• Personalized medicine: Will it bend the cost curve down — or up?

    Personalized medicine is expected to revolutionize health care by developing new molecular tools to precisely diagnose and treat the problems of individual patients. But how will these new technologies affect the cost of care? Changing the cost of care will in my view require more than just changing how medical care is paid for: We also have to develop more efficient treatments and delivery systems. The story of a new treatment for cystic fibrosis (CF) may give us some idea about what will be possible in the near term.

    Diagram of the mechanism for the transport of chloride across a cell mechanism.

    Diagram of the mechanism for the transport of chloride across a cell mechanism.

    CF is a rare (1 in 4,000 US children) genetic disorder of the gene for the protein ‘cystic fibrosis transmembrane conductance regulator’ (CFTR). Defects in that protein cause abnormal transport of chloride and sodium across an epithelium. This in turn causes dysfunction in the lungs, pancreas, liver, and intestine. In particular, it results in the development of thick mucous secretions in the lung. These secretions become a site for infection. The infections cause diseases such as pneumonia, leading to reduced life expectancy.

    Previously, caregivers could only address the symptoms of CF, for example by keeping the patient’s lungs clear, keeping the patient fit, and by aggressively treating the infections. They could not address the disease at the cellular level. However, clinicians have been extremely successful at managing symptoms and life expectancy at birth for US patients has increased from just months in the 1950s to over 50 years today. Much of this has been achieved through close cooperation between CF caregivers and CF families.

    The cover of Science announcing the discovery of CFTR.

    The cover of Science announcing the discovery of CFTR.

    There is also great interest in attacking the disease at a molecular level. The CFTR gene was discovered in 1989. Since then, there have many efforts at developing either a gene therapy (without success) or a medication that can address the epithelial transport problems at the origin of the disease’s causal chain. Among many challenges is that there are more than 1500 mutant alleles, so in effect, the already small ‘CF population’ is actually a cluster of many tiny genetic subpopulations.

    One research group picked the mutation G551D-CFTR, which occurs in 4-5% of CF patients. With funding from the CF Foundation, they tried over 600,000 compounds and eventually developed a drug named Kalydeco (ivacaftor). Results of a clinical trial were published in the New England Journal in 2010 and FDA approval was obtained in January of 2013. So far this effort has been a success, producing a significant improvement in lung function for patients. So, just as the CFTR was one of the earliest disease genes to be mapped, Kalydeco is one of the first successful medications produced by genomic medicine.

    That is the good news. This bad news is that it took 24 years and tens of millions of dollars to get from the discovery of the CFTR to the FDA approval of a drug. Moreover, this drug was designed for a mutation found in only a small fraction of the population of an already rare disease. Most importantly, the drug reportedly costs $300,000 / year, in part because the market is likely small. (These costs may be offset in part if the patients taking it have fewer hospitalizations or are less likely to need a lung transplant.)

    Future drug development in personalized medicine might get cheaper. Still, precisely because the treatments are targeted at phenomena at the level of specific harmful mutations, they are not just personalized but practically bespoke, and correspondingly pricey. The CF story therefore gives us reason to fear that personalized medicine will make health care more rather than less expensive

    So what should we do? I am hopeful about personalized and genomic medicine: I am part of a group that has been working on the development of gene expression biomarkers for depression. More importantly, even if genomic medicines prove to be very expensive that may be okay if they are also very effective treatments. What matters is not the cost but whether we get real value for the money.

    Having said that, if we are going to bend the cost curve down, we will probably need to do it in the face of upward health cost pressure from personalized medicine. But I think the big lesson from cystic fibrosis is how much can be accomplished through low-tech but intensive management of a chronic disease. We need to find better ways for patients to take care of themselves and better ways for clinicians to help them do this. Finding better ways to build partnerships between clinicians and families does not have the scientific glamour of molecular medicine, it will not garner Nobel prizes, and it will not create profitable intellectual property. But it can be extraordinarily effective at saving lives.


    • Any way I think about it, personalized medicine is going to be costly. Just look at the cost of target therapy for cancer, eg. Herceptin, approved by FDA in 1998, still cost $70,000 a year. When personalized medicine does work, patients live longer, which also tends to generate higher cost for care. As Bill said, it eventually comes down to a cost/benefit analysis, and we’d better get our money’s worth for it.

      I think we can still do a lot better with traditional, intensive management. The extension of longevity in CF patients in the past was mostly due to aggressive treatment of CF symptoms, eg. airway clearance, supplemental nutrition, antibiotic use for infection. Establishment of coordinated CF centers around the country probably helped a lot. Another example came to mind is cancer. We now have pretty good 5-yr survival for ALL with combined chemotherapy even before we know the mutations that drive it.

    • the drug reportedly costs $300,000 / year

      But the great thing about drugs is that they eventually go off patent. So spending may rise for these patients but will later fall.

      • With such a small patient population, it is highly unlikely that any generic company will choose to enter the market. If one did, it would surely charge $290,000 a year to try and get some market share while maximizing profits.

    • In a similar vein, I have a relative who has treatment resistant depression. Her neurologist reported that he had heard of a study that showed benefits for Deplin (l-methylfolate) (this is NOT spam, I am NOT a spambot) on patients with a particular gene mutation with treatment resistant depression. I’m not sure which article that is, but I did find an article that discusses Deplin supplementation.


