FDA: Moving to a Safety-Only System

by on July 30, 2011 at 7:35 am in Economics, Law, Medicine | Permalink

It now costs about a billion dollars to develop a new drug which means that many potentially beneficial drugs are lost. Economist Michele Boldrin and physician S. Joushua Swamidass explain the problem and suggest a new approach:

Every drug approval requires a massive bet—so massive that only very large companies can afford it. Too many drugs become profitable only when the expected payoff is in the billions….in this high-stakes environment it is difficult to justify developing drugs for rare diseases. They simply do not make enough money to pay for their development….How many potentially good drugs are dropped in silence every year?

Finding treatments for rare disease should concern us all. And as we look closely at genetic signatures of important diseases, we find that each common disease is composed of several rare diseases that only appear the same on the outside.

Nowhere is this truer than with cancer. Every patient’s tumor is genetically unique. That means most cancer patients have in effect a rare disease that may benefit from a drug that works for only a small number of other patients.

…We can reduce the cost of the drug companies’ bet by returning the FDA to its earlier mission of ensuring safety and leaving proof of efficacy for post-approval studies and surveillance.

Harvard Neurologist Peter Lansbury made a similar argument several years ago:

There are also scientific reasons to replace Phase 3. The reasoning behind the Phase 3 requirement — that the average efficacy of a drug is relevant to an individual patient — flies in the face of what we now know about drug responsiveness. Very few drugs are effective in all individuals. In fact, most are not effective in large portions of the population, for reasons that we are just beginning to understand.

It’s much easier to get approval for drugs that are marginally effective in, say, half the population than drugs that are very effective in a small fraction of patients. This statistical barrier discourages the pharmaceutical industry from even beginning to attack diseases, such as Parkinson’s, that are likely to have several subtypes, each of which may respond to a different drug. These drugs are the underappreciated casualties of the Phase 3 requirement; they will never be developed because the risk of failure at Phase 3 is simply too great.

Boldrin and Swamidass offer another suggestion:

In exchange for this simplification, companies would sell medications at a regulated price equal to total economic cost until proven effective, after which the FDA would allow the medications to be sold at market prices. In this way, companies would face strong incentives to conduct or fund appropriate efficacy studies. A “progressive” approval system like this would give cures for rare diseases a fighting chance and substantially reduce the risks and cost of developing safe new drugs.

Instead of price regulations I have argued for more publicly paid for efficacy studies, to be produced by the NIH and other similar institutions. Third party efficacy studies would have the added benefit of being less subject to bias.

Importantly, we already have good information on what a safety-only system would look like: the off-label market. Drugs prescribed off-label have been through FDA required safety trials but not through FDA-approved efficacy trials for the prescribed use. The off-label market has its problems but it is vital to modern medicine because the cutting edge of treatment advances at a far faster rate than does the FDA (hence, a majority of cancer and AIDS prescriptions are often off-label, see my original study and this summary with Dan Klein). In the off-label market, firms are not allowed to advertise the off-label use which also gives them an incentive, above and beyond the sales and reputation incentives, to conduct further efficacy studies. A similar approach might be adopted in a safety-only system.

Addendum: Kevin Outterson at The Incidental Economist and Bill Gardner at Something Not Unlike Research offer useful comments.

mdb July 30, 2011 at 7:46 am

2 problems with this. The first, what person would sign for a study, if the drug is available? No one with a serious disease. Second, This just perpetuates the myth the government can protect a desperate person from wasting money on an ineffective treatment – which is really the worst outcome if it is safe.

Marian Kechlibar August 1, 2011 at 9:04 am

So, who signs for a study now? Or are the people forced by FDA at gunpoint?

TomStarke July 30, 2011 at 8:01 am

Without an effective national disease registry capturing a data-based understanding of each disease treated, the effect of these “unproven” drugs won’t yield value. WIth our current health data systems, implementing this proposal will only lead to more drugs prescribed, more profits, and less understanding.

Rich Berger July 30, 2011 at 8:07 am

I read the first two comments and my reaction was: what the heck are they talking about?

J July 30, 2011 at 12:14 pm

Some people have difficulty understanding that profits are the reason the drugs that protect them from disease and premature death exist.

I like the idea of the FDA researching and advising on the safety and efficacy of drugs, and have no problem with insurance companies basing coverage on those recommendations. But if I’m spending my own money, it’s none of the government’s business what I’m taking, and no government agency should be permitted to restrict the sale of any drug, with the possible exception of things like antibiotics, in which effectiveness can be compromised with overuse.

“Third party efficacy studies would have the added benefit of being less subject to bias”

Baloney. They’d have different biases, but they’d be just as biased. And the entities doing third party efficacy studies, including the FDA, should have the same liability exposure drug companies are subject to if they’re wrong.

Doctorsh August 5, 2011 at 8:02 am

Agree with J

asdf July 30, 2011 at 8:24 am

implementing this proposal will only lead to more drugs prescribed, more profits, and less understanding.

No. It reduces the barrier to entry in the market and will over time lead to less profits. It also means that it will be possible to go after smaller markets, like $1B or even $100M markets rather than $10B+ markets.

As for understanding, it means many more field trials of compounds. Which is still the only way to actually gain understanding of what a compound does. We’ve already seen that this works with nutritional supplements: things like creatine and omega-3 fatty acids are the wheat that is sorted from the chaff through a market process.

Basically, if you want to end the war on drugs, you should be in favor of also ending the war on drug companies.

J Thomas July 30, 2011 at 8:35 am

As for understanding, it means many more field trials of compounds. Which is still the only way to actually gain understanding of what a compound does.

Yes! More field trials, on much larger numbers of people, run by people who are not punished if they report that the drug fails to live up to its hype. That’s vital.

We’ve already seen that this works with nutritional supplements: things like creatine and omega-3 fatty acids are the wheat that is sorted from the chaff through a market process.

WTF? You believe that marketing has sorted the wheat from the chaff in this area?

J Thomas July 30, 2011 at 8:32 am

A combination of these ideas may be useful.

First, it’s absurd to depend on pharmaceutical companies to run their own studies of safety and efficacy. This is begging for fraud. The government-funded health care system should do that.

Second, without an effective system of testing for efficacy, we get expensive and ineffective drugs. Which we obviously do. Making giant pharmaceutical companies pay a fortune to get the right to market an effective or ineffective drug, is not a good thing.

The obvious solution is to use Medicaid patients as test subjects. They get free medical care, and when they volunteer for free medical care they volunteer to sometimes test experimental products. Each experiment gets called off immediately if it appears to produce worse results than standard treatment. New treatments that pay off, pay off immediately for Medicaid patients — they are the first to benefit. The government is already tracking them and their treatments. It’s win/win/win/lose. Everybody wins except the evil people who profit by the old system.

Also, we could have sanitized results available on the Internet, so anybody could draw their own conclusions about prognosis.

