategy for the next decade is largely the work of picking these three up and wiring them into how the system runs.
A payer that buys molecules is optimising a component and calling it a system. Take interstitial lung disease, the exemplar I keep returning to. The antifibrotic is the visible, negotiable, commoditisable object — so it is the thing the apparatus assesses, prices and rebates. But the molecule is a minority shareholder in the patient's outcome. The years of diagnostic odyssey before anyone names the disease, the late referral, the unmanaged exacerbation that puts someone in hospital, the monitoring that does or does not catch progression — that is where the cost lives, and where the outcome is won or lost. Price the drug to the floor and you have disciplined perhaps a fifth of the value at stake while leaving the other four-fifths untouched.
This is the reductionist error stated in market-access terms: a complex system's behaviour does not decompose into the sum of its optimised parts. A pathway is not a list of purchasable components; it is a set of relationships, hand-offs and delays, and the leverage sits in the relationships, not the components. A payer strategy worth the name therefore moves the unit of contracting up a level — from the molecule to the journey. Pay for diagnosis-to-outcome, not drug-to-response. Share risk across the pathway, not just on the line item. Let the value-based agreement span the referral and the monitoring, because that is where it can actually move the number.
Price the drug to the floor and you have disciplined a fifth of the value while leaving the other four-fifths untouched.
There is a second prize here, and it answers the trap from the last part directly. If competition happens at the level of the molecule, commoditisation eventually starves innovation. If it happens at the level of the pathway, innovation has somewhere else to go — into diagnostics, digital pathway tools, care redesign, the things that move the four-fifths. Pathway thinking does not just buy better; it relocates the innovation race to ground where a race to the bottom does less damage.
The dominant texture of the payer–manufacturer relationship today is opacity by design. Confidential discounts. Managed-access agreements behind non-disclosure. Commercial-in-confidence net prices that even the system's own analysts cannot see. China's new commercial-insurance lane prices the negotiated discount confidentially; the UK's patient-access schemes are routinely commercial-in-confidence. Each opaque deal is locally rational and systemically corrosive, because opacity is the medium in which gaming grows and trust dies.
A complexity-informed payer treats the health system as a partner in a shared, shifting problem rather than a counterparty to be extracted from. That means working with the cards face up: shared horizon-scanning, jointly designed evidence generation, open real-world data on what the pathway is actually doing. The embryo of this already exists — the EU's joint scientific consultation lets developers and assessors agree the evidence question before the trial is run, which is collaboration before commitment rather than litigation after it. The strategic move is to make that the norm rather than the exception, and to extend it from evidence into pricing and pathway design.
The deeper reason is what Nora Bateson calls warm data — the information that lives in the relationships and context between the parts, not in the parts themselves. Confidential bilateral deals freeze that information in silos; transparent collaboration keeps it flowing, which is the only condition under which the pathway-level risk-sharing of the first move becomes possible. You cannot share risk with someone you are hiding the numbers from. Transparency is not a virtue bolted on for its own sake; it is the precondition for everything else the strategy needs to do.
We say the patient is at the centre, and then we seat them as the subject of a negotiation between two institutions rather than a party to it. The third capability is to give patients a structured, standing voice in the value questions that decide their care — not a satisfaction survey after the fact, but a hand on the decision before it is made. And the striking thing is that the most ambitious version of this was constructed two decades ago, here, and then allowed to fade.
In 2002 NICE established a Citizens Council — thirty members of the public, stratified to reflect the adult population of England and Wales by age, gender, socioeconomic status and ethnicity. It was the first body of its kind in the UK, and the appetite was real: more than four thousand people applied for the thirty places. It ran citizens'-jury style — members heard expert witnesses on a hard question, deliberated over several days in open session, and reported out to public comment.