      This doesn’t appear to be a well-known treatment. But it is $144 for 90 15mg capsules through the manufacturer. That’s not the cheapest, and it’s not covered by insurance. But it does raise the possibility that not all forms of personalized medicine will be stratospherically expensive.

      That said, though, the clear economic incentive is for the industry to develop the $300k personalized medicine solutions. So maybe we need to establish a prize system for specified medical problems, as well as a patent system (Matt Yglesias says we could do just as well abolishing patents entirely right now but that’s another story).

    • I’m hopeful that personalized genetic medicine will result better treatments with existing drugs. Cancer drugs only help some patients. If we could figure out which drugs help which patients, getting the right drug would require a lot less trial and error and a lot less unnecessary side effects. Even painkillers only help some patients. Aspirin + Tylenol + caffeine works better for migraines than any one or two of those on average across large populations. But is it really that combination, or is it that some folks respond to A+T, others to A+C, and still others to T+C, and you have to give all three to get everyone treated.

      The problem, of course, is that past drug trials didn’t sort for genetic differences, so there’s a near infinite amount of work to be done in rerunning trials on more carefully targeted patient populations.

      Unfortunately, there’s no real savings in there, just better treatment.

      Another thing here, is that as a Comp.Sci. major, the idea that 90% of human DNA was junk was insane from the start: there has to be control information as well as protein information in the DNA. (A corollary here is that it’s no surprise that we’re not finding very many diseases controlled just by problems in protein coding; basically all the easy (i.e. protein genetic glitches with strong effects) ones will be visible in the effects of the problems and the massive genome searches will only find minor things.)

      So as the biologists begin to get their heads around this idea, there will be a lot of new science coming out. The next 20 years should be fun.

    • One of the many problems of Mr Gardner’s cost analysis is he appears to be drawing on current or even retrospective examples (“it took 24 years and tens of millions of dollars to get from the discovery of the CFTR to the FDA approval of a drug.”), and cherry-picking examples of $300,000 drugs to arrive at possible cost outcomes of a field that is just coming into fruition.
      One real world, current example that refutes his analyses: we are about to release tests that can identify the entire range of driver mutations in acute myeloid leukemia. As few as six months ago this would have been impossible – these mutations were not even identified until the last year or so.
      Further, all these critical driver mutations will soon be identified for a few thousand dollars (as opposed to $300 per mutation) and many linked to drugs already approved for other diseases.
      New current thinking, which Mr. Gardner is clearly unaware, demonstrate that cancers such as breast cancer should not even be evaluated in the limited context of “breast cancer”; rather, they are more properly evaluated in the context of the driver mutations that constitute the genetic basis of the heterogeneous group of diseases that are cancer. In sum, a colon cancer drug may be the best and most targeted treatment for a patient with breast cancer, prostate cancer, or lung cancer.
      Still further, the real promise of cost savings from personalized molecular medicine is the ability to intervene early to head off the onset of chronic diseases (e.g., diabetes), which are epidemic and cost taxpayers tens of thousands of dollars per patient, per year. This facet alone can save taxpayers literally billions of dollars per year in healthcare costs and improve quality of life for millions of Americans.

      • Jeff/Invivoscribe: You mention the “real promise” is to treat chronic diseases like diabetes earlier. We already know how to detect many diseases early–the economic and social costs are driven by non-compliance, lifestyle, and other factors, not the lack of intervention. A genetic test is not going to change people’s motivation, habits, or time preferences.

    • “Personalized medicine” is a bit of a marketing term. What the issue is mainly about is genetic risk groups and targeting interventions (mostly drugs I assume) at certain groups. It is not about individuals somehow having uniquely created medicines just for them, which is how it’s often portrayed in the media. BRCA testing is a reasonable example of cost-effective personalized medicine.

      My main concern about this development is the massive growth in skill that will be required to select a course of X treatment for a given set of patients and manage their therapy, while presumably being aware of other known individual risks for patients. We are going to be way too overconfident in our tendency to leave this to individual clinicians, and would be better off moving more of this kind of treatment planning into Watson-style algorithms.

    • “What matters is not the cost but whether we get real value for the money”?!!!! It is exactly this kind of thinking that inspires ever greater investment into the ever-growing profits of biomedical research. Cost control is always postponed in the name of saving lives.

      I submit that we will save more lives when we stop inventing more medical care. More medical care simply puts health insurance further out of reach for more people. And we all know that the uninsured die younger.

      Let’s stop deluding ourselves that new medical care saves lives. The new expense is killing us.

    • When I worked in the Pharma industry, we often got into discussions about the future of personalized medicine. Its not just individualized treatments. Its also the ability to predict whether a given medication will work for a particular patient. Should we start this person off on ibuprofen, or would naproxen be better for her?

      From an industry perspective, the upside is that R&D could be dramatically cheaper. If you were highly certain that a drug would work in a given patient, trials could be much smaller. If you could predict which patients would suffer dangerous side effects, trials could be much shorter in duration.

      But the marketing people would never go there. They couldn’t, for the life of them, figure out how the company could make any money in such a fragmented marketplace. So we never really invested any serious R&D money in this kind of technology.