Lansbury’s point is mostly bogus. When we think a drug is effective for a small fraction of patients but we have no way to guess which ones, it’s irresponsible to give it to all of them and bill them large sums for it. Particularly when it’s so easy to fool ourselves about the rare successes. We should be spending more of our efforts at finding out which patients are which. And we do less of that because there’s less money in it.

Pharmaceuticals and free markets have an impedance mismatch. It’s too expensive to find out what works, and the rewards for faking it are too great.

Jan July 30, 2011 at 12:35 pm

I don’t think having people “volunteer” for subsidized health insurance that includes exposure to potentially dangerous treatments is the obvious solution. Using one particular group, especially a low-income subpopulation, for research purposes in exchange for medical insurance would never be allowed under human subjects research protections. Subjects must always agree to participate in research independently and have the option to receive the standard of care, no matter who is paying for their treatments.

J Thomas July 30, 2011 at 3:08 pm

Using one particular group, especially a low-income subpopulation, for research purposes in exchange for medical insurance would never be allowed under human subjects research protections.

Perhaps we should agree to disagree about this. On the one hand we have a whole lot of Republicans who say that poor people don’t deserve to have medical care. On the other hand we have you who says that the medical care they receive must be ethically pure.

Orange14 July 30, 2011 at 9:14 am

I worked in the biopharmaceutical industry for 27 years prior to retirement (whether this qualifies me as an expert is arguable since anyone responding to blogs can make points that are good bad or indifferent. Here are some observations.

Relying on off label evidence is risky since there are a number of excellent examples of companies abusing this by engaging in illegal acts judging by the significant number of multi-million $$ settlements (classic example is Neurontin which never worked in the off label indication yet had huge sales because of illegal marketing).

Companies can already do what Boldrin and Swamidass suggest and sell investigational drugs through the Treatment IND regulations. This permits companies to recoup costs as they continue to do the clinical trials and the regs have been around for over 20 years. Want to know how many companies have used this? ZERO, because they have to open their books to the FDA and show that they are not profiting from the sales. Companies do not want reveal what it costs to manufacture drugs (even though it’s relatively easy to back out costs).

Regarding cancer drugs – this does present a conundrum IF the assumptions are correct that there will always exist a subset of people who are helped by targeted therapies. On this point the evidence is still unclear. Judgements about what is acceptable therapy still need to be worked out. Many of the therapies do not offer a cure but only a delay in progression of of symptoms or slightly increased survival (both of which are linked. We don’t have the tools yet (except for a couple of cases) to identify the underlying genetics of who will be helped or not. In the absence of this, the only reasonable way forward is to reimburse only for successful treatment. This then makes the cost of treatment large since companies will be maximizing profits on a small base of patients out of the large number that are treated. You can already see this in the pricing of new cancer therapies for small patient populations (Provenge for prostate cancer being one good example).

Landsbury’s argument about Parkinson’s Disease is off the mark. The key issue here (as it is with depression and Alzheimer’s Disease) is that we lack good biomarkers for target study and clinical evaluation. As a result we are reliant on subjective endpoints which are more difficult to measure. It may be that for some medical conditions we never reach this point but there are huge efforts going on right now (a major industry/academic/FDA initiative is underway right now to look for such biomarkers and about $60 million is being spent on brain imaging of Alzheimer’s patients in a effort to figure this out).

Rahul July 30, 2011 at 9:17 am

Nice comment!

AnonLawStudent July 30, 2011 at 11:52 am

“The key issue here (as it is with depression and Alzheimer’s Disease) is that we lack good biomarkers for target study and clinical evaluation.”

Which is one of the things that large scale use of a nominally* marginally-effective drug could help figure out. (*) I say nominally, because the overall statistics would show nominal benefit, but that benefit may be very high for individuals with the relevant biology.

AnonLawStudent July 30, 2011 at 11:53 am

make that “show marginal benefit”

Bill July 30, 2011 at 4:34 pm

Great comment.

Particularly: “We don’t have the tools yet (except for a couple of cases) to identify the underlying genetics of who will be helped or not. In the absence of this, the only reasonable way forward is to reimburse only for successful treatment. ”

I’ve taught a graduate pricing class and invite industry folk in to speak, in this case a small pharma company’s VP of marketing. He would agree with you because if you have an effective drug, as a small company, you can target the population for which it will work, or, if not, blanket a larger population and receive payment when it does work. Their problem is competing against large Pharma and its marketing, which often relies on doctor memory of names and marketing, and not results.

Michael Cain July 30, 2011 at 9:56 am

How do patents on drugs fit into this? Under a no-efficacy-required regime, am I allowed to patent molecules on the chance that they might be effective against some disease or condition? Do patents go to the party that conducts an efficacy study? Or do we toss drug patents and the monopoly pricing that goes with them completely?

Colin July 30, 2011 at 10:38 am

@Michael Cain: You misunderstand. The patent system and the FDA system are currently independent, and they would likely remain so under the proposed system. There are many compounds that have been patented, for drug use, that turned out not to be useful as drugs. Reforming the patent system is another issue, however the two lines need to be addressed separately, and within their own universe, so as to avoid unintended consequences in other areas.

bk, md July 30, 2011 at 10:24 am

“Nowhere is this truer than with cancer. Every patient’s tumor is genetically unique. That means most cancer patients have in effect a rare disease that may benefit from a drug that works for only a small number of other patients.”

1) this is not true and not necessarily relevant.

2) if true, it also means that the drug may very well hurt a certain percentage of the people who take it. how do you know which patients the drug will benefit, which patients the drug will hurt and which patients the drug will have no effect on? how do you know the relative percentages of each in order to allow the patient to give fully informed consent? A respective split of 5/90/5% is very different from 5/5/90% even though the same number of patients benefit.

Memnon July 30, 2011 at 8:28 pm

Exactly. This is the key fallacy of the Boldrin-Swamidass approach.

jk July 30, 2011 at 11:02 am

Companies are not pursuing high R&D costs drugs also due to free-riding manufacturers overseas.

Orange14 July 30, 2011 at 11:42 am

Can you point to specific examples? Otherwise this is just an inane statement. The more central point is that drug development has become increasingly difficult as almost all the low hanging fruit has been harvested. In addition, for many medical conditions we already have sufficient therapies and anything new will only be of incremental benefit (if even that). For chronic conditions such as asthma and allergy, high blood pressure and cholesterol, and some others there is almost no R&D going on since the current therapies are sufficient. Remember that these are all chronic conditions and companies have made a lot of money because large numbers of patients need these drugs every day. The financial calculation is of course different for oncology drugs since both patient populations and treatment times are smaller. Compare the size of the patient pool with high blood pressure versus those who have melanoma; dramatic difference. Alzheimer’s Disease (and other dementias) is of course different with an increasing patient population and no good treatments. Problem is how to come up with a drug that really works.

bulldog July 30, 2011 at 11:05 am

This idea is insane. Just insane.

