Personalized Medicine

Patient X was rushed to the hospital for emergency surgery. As she entered the hospital she said to the anesthesiologist, “You may not want to use suxamethonium on me.”

“Have you had a previous reaction?” inquired the anesthesiologist.


“Ah, a family member must have had a reaction.”


“Why then are you concerned about this drug?”

“I’ve had a good portion of my genome sequenced,” the patient replied, “and I found that I have a genetic variation in the enzyme that breaks down suxamethonium and am part of the 5% of patients who respond unusually to this drug. I thought you should be aware of this information.”

The flabbergasted anesthesiologist wondered how long it would be before more of her patients came prepared with their own genetic code.

I made up the details of the conversation above, but otherwise the story is true. The patient was a customer of 23andme, a service that for around $200 will give you information on about half a million sites on your genome, how you differ from other people at those sites, and which of your variations are associated with various diseases, behaviors and capabilities.

The costs of sequencing are falling so rapidly it will soon make sense for everyone to carry their entire genetic code with them on a USB drive (23andme only identifies part of the code). In 2001 it cost Craig Venter $100,000,000 to sequence the first human genome (his own.) Today, it costs just $16,000; in a few years, it will cost less than $1,000–a 100,000-factor decrease in costs in less than two decades!

That’s me from a piece called The FDA and Personalized Medicine written to help launch a new blog from the Manhattan Institute, Medical Progress Today. I go on to argue that if we are to take advantage of the new possibilities for personalization “we must move the FDA away from pre-market gatekeeping and towards post-market surveillance and information provision.”

David Henderson, Rita Numerof and Paul Howard all comment.


You start to tap out the number of cheap compounds you can give to an entire population and get high benefits and low negative outliers. You are reaching the point of diminishing returns on broad-based drugs from cheap to synthesize small molecules, so one of the next possibilities is personalization and small run more expensive compounds (e.g. complex molecules, proteins, antibodies, synergistic combinations, etc).

Doctors will have to adjust to a new world where they must still know the generalities, maybe moving toward greater understanding of drug interactions (both positive and negative), but now their patients handle more of their own personal specifics. Many things will change and if that requires the FDA to change, then that's the way it is. So be it. Requiring a drug that is never intended to be used in the whole population to be able to be used in the whole population shall we say, suboptimal. Alex didn't invent that, he's just informing you of how reality works under certain circumstances..

Why can't computers do the generalities?

Good points. Gotta remember to add this to my to-do list.

File under There is no Great Stagnation?


"File under There is no Great Stagnation?"

File under "proof" of The Great Stagnation. Simple, low molecular weight compounds that are easy to develop and apply to the vast majority of the population yield large economic gains. Think Penicillin, Tetracycline, Valium, Lipitor, steroids, etc.

Personalized medicine fits the new model of ever more marginal improvement with exponentially rising costs. Type "medicine diminishing returns" into Google and you will see what I mean. For one example, see "Have we discovered it all? Billions are spent on medical research, but we have entered an era of diminishing returns" ( Quote

"And yet the more generous these financial resources become, and the more daunting the tidal wave of new facts and knowledge, the more striking is the paradox that the rate of medical innovation – in terms of discoveries that really make a difference – was enormously greater 30 or 40 years ago, when the scale of research funding was a fraction of what it has become. That era saw the advent not just of virtually every class of drug in common use today, but of open heart surgery, hip replacements, dialysis, transplants, test-tube babies, and much else besides."

I'm a geneticist and I make genetic tests for disease predisposition. The devil is in the detail, check out this:

The problem is that each test has a false-positive rate and 23andme (and the like) do thousands of tests at once. So when you do 23andme you are (on average) expected to get a bunch of false positives.

There's a good chance that health-providers will within 10 years ignore genetic findings because of the ridiculous compound false positive rates.

What we do now is test the drug in the population, we restrict the size enough to make us feel better, then we give the drug a go/no go and haven't really learned anything.

I meant the other type of 'compound'.

Yah. But now we don't do any genetic screening. Or, rather, we do our genetic screening by dumping the drug on the population and seeing who dies. The problem with false positives I suspect comes into play with what I call the fallacy of precision we think we have in the medical industry.

Alan, I think you are absolutely right that "the devil is in the details," but I think you underestimate (vastly) the potential for the next generation of research to go far beyond what we do now. In your lifetime, I suspect, science will know not only the statistics but also the *reasons* for what happens when drugs meet bodies. Then there will be a totally different model of what to do with the genetic information.

Ken, your optimism is admirable, but we have been on the cusp of really understanding "what happens when drugs meet bodies" for years now (at least according to the popular press), and basically the more physicians know, the less certain we become. How far away are we from understanding what even a fraction of our genes really do, in isolation? Go ponder that question (the answer will depress you), then compute the statistics of how long it will take to really understand how those millions of genes interact in their literally infinite combinations.

Most non-scientists and non-research physicians don't have a clue just how inexact the state of science is in medicine and pharmacology, and this leads a layman to tell a geneticist that we'll have Star Trek's medical tricorder in a generation.

Where I'm really skeptical is not that medical and research progress will be made. It will. But the idea that this will lead to 'personalized medicine' that will be dramatically better, and cheaper, than what we have now is unbelievably naive, and flies in the face of the history of modern medicine. I have no doubt that reforming the FDA will lead to more and cheaper drugs, and that in a generation more people will be taking more drugs than they do now,. But I don't expect our quality of life or health to be an order of magnitude different, or our costs to be a much smaller than they are now because of that.