Crucially, it was asked the right kind of question. Not "should NICE fund this drug" — that is a technical matter dressed as a moral one — but the genuine value trade-offs underneath: the rule of rescue, the weight to give age, whether rarity alone should buy a higher price, what "clinical need" even means. Its conclusions fed NICE's Social Value Judgements document, the principles its committees were then bound to apply. And the public, deliberating, produced answers that were more nuanced and in places tougher than the advocacy-driven shouting match — concluding, for instance, that rarity by itself does not justify a higher cost per QALY; severity and the scale of benefit have to be in the frame too.
It was not perfect, and it is worth being honest about why it lapsed. Some of the questions put to it were too loose to answer cleanly, a criticism aired in The Lancet at the time. It fed only the broad framework, never live decisions, so its influence was real but diffuse. It was expensive. And, fatally for its longevity, it was a set-piece — an annual seminar rather than a continuous channel. It was an event, not an operating system, and events are easy to cancel.
So the future move is not to rebuild the committee. It is to build the pattern the Council pointed at — representative, deliberative, transparent — but continuous, and woven into the pathway rather than convened once a year in a hotel. New patterns for patients to have a voice with their health systems: standing deliberative panels with a real route into contracting; patient-reported outcomes that actually feed the value-based agreement rather than decorating the appendix; and — to close the loop with the last part — citizen input into what the request for proposal even asks for. The patient should help write the brief, not merely receive the verdict.
In the language of an ethics of care, after Tronto, this is the missing fourth phase. Attentiveness, responsibility and competence a good payer can manage on its own. Responsiveness — knowing whether the care actually met the need, from the only people who can tell you — requires the care-receiver in the loop. A payer strategy without patient voice is care without responsiveness: it can buy efficiently and still have no reliable way to know whether it met the need it was buying for.
The temptation will be to pick one. Run a pathway pilot, or publish some prices, or stand up a patient panel, and declare the box ticked. But the three moves are load-bearing only together, and the reason is structural.
Pathway contracting without transparency just relocates the opacity one level up — now it is the whole journey that is priced behind a non-disclosure instead of the molecule. Patient voice without pathway thinking gives people a say over a line item when their real stake is the entire journey through the system. Transparency without patient voice is simply two institutions agreeing in daylight, over the patient's head, and calling it progress. Each move needs the other two to mean anything; assemble all three and you have a payer that can see the system, work inside it honestly, and hear the people it is for.
This is, finally, a Cynefin point. The value question is complex, not merely complicated, and the persistent error of the last two decades has been to attack a complex problem with ever-more-refined complicated tools — a cleverer QALY, a finer threshold, a more elaborate appraisal. You do not resolve a complex problem with a better algorithm; you resolve it with better process: deliberative, transparent, continuous, pathway-wide. The instrument has to match the domain it is used in, and the domain here was never the kind that yields to a sharper calculation.
The payer of the next decade is not, at root, a harder buyer. It is a better convener. Its strategy is less about the price discipline it already has in surplus and more about three capabilities it currently lacks — seeing the pathway, working in the open, and hearing the patient. The first relocates value to where it actually sits. The second makes the risk-sharing that the first requires possible at all. The third tells the system whether any of it worked, from the only people entitled to say.
We glimpsed the third capability twenty years ago, built it with some courage, and set it down because it was inconvenient to keep. The work now is to pick it back up — and this time to wire it into how the system runs, as an operating system rather than an annual seminar. Do that, and the question that opened the last part stops mattering. The point was never for the payer to win. It was for the payer to build the thing that lets everyone — innovator, system, and patient most of all — lose less.
The first Joint Clinical Assessment has landed — a triumph of machinery over question
The first Joint Clinical Assessment has landed. Like the legendary supercomputer that spent an age computing the Answer and returned a flawless, baffling number, it is a triumph of machinery over question.
There is a scene every reader of Douglas Adams remembers. A civilisation builds a computer of unimaginable power and asks it the ultimate question — the meaning of life, the universe, everything — then waits seven and a half million years for the result. The machine, which has done exactly what was asked of it, flawlessly, returns a single number. Forty-two. And the assembled philosophers realise, slowly and with horror, that the fault was never in the computer. It computed beautifully. They had simply never worked out what the question actually was.