Old oncologist July 30, 2011 at 11:08 am

There are no games that can’t be gamed.
Under this system, a drug maker tests a drug in a hundred patients. Obviously, some patients do better than others. Next the drug company marketing branches will claim that half the patients treated had above average outcomes!
During the early days of oncology, I heard many presentations that could be distilled into the above claims. Oncology has become more sophisticated, but the stakes have also risen, and the drug makers know that billions are at stake. How much do drug makers spend on research versus marketing?
I was amazed by the recent controversy regarding Avastin in breast cancer. Avastin is largely ineffectual in this disease based on the evidence, even the most favorable study showed marginal results. The drug was painted as some great breakthrough, and Genentech as some small bold savior.

Dan in Euroland July 30, 2011 at 11:17 am

How do you game Chess?

Mike July 30, 2011 at 12:30 pm

Chess has a solution. All finite games of complete information do.

Chess Fiend July 30, 2011 at 3:53 pm

My home computer is as strong or stronger than the world champion. You game it like any other game, you cheat!

Orange14 July 30, 2011 at 11:45 am

Quite right. The bigger issue with oncology drugs is whether or not we accept a marginal increase in survival as a realistic endpoint and the related cost that goes along with it. I remember a New England Journal of Medicine article of several years ago where the authors looked at current colon cancer therapy and survival. The cost of the drug cocktail was in the neighborhood of $120,000/year for 2-3 extra months (though the data on survival was somewhat questionable).

Rahul July 30, 2011 at 2:26 pm

The decision about whether $120,000 / yr makes sense or not should rather be made by individual insurers, HMO’s, or (rich) people paying out of pocket. FDA making that call is iffy. Who’s to say if some rich person would think $120k per year as cheap for extending his life?

Even if there is marginal improvement the FDA should give the ok and leave the cost decisions to others.

quigley July 30, 2011 at 1:26 pm

Old Onc and I agree completely except the avastin example is not shocking. Only common.

Jan July 30, 2011 at 12:20 pm

First, the cost of developing a drug is nowhere near $1B. Industry-backed estimates of development costs have been debunked numerous times. Pharmaceuticals are still one of the most profitable industries in the US.

The thesis of TGS applies here, too. It is getting tougher to discover new therapies because the low hanging fruit has been snatched up and cashed in on. And the ROI problem that has always been there for rare diseases remains. There is a tiny fraction of drugs that will only work on those with certain genotypes—the personalized medicine therapies. Genetic variation between humans is about 0.1%.

That said, having FDA revert to a safety-only mission is a terrible idea. It will not save money. Actually, it would increase health care spending. As it stands now, Medicare, Medicaid and other insurers, pay for almost any FDA-approved product. Removing the efficacy requirement means the market will be flooded with drugs that insurers will be paying for without any evidence they do people any good. Industry will be delighted to push these drugs via advertising campaigns that cost more than the drug development itself. They may even show doctors fancy charts summarizing studies they’ve conducted that demonstrate excellent results. (The studies will not be subject to FDA scrutiny, or the scrutiny of any regulators beyond those ensuring human subjects protection. Unless, of course, the go to Nigeria to conduct trials).

The reason we cannot allow FDA to leave studies of effectiveness to companies is because Medicare and Medicaid are too politicized to refuse payment for products they are convinced they need, even if there is a lack of evidence. Even the minor investment in a comparative effectiveness research institute this administration has made through health care reform has been so attacked by various lobbies that its impact will be nill—all before it even began work.

There is no appetite to devote any more government resources to effectiveness research. And, politically, public and private insurers just can’t say to paying for any treatment that comes on the market. The FDA is the only way to ensure we don’t unnecessarily pay for therapies that don’t work.

Colin July 30, 2011 at 1:11 pm

“Debunked numerous times”?

DiMasi, Hansen, Grabowski (2003) “The price of innovation: new estimates of drug development costs,” Journal of Health Economics; Vol 22, pp. 151-185.

Argues that new drug development costs ~$800 Million.

Adams CP, Brantner VV (2010) “Spending on New Drug Development” Health Econ. 19: 130–141 (2010)

Using public data, argues that the $1 Billion figure is an underestimate.

It’s been re-bunked many times also.

quigley July 30, 2011 at 1:33 pm

Colin-
As someone who works with pharma daily, these numbers are hogwash and simply part of marketing. The drug manufacturers do spend millions upon millions in phase 2 and 3 trials that are also marketing in so far as they are conducted at academic institutions who train young MDs to use the study drugs.

The basic science is largely conducted via academic research that the populous has already paid for via NIH funding. There are exceptions, but 90+ % of pharma’s budget can easily be viewed as marketing. Get to know top folks in pharma well, and they’ll admit it (not on the record).

Rahul July 30, 2011 at 2:43 pm

Can’t marketing be also considered as a “cost”? Without it there’d be no sales and hence no recouping the expenditure. I’m just being the Devils Advocate for a second. Is there any reason to keep marketing separate?

quigley July 30, 2011 at 3:09 pm

Yes. Because pharma mis-labels marketing costs as R&D.

Orange14 July 30, 2011 at 4:18 pm

quigley writes: “Yes. Because pharma mis-labels marketing costs as R&D.” This is absolutely not true. Companies report out marketing and R&D on separate lines in their SEC filings. Read the 10K filings with the SEC!

@Rahul – you are correct; it is an expense and itemized as such.

J Thomas July 30, 2011 at 7:29 pm

This is absolutely not true. Companies report out marketing and R&D on separate lines in their SEC filings.

??

He says they lie about it. You say they report it. Is there a disagreement here?

Rahul July 31, 2011 at 12:49 am

@J Thomas:

quigley said pharma put marketing cost under R &D.
Orange14 said they report it separately.

Yes, clearly a disagreement.

J Thomas July 31, 2011 at 10:04 am

Rahul, if I understand the argument correctly, Quigley is saying that a lot of what drug companies call R&D is really marketing. But they mislabel it and count it as R&D cost.

While Orange14 says that companies put two separate items in their SEC filings, one for marketing and one for R&D.

I don’t see any disagreement here. There really are two separate numbers they report. And they could easily be misrepresenting those numbers.

It’s like, if I said “Nixon was a crook!” and you said, “But Nixon himself went on TV and announced he wasn’t a crook.”.

That’s what it’s like.

quigley July 31, 2011 at 1:47 pm

J Thomas reconciles the disagreement between Orange14′s and my positions the way I would.
You have to understand the motivations behind the “research” that pharm does on drugs they’ve already developed.

Yancey Ward July 31, 2011 at 6:17 pm

Quigley is wrong. In fact, I am willing to bet he doesn’t even know where to find actual marketing costs in any company since I have seen others with his same views consistently misidentify marketing costs with other costs involved in running a business. Indeed, I would not even be surprised to find that he is conflating several different terms such as marketing, advertising, and other items.

Jan July 30, 2011 at 3:05 pm

And re-de-bunked again in 2011.

Both studies estimate factor in “opportunity cost” in calculating the figures–half of the estimates are profits foregone. This is a curious technique. When I estimate my costs for a weekend trip I don’t count the income I didn’t receive that weekend because I didn’t take a part-time job for Saturdays.

