What this covers

Vinayak Prasad, an oncologist and author of Malignant, joins Russ Roberts on EconTalk to examine why cancer drugs are expensive, ineffective, and unlikely to change without structural reform. The conversation treats cancer pharmaceuticals as a case study in broken incentives: most approved solid-tumor drugs extend life by a median of 2.1 months while costing $10,000–$16,000 monthly, yet Medicare must pay any FDA-approved price and cannot negotiate. Prasad argues the problem is not individual malice but the absence of a functioning market—companies design their own trials, conflicts of interest pervade every decision point, and patients bear no direct cost. The fix, he contends, is to realign incentives so that doing the right thing becomes profitable rather than relying on the character of good people.

The conversation moves across the machinery of drug approval and pricing. Prasad traces how surrogate endpoints—measures like tumor shrinkage defined by a 1970s experiment with calipers and foam rubber—became codified as proxies for meaningful benefit without evidence they predict survival or quality of life. He documents how control arms are weakened, trial populations skewed younger and healthier than real patients, and post-market confirmatory studies abandoned, so approved drugs often deliver far less in the world than in the trial. The money flows create a web of Conflict of Interest Bias: patient groups, thought leaders, journals, and professional organizations all receive pharmaceutical funding, leaving almost no independent voice pushing back on price or demanding evidence of true benefit. Cost-per-QALY figures routinely hit $900,000 or higher—far above thresholds even permissive systems will defend. Prasad and Roberts explore whether speed-to-market and lower evidence standards help patients or primarily serve industry, and whether patients can reasonably be expected to resist drugs their doctors recommend when the system itself strips away the feedback that makes resistance rational.

Sharpest takeaway

Prasad argues that most cancer drugs cost too much and deliver too little because the system is not a real market—Medicare must pay any FDA-approved price, conflicts of interest pervade every decision-maker, and trials are designed by the very companies that profit—so the fix is changing incentives, not blaming individuals.

  • The median overall-survival benefit of approved solid-tumor drugs is only ~2.1 months, while prices run $10-16k/month, yielding cost-per-QALY near or above $1 million.
  • Medicare is legally required to pay for any FDA-approved cancer drug at the company's chosen price and cannot negotiate, removing all downward price pressure.
  • Companies design and conduct their own registration trials, creating systematic incentives to game endpoints, control arms, and dosing.

The argument · threads8 threads · 41 claims
0.72

Most approved cancer drugs deliver marginal or unproven survival benefit at unsustainable cost per quality-adjusted life year.

5 pointscentrality 5/5
  • Although exceptional success stories like Imatinib, Rituximab, and Trastuzumab exist, an empirical and broad look at cancer drugs shows they offer marginal benefits—the median improvement in overall survival across 71 consecutively-approved solid-tumor drugs was 2.1 months—while costing $12-16,000 per month for indefinite, non-curative treatment, producing very poor value.

    “They looked at 71 consecutively-approved drugs for solid tumors. Whey found that the median improvement in overall survival is 2.1 months.”

  • Because cancer drugs offer marginal benefits in ideal settings while costing enormous amounts, cost-per-quality-adjusted-life-year figures reach stratospheric levels—around $900,000 for some colon cancer drugs and effectively infinite for drugs like Aflibercept that cost more than alternatives while delivering nothing more—far above the $60-70k UK ceiling or the $200-250k upper limit some US economists endorse.

    “we see dollars per quality that approach a million dollars. Regraph says it is $900,000 per quality-adjusted life year for colon cancer.”

  • Idealized clinical trials enroll younger, healthier patients with fewer comorbidities than real-world patients, so trial survival overstates real-world benefit; in some cases real-world patients on the active drug live half as long as control-arm patients on a sugar pill in the trial, meaning the true absolute benefit of these agents is even smaller than the published figures.

    “I give some examples where people in the real world taking the active drug live half as long as people in the control arm, taking the sugar pill.”

  • Some cancer drugs are not merely marginal but actually ineffective—they appear marginal only in the contrived trial setting where they are tested against a weak control arm that is not standard of care, with a flawed design, in an atypical population—so in real-world use they probably do not work at all.

    “it's not just that they're marginal. I actually think that they're ineffective. That they really don't work.”

  • Of 100 consecutive oncology drug approvals, only about a third improve overall survival or quality of life; a third merely delay progression-free survival, which only gives the illusion of being meaningful because it rests on arbitrary radiographic cut-offs; and a third are approved on single-arm studies with no control arm using response rate—so approving drugs on these surrogate endpoints means many likely do not improve survival or quality of life.

