No sooner do I say that perhaps it's time to stop blogging about health care short of interesting economic side issues, then Julian Sanchez raises an interesting economic side issue:

Do we believe that? That every American has basic right to extremely expensive drugs that provide very little benefit? It's one thing to say there's a shared obligation not to let people suffer or die when we know how they could live many years longer, or in much less pain. I find it a whole lot less compelling to suggest that people are entitled to public provision of, say, Tacerva--which the Times article says was approved to treat pancreatic cancers because it improves survival time by a whopping 12 days at a monthly cost of $3,500. Another is good for an additional month and a half on average, at a per-patient cost of $50,000. Is it only people who favor dismantling Medicare who might think that this goes beyond what people must have as a matter of basic justice?

I think one of the genuine problems that liberals will face if they get the comprehensive system they want with some sort of cost-effectiveness review board, is that cost-benefit analyses break down at the margins of life.  I'm not concern trolling here; it's just an intellectually fascinating, and politically thorny, problem, which is why Ezra Klein wrote the original post that Julian linked.

For starters, I'm not sure that it's meaningful, in discussing a medical treatment, to say that a drug, improves survival time by a whopping 12 days.  Here's what the New York Times wrote:

And even some of those drugs offer only a few months at most of extra life or tumor stabilization despite prices that often reach thousands of dollars a month. The drug Tarceva, which costs about $3,500 a month, was approved as a treatment for pancreatic cancer because it improved survival by 12 days.

The battle to treat cancer has become, as a commentary in a leading journal put it, a "grinding war of the trenches."

I think that when we talk about it this way--and this is how we talk about so many diseases--we're prone to composition fallacy.  Which is not to say that this is how Julian is thinking about it, but that this is how such statistics are often taken when policymakers discuss them among themselves. 

Twelve days indeed does not sound like a lot.  The problem is that the drug does not improve each patient's life expectancy by 12 days, in which case yeah, it doesn't sound like such a great deal.  It gives some people extra months of lives, probably slightly shortens some lives because most serious drugs have iatrogenic effects, and does less for other people.

If you frame the question as "do you want to pay $12,000 to get a 20% chance of an extra two months?" you probably get a different answer than if you ask whether someone wants to spend it on an extra twelve days.  Particularly when you consider that the mean survival time for metastatic pancreatic cancer--60% of all cases--is four to six months from diagnosis.  For another 30%, who have locally advanced cancer, mean survival is 8-9 months.  Twelve days, or two months, doesn't sound like so much to me, when I'm in my thirties and, as far as I know, cancer free.  But if we were talking about a 20% chance of increasing my lifespan somewhere between 25% and 50%, I'd probably want to pay for Tarceva.  I might even start sobbing and call my local television station if my doctor, or insurer, denied it.

Expected value calculations just don't work very well when you're facing imminent death--a criticism that liberals justly make of libertarians who claim that the only problem with the health care market is third-party payment.  We get away with doing these sorts of cost benefit analyses for environmental regulations and highway safety issues because they don't involve directly denying a specific person a treatment that has a decent chance of extending their life.  Even countries where the political culture is much friendlier to bureaucratic limits have ended up backing down on these sorts of decisions in the face of exactly the kind of outrage American wonks are getting from seniors when we propose not just showering them with money every time they ask. 

I think you could make a coherent argument that in fact people who have just found out they have imminently fatal diseases should get more money spent on comparatively small gains in lifespan, for the same reason that I'm potentially willing to incur a large deadweight loss in order to transfer $500 a month from the middle class to welfare mothers who earn a few thousand dollars a year, and not potentially willing to incur the same deadweight loss transfer $5,000 a month from Warren Buffett to me.  I'm not saying that this is the standard we should adopt.  But I actually think that it has a fair amount of emotional resonance, which is why appeals from cancer patients are so effective when they're denied marginal treatments.

There are a few other considerations that I think we have to at least consider adding into any sort of cost-benefit analysis.  My understanding of cancer treatment is that it is incredibly incremental.  Doctors improve survival rates not with one or two eureka! drugs, but by grinding out often tiny incremental improvements:

Most scientific progress is incremental and unspectacular, and some spectacular successes have been progressive but, some would say, scientifically mundane. The cure rate for childhood leukemia went from virtually zero to 80% over 50 years with only two novel ideas. The use of drugs in combination instead of singly, adapted from the successful antibiotic treatment of tuberculosis, and the application of treatment of pre-symptomatic leukemic involvement of the central nervous system. Even those innovations failed to cure any children at first.

But a systematic series of clinical trials, that tested variations in the sequence, doses, and combinations of available therapy, led progressively to this amazing success. This occurred despite having no effective new drugs introduced after 1970. All of these studies made scientific sense, were doable, and were funded by grants from the NCI.

A twelve day improvement in lifespan may be part of a package that doubles or triples pancreatic cancer mean survival times.  I'm not saying that this is necessarily true of Tarceva--some drugs don't work so well.  But 12 days probably isn't the right way to look at the problem.

There are the next drugs that we may want to encourage--unfortunately, we don't know that a drug won't give us spectacular lifespan enhancements until it's cleared Phase III trials.  Reducing the return on the drugs that offer modest improvements reduces the number of drugs brought to market, reduces the number of eureka drugs, yada yada.  That tradeoff may be worth it.  But it should be in your calculation as a good social planner, even if only to worry about how you are going to replace the lost R&D.

There's also a question about how many months of life we're expected to bear this cost.  A very expensive drug for pancreatic cancer still probably doesn't cost us that much money, because of those ugly, ugly mean survival times.

And finally, a core problem that comparative effectiveness research is going to have to face is that mortality is much easier to measure than morbidity.  This can cut either way--we all hear the horror stories of old people senselessly hooked up to machines to keep them miserably alive for a few more days, or cancer patients taking punishing chemo regimes that offer little hope of much prolonging their life, and a near-certainty of immiserating what time they have left.  But chemotherapy and radiation regimes, even quite expensive ones, can also be palliative care, shrinking tumors that cause pain or other sorts of misery.  Maybe it's not worth $3,500 a month to approve Tarceva in order to get patients an extra 12 days.  But it might be worth it to give patients an extra 12 days, and a 10% decrease in their cancer symptoms.

Again, I'm not saying that this is necessarily the case with Tarceva for pancreatic cancer--but it does seem to offer palliative as well as life-extending benefits in other forms of metastatic cancer with a pretty attractive side effect profile--most common effects were diarrhea and rash.  These are things that doctors can go over with patients when they're making a decision.  I think it's harder for a comparative effectiveness research board to determine whether the tradeoff between digestive upsets and skin disorders, and more rapid progression of a really very unpleasant cancer, plus whatever the extension in mean survival time, is worth $3,500 a month for 4-8 months.

Obviously, Julian is only considering what the New York Times wrote, not what it didn't.  And I suspect he's primarily interested in the theory of justice rather than the policy theory.  But for the wonks who will actually be deciding the application of this theory of justice, things rapidly get more complicated.  Especially when you add politics into it.  Because I suspect that while in theory "we" don't want to spend $3,500 a month to buy a measly twelve extra days of live, in practice we'll find that "we" actually do want to spend $3,500 a month to give a frightened cancer patient a slightly better quality of life, and a decent shot at a few extra months to say good-bye to her family.