Amazingly, Light and Warburton estimate something around $50 million per drug using the same data as Tufts and Adams’ studies.

Donald W Light, and Rebecca Warburton. Demythologizing the high costs of pharmaceutical research
BioSocieties (2011) 6, 34–50. doi:10.1057/biosoc.2010.40; published online 7 February 2011

Congressional Budget Office. Research and Development in the Pharmaceutical Industry. October 2006.

As a side note, the researchers at Tufts responsible for the highest profile studies receive most of their funding from the drug companies. That in itself does not mean the studies are bad, but when other academics point to serious deficiencies in their methodology, one has to question the findings.

Yancey Ward July 31, 2011 at 6:12 pm

Both studies estimate factor in “opportunity cost” in calculating the figures–half of the estimates are profits foregone. This is a curious technique. When I estimate my costs for a weekend trip I don’t count the income I didn’t receive that weekend because I didn’t take a part-time job for Saturdays.

Then you would be the one doing it wrong, and that is even conceding the rest of what you wrote is correct, which I find very doubtful for a lot of other reasons. When a company has a pool of capital invest, part of the cost of taking a risky choice is that that pool could have been invested in relatively safe investments with near guaranteed returns. Not taking that path is a cost, whether you realize it or not.

quigley July 30, 2011 at 1:34 pm

I understand Jan’s concerns.

Dan Weber July 30, 2011 at 3:56 pm

As it stands now, Medicare, Medicaid and other insurers, pay for almost any FDA-approved product.

Doesn’t that seem like the bigger problem?

If my neighbor buys whatever car I buy, and me buying a Ferrari bankrupts him, that doesn’t mean I shouldn’t buy a Ferrari. It means my neighbor should get a better selection algorithm.

J Thomas July 30, 2011 at 8:06 pm

The federal government interferes in health care in lots of ways.

To some extent health care for rich people is subsidized.

The government does a lot to prevent substandard medical care. Anybody who’s caught practicing medicine without a license can get in big trouble.

Meanwhile, if you choose an MD and then you don’t like the service he provides, you can sue him in a government court for malpractice.

I say, if we want anything like a free market in medical care, we need to get the government out of it in lots of ways. If we don’t let poor people get good medical treatment, for gods sake let them get medical advice from their astrologers. Let them buy whatever patent medicines they want.

If you have a Ferrari, don’t tell poor people they have to have a Ferrari or walk. Let them buy Yugos or whatever they can afford.

It will cause trouble any way we do it. Like the folk song says, “If health was a thing that money could buuuuuuyyyyy, the rich would live, and the poor would diiiiyyyyyiiiiii”

Mike July 30, 2011 at 12:28 pm

Using “rare” to describe diseases that affect few people and “rare” by saying each aspect of the disease is unique is equivocation. Humans are “rare” because we are all unique? I’m rare. You’re rare. People with dark skin and blue eyes are rare. But humans are not rare.

Nutritious foods are both “safe” and “effective” at keeping almost all humans alive, except for the few who have strong allergies or intolerance. We essentially have a safety-first system for food. Some foods have no nutritional value. For centuries, people thought nutritious tomatoes were poisonous. The process of determining what was both edible and nutricious developed from hundreds of thousands of years of billions of people putting things into their mouths. Science speeds up that process a bit.

The key to drug manufacturing is the same as anything else – bang for the buck. Why would any company (or a government) want to waste precious time and resources on something that afflicts few people? If it’s a mutation, seek out the mutagen. If it’s genetic, then cut off the offending genetic strain or do nothing. For rare cases, it sucks to be you. All life ends in death, so I don’t know why everyone thinks they have an entitlement to vast amounts of resources to prolong their own miserable existence. We save the most people we can with the least resources possible.

The ideal drug would be a poison that could kill only cells with a specific genetic code. Each cell is unique and cancer cells share part of the genome of healthy tissue. So the trick is to find the longest chain which is common to the disease cells but not healthy cells. It’s a math problem. As you kill off disease cells, you could lengthen the chain to get the stragglers, but there are diminishing benefits and rising costs as you begin to affect more healthy cells and miss bad ones.

The ramifications of this technology are shocking though: a poison that could kill only one person, or black people, or people with red hair, or southpaws. It would bring new meaning to “genocide.”

TGGP July 30, 2011 at 12:31 pm

Possibly of interest, Mike Darwin on the rapid scientific progress made by The AIDS Underground.

Hat-tip to gwern.

quigley July 30, 2011 at 1:18 pm

I assume this this a joke.

TGGP August 1, 2011 at 1:34 am

Nope. Not saying they were effective on a per-dollar basis, but a highly motivated group managed to focus on applicable research and was able to dramatically increase life expectancies, as shown in the graphs.

figleaf July 30, 2011 at 1:11 pm

“Third party efficacy studies would have the added benefit of being less subject to bias.”

Yes! This is exactly why independent ratings agencies and insurance companies like AIG prevented the current financial meltdown! Because third-party companies never, ever have any incentive to be anything but dispassionate and objective!

Oh wait!

The only possible benefit to a safety-only approval criteria is that as it stands many drugs that provide only minute, haveta-squint-just-right-to-see-it improvements are routinely approved, even when known side effects (not to mention expense) is higher than the better established and understood drugs (aspirin being only the most unfair example) developers hope to replace. The only possible downside would be that, free of any obligation to prove effectiveness manufacturers could dedicate all their resources to finessing safety assessments.

figleaf

p.s. Seriously, as a small-l libertarian I really did once have incredible confidence that ratings agencies and insurers could replace whole sectors of regulation. My hope that auditors would do the trick went down with Arthur Anderson. Given the extraordinary amounts of money squandered on echinacea, zinc, and vitamin c for colds in the over-the-counter “herbal” market even after repeated demonstrations that they have zero effectiveness I have no confidence that third-party safety or effectiveness reports will be even remotely… well… safe or effective.

J Thomas July 30, 2011 at 3:48 pm

“Third party efficacy studies would have the added benefit of being less subject to bias.”

Yes! This is exactly why independent ratings agencies and insurance companies like AIG prevented the current financial meltdown! Because third-party companies never, ever have any incentive to be anything but dispassionate and objective!

The hope is, you get researchers who want to do a good job at finding the truth more than they want to become rich. We do have a lot of epidemiologists who’re that way. Maybe if they had more opportunities to take bribes they would be less so, or would be replaced by people who want the jobs for the bribes.

But when the research is funded and directed by people who most definitely want to get rich, it gets harder for researchers to do the right thing.

So one thing we definitely should do, is compile records on all the Medicare/Medicaid patients. What possible reason is there not to? It wouldn’t be controlled studies, but it would give a sense of something.

Like, in clinical trials various contraceptives were 99% effective. But in actual practice they are usually only about 90% effective. That’s a difference that’s worth noting.

quigley July 30, 2011 at 1:23 pm

Alex-
If you need a clinician/researcher to help your cause. Let me know via email.

