What happens when a drug works — but only for one person?

Really, really intriguing piece at Nature News by Heidi Ledford. It's all about a class of patients called "exceptional responders" — aka, the people who got a benefit (sometimes a big one) from a medication or treatment that otherwise failed the clinical trial process. When we do clinical trials, we're looking at group averages. We want to know whether a drug performed better than placebo when administered to lots of people. Sometimes, though, drugs that can't do that do seem to have a positive effect for a few lucky individuals. Now, scientists are trying to figure out why that is. What makes those people special? And how should this change the way we do research?


  1. A friend of mine recently gave a PhD defense.  He was looking for causes in increased HIV infectivity after altering some hormones that stimulate proteins secreted by cells in the uterus.  They tested 6 different protein-hormone combinations, and it’s the state of that art.  Later in his talk, he mentioned that there are over 500 known proteins in that area.  It’s on that level.  We are 1% in the know, and 99% in the dark.

  2. This is where I think most Woo-woo falls. A technique is not scientifically proven but seems works for some people; be it homeopathic dilutions, spiritual healing or chiropractic medicine, they can’t all be delusional can they?

          1. I don’t quite follow your question either. Do you mean what’s the difference between magical thinking and seeing (or modeling) an effect as occurring due to chance?

          2. An effect occurring due to chance vs. an effect occurring after magical thinking.  Seemed pretty clear to me, but there it is, spelled out.

    1. I don’t think it’s the same as woo. It is possible to have evidence that a treatment has worked for one person, even if it hasn’t been shown to have a significant average effect for a larger population. A combination of temporal correlation, plausible mechanism, and unlikelihood of response in the absence of treatment can separate something that actually “seems to work for some people” from situations where that phrase is a euphemism for “it looks like a placebo effect but I want to be polite”.

      1. The word woo doesn’t describe a kind of therapy; it describes the speaker’s contempt. There is no objective difference between a therapy that only works because woo and a therapy for which no explanation has yet been found.

        Suppose there’s a compound that sometimes seems to cure something. If that compound is only 10% more effective than placebo, it can never get FDA approval.  Is it science, or is it woo? What if the compound is called 1,3,5 alpha-techno-enzymase? What if it’s called St. Judas Goatweed? What if those are both the same compound?

        1. I don’t understand. Why could it never get FDA approval if it’s demonstrably more effective than placebo? As for terminology, I think I see what you’re getting at — sometimes we can be confident that something works without understanding the mechanism. I agree with that. But it’s also true that some things don’t work even though a lot of people think they do. You’d agree with that, right? It’s not so important to me what names we give these categories.

          1. I could be wrong, but I think the FDA demands efficacy much greater than placebo+10. For example, there’s a popular prescription that helps people quit smoking. It was found to also be effective in treating addictions like cocaine and heroin – with inconsistent results. So the FDA can’t approve it, but many doctors will prescribe it anyway.

            I’ve always wondered: if the placebo effect is real, why can’t a doctor prescribe one? I mean, they prescribe antibiotics for a virus regularly, and that’s pure placebo. But there should be a scary medical name – 1,3,5 alpha-techno-enzymase – that’s really a sugar pill.

          2. I hadn’t heard before of a minimum for how effective an effective drug has to be for FDA approval, but you could be right. Of course, the less effective a drug is, the harder it is to demonstrate that it’s effective at all.

            To make a placebo work a doctor has to mislead the patient to some extent. That’s why it’s considered unethical — patients are supposed to be able to trust their doctors. Some doctors are arguing that it’s okay as long as you say something factual, such as, “Here’s a treatment that studies show can be effective for treating symptoms such as yours.” I think it’s still unethically misleading to leave out that last, crucial, “…because it’s a placebo and I’m avoiding telling you that in a way you’ll understand.” There’s been some coverage of this debate in the press: https://www.nytimes.com/2008/10/24/health/24placebo.html

          3.  In answer to allenmcbride… one of the most interesting things about placebos is that the doctor /doesn’t/ have to deceive the patient. Placebos, for whatever weird reason, generally work just fine when the patient knows it’s a placebo. If I remember right, there isn’t even much of a drop in their level of effectiveness.

            Made for some interesting conversation in anthropology of medicine class.

          4. [Reply to Daemonworks:] I’m curious and skeptical. How did they measure whether the patient knew it was a placebo? Like the example I gave of what a doctor might say to a patient on giving a placebo, there are many ways to say words that add up to “this is a placebo” without successfully communicating the heart of the message to someone with little scientific background. But maybe the doctor said, “This is a sugar pill. There is nothing in here but ordinary sugar. It will not help you physically. But for psychological reasons, just the act of taking a pill might help you.” If they said that and the placebo still worked, *then* I’d be impressed.

          5. FDA as a rule doesn’t end up approving natural substances as drugs, even if there is proof of efficacy.  Items that can’t be patented won’t get backing of a drug company to spend the billion or so required for drug testing.  Ginger for nausea is the typical example or an unpatented plant generally accepted as having efficacy, taxol is the patent exception.

          6. I think a lot of that is that ‘ginger’ isn’t a particularly useful prescription. There are a number of varieties of ginger. Which one (if any) are effective for particular symptoms? How about the conditions that it was grown in? Does it need to have a particular mix of environment for the end product to be usefully effective or have a predictable dose?

            When you get prescribed a FDA approved medicine, you know exactly what you are getting in a chemical sense. The ratio, dose, strength, everything. When your medicine is picked out of the grocery store produce isle you can’t be entirely sure about things like what country it was grown in or what subspecies it may be. These are things that tend to make the FDA nervous.

          7. If it made them nervous they wouldn’t consider it as safe. It’s not recommended as a drug because that’s not the sort of thing they determine. Take a gram a day is not that risky a recommendation.  May aggravate you if you have gallstones.  You can grow it yourself if worried about where it’s grown. If you want get hipster and only take a subspecies that I probably never heard of, you can so indulge yourself.  I do worry about where and how it’s grown, but I do so about everything I eat.  I eat considerably more than a gram a day of other plants.

          8. The FDA does end up approving natural substances as drugs because as far as the FDA is concerned, they’re not drugs. Things like herbal remedies and foodstuffs are regulated completely differently, and there’s no need to demonstrate effectiveness. If someone wants to drink ginko tea because their granny always told them that was good for a stomach ache, the government sees no point in regulating that. This is why a number of recreational drugs were sold openly on the market as herbal supplements until laws were created specifically to lassify them as illegal drugs.

            Back on point, a doctor could advise some sort of herbal supplement as part of a healing regimen if he had good reason. The $64,000 question is whether, for a given ailment, that’s considered treatment enough to cover the doctor’s liability. Generally, doctors are pretty wary of this sort of thing. A doctor experienced in such things might recommend a cheap herbal approach when possible, with an “as needed” prescription to back it up if that doesn’t work.

            There has to be a good deal of research to back up an herbal approach. Even if there’s research out there to back it up, it’s not entirely reasonable to expect your doctor to be aware of it. The notion that you can reasonably expect your doctor to be well-versed and competent in all medical knowledge in this day and age is pretty
            ridiculous, but it’s an idea that people cling to nonetheless.

        2.  There is a difference between science and woo. Woo are things that have been proven to NOT WORK. Science is something that hasn’t been proven yet to work. There’s a very significant difference.

          1. nope. you can’t *prove* that something doesn’t work. you can’t prove a negative anything. basic logic.
            you can show that you can’t, personally, get it to work – but how ever hard you try you can’t defend against someone saying it would work under other circumstances.

          2. I agree that a lot of alternative medicine doesn’t work.  But I don’t think we should dismiss it out of hand.  If I’m not mistaken, the mechanisms behind most mainstream interventions are not well-understood either.  We don’t really understand how antidepressants work.  And many common surgical procedures are unproven. Our scientific medicine is not as scientific as we would like to think.

    2.  The difference here is that woo “seems” to work for people it doesn’t actually work for. It’s not just that its anecdotal, it’s that the evidence is against it in every conceivable way.

      For this, though, the evidence is there. That’s what they’re looking at, and trying to learn more from. It’s not captured by traditional medical studies, but traditional medical studies make a lot of assumptions and are dealing with something incredibly complex.

      Imagine, if you will, that instead of testing people with a given disease, when testing the cure, they tested EVERYONE. Suddenly, even a cure with a 100% success rate wouldn’t have a statistically meaningful benefit.

      I suspect that’s the case for situations like these – the studies will end up being seen as too broad, the illness actually a broad category that doesn’t match the underlying details we want to test, and the benefits are being lost in the noise – except for exceptional cases like this, outliers, where it’s too hard to ignore.

      They still aren’t hard enough evidence by themselves that the treatment works. I could have been due to something unrelated. But they are certainly an avenue worth pursuing, research-wise, because such outliers indicate we’ve failed to understand the underlying problem correctly when we ran our tests, and a more precise test might yield meaningful and beneficial results.

      Woo, however, well… there’s a reason no one is actually researching this stuff for woo, and why woo-pushers insist that it’s right for most people despite the evidence. Because these isn’t really anything like woo.

      1. I think you may have missed my point. I was trying to say that when Woo does work for someone – and I grant that many/most (but not all) people who claim benefit from woo, don’t in fact. But when it does work, perhaps that person is an Exceptional Responder. Like a unicorn. 

        Perhaps, like the article it (the woo of choice) only works on some minority of people; they go on to sing the praises of whatever woo it was and others try it. Sure, most of them get no actual benefit but enough get something that it sticks with us. 

        You are probably right though.

  3. Eventually, many new drugs are going to be “engineered” to work for one person based on his or her genome.  That could have a big impact on the pharmaceutical industry, changing their R&D processes and reducing the number of “blockbuster” drugs that bring in billions of dollars each year during their patent-protected lifetime.

  4. Here’s a similar story about an experimental antidepressant:
    Even after decades of research and technological advancement, pharmaceutical design is still basically a crapshoot.

  5. Allen Roses (Duke/GSK) has been talking about exactly this issue of refining pharmacogenomics for clinical trials for over 10 years, and he’s been writing very detailed proposals about how to do it.

    I’ve seen people present his ideas as if they’d thought of it themselves for about 5 years.  Seeing an article with 1 citation and a picture of some guy from McGill is not surprising. 

  6. When a drug works for only one person, my layman guess is that they should do some extensive mapping of that person’s DNA, particularly in cases of cancer at this point.

  7. Just one question: shouldn’t a new drug be tested against the best in
    the market for that illness (if any)? Testing against placebo is one of
    the tricks to pass more easily a new drug (it might be better than
    placebo, but not better than a product already in the market) 

  8. I see this as more a question of sampling error.  If a drug works only for one individual, I would suspect a placebo effect.  But it is quite possible that the individual may have the same symptoms as others in an experiment, but their symptoms may not be caused by the same problem.  If you were to sample more people, you might find others with the same response.  There is always the problem of how many people you can afford to test, and is it enough of the population to be economically feasible to release a new drug for such a small group of people.

  9. Dr.s are familiar with drugs that don’t work for individuals (or even differences between ethnic races).  

    A big question, related, is what about *sensitive people* exposed to things (radiation, low level chems) that most people tolerate well.  Are those sensitives accounted for in public-health/risk calculations? 

    (I don’t mean “wifi sensitives” :-) I mean folks with real genetic, enzymatic differences, typically hidden, that make them more succeptible to damages or react differently than your average.)

  10. The thing is that — while the exceptional responders presented here are statistically exceptional in that a drug that doesn’t work for most people does work for them — the fact that a given drug will not have entirely similar responses across a wide range of people is not news. It’s very well-documented and -known, even if the mechanisms of the difference are not well-understood for a given drug. For example, a narcotic painkiller like Talwin might make a large swath of people nauseous, some people not. What’s interesting about exceptional responders is the dramatic difference between the response of groups, and low proportion of the responsive group. It’s just a matter of degree.

    It’s something to consider because, in general, our healthcare systems are not at all set up to cope with personalized medicine. Minimizing liability is the driving factor behind almost every decision, which discourages doctors from personalizing care as much as possible, and encourages regulators (at least in the U.S.) to reject new treatments whenever adverse effects are found. On a grander scale, this sort of legal environment exists because human beings are incredibly bad with risk management. You’re a lot more likely to get a combustion-based carcinogen delivery system onto the market with only a warning “this cigarette may kill you” than you are to get a new medicinal product with only a 95% success rate onto the market with the warning “this defibrillator may not save your life.”

    Most trials also aim for a very shallow pass/fail for sort of testing. By the time it gets to that stage, it’s assumed you are no longer trying to characterize the problem, but rather know the group you’re trying to help, know the outcome you’re trying to achieve, and have a good understanding of the attendant side effects. The burden of identifying groups which may or may not benefit falls onto
    the researchers before the trial. If they can’t do that, the regulators
    certainly won’t make the attempt. The more factors that interacted in a complex dance to create the disorder, the more likely you are to have only a handful of people who are cured by a drug targeting one of those factors. It’s like trying to plug a leaky boat with only a single plug, regardless of how many holes are actually drilled into the hull.

Comments are closed.