I have been thinking about that scene since the European Commission published the first Joint Clinical Assessment in June 2026. Not because the JCA is absurd — it is the opposite of absurd; it is meticulous, expensively built, and staffed by serious people doing careful work — but because it has the same shape. An apparatus of extraordinary sophistication has been constructed to compute an answer, and in its first run it computed one, rigorously. The question of whether that was the right thing to be computing for the problem we actually have is the part nobody has slowed down to ask.
The inaugural JCA assessed tovorafenib — Ipsen's Ojemda — for paediatric low-grade glioma, a slow-growing brain tumour in children. Ireland's NCPE acted as assessor, Germany's IQWiG as co-assessor; the marketing authorisation came on 22 April 2026, the report was endorsed eight days later, and published on 9 June. By the standards the regulation set itself, this is a genuine milestone: one clinical assessment, prepared in parallel with the EMA review, available to all member states at once. Real harmonisation. Real transparency. Grant all of that without reservation.
Now read what the machine actually returned. The assessment was split across multiple populations and built out into six, then seven, PICO questions — the population–intervention–comparator–outcome frames that define what is being compared with what. And for much of that scope, the rigorous conclusion was a careful account of what could not be concluded. For one PICO the only available evidence came from unanchored indirect comparisons carrying, in the report's own framing, major uncertainties whose results should not be read as causal. For others the developer could not find a fit-for-purpose comparator at all, so the single-arm data on the drug were not even discussed, because they say nothing about a relative effect. One PICO's results were excluded outright for insufficiency.
A first-in-class drug for a disease with almost no comparators, run through seven comparative questions to produce an immaculate map of what cannot be compared.
This is not a failure of the assessors. It is the system performing exactly as designed and, in doing so, revealing what it was designed for. Tovorafenib treats a rare childhood cancer where there is barely a comparator to anchor to; the modern evidence base for such a product is, almost necessarily, single-arm. Asked the comparative-effectiveness question seven ways, the machine answered seven ways, with precision, and the answer for much of the indication amounted to: we cannot say. Forty-two, rendered in MAIC and confidence intervals.
The deeper issue is a Cynefin one, and it is not a quibble about a single rare disease. It is that the JCA applies the playbook of the complicated domain to a problem that lives in the complex one.
A complicated problem has a right answer, reachable by expertise and standardisation. Engineering a bridge is complicated: hard, specialised, but knowable, and the same analysis works in Lisbon as in Helsinki. A complex problem has no single right answer because it is made of many interacting agents and irreducible context — twenty-seven health systems with different standards of care, different pathways, different epidemiology, different things they are willing to pay for and different reasons. Market access is complex in exactly this sense. The JCA treats it as complicated: one centralised analysis, one method, one endorsed report, the assumption being that the national variation was duplicated effort to be engineered away rather than real difference to be worked with.
That assumption is the category error, and everything downstream follows from it.
When you apply a complicated-domain solution to a complex problem, it does not fail loudly. It fails by multiplying. The diagnostic sign is always the same — the apparatus meant to simplify starts generating more of the very thing it was built to reduce.
Watch the PICOs. To honour the genuinely different questions of all member states inside one assessment, the consolidated scope does not converge on a single comparison; it fans out. Ten or more PICOs per product is now the working expectation, and as the count climbs the report becomes harder, not easier, for any one national body to interpret. The harmonisation machine manufactures the heterogeneity it was supposed to abolish — because that heterogeneity was never noise in the map. It was structure in the territory.
The other symptoms are the same disease. Duplication is not actually removed: the JCA covers relative clinical effect only, while economic evaluation, the additional-benefit judgement, pricing and reimbursement all stay resolutely national — and because the regulation cannot compel follow-up data, national bodies may simply re-open assessments later as evidence matures, quietly reintroducing the duplication the whole edifice was meant to end. Uptake of the joint report by national agencies is, in the end, voluntary. The developers who hold the evidence sit largely outside the formal scoping of the questions. None of these is a teething problem to be ironed out. Each is the complex problem reasserting itself against a tool built for a complicated one.
And the response to the strain, predictably, is more machine. The dossier must arrive within a hundred days of the consolidated PICOs, in parallel with the EMA submission, stretching the same teams across both. The 2026 work programme scales from roughly ten assessments to around fifty; high-risk devices enter from June 2026; orphan medicines in 2028; every new medicine by 2030. Around the edges grow AI tools to predict the PICOs, more scientific-consultation slots, more guidance, more cross-functional alignment. Every one of these is a sensible local fix. Not one of them touches the category error. They make the machine more elaborate while leaving the question it is computing exactly where it was.
This is the quiet way complicated solutions to complex problems go wrong. They do not produce a visibly wrong answer you can point at and reject. They produce an ever more intricate, ever more expensive apparatus that is immaculate on its own terms and drifting steadily away from the thing that mattered — and the institutional reflex is always to refine the apparatus rather than to ask whether it is aimed correctly. In the book, the bypass gets built, the paperwork was technically on display, and the question of whether anyone actually wanted the road never quite comes up. You can feel the system becoming more complicated. Never mind the cost.
The sharpest cost is not the budget. It is the set of questions the JCA cannot, by construction, address — and which happen to be the ones that decide whether a patient actually benefits.
The JCA scores the molecule's relative effect. Whether a child in a given member state is diagnosed at all, referred in time, and reaches a centre able to deliver the drug — the pathway, where four-fifths of the outcome and cost actually live — is nowhere in the assessment. A perfect comparative number changes nothing for a patient the system never reaches.
Frontier medicine — rare disease, first-in-class, cell and gene therapy — increasingly runs on single-arm evidence, because that is what the science and the ethics of these diseases permit. A frame whose native demand is the anchored comparison keeps returning "not estimable" for precisely the most innovative products, and quietly pressures developers to design trials for the assessment rather than for the patient.
Resilience, sovereignty, security of supply, affordability — the second value logic now reshaping European access through the Critical Medicines Act — sits entirely outside the JCA's remit. It answers a clinical question while the strategic questions that increasingly govern whether Europe makes and can afford a medicine are decided in another building.
A genuinely complexity-informed approach would not try to compute one clinical answer for twenty-seven systems and call the variation a problem. It would probe and sense rather than analyse-and-prescribe: transparent evidence-sharing across borders, real-world data generated in the different contexts where the drug is actually used, national pathways allowed to adapt rather than forced to converge, the question framed with the developers and patients who hold the relevant knowledge rather than around them, and an honest acceptance that the answer is plural and emergent — not a single endorsed PDF. The JCA need not be abolished to make room for this. It needs to stop being mistaken for the whole of the task.
The lesson of that supercomputer was never about the number. It was that a sufficiently powerful machine will give you something immaculate and beside the point if the question was never properly formed — and that the temptation afterwards is always to build a second, bigger machine to work out what the question should have been. The JCA is good, on its own terms, at a complicated thing. The danger is that being good at the complicated thing becomes the reason not to face the complex one: that the energy, budget and political capital pouring into perfecting the assessment machine are energy, budget and capital not going into pathway, innovation and industrial strategy — the things that actually gate whether a European patient benefits.
Three parts, one argument. The payer will not simply win, because a complex adaptive system bounds any single logic that tries to run to completion. What the payer should build is therefore not a harder tender but the capacity to see the pathway, work in the open, and hear the patient. And the warning of the first Joint Clinical Assessment is what happens when we forget all of that and reach, once again, for a complicated machine to settle a complex question.
Forty-two is a fine answer. It is precise, it is defensible, it survived seven and a half million years of computation. We just have to be honest about whether it is the answer to the question we have — or only to the one the machine was able to ask.