Fact idea that needs a good re-visit is how much money pharm truly spends on “R&D” vs. marketing. Look there for the real costs. Pharma is spending much much less on R&D than they would have you believe. The gabapentin example is a good one. Merck’s R&D was run by their marketing arm.

Orange14 July 30, 2011 at 1:54 pm

The Tufts studies include the cost of capital that could otherwise be deployed in other ways as per the long lead time of development of a new drug (the Tufts researchers have defended their numbers on numerous occasions and one can always argue against their approach). The fastest drug development that I’m aware of was the Merck protease inhibitor for HIV. From inception to the market was about five years which is really fast (they also were helped by a very fast FDA review). Unfortunately most non-critical care drugs don’t get developed as fast and the issue remains how to capture the profit over a relatively short patent period. Are you confusing gabapentin and Merck? The drug was originally developed by Parke-Davis and Pfizer inherited both the drug and illegal marketing when they acquired the company (the acquisition was primarily to get Lipitor just as the same way they bought Searle to get Celebrex). Merck’s R&D has never been run by their marketing arm (I know the folks working there) though marketing is involved at some level (as is the case with all big pharma companies).

quigley July 30, 2011 at 3:17 pm

You’re right. I got the wrong company.

libertarian_adi July 30, 2011 at 3:08 pm

Abolish the FDA! It is in the self-interest of the manufacturer to not put out an ineffective drug. If a corporation does put it out, tort law takes care of it. What makes people think that a bureaucrat in Washington — who doesn’t know you or me — has the incentive to make things better for us? If you are a bureaucrat and you approve a bad drug, you will be out of a job and probably be in the news. But if you delay the approval of a good drug, which could have saved tens of thousands of lives each year, by five or six years, none of the people who died as a result or their relatives will know about it. The FDA has killed hundreds of thousands of people through the drug lag problem. People only see the deaths which make the newspaper due to Thalidomide. It’s a cognitive bias.

And don’t even start with the “tort law makes it costly” crap. The monopolistic regulations of the FDA cost more to the average American — both in dollars and human lives.

J Thomas July 30, 2011 at 4:16 pm

Libertarian_adi, what you say makes sense. I think we should establish a libertarian nation somewhere with many millions of people in it, and do completely without drug or medical laws of any kind, and do large-scale comparative trials to study how it works.

It might be worth doing, but we need a few generations or at least a few decades of controlled tests first, to make sure it works.

JordanT August 1, 2011 at 7:01 pm

That’s not what happened pre-FDA, and it’s why the FDA was created in the first place. Companies selling unproven drugs or elixirs that ended up being toxic. The laws were strengthened when a company killed over 100 people because they changed the solvent their drug was in from ethanol to diethylene glycol (antifreeze, and toxic to humans). There were also a host of ineffective drugs, hence the term “snake-oil salesmen”‘

Take a quick glance in history and you can see why the FDA was created, companies were selling ineffective and sometimes dangerous drugs to make a quick buck.

mulp July 30, 2011 at 3:12 pm

On the cost of comparative effectiveness, with large electronic databases, the costs should be under a million, and probably under a $100K to produce a statistically significant result better than a hundred million dollar random drug trial effectiveness study. (I’m speaking to the speed of progress in data mining and the certainty of convergence on national digital medical records standards leading to global data mining. I’m guessing the data mining in a decade will begin to merge consumption patterns with medical records – in some nation, China/Taiwan/Israel maybe, the medical records will use the same IDs as are used for purchases of food and other goods.)

See SciAm: http://www.scientificamerican.com/article.cfm?id=the-best-medicine-july-11

This methodology was and will be used by classic HMOs, like Kaiser, which are now called ACOs after insurers destroyed the HMO brand, and most HMOs. I was a member of an HMO, destroyed about two decades ago, that using 80s limited data mining, was changing the recommendations to its staff of doctor employees using both internal and external comparative effectiveness studies. They were the first to discount PSAs after being one of the first to adopt regular PSA screening – they could see in their own data the misdirection of the PSA tests within a few years, confirmed elsewhere. I know this because my doctor updated me each year for five years when I updated my family history right after the PSA test was approved – my dad’s diagnosis.

The SciAm article reports on a $120 million controlled study that reported thiazide diuretics were the first line treatment for essential hypertension – the study was begun in the 90s, a conclusion that my European trained doctor told me in 1978-9 based on studies done there by the national health systems. He also told me the next line were beta and ACE inhibitors. But hey, they used government run comparative effectiveness studies to set the standard of care back then.

Dan Weber July 30, 2011 at 4:04 pm

But the customers — whether that means the clients of the ACO or the seniors for Medicare — hate being told they aren’t getting something they think they ought to.

If, as a people, we were mature enough to understand effectiveness research, it would work. But if we were mature enough for that, we’d already be there.

J Thomas July 30, 2011 at 4:55 pm

Release the sanitized data to the public, and a whole lot of people will make amateur efforts. They will fervently believe in it. They will disagree a lot with the professionals, but the idea will be firmly fixed in the popular imagination.

Orange14 July 30, 2011 at 4:28 pm

You are correct and Kaiser does some of the best research out there as they want to make sure that they pay for performance. The problem is that the databases here in the US are not nearly as homogeneous as the databases in Europe where there are national health systems. The Euro databases also quickly disproved any link between childhood vaccines and autism but the autism lobby of Jenny McCarthy didn’t want the hear that truth. The other point that you didn’t pick up on is that you need significant use data to be able to make prescribing decisions like the one with the thiazides so you need a lot drugs to be prescribed both within that class and future classes. Sure we know this today but it wasn’t as clear some years ago when the newer drugs came on the market. We could have stopped the development of statins when Mevacor was approved but then would have missed the better drugs in the same class. It’s pretty complicated stuff and the simple intuition is not always correct.

Personally I think we need a lot more comparative research but there is a knee jerk reaction against this. NIH doesn’t want to do this because they believe in funding basic research so we are left with the little bit that goes to AHRQ (which will probably suffer in the Federal deficit reduction still to come). With the Medicare Part D benefit there will be some good data coming in from that.

Bill July 30, 2011 at 4:42 pm

Agree with you again.
I particularly like some HMOs which have formed foundations to measure effectiveness of various drugs and protocols.
Sometimes there are even private solutions. Too bad, though, that when there is a public benefit we cannot agree on what government should do.

quigley July 31, 2011 at 1:54 pm

I agree with Orange14 here.
Comparative effectiveness research is needed. It is being funded through a new, though small, independent agency set up by HHS. I saw FDA/CMS/NIH presentations on this in the spring. NIH is on board and interested in funding more comparative effectiveness studies and CMS too. FDA is spinning mindlessly in the wind.

Mike Huben July 30, 2011 at 4:41 pm

I love these libertarian underpants gnomes plans to solve the world’s problems.

The plan is:
1. Remove requirement to test drugs for efficacy.
2. ?
3. Vastly more reliable knowledge of drug efficacy.

