(Quelle: http://www.psychologytoday.com/blog/side-effects/201207/drug-trials-and-data-based-medicine-interview-david-healy)
Dr. David Healy is an internationally renowned psychiatrist, psychopharmacologist, scientist, and author. A professor of Psychiatry in Wales and former Secretary of the British Association for Psychopharmacology, he is the author of more than 150 peer-reviewed articles and 20 books, including The Antidepressant Era and The Creation of Psychopharmacology, from Harvard University Press; Let Them Eat Prozac from New York University Press; Mania: A Short History of Bipolar Disorder from the Johns Hopkins University Press; and, most recently, Pharmageddon, from the University of California Press. He was responsible for submitting the key document that led to New York State's successful fraud action against GlaxoSmithKline, a key plank in the Department of Justice's recent case against the drug-maker.
David, thanks for answering a few questions about your latest work. With a team of other medical specialists you recently launched a new website, RxISK, that provides a wealth of user-friendly drug-related information for doctors and patients. What distinguishes RxISK from other sites listing medical information and what were some of your aims in launching it?
Chris, we’re trying to provide much-better descriptions of drug-related side effects, including by getting patients and doctors to work as teams and by using a series of causality algorithms that help establish when there’s a link between a treatment and a problem.
Other sites providing medical information tend to offer one of two things: either they summarize clinical trials, most of which are ghostwritten and where the complete data are unavailable, or they list adverse event data from agencies like the FDA, which is of poor quality and increasingly regarded as anecdotal. Regulators and academics now have a track-record of acknowledging significant hazards on drugs 10-20 years after patients and others first draw attention to them. The reason that happens is because agencies like the FDA have degraded adverse event reporting.
I expect we’ll get reports from people like you and your readers outlining a new problem on some new drug that clears when the drug is stopped and maybe reappears if it’s restarted. Companies and academics will then scream blue murder—this is just anecdotal.
My response will be to ask whether that report is more or less likely to be correct than clinical trial data run by companies whose data are hidden, and where the patients sometimes don’t even exist. Even where the data prove beyond doubt that the drug causes the problem, academics unfortunately will still line up to deny that the drug could cause it.
You’ve written on your blog that “evidence-based medicine and RCTs [random controlled trials] are supposed to help us control the pharmaceutical industry.” Yet “RCTs are simply not the answer to determining cause and effect,” you go on to say, because they’re “quite likely to hide rather than reveal a problem like antidepressant induced suicidality.” As one of the first researchers to draw attention to the now well-publicized suicide-inducing side effects of many antidepressants, you’re clearly in a position to answer: how in fact do RCTs hide such information?
There are a few ways that RCTs can hide effects. First, the process doesn’t encourage anyone to look closely at particular things that happen on a drug—the focus is instead on the group and on average effects. That’s true of all trials. In company trials there are more specific problems like miscoding, where suicidality becomes “nausea” or “emotional lability” or even “treatment non-responsiveness.” There is also the problem of mislocation—patients on placebo end up being given problems they never had—and of nonexistent patients, who don’t of course have adverse events.
Beyond that, there are more sophisticated tricks that companies can and do play—such as claiming that increased rates of a problem on a drug are not really evidence of an increase in rates if the data are not statistically significant. In this way, companies have hidden many more heart attacks on Vioxx and Avandia or suicidal acts on SSRIs than have been hidden by miscoding or mislocation.
Isn’t what you’re describing tantamount to fraud? I’m all in favor of clinical trials—if done right, wouldn’t they give us the correct answer?
Actually no, when it comes to adverse events, trials almost never get the right answer.
Let’s assume in a trial that we have 3,000 depressed patients on Paxil who had 10 suicidal acts and 1,750 on placebo who had 0 suicidal acts. Paxil clearly causes suicidal acts here. Now let’s take 200 depressive personality disorder patients on Paxil who have 30 suicidal acts and 200 depressive personality disorder patients on placebo who have 25 suicidal acts—again, that’s an increased rate of suicidal acts on Paxil. But add these two increases together and you end up with a reduced rate of suicidal acts on the SSRI compared to placebo—40 suicidal acts in 3,200 patients is less than 25 in 1,950.
Hey presto—problem gone. Exactly the same thing can happen in every clinical trial where we don’t fully understand the condition we’re treating—which is, frankly, most conditions from back pain to diabetes to psychosis. We mix patients who superficially appear the same but who in fact have different conditions.
That is just one trick that no-one ever mentions—I’ve laid out several more on davidhealy.org.
Is there any way to overcome such tricks and masking problems?