Medical Progress Today is new? I could swear I have some of their podcasts somewhere on my playlist.

In other news, my family still can't afford $1000 per person 5 years from now, because there won't be any jobs, and the insurance companies won't cover anyone. So, health tech is becoming increasingly irrelevant even in America.

Your personal lack of optimism about what you will be able to afford in 5 years has nothing to do with the relevance of health tech in America. If the cost truly drops to $1000 per person 5 years from now, the number of people who can afford that - regardless of whether or not you expect to be among them - will still be much larger than those who can afford $16000 per person today. And I would encourage you to raise your own expectations as well about what you will be able to accomplish in the next five years...

The Great Stagnation rears its ugly head again. Seems unlikely this will be properly reflected in the CPI. Oops, that's Alex posting, not Tyler, I always need to check that when I have this kind of reaction to a post.

Looks like Alex Kossoy beat me to it.

@Alex: the false positive/false negative rates of these tests are indeed a problem, but knowing the information, and that the error rates are small (for a given locus) does increase the decisionmaking power of the whole system. Yes, each result-set can be expected to include errors, but the error is unlikely to cover the entire result set (barring techs mixing up samples!). In the demonstrated case, the doctor avoided succinylcholine in favor of a different anesthetic. If the test was in error, and succinylcholine was indeed safe? No biggie. If the invese occurred, apnea is a manageable and known complication. What if the complication would be uncontrollable? Then repeat the test for that locus. If time isn't an option, then the doc has to make the call, but I'd steer clear of anything that was negatively indicated, as the power of the test has increased my knowledge, even with error (Bayesian formula here).

I meant @Alan!

"I made up the details of the conversation above, but otherwise the story is true. "

Huh? Did you make it up, or is it true? If by "true" you mean is "made up but hypothetically possible," that isn't quite the same thing.

What the author is saying is that a patient did indeed inform their doctor about her potential drug reactions based on knowledge from her genetic profile. But this real-life conversation between the patient and her doctor was obviously unrecorded, so the author imagined how it might have went. The author dramatized a real event.

Wait a tick: why are you assuming that an anaesthesiologist wouldn't be up-to-date on possible drug interactions and their warning signs? Or that, given the tremendous training required to be one, they wouldn't be able to come up with a treatment regimen on the fly that avoided use of any one drug?

Oh, right, you were being realistic about the medical cartel ;-)

Modern medicine is based on the assumption that people who are different are the same.

i would say having portable storage of their genetic code could have effects on social issues. if its made so that it can sync with anyone else's genetic code, it could make quick genealogy tests. it would become trivially easy for fathers to test their children for paternity. it would also help avoid the growing instances of accidental incest.

Where did you find out about this new rising trend in accidental incest? I have to say your post is the first I've heard of it being an increasing problem...

This is really cool. I wonder who's working on a standardized (XML-based?) way to store this information on the USB key so that it's interchangeable and usable by the hospitals.

Judging by the success of past/current e-health initiatives in Ontario and the UK, it probably shouldn't be a large integrated contractor that takes it on. It should be the online open-source community. But no doubt the U.S. govt. will still spend a billion dollars for Raytheon to develop a useless system.

something tells me that a patient seeing an anesthesiologist upon admission into an ER won't be saying much of anything.

One barrier to having your genome sequenced is that insurance currently relies on us not knowing the likelihood of getting various diseases. Once we do then those prone will over insure (if you know for sure that you will die in 5 years from a disease with no symptoms at the moment you might well insure to leave your family secure) and those not will under insure which will make it very difficult for the industry to survive. One response will be to require anyone taking coverage to say that they have not had any genetic tests that reveal a greater propensity to diseases with forfeiture of coverage or benefits if it can be shown they lied. The other alternative is to prevent adverse selection by requiring everyone to buy life insurance and thus keeping the pool nice and deep but it's difficult to see how this could be done.

This is exactly why it represents the Next Investing Boom.

Hello everyone. I am enjoying reading this because I actually was patient X about whom the blog was written. The dramatisation of the conversation is not a million miles away from the reality, which actually went a bit more like this. Its important to remember that the background state of mind was acute pain due to a really horrendous swelling of the face.
[on admission, night before operation]
Anesthetist (A): Have you or any of your family members ever had any adverse reactions to general anesthetics?
Me (X): no, I haven't had an anesthetic for a long time and I don't think my family have had any problems. By the way, these pain drugs are great, they really work
[morning of operation]: A: I know we have asked you this before but have you or your family ever had any adverse reactions...
(X) no I don't think so. please can I have some more pain drugs.
[a little later before the operation] :A: just checking that you haven't had any adverse reactions to GAs?
X: Actually now you mention it, I have been genotyped using 23andme, and I seem to remember there was something in the drug reaction content that was relevant..can't quite remember, why don't we take a look?
[A and X then go to the ward computer and log on to the 23andme website to retrieve result and read about the relevant mutation]
A: that is amazing. Not only is it fascinating clinically, but I have never had a patient present before with knowledge of their own genetic data. [understands and explains implications of mutation if it were to be clinically confirmed]. the clinical test is time consuming and we are going in to operate, luckily there are other reasons for us to use a specific other anesthetic but if we hadn't chosen that route in any case this would have been very important.

History does not relate what would have happened if this was the only factor to indicate an alternative anesthetic and the need to clinically validate the mutation (23andme is not done 'for clinical purposes').

Comments for this post are closed