I agree with Tyler that more publicly funded studies are the way to go: finance them with a tax on off-label prescriptions. Indeed, with such a tax it might be possible to eliminate the requirement for stage 3 testing by drug companies, and let them follow the economic incentives whether to test or not.

It is expensive to produce information, but information about effective treatments is very valuable to patients, and drug companies cannot squeeze all that marginal benefit out of patients.

mdb August 1, 2011 at 8:03 am

Off label usage.

What is the empirical justification for prescribing a treatment in a non-clinical trial setting that has no proven benefits? July 30, 2011 at 4:58 pm

The scenario you describe from Boldrin and Swamidass of regulated prices until effectiveness is proven would never happen in real world American politics. That doesn’t necessarily mean that it is a bad idea, just that practically speaking if the FDA stops testing efficacy, neither prices nor efficacy will be regulated. Instead, it would be an opportunity for drug companies to make a boondoggle on drugs that are “safe” but ineffective (I put “safe” in quotations because what does it really mean for a cancer drug to be safe but ineffective?), or no more effective than a cheap generic that is already on the market.

That’s an aside, my real question for you is what is the rational basis for prescribing a drug the efficacy of which has not been tested? What is the empirical justification for prescribing a treatment in a non-clinical trial setting that has no proven benefits? My answer would be that you would have to believe that the prior probability of success of an unproven drug from a company that has a market incentive to sell the drug regardless of its medical benefits is greater than the proven probability of success of an existing treatment. For the vast majority of drugs, I doubt this is the case. For the vast majority of drugs, it makes no sense whatsoever to allow drug companies to sell an unproven drug (except if you’re a shareholder of Pfizer, which I am, and so I thank you for your ridiculous suggestions that would nonetheless enrich me).

The only context where I find what you are suggesting at all plausible is when all available treatments have failed for a particular patient. In most cases, however, there is no reason to think that the unproven drug is more effective than the proven drug, indeed there is good reason to believe the opposite. And in these cases where it just might make sense to prescribe an unproven drug, I don’t trust doctors and I certainly don’t trust the medical system to make this rationally informed decision that is based on the patients interests rather than their own bottom line and the bottom line of the drug company that has spent millions on propagandizing (I do use the word here in the pejorative sense here because we’ve defined the drugs in question to be unproven, so what the hell are they telling doctors and patients other than pure profit motivated bullshit?) on the glory of their new drug.

Finally, Peter Lansbury’s discussion (at least the portion you quoted) of drugs that are only effective in a small portion of the population makes it sounds like the FDA is being unreasonable with their “statistical barriers.” There is a very good reason why a drug that is only effective in a small portion of the population is more difficult to get through a clinical trial: it is much harder to distinguish from a drug that is only effective in none of the population. Increasing the size of a clinical trial to have enough statistical power to detect such effects is indeed costly, but the idea that the solution is anything but to increase the sample size in order to detect the effect is absurd. Because for everyone of these large effect in a small population drugs, there is a no effect in the whole population drug. And if you can’t distinguish the two, there is no rational basis to prescribe either drug if your motivation is the patient’s health.

Orange14 July 30, 2011 at 5:35 pm

Good comment. There two other issues that are directly linked as well. The first is the ethical issue of whether a company should be permitted to market a drug for which efficacy has not been assured. Under the current laws and regulation this is prohibited except under a Treatment IND that I mentioned in an earlier response. There would have to be a change in the ethics regulations to make this possible. The second and greater problem is the increased legal liability that a company would potentially incur were they to market a drug with unknown efficacy. Would patients have standing if the drug didn’t work? I suspect that no company would be willing to go down this road without some type of informed consent on the part of the patient. What happens when the company learns that the drug really doesn’t work and there are lots of patients taking it? Additionally, does anyone think that insurers are going to reimburse for such drugs? If they don’t, the drug would have to be sold cheaply or only to those who could afford it (which would limit the number of patients). If it’s sold cheaply, the company really boxes themselves in once the drug is approved with full efficacy claims. Do you really think a company could jack the prices up ten fold? No they would not, just look at the problems Genentech (now Roche) got into when Avastin was shown to help with macular degeneration. The company quickly brought out a new formulation with a 10 fold price increase and ophthalmologists refused to use it accusing the company of price gouging. The company claimed the drug was different and NIH then undertook a study that showed, guess what? The drugs were identical in terms of efficacy.

J Thomas July 30, 2011 at 8:30 pm

What is the empirical justification for prescribing a treatment in a non-clinical trial setting that has no proven benefits?

My thought is to have a large population that is sort-of clinical trials. Give the people who suffer from some particular malady the standard treatment. After the standard treatment gives inadequate results once, then try your first alternative. Keep using the alternative until it fails once, then switch to your second alternative. Use that until it fails once. I would suggest trying no more than three alternative treatments at a time. When the alternatives have all failed once then switch back to the standard treatment and continue it until it fails once.

This way the treatment which in practice works best gets used the most even while the trials continue. Alternatives get more chances to be used, the more often the standard treatment fails. Patients get the standard treatment most of the time, provided it gives good results most of the time.

I don’t see how we can do better.

Do you really believe that the prior probability of success of an unproven drug from a company that has a market incentive to sell the drug regardless of its medical benefits is greater than the proven probability of success of an existing treatment? July 30, 2011 at 10:21 pm

If I understand what you’re suggesting, treatment would be assigned based on whether or not a given drug worked on the previous patient, but each patient would only receive one particular treatment. Under the scenario you describe, the information from the “sort-of clinical trial” would be of little value because the trial would be neither double-blind nor would there be a placebo. It is also entirely possible that the most effective treatment would not be used the majority of the time. For example, if the standard treatment was effective 25% of the time, while the other two treatments were completely ineffective, performing no better than a placebo would, a majority would end up taking one of the two inadequate treatments under your scenario.

Despite lacking the rigour of a clinical trial that allows for the establishment of whether the treatment has a causal impact on the malady, your scenario has all of the expensive costs that come with a clinical trial because it requires you to recruit patients who are willing to be semi-randomly assigned a treatment, including one’s which have no proven value.

If instead you meant that one individual would cycle between many drugs, this would also be of little value because what you want to know is what effect x drug has on y malady, not what effect x drug has on y malady given that drug z has already failed. It would also lack a plocebo, would not be double blind, and would still have many of the same expenses as a clinical trial.

Incidentally, I offered an answer to my own question that you failed to address:

“What is the empirical justification for prescribing a treatment in a non-clinical trial setting that has no proven benefits?

My answer would be that you would have to believe that the prior probability of success of an unproven drug from a company that has a market incentive to sell the drug regardless of its medical benefits is greater than the proven probability of success of an existing treatment. And as I’ve pointed out, this would rarely be the case.”

Do you really believe that the prior probability of success of an unproven drug from a company that has a market incentive to sell the drug regardless of its medical benefits is greater than the proven probability of success of an existing treatment? Because that’s what you have to believe to think Alex’s proposal is a good idea and that would seem like nonsense to me.