Yes, actually, there is. One way is to do trials in healthy volunteers—these are the true drug trials. Companies do these but rarely publish them. There’s no register of these trials and no data are made available, though there’s no issue of clinical confidentiality involved. Given that these trials tell us so much—10 years before Zoloft came on the market, for instance, they indicated that the drug made healthy volunteers suicidal—it’s a huge scandal that these data in particular are buried.
Speaking out against trials won’t of course make you popular.
It certainly doesn’t. But the most painful thing of all is it puts me at odds with almost everyone who should be a natural ally: those who are committed to Evidence-Based Medicine, some of whom agree with what is being said, or who even claim that I am saying nothing new, but who really don’t want to see RCTs questioned in public and can get unpleasantly angry when they are.
Is there a chance that RxISK will represent data from other countries, to establish a truly global perspective on adverse effects from medication?
Absolutely—RxISK will have data from every country under the sun and will also breakdown the data by locality so, for instance, people in Chicago will be able to see what side effects are being reported on which drug in their area. This may well be of interest to journalists who want local stories—if in this case a large urban area can be called local.
You’ll also be able to follow your side effect across time—how common it becomes, where it’s being reported most, who gets it—men, women, young, old, etc. Given the input of hundreds of thousands of people following these things, I expect the visualization involved will help researchers come up with good ideas about what might actually be going on.
RxISK dovetails with—indeed, is clearly a practical extension of—the argument of your latest book Pharmageddon, that medicine has become increasingly “pharmaceuticalized” since the 1950s—oriented to fierce marketing campaigns that cherry-pick data, overhyph the overall benefits of drugs, and mask their very real hazards. This is obviously a timely argument in the States right now, with the Supreme Court’s decision on the Affordable Care Act and the debate about how to trim costs without affecting care. What (beyond your new site) are some of your recommendations for reforming healthcare and improving drug safety?
Well, the website is a bottom-up approach, a wisdom-of-crowds or bidet approach.
There are also top-down or shower approaches that could help. Our current problems stem tragically, in the proper sense of that word, from a system we put in place 50 years ago, following the thalidomide disaster, in an attempt to prevent such a problem happening again.
There are three components to the system—the patent status of drugs; the prescription-only status of drugs; and the issue of demonstrating efficacy through controlled trials. All need review to see whether some tweak to the system might produce better results than we are getting now.
Despite recent congressional attempts to create greater transparency over pharmaceutical decisions, such as Sen. Chuck Grassley’s Sunshine Act, you clearly are skeptical in the book that they’ve had much good effect as reforms. Part of the problem is clearly the FDA’s reliance on RCTs and a presumption that, as you put it, “takes for granted that data don’t lie.” What’s wrong, in your opinion, with the way the FDA currently interprets data?
Well, let’s run a thought-experiment and bring alcohol or nicotine on the market as antidepressants.
To do this, we don’t have to show lives saved or people returning to work—we only have to show a change in score on rating scales that may be sensitive to the anxiolytic or sedative effects of alcohol.
Next, we need do only a few studies in which alcohol beats placebo on our rating scale. If in most of our studies it doesn’t beat placebo, then these are discounted and FDA is happy for us to conceal that.
In our trials, placebo might account for 80-90% of the effect of alcohol but the FDA is fine with us leaving the public with the impression that 100% of the apparent benefits of alcohol for depression stem from the alcohol, with no contribution from placebo.
Better again, we can outsource our studies. Let’s say in our key study that alcohol proves no better than placebo in 30 U.S. centers, but is dramatically better than placebo in 2 Mexican centers, so, when added to the mix, alcohol marginally beats placebo. The FDA let’s us do this and the published article will make no mention that alcohol only works in Mexico.
What about side effects—could the FDA do more to improve drug safety?
The FDA couldn’t do worse. Our alcohol studies only have to last six-to-eight weeks and, as most of us know, few of the problems that might be expected from alcohol emerge in a six-to-eight week period.
If there’s any hint of liver problems in our trial, the FDA and academia are likely to attribute that to the depression for which the person is being treated. Even though the entire medical literature up till then might not have a scrap of evidence that depression causes liver dysfunction, within weeks companies have the ability to get a significant proportion of the medical profession to agree that it’s well-known that depression causes liver dysfunction.
Something else that’s extraordinary from a safety point of view is this: Several different companies can file for patents on whiskey, gin, brandy, wine, or port, or even to distinguish Irish whiskey from Scottish scotch. Their combined marketing can encourage doctors to put patients on combinations of whiskey, gin, brandy, and port and to keep their patients on these combinations for extended or indefinite periods of time.
If you or I had the power that Pharma has, we’d be able to get independent guidelines to endorse alcohol for depression, making it almost mandatory for doctors to use it.
Where are the AMA and APA in all of this?
This is where things get weird. The major difference between alcohol and Lexapro or Abilify lies in a curious inversion of the stranger-neighbor phenomenon. Rather stupidly, we are wary of strangers but comfortable with neighbors, even though we’re most likely to be harmed by neighbors or relatives.
Now, alcohol should be the familiar neighbor and SSRIs the dangerous stranger. But in fact we treat alcohol as a dangerous stranger, ripping a glass of wine out of the hands of a pregnant woman, while we regard SSRIs as something that can only do good even though these drugs are prescription-only precisely because we have every reason to think they will be riskier than alcohol—which we’re still basically happy to let people manage for themselves.
Doctors, you see, provide a risk-laundering service to companies. In fact, making drugs available through doctors is a way to hide significant hazards such as liver failure or lung cancer, on average for 10-to-15 years from the time people begin to report them first, and claim that their liver failure or lung cancer stems from the treatment.
Indeed, even after FDA puts a black box warning on alcohol or nicotine, most doctors will still deny that this risk happens.
In Pharmageddon, you describe compellingly the dilemma that general practitioners and psychiatrists face today, given the many targets they’re given and the guidelines they’re told to follow relative to their immediate issue of responding quickly, effectively, and safely to patient needs. I don’t know how closely you’re following the UK and U.S. debates about DSM-5 and ICD-11, editions that obviously will be central to determining future treatment patterns and goals, but how in your opinion can caregivers work around that dilemma, even diminish it in their work?
I think DSM-5 and ICD-11 are not at the heart of the problem. One reason for thinking this is that the key problems apply to all of medicine rather than to just mental health. DSM-5 is an example of a measurement technology, like DXA scans or peak flow meters, that generates problems for doctors to which a drug becomes an answer.
The deeper problem, as I mentioned above, is the combination of product patents, prescription-only status, and the use of clinical trials as a means of determining efficacy—in particular, when the data from those trials are not made available. This creates a perfect product for companies, with a perfect consumer (doctors) and the perfect raw material (trials), which industry can manipulate to mean whatever they want them to mean. It all adds up to the perfect market, or the perfect perversion of a market, depending on your point of view.
Dr. David Healy is an internationally renowned psychiatrist, psychopharmacologist, scientist, and author. A professor of Psychiatry in Wales and former Secretary of the British Association for Psychopharmacology, he is the author of more than 150 peer-reviewed articles and 20 books, including The Antidepressant Era and The Creation of Psychopharmacology, from Harvard University Press; Let Them Eat Prozac from New York University Press; Mania: A Short History of Bipolar Disorder from the Johns Hopkins University Press; and, most recently, Pharmageddon, from the University of California Press. He was responsible for submitting the key document that led to New York State's successful fraud action against GlaxoSmithKline, a key plank in the Department of Justice's recent case against the drug-maker.
David, thanks for answering a few questions about your latest work. With a team of other medical specialists you recently launched a new website, RxISK, that provides a wealth of user-friendly drug-related information for doctors and patients. What distinguishes RxISK from other sites listing medical information and what were some of your aims in launching it?
Chris, we’re trying to provide much-better descriptions of drug-related side effects, including by getting patients and doctors to work as teams and by using a series of causality algorithms that help establish when there’s a link between a treatment and a problem.
Other sites providing medical information tend to offer one of two things: either they summarize clinical trials, most of which are ghostwritten and where the complete data are unavailable, or they list adverse event data from agencies like the FDA, which is of poor quality and increasingly regarded as anecdotal. Regulators and academics now have a track-record of acknowledging significant hazards on drugs 10-20 years after patients and others first draw attention to them. The reason that happens is because agencies like the FDA have degraded adverse event reporting.
I expect we’ll get reports from people like you and your readers outlining a new problem on some new drug that clears when the drug is stopped and maybe reappears if it’s restarted. Companies and academics will then scream blue murder—this is just anecdotal.
My response will be to ask whether that report is more or less likely to be correct than clinical trial data run by companies whose data are hidden, and where the patients sometimes don’t even exist. Even where the data prove beyond doubt that the drug causes the problem, academics unfortunately will still line up to deny that the drug could cause it.
You’ve written on your blog that “evidence-based medicine and RCTs [random controlled trials] are supposed to help us control the pharmaceutical industry.” Yet “RCTs are simply not the answer to determining cause and effect,” you go on to say, because they’re “quite likely to hide rather than reveal a problem like antidepressant induced suicidality.” As one of the first researchers to draw attention to the now well-publicized suicide-inducing side effects of many antidepressants, you’re clearly in a position to answer: how in fact do RCTs hide such information?
There are a few ways that RCTs can hide effects. First, the process doesn’t encourage anyone to look closely at particular things that happen on a drug—the focus is instead on the group and on average effects. That’s true of all trials. In company trials there are more specific problems like miscoding, where suicidality becomes “nausea” or “emotional lability” or even “treatment non-responsiveness.” There is also the problem of mislocation—patients on placebo end up being given problems they never had—and of nonexistent patients, who don’t of course have adverse events.
Beyond that, there are more sophisticated tricks that companies can and do play—such as claiming that increased rates of a problem on a drug are not really evidence of an increase in rates if the data are not statistically significant. In this way, companies have hidden many more heart attacks on Vioxx and Avandia or suicidal acts on SSRIs than have been hidden by miscoding or mislocation.
Isn’t what you’re describing tantamount to fraud? I’m all in favor of clinical trials—if done right, wouldn’t they give us the correct answer?
Actually no, when it comes to adverse events, trials almost never get the right answer.
Let’s assume in a trial that we have 3,000 depressed patients on Paxil who had 10 suicidal acts and 1,750 on placebo who had 0 suicidal acts. Paxil clearly causes suicidal acts here. Now let’s take 200 depressive personality disorder patients on Paxil who have 30 suicidal acts and 200 depressive personality disorder patients on placebo who have 25 suicidal acts—again, that’s an increased rate of suicidal acts on Paxil. But add these two increases together and you end up with a reduced rate of suicidal acts on the SSRI compared to placebo—40 suicidal acts in 3,200 patients is less than 25 in 1,950.
Hey presto—problem gone. Exactly the same thing can happen in every clinical trial where we don’t fully understand the condition we’re treating—which is, frankly, most conditions from back pain to diabetes to psychosis. We mix patients who superficially appear the same but who in fact have different conditions.
That is just one trick that no-one ever mentions—I’ve laid out several more on davidhealy.org.
Is there any way to overcome such tricks and masking problems?
Yes, actually, there is. One way is to do trials in healthy volunteers—these are the true drug trials. Companies do these but rarely publish them. There’s no register of these trials and no data are made available, though there’s no issue of clinical confidentiality involved. Given that these trials tell us so much—10 years before Zoloft came on the market, for instance, they indicated that the drug made healthy volunteers suicidal—it’s a huge scandal that these data in particular are buried.
Speaking out against trials won’t of course make you popular.
It certainly doesn’t. But the most painful thing of all is it puts me at odds with almost everyone who should be a natural ally: those who are committed to Evidence-Based Medicine, some of whom agree with what is being said, or who even claim that I am saying nothing new, but who really don’t want to see RCTs questioned in public and can get unpleasantly angry when they are.
Is there a chance that RxISK will represent data from other countries, to establish a truly global perspective on adverse effects from medication?
Absolutely—RxISK will have data from every country under the sun and will also breakdown the data by locality so, for instance, people in Chicago will be able to see what side effects are being reported on which drug in their area. This may well be of interest to journalists who want local stories—if in this case a large urban area can be called local.
You’ll also be able to follow your side effect across time—how common it becomes, where it’s being reported most, who gets it—men, women, young, old, etc. Given the input of hundreds of thousands of people following these things, I expect the visualization involved will help researchers come up with good ideas about what might actually be going on.
RxISK dovetails with—indeed, is clearly a practical extension of—the argument of your latest book Pharmageddon, that medicine has become increasingly “pharmaceuticalized” since the 1950s—oriented to fierce marketing campaigns that cherry-pick data, overhyph the overall benefits of drugs, and mask their very real hazards. This is obviously a timely argument in the States right now, with the Supreme Court’s decision on the Affordable Care Act and the debate about how to trim costs without affecting care. What (beyond your new site) are some of your recommendations for reforming healthcare and improving drug safety?
Well, the website is a bottom-up approach, a wisdom-of-crowds or bidet approach.
There are also top-down or shower approaches that could help. Our current problems stem tragically, in the proper sense of that word, from a system we put in place 50 years ago, following the thalidomide disaster, in an attempt to prevent such a problem happening again.
There are three components to the system—the patent status of drugs; the prescription-only status of drugs; and the issue of demonstrating efficacy through controlled trials. All need review to see whether some tweak to the system might produce better results than we are getting now.
Despite recent congressional attempts to create greater transparency over pharmaceutical decisions, such as Sen. Chuck Grassley’s Sunshine Act, you clearly are skeptical in the book that they’ve had much good effect as reforms. Part of the problem is clearly the FDA’s reliance on RCTs and a presumption that, as you put it, “takes for granted that data don’t lie.” What’s wrong, in your opinion, with the way the FDA currently interprets data?
Well, let’s run a thought-experiment and bring alcohol or nicotine on the market as antidepressants.
To do this, we don’t have to show lives saved or people returning to work—we only have to show a change in score on rating scales that may be sensitive to the anxiolytic or sedative effects of alcohol.
Next, we need do only a few studies in which alcohol beats placebo on our rating scale. If in most of our studies it doesn’t beat placebo, then these are discounted and FDA is happy for us to conceal that.
In our trials, placebo might account for 80-90% of the effect of alcohol but the FDA is fine with us leaving the public with the impression that 100% of the apparent benefits of alcohol for depression stem from the alcohol, with no contribution from placebo.
Better again, we can outsource our studies. Let’s say in our key study that alcohol proves no better than placebo in 30 U.S. centers, but is dramatically better than placebo in 2 Mexican centers, so, when added to the mix, alcohol marginally beats placebo. The FDA let’s us do this and the published article will make no mention that alcohol only works in Mexico.
What about side effects—could the FDA do more to improve drug safety?
The FDA couldn’t do worse. Our alcohol studies only have to last six-to-eight weeks and, as most of us know, few of the problems that might be expected from alcohol emerge in a six-to-eight week period.
If there’s any hint of liver problems in our trial, the FDA and academia are likely to attribute that to the depression for which the person is being treated. Even though the entire medical literature up till then might not have a scrap of evidence that depression causes liver dysfunction, within weeks companies have the ability to get a significant proportion of the medical profession to agree that it’s well-known that depression causes liver dysfunction.
Something else that’s extraordinary from a safety point of view is this: Several different companies can file for patents on whiskey, gin, brandy, wine, or port, or even to distinguish Irish whiskey from Scottish scotch. Their combined marketing can encourage doctors to put patients on combinations of whiskey, gin, brandy, and port and to keep their patients on these combinations for extended or indefinite periods of time.
If you or I had the power that Pharma has, we’d be able to get independent guidelines to endorse alcohol for depression, making it almost mandatory for doctors to use it.
Where are the AMA and APA in all of this?
This is where things get weird. The major difference between alcohol and Lexapro or Abilify lies in a curious inversion of the stranger-neighbor phenomenon. Rather stupidly, we are wary of strangers but comfortable with neighbors, even though we’re most likely to be harmed by neighbors or relatives.
Now, alcohol should be the familiar neighbor and SSRIs the dangerous stranger. But in fact we treat alcohol as a dangerous stranger, ripping a glass of wine out of the hands of a pregnant woman, while we regard SSRIs as something that can only do good even though these drugs are prescription-only precisely because we have every reason to think they will be riskier than alcohol—which we’re still basically happy to let people manage for themselves.
Doctors, you see, provide a risk-laundering service to companies. In fact, making drugs available through doctors is a way to hide significant hazards such as liver failure or lung cancer, on average for 10-to-15 years from the time people begin to report them first, and claim that their liver failure or lung cancer stems from the treatment.
Indeed, even after FDA puts a black box warning on alcohol or nicotine, most doctors will still deny that this risk happens.
In Pharmageddon, you describe compellingly the dilemma that general practitioners and psychiatrists face today, given the many targets they’re given and the guidelines they’re told to follow relative to their immediate issue of responding quickly, effectively, and safely to patient needs. I don’t know how closely you’re following the UK and U.S. debates about DSM-5 and ICD-11, editions that obviously will be central to determining future treatment patterns and goals, but how in your opinion can caregivers work around that dilemma, even diminish it in their work?
I think DSM-5 and ICD-11 are not at the heart of the problem. One reason for thinking this is that the key problems apply to all of medicine rather than to just mental health. DSM-5 is an example of a measurement technology, like DXA scans or peak flow meters, that generates problems for doctors to which a drug becomes an answer.
The deeper problem, as I mentioned above, is the combination of product patents, prescription-only status, and the use of clinical trials as a means of determining efficacy—in particular, when the data from those trials are not made available. This creates a perfect product for companies, with a perfect consumer (doctors) and the perfect raw material (trials), which industry can manipulate to mean whatever they want them to mean. It all adds up to the perfect market, or the perfect perversion of a market, depending on your point of view.