    “One third of them, we know they improve overall survival... One third of them, we know they delay something called progression... And a third of them are approved on the basis of studies that have no control arm.”

0.65

Removed cost-sharing and broken market mechanics eliminate downward price pressure on drugs.

3 pointscentrality 5/5
  • Almost no one would pay $100,000 or $10-12k/month out of their own pocket for an extra two months of low-quality, side-effect-laden life; people accept these drugs only because a constructed system means society pays—through taxes, premiums, and lost wages—so the absence of direct cost removes the feedback that would otherwise constrain demand.

    “if you were actually paying for it, very, very few people would choose to do that with their own money. But, we live in a system that has been constructed in a way that you are not paying for it.”

  • Cancer drug prices are set at what 'the market' will bear, but that market is a broken, convoluted system exerting almost no downward force on price—not because of patent monopoly, which would be acceptable, but because Medicare is legally required to pay for any FDA-approved drug at the company's chosen price and cannot negotiate, so the only real constraint is social pressure (the threat of being on '60 Minutes').

    “nothing stops a company from coming tomorrow with an approved drug and saying it's going to be $2 million a pill. I think literally the only thing that stops them from doing that is the fact that they're going to be on 60 Minutes if they do that.”

  • If patients truly had skin in the game in a genuine free market, they would not be willing to pay so much money for such marginal drugs with so much uncertainty, so the only reason the current pricing persists is the arbitrary construction of rules that removes downward pressure—though the challenge for the skin-in-the-game view is how to implement it for a vulnerable population that often cannot afford out-of-pocket costs.

    “if we really did have a true, you know, free market where people had a lot more skin in the game, I just don't believe people would be willing to pay so much money for such marginal drugs”

0.61

Pervasive industry funding and conflicts of interest throughout the approval and clinical ecosystem corrupt decision-making.

6 pointscentrality 4/5
  • Industry research funding to universities often exceeds the true cost of the research, and the leftover 'slush' gets recycled within institutions so that prominent group leaders build war chests of tens of millions of dollars that fund their labs, staff, luxurious business-class travel, and assistants—making industry collaboration materially comfortable for academics beyond direct consulting payments.

    “leaders who run certain groups that are very prominent, they may build up war chests of tens of millions of dollars of funding, which can fuel their laboratories, lead to them hiring staff, have luxurious business class travel”

  • Although FDA reviewers are legally barred from current financial ties to industry, 60% of them go on to work for industry after leaving the agency, creating a revolving-door incentive to go easy on the very companies they may later join—analogous to taking it easy on a future employer.

    “60% of them go on to work for the industry when they leave the industry. And that's the so called revolving door politics”

  • An FDA Drug Advisory Committee meeting works like a courtroom—the FDA gives an opening statement, industry mounts a defense with a conflicted expert witness, panelists act as jurors, and the public asks questions—but unlike a courtroom every single participant has heavy financial ties to the pharmaceutical industry, undermining the impartiality the structure is supposed to provide.

    “the difference between a courtroom is the money. And in the book I trace that every single person, every single entity that is speaking, has heavy financial ties to the pharmaceutical industry.”

  • The money flows from pharmaceutical companies to patient advocacy groups (who stay quiet about drug prices), to uncritical 'thought leaders' who are personally and professionally conflicted, to journals deriving revenue from study reprints, to professional organizations with large undisclosed industry revenue—so nearly every voice that could honestly represent patients is instead speaking on behalf of industry, leaving almost no pressure toward reform.

    “all of the players in this dialogue are really kind of speaking on behalf of the industry, or at least part of them are speaking on behalf of the industry. And there is very little pressure the other way, towards sort of serious reform.”

  • In medicine it is accepted that the first author of a New England Journal or JAMA paper—a career-defining achievement—may not have written much of it, because the paper is drafted by corporate-funded medical writers who concisely communicate trial results but also produce spun, hyped sentences; this gift-authorship is ethically troubling and would be considered absurd in fields like philosophy or economics.

    “an author who is the first author of a New England Journal paper... that person may not have written much of the paper, that the paper is drafted by a corporate funded medical writer.”

  • Although pens and occasional dinners are trivial sums, a large literature shows even these small gifts are linked to prescribing patterns and brand-name prescribing for certain drug classes, meaning industry gets a measurable return on investment from seemingly insignificant marketing.

    “there's a large literature that shows that even those trivial sums of money have been linked to prescribing patterns, brand-name prescribing for certain classes of medications”

0.51

Trial design and conduct by conflicted sponsors creates irresistible incentives to game results in subtle, legal ways.

3 pointscentrality 4/5
  • When the difference between a p-value of 0.49 and 0.51 is $10 billion in expected revenue, companies will do everything legally permitted to win: choose the weakest comparator drug, run trials in settings with no further therapy after progression to exaggerate differences, enroll younger healthier populations, and skew dose-reduction rules to penalize the competitor—so as long as something is not explicitly prohibited, the incentive guarantees the thumb goes on the scale.

    “if the difference between a -pvalue of 0.49 and 0.51 is $10 billion in expected earned revenue--that's the incentive difference for the company--that, even within all of the rules of clinical trials, you will do everything you can to win.”

  • Asking the drug sponsor to design and conduct its own study is like asking a painter to judge the painting contest in which they submitted a piece—the temptation to game the system and put a thumb on the scale is too hard to resist, so the conflicted party must be removed from trial design and conduct.

    “Asking the sponsor to design and conduct their own study is like asking the painter to judge the painting contest for which they've submitted a piece of artwork.”

  • Instead of the conflicted company designing, conducting, and reporting its own registration trial, the company should pay a much larger fee (e.g. $32-52 million instead of $2 million) and supply the drug, while the FDA or an impartial third party designs the control arm, population, post-protocol therapy, and statistical plan and runs the trial; once the Data and Safety Monitoring Board declares success, the drug is approved instantly with no back-end review, which removes the conflict, prevents gaming, and could even save a year of regulatory time.

    “you design the control arm, the clinical trial, the population, the post-protocol therapy. You design all that and you run the trial... the moment that DSMB calls you and says this trial is successful, you approve the drug.”

0.48

Surrogate endpoints like tumor shrinkage lack evidence they predict longer or better life.

2 pointscentrality 4/5
  • A surrogate endpoint is an endpoint a patient did not know was important until a doctor told him it was—just as patients did not care about their LDL cholesterol or blood sugar until told these mattered—and similarly patients do not intrinsically care whether their tumor shrinks 30%, an arbitrary definition of being a 'responder.'

    “A surrogate endpoint is an endpoint a patient didn't know was important until a doctor told him it was.”

  • The cut-offs defining tumor 'response' (a 30%+ shrinkage in the sum of tumor diameters) derive not from evidence that they predict longer or better life, but from a 1970s Charles Moertel study in which 16 oncologists measured marbles through foam rubber with calipers; the cut-offs reflect the smallest change two oncologists could reliably discriminate and have simply been codified and passed forward through historical accident.

    “the cut-offs that we use to this date derive from the amount of change in a marble that can be reliably discriminated by two oncologists measuring through foam rubber with a caliper.”

0.44

Gradual price escalation and policy failures like the UK Cancer Drug Fund show the system persists despite obvious inefficiency.

9 pointscentrality 2/5
  • Once a launch price is established, companies raise it incrementally—10, 15, 20% one year, 4% the next—with no underlying rationale beyond avoiding public outrage, exploiting the fact that gradual increases do not provoke the backlash a single large jump would (the frog does not jump out of slowly heated water).

    “The frog doesn't jump out of the pot when you heat it slow, heat the water slowly; and that's what they're doing.”

  • The failure in cancer medicine is structural, not personal: pharmaceutical employees and academic physicians are good people who genuinely want to cure cancer, but the structure of the system creates incentives that reward ineffective or marginal drugs at tremendous revenue, so blaming individuals yields no traction—reform requires changing the incentive structure so doing the right thing is the incentivized thing.

    “it's the structure that has allowed the incentives to incentivize ineffective or marginal drugs at tremendous revenue. That's the failure. It's the structure of the system.”

  • A well-intentioned policy that let expert panels create registries of off-label drug use—originally sensible for older, cheap, off-patent cytotoxics with no sponsor to seek formal approval—has been captured by makers of $100-200k drugs who gain market share through expert recommendations, since Medicare is wedded to pay for off-label uses at certain evidence levels and cannot say no even when the evidence is flimsy or absent.

    “we have wedded Medicare to pay for off-label drug use as recommended by experts at certain levels of evidence. And Medicare here, too, cannot negotiate price; cannot say no.”

  • Rather than focusing patients on a median survival figure, a more helpful approach is to present the 20th and 80th percentile outcomes both without and with treatment—e.g. 'without treatment you'd live perhaps two to eleven months, with treatment perhaps four to seventeen months'—which gives individuals a clearer basis for making their own choices.

    “What is the 20th and 80th percentile without treatment and with treatment?... I find that that can kind of be a helpful way for individuals to make these choices.”

  • Because the US accounts for half or more of cancer drug spending and approves drugs based on surrogate endpoints, other regulators like the European Medicines Agency follow the FDA in lockstep on endpoints, approvals, and trial requirements, and other nations' price negotiations are anchored to the astronomical US starting price—so it is very hard to learn lessons from other countries because the US is such a dominant player that it sets the terms for everyone.

    “What the United States does around cancer drugs, how we decide to approve drugs, how we pay for drugs--this prevents any other nation from straying too far from what we do.”

  • Unlike most cancer drugs that offer only incremental benefit with no durable remissions regardless of where a patient falls on the prognostic spectrum, certain classes such as immunotherapy may have a subset of patients who demonstrate a longer-than-average, more durable benefit—a genuine right-hand tail.

    “that might be true for certain classes of medication, like immunotherapy... that there's a subset of people who demonstrate a longer-than-average benefit.”

  • A hazard ratio is the instantaneous risk of an adverse event at any point in time, serving as a metric of the relative benefit a drug provides; an 80% benefit applied to an average survival of six to eight months produces the roughly 2.1-month absolute survival figure.

    “hazard ratio is like the instantaneous risk of an adverse event at any point in time. It's sort of a metric of the relative benefit a drug provides.”

  • The UK normally prioritizes cancer drugs by cost-effectiveness, but about a decade ago political pressure led it to create a Cancer Drug Fund—a special pot of money to pay for drugs that fail traditional cost-effectiveness thresholds—which the Lancet called intellectually bankrupt from the outset because it contradicts the goal of maximizing health outcomes with finite money; the fund ballooned out of control and was eventually dissolved, showing you cannot pay an infinite price even in the UK system.

    “the moment they created such a fund, it's already intellectually bankrupt... it flies in the face of the entire idea that with what money you have, you ought to maximize health outcomes”

  • Rather than banning conflicts of interest or imposing harsh punishments, the better approach is to make a non-conflicted career path more successful by reserving certain privileges—serving on guideline panels, writing editorials and review articles—as badges available only to physicians who avoid industry ties, thereby incentivizing people to stay clean.

    “I actually wouldn't ban it. I would incentivize people not to have it.”

0.34

Regulatory approval speed and evidence hurdles are treated as separate goods but lower bars increase uncertainty without actual acceleration.

2 pointscentrality 3/5
  • It would be acceptable to allow uncertainty at market entry in exchange for speed if that uncertainty were later resolved and ineffective drugs were withdrawn, but the FDA almost never removes drugs from the market and has a poor track record of ensuring post-market confirmatory studies get done; when they are done, 20% of the time the drug is converted to full approval on the same surrogate endpoint, which doubles down on the uncertainty rather than eliminating it.

    “20% of the time, they convert and accelerate it to a full approval based on the same exact surrogate endpoint, which really doesn't eliminate uncertainty. It just doubles down on the uncertainty”

  • There are two ways to speed drugs to market: streamlining trials to accrue more patients per unit time, which is genuine speed, or lowering the evidence hurdle for approval, which is not just speed but added uncertainty—approving a drug that may or may not work—and empirical work suggests lowering the bar does not actually improve speed as much as assumed because of the technical nature of surrogate endpoints.

    “Speed, people use that primarily to mean lowering the evidence hurdle for drug approval. That isn't just speed, that's also uncertainty.”

0.34

Paying for low-value cancer drugs diverts finite health resources from higher-value interventions affecting life expectancy.

2 pointscentrality 3/5
  • Because there is only finite money for health, paying for low-value cancer practices means actively not paying for higher-value practices—blood pressure reduction, cardiovascular care, primary prevention—that would deliver far more per dollar, so the trade-off contributes to falling US life expectancy relative to other industrialized nations despite access to all the new drugs.

    “What we're doing in medicine is not paying for these low value practices: We are actively not paying for higher value practices.”

  • When the UK denies access to a drug, the hardest case is a modest benefit at huge price, but far more often the drug has no benefit or wholly uncertain benefit with certain, unpleasant toxicity; there are worse ways to die than dying of cancer, and patients sometimes die worse from taking experimental or uncertain drugs, so part of policy's role is to protect sick, vulnerable people from being driven to desperate, barbaric choices that offer no possibility of benefit.

    “there are people who die from taking experimental drugs or drugs with uncertainty, that die worse than they would die if they didn't take the drug.”