J Thomas July 31, 2011 at 11:24 am

[your name here], you make good arguments. I kind of agree with you.

your scenario has all of the expensive costs that come with a clinical trial because it requires you to recruit patients who are willing to be semi-randomly assigned a treatment, including one’s which have no proven value.

My natural thought is to use the entire Medicaid sample for this. Many of the expenses are already being paid. Jan above has already voiced ethical objections to the idea.

If instead you meant that one individual would cycle between many drugs, this would also be of little value because what you want to know is what effect x drug has on y malady, not what effect x drug has on y malady given that drug z has already failed.

Yes. When you change treatments in midcourse you don’t know what the result means. I would suggest that each presiding MD be allowed to switch treatments if he decides the current treatment is not working, and that this should count as a failure of the first treatment. Or perhaps as a misfire. There’s room for bias here that I’m not sure how to remove.

It would also lack a plocebo, would not be double blind, and would still have many of the same expenses as a clinical trial.

Medicaid patients probably would not think of themselves as being in a clinical trial, and would not be told when they got an experimental treatment. They might get just as much placebo effect from the standard treatment. It’s a flaw that it wouldn’t be double-blind, agreed.

It is also entirely possible that the most effective treatment would not be used the majority of the time. For example, if the standard treatment was effective 25% of the time, while the other two treatments were completely ineffective, performing no better than a placebo would, a majority would end up taking one of the two inadequate treatments under your scenario.

Let me try the other extreme first. A treatment is 95% effective, but the new alternative does nothing. Then the large majority of patients will get the standard treatment. You’ll probably get close to a hundred who get the standard treatment before one gets the alternative, and close to a hundred before another gets the alternative, and so on. And it won’t take very many failures before the alternative gets shut down. Because it will be obvious that it doesn’t work as well.

Now try your example, but with just one alternative. The standard treatment is 25% effective, the alternative 0%. You get 2 or 3 or 4 patients with the standard treatment, and then 1 with the new treatment. 2 or 3 or 4 more with the standard, and another with the new. It takes very little time before you get enough tries to tell that the new treatment doesn’t work. It takes more tries to prove that it isn’t as good as the standard treatment than it does for the 95% case, because the standard treatment is not very good either.

The thing is, when your best treatment is only 25% effective, you don’t have a big responsibility not to look for something better. The better the existing treatment, the less reason to look for something better still. And this approach reflects that.

“What is the empirical justification for prescribing a treatment in a non-clinical trial setting that has no proven benefits?

My answer would be that you would have to believe that the prior probability of success of an unproven drug from a company that has a market incentive to sell the drug regardless of its medical benefits is greater than the proven probability of success of an existing treatment.

I agree with you. If MDs are going to use the unproven expensive drug as much as they can be marketed to do it, with no follow up, that’s a bad thing.

However, I also don’t like the alternative of letting the company that has an incentive to sell regardless, be the one who does the testing to prove that it works, and then sell for whatever the market will bear to recoup their claimed costs.

So my thought is to test it. In as much as possible the same conditions it will be used after it is approved. In reality, patients sometimes forget to take their meds, and then double the dose next time, they do all sorts of things that clinical trials tend to avoid. If you want to see the actual effect of the drug then test it in real conditions and not controlled conditions. And collect real results. If you have Medicaid results available, it’s no tremendous expense to look back at results 5 years later, or 10 years later. You might find lots of (often spurious, but maybe worth testing) correlations.

You must trust the drug company enough to believe their drug is worth testing. They have an incentive to get you to test it, but mostly their payoff only comes when they are right. They lose credibility when they give you ineffective drugs to test. If you don’t believe their drug is worth testing, then test something else instead.

If somebody with no budget can convince Medicare to test a treatment that cannot be patented, they can test it. There isn’t much of that going on today, is there? I didn’t mention a plan for choosing which alternatives to test first. You could probably come up with a good one, and I don’t know how to get your good plan put into use. That’s another flaw.

You can argue that Medicaid should be done purely for the benefit of the patients. But there’s a fair chance that Medicaid done purely for the patients’ benefit will be eliminated. If they are a central part of medical research, I hope that chance becomes smaller.

PS July 31, 2011 at 2:21 pm

J Thomas,

“Medicaid patients probably would not think of themselves as being in a clinical trial, and would not be told when they got an experimental treatment. ”

Economists, or readers of economics blogs anyway (I doubt you are formally trained as an economist), need to do some serious soul searching. What is wrong with you that you think it is ok to exploit poor people for a medical experiment because they’re poor? And I don’t mean exploit in any Marxist sense where even if someone is fairly paid for their participation they are being exploited because they can be paid less because they’re willing to take less as they’re poor. I mean exploit in a classically liberal sense of not having a contract. Of not having informed consent waivers, which is in effect a contract that you are proposing doing away with. When economists want to avoid contracts with poor people, they’ve crossed the Rubicon from soulless model builders to soulless dementors who want to remove all that is good in the world. Slight hyperbole, of course, but seriously, you need to think where your moral compass went broke and how to get it fixed. And no, you can’t buy a new one, you need to fix yours.

“There’s room for bias here that I’m not sure how to remove.”

There is only one way to remove it: a double blind placebo controlled experiment, otherwise known as a clinical trial.

“Let me try the other extreme first. A treatment is 95% effective, but the new alternative does nothing.”

You’re cherry picking. I am allowed to pick a counterexample because I only need to show the flaws in your system, you are not because you need to show that it works. The main point is not the numbers, it’s the fact that you have no idea how effective a new treatment is until it goes through an efficacy trial, so you have no basis to prescribe it because it could very well be worthless. Prescribing the untested new drug outside a clinical trial has no rational basis and is as such highly unethical (before we even get to your fucked up notion of not telling a patient that a drug is experimental).

“They might get just as much placebo effect from the standard treatment. It’s a flaw that it wouldn’t be double-blind, agreed.”

You have been failed by every science and social science teacher you have ever had. And for that I apologize. Take comfort in the fact that you are not alone. Almost no one in the general population understands experiments, the concepts of treatment, control and placebo, or causality. This includes almost all of your teachers prior to getting to college. But for the love of reality, you cannot measure a placebo effect from an actual drug alone because it is not a placebo, it is an actual drug! Yes, someone may get a “placebo effect” from an actual drug, but you cannot measure what portion of the effect of a drug is purely a placebo effect unless you actually administer a placebo.

To know whether or not a drug works you need to:

Randomly assign treatment and control (if treatment assignment is non-random, the doctor can consciously or subconsciously select healthier patients for the treatment and more sickly patients for control.)
Use a placebo as a control (Does Viagra really give a man an erection or is it all just psychological? You can’t know unless the control group is also popping a blue bill)
Double Blind (Blind to the patient, otherwise they know the placebo is just a placebo and that the treatment is really the treatment. Blind to the doctor, otherwise they will treat the patients in the treatment group differently than those in the control group).

Without this, you don’t know a damn thing and you have no business giving patients outside a clinical trial the drug.

Finally, you are right that it makes no sense to let drug companies run their own trials. However, the solution there is not less regulation, but more regulation (gasp), possibly making the drug company pay for the FDA/NIH to run the trial.

-PS

P.S. Don’t forget to fix your compass.

J Thomas July 31, 2011 at 8:54 pm

What is wrong with you that you think it is ok to exploit poor people for a medical experiment because they’re poor?

How likely do you think it is, that the US government will continue to give free medical treatment to poor people?

You are not arguing what can be done in reality, you are arguing that my proposal is not as good as the ideal ethical case which will not happen. I’m not sure what to say about that, except that you are talking about preventing a whole lot of good from happening.

“Let me try the other extreme first. A treatment is 95% effective, but the new alternative does nothing.”

You’re cherry picking.

As are you. You can look at the worst case and say it’s so bad the result should never be allowed. I want to look at the best case and the worst case both, and ideally all the cases in between. I say your cherry-picked example is not so bad, and my cherry-picked case is fine too.

The main point is not the numbers, it’s the fact that you have no idea how effective a new treatment is until it goes through an efficacy trial, so you have no basis to prescribe it because it could very well be worthless.

Well see, your bias is blinding you somewhat. You want to think in terms of having efficacy trials, and after the trials it’s OK for ignorant MDs to prescribe it. I want to track the results over the entire population of patients who use it. But I’ll settle for just tracking Medicaid until we can get the rest too.

Efficacy trials show what works in the artificial situation of efficacy trials. Tracking the results for the whole population shows what works in practice.

But for the love of reality, you cannot measure a placebo effect from an actual drug alone because it is not a placebo, it is an actual drug!

What I want to check is whether the new treatment is better than the old treatment. To the extent that they get the same placebo effect, an improved treatment is an improvement. To the extent that the new treatment will continue to get a better placebo effect in actual practice, that makes it a better treatment. And the closer we get to actual practice, the more likely that the effect carries over. If you suspect that the standard treatment is mostly placebo, then it makes sense to test it against a cheaper placebo. But unless the standard (possibly placebo) treatment takes too many resources, why not continue it until you find something better?

To know whether or not a drug works you need to:

Randomly assign treatment and control (if treatment assignment is non-random, the doctor can consciously or subconsciously select healthier patients for the treatment and more sickly patients for control.)

My approach is to number patients by order of admission, and when one treatment fails, assign an alternative treatment to the next patient to turn up, and all further patients until the alternate treatment fails once. This might correlate with something or other, but perhaps the idea could be improved to reduce such possible correlations. If individual Medicaid MDs get to everride the treatment choice, that’s another source of bias. I admit there are details I don’t have worked out.

Double Blind (Blind to the patient, otherwise they know the placebo is just a placebo and that the treatment is really the treatment. Blind to the doctor, otherwise they will treat the patients in the treatment group differently than those in the control group).

I am not so much interested in finding out whether treatments are placebos, as I am in finding better treatments. For that purpose, we don’t have to fool patients into thinking their placebo is just as good as a real treatment. We give them one of two treatments, and nobody knows which treatment is better. If we knew which treatment was better we would give them the better treatment. I hope that would somewhat mitigate the chance that MDs would find out what treatment they are providing. Sometimes, after all, it isn’t just a question of two blue pills but a difference that can’t be hidden from the MDs.

Patients are a different can of worms. There’s no reason to think that patients who volunteer for clinical trials are representative of normal patients. Your ethics interfere with your search for truth. Patients who volunteer for clinical trials, who think there’s a 50% chance they are getting a placebo and who deeply care about that possibility, are every different from normal patients.

I say, discard that model except when it’s particularly useful. To improve medical practice, the placebo effect is worth as much as nonplacebo effects. Maximize it to the extent you can. If one treatment is better because it gets better placebo effect, then that treatment is better. We want treatments that on average get better results in practice.

The better the result we already get, the less we need to look for better results still. So we try alternatives slowly.

The less successful the standard treatment, the more effort should go into finding something better. Try alternatives often.

If you have a treatment with a 25% success rate, your obligation to give patients that treatment instead of something plausible but unproven is not very large.

I am glad to see that you understand about doing scientific experiments. A whole lot of people don’t. I hope you will consider my ideas and look for ways to improve them. The clinical trials method does not work well enough. We need something better, and quickly. I am convinced that the approach I outlined is heading in the right direction, but it has a couple of flaws and some big holes. I did not suggest how to choose which alternate treatments to try first. Individual MDs could interfere in ways that could bias results. It’s vaguely possible that a new treatment might get a placebo effect when it’s new that it would lose when the new wore off.

Also, this approach does better toward improving treatment than it does toward determining mechanisms. For treatment results, placebo is as good as anything else. If it works, use it. But when you are groping with the science to create a better treatment, placebo effect completely gets in the way.

“One man’s confounding variable is another man’s experiment.”

Barnley B July 30, 2011 at 11:13 pm

Milton Friedman argues for that policy in this video: http://www.youtube.com/watch?v=dZL25NSLhEA

Andrew' July 31, 2011 at 5:14 am

It’s getting more and more obvious we need more of that $billion invested on the back end of drug individualization rather than the front end of universality.

As I like to say, there is nothing as discriminatory as disease – the only thing that affects us all is aging (and that’s another story).

Andrew' July 31, 2011 at 5:19 am

“This statistical barrier discourages the pharmaceutical industry from even beginning to attack diseases, such as Parkinson’s, that are likely to have several subtypes, ”

Is this all because we require MDs to be stupid? We can’t subtype something we call Parkinson’s, therefore all drugs must address all the Parkinsons? It can’t be, can it?

David July 31, 2011 at 2:12 pm

Japan has had this kind of safety-only regulatory regime for decades. I don’t think that Japan can be held up as a example of how we want the pharmaceutical market to work.

I agree that more publicly funded studies, especially comparative efficacy studies, would be a step forward. I think that creating an infrastructure to efficiently decide which of these extremely expensive studies should be funded is a very big problem, maybe even a nightmare. Add to that challenge, the difficulties in getting public institutions to agree on study design, entrance criteria, etc., and I am pessimistic that there can be any large expansion in publicly funded studies in the near future.

Also, the comments earlier about the problems in getting patients to participate in such studies should not be undervalued. There are hundreds of years of experience that shows that desperate patients are happy to buy into snake oil.

Old oncologist July 31, 2011 at 4:38 pm

Some late further thoughts ( esprit d’escalier gets more common as one gets older).
What does the term “safe” actually mean? Is cisplatin a “safe” drug? Is high-dose chemotherapy with autologous stem-cell rescue “safe”?
Safe can only mean having a favorable risk/benefit ratio. Then cisplatin is spectacularly safe in testicular cancer, and high-dose chemo is immensely safe in relapsed Hodgkins. If you don’t have efficacy data, then you don’t know the benefit.

Comments on this entry are closed.

Previous post:

Next post: