(Editorial Note: This is the fifth in the Lasagna series of posts that began with Not So Bad Pharma, April Fool, Tragedy of Lou Lasagna, Empire of Humbug: Bad Pharma and will continue through to Brand Fascism and Witty A: Report to the President.)
In 1954, soon after his article with Beecher put the placebo on the map, Lou Lasagna was recruited from Harvard to Hopkins. Beecher pleaded with him to stay in Boston;
“Now that I am faced with the unhappy fact that you are gone, I find it more depressing than I can say… If your future career is anything like your performance here it is bound to be spectacular.”
Lasagna ended his days back in Boston, leading the Center for the Study of Drug Development, whose brief was to foster the development of specific treatments for specific conditions. In 1980, he was made president of the American College of Neuropsychopharmacology. The placebo had vanished from his working life.
Michael Shepherd had at one point been expected to take over the lead of Britain’s Maudsley Hospital when Aubrey Lewis retired but his style was too much at odds with the developing medico-pharmaceutical complex for this to be possible. He had been Vice-President of the World Psychopharmacology College (CINP) but by the mid-1960s he was persona non-grata within psychopharmacology.
He turned to primary care research and accidentally created a market for the SSRIs in the process. He grew increasingly opposed to the notion of specificity in medicine, and turned to exploring the placebo.
The two men met for the last time in 1992 at the 50th American Psychosomatic Society meeting in New York. Lasagna was chairing, Shepherd speaking. His topic – The Placebo (see Empire 1). [Shepherd is on the left]
When the lecture finished, Lasagna said: “this paper is now open for questions”. Nothing happened. Half a minute elapsed, a minute elapsed and nobody said anything at all. In Holland this would be normal – in America it was plain bizarre. Lasagna then said that in the circumstances they would have to move on to the next paper but he couldn’t refrain from commenting on this unnatural silence.
“There are 3 possible explanations” he said. “First you were all asleep and therefore you heard nothing. Secondly it was so bad that since this speaker has come 3,000 miles you didn’t want to embarrass him. Third it is genuinely so original and new that you don’t quite know what to make of it. I’ll leave you to decide which it was”.
What had Shepherd said?
Mirage in 1962
The key point is that little was known about RCTs or effectiveness in 1962. RCTs appeared in a late night political haggling session as a simple solution to an extraordinarily complex problem. But to paraphrase HL Mencken, like Lasagna also based in Baltimore, every complex problem has many simple solutions – all of them wrong.
Coming as he did from a position of running studies on internal states such as pain or on sleep induction, the inclusion of controlled trials late night in ’62 may not have hit Lasagna as problematic – even though in his hands imipramine, a far more potent antidepressant than any SSRIs, could not be shown to work in a controlled trial.
Analgesic and hypnotic effects were determined by means of rating scales. These effects were demonstrations of efficacy rather effectiveness against a pathological state. A demonstration of analgesic efficacy is consistent with FDA regulations that a drug be demonstrated to have an effect on a structure or function of the human body and analgesics are licensed on this basis to this day, as are hypnotics, oral contraceptives and a range of other treatments.
It would have been entirely possible to bring antipsychotics, benzodiazepines and SSRIs on the market by demonstrating their tranquilizing or anxiolytic properties. This focus would have put a premium on later teasing out what is distinct about the anxiolysis in each case – something that clinicians today quite astonishingly still cannot do although the experience you or I have of these drugs is entirely distinctive – we can easily tell them apart.
In contrast the SSRIs have performed so miserably in effectiveness trials that there is intense debate about whether they work. Half of the trials undertaken show a rating scale only differentiation from placebo. The other half shows nothing.
If in contrast, SSRIs had been licensed on the basis of an anxiolytic effect, the question of how and when to use these drugs as a therapeutic principle for depressive or anxious states would have been a separate matter that could have been tackled by the medical profession after these drugs came on the market.
Instead the effectiveness criterion has made medical conditions into a target for treatment and in so doing it set these conditions up for capture by pharmaceutical companies and their marketing departments. The effectiveness criterion created the inevitability of what is now called disease mongering.
To have to fish for possible effectiveness in some heterogeneous clinical condition is not what controlled trials were designed for. Controlled trials discover nothing. Even worse, the effectiveness claims for many drugs rest on minor surrogate marker or rating scale changes in trials that are powered to ensure that clinically insignificant changes achieve statistical significance. It would be entirely possible to prove Snake Oil was effective on this basis.
Meanwhile in the real world
Meanwhile, in the real world, there are in fact more dead bodies in the active treatment arms of most blockbuster studies than in the placebo arms. The claim that these surrogate or rating scale changes demonstrate effectiveness therefore as a matter of logic does not stack up.
Aside from dead bodies, we know that many of those who “benefit” will have enduring problems from physical dependence and other legacy effects from treatment. This is not effectiveness as the 1962 regulations envisaged. We have ended up instead in a world in which hints of an effect are used to gain market entry for drugs, and when we use these drugs mindlessly on the assumption that that they are effective, then if the RCT evidence is to be believed we produce greater disability in the long run than if we never had them. The effects may turn out to be worse than if we had encouraged chronic alcohol intake.
The only thing that stands in the way of this outcome at the moment are the dwindling number of doctors who pay no heed to guidelines and try to be discriminating in their prescribing, and the patients who simply don’t take the treatments when their experience tells them this is not for them.
So far the argument is consistent with a view that controlled trials are gold standard processors into which we have fed garbage and have got garbage back – And that many people have a great capacity to greet garbage as gold.
In fact, we are cursed like Midas and everything to do with this gold standard, rating scales and surrogate markers, has produced an informational reductionism and rigid standardisation that is destroying medical care.
Little noticed is the fact that the RCT processor malfunctions badly in a number of ways.
Whenever a drug and an illness produce superficially similar effects from pain, to nausea, to suicide, it is possible to hide drug-induced problems. Antidepressant trials can be “gamed” so that drugs known to cause suicide will show up as preventing suicide rather than causing it. Perverse outcomes like this are built into RCTs (See The Best Bias that Money can buy, Heads we win, tails you lose).
In all these cases, the Christmas tree lightbulb test is better than clinical trials ever could be. When the lightbulbs don’t work after they have been laid up for a year, you unscrew each in turn, until unscrewing one leads the lights to come on, screwing it back in causes them to go off again and removing it turns them on. In this case, we have as definitive a result as Martin Teicher had with Prozac and suicide in 1990. Lilly claimed their clinical trials showed no problem. Teicher’s observations can now clearly be seen to be much more valid than Lilly’s.
When we get clearcut results from celebrated trials such as the Women’s Health Initiative study of HRT, it’s often because these are conducted in essentially healthy volunteers so that excess rates of breast cancer stand out – this was a drug trial rather than a condition trial.
Confused outcomes also happen when the clinical condition being investigated is heterogeneous as for instance in back pain, depression or asthmatic states or even Parkinsonian syndromes. In the case of SSRIs and suicide, GSK were able to use the different background rates of suicidal acts in borderline personality disorder patients compared to major depressive disorder patients to hide the suicide risk on paroxetine. The same can be done on purpose or by accident for many if not most medical conditions.
Finally in the case of the antidepressants, although in Lou Lasagna’s hands imipramine didn’t beat placebo, it is relatively straightforward to design trials to get imipramine to beat SSRIs – all that is needed is to pick a group of more severely depressed patients. But in this case, what would FDA license imipramine for? For severe depression only or for all depressions and what consequence if any would there be for the licensing of SSRIs?
Panalba & Combination Treatments
It was the issue of relative efficacy that blew Lasagna’s career off track in the fiercely contentious Panalba case in 1969.
Panalba, as mentioned (see Tragedy) was a combination antibiotic. FDA wanted to pull it off the market – on the basis of efficacy not safety. They claimed that the test of effectiveness for a combination antibiotic was a demonstration of effectiveness over and above the effectiveness of each of the individual components.
If we compare the regulation of drugs with the regulation of butter, FDA never takes a position on the comparative merits of butters. In similar fashion, FDA does not now allow, and likely never will allow, comparative effectiveness claims to stand within a therapeutic domain.
But in the case of combination antibiotics, FDA chose to remove all such treatments. Lasagna publicly contested the issue, not as a believer in Panalba, but on the point that the basis for the decision appeared arbitrary. He argued it was bad medicine to use Panalba. But was it up to Medicine or up to the FDA to sort this out? And, although philosophically against Panalba, he wasn’t prepared to completely deny some place for clinical wisdom in its use.
Panalba was an obvious combination drug. But many drugs embody several therapeutic principles – they are cocktail drugs rather than combination drugs. Imipramine contains an alerting norepinephrine principle, an anxiolytic serotonergic principle, a euphoriant anticholinergic principle and a sedating anti-histaminic principle. In just the same way some antibiotics embody more than one antibiotic principle. Indeed Prozac has antibiotic effects. When it comes to the licensing of cocktail compounds like this, FDA does not and cannot require a demonstration of greater effectiveness.
In fact a good case can be made that it is imipramine’s cocktail nature that makes it difficult to run an RCT on it. It has an anxiolytic effect, but not so marked as Prozac, an alerting effect but not so marked as a stimulant, and a sedative effect but not so marked as an hypnotic. This cannot be brought out in milder depressions but imipramine is clearly better than Prozac in severe depressions and in clinical practice a discriminating clinician can deploy its many facets to good advantage, regardless of what a purist might prefer. Good practice with this drug and many other drugs will never be RCT based.
If so for imipramine, to return to Lasagna’s point, what about Panalba? Imipramine is just one drug – but FDA have since licensed many combination treatments from Fen-Phen to Symbyax (Prozac and Zyprexa) none of which met 1969 standards for efficacy.
Compromising Safety
When it comes to safety the problems get more acute. The effectiveness requirement is enshrined in regulation for the contribution it makes to safety. If all drugs are poisons then there is a better risk-benefit trade off if we have some confidence there is a potential benefit to trade against. The wording of the regulations continues to put safety first – safety and effectiveness.
But in using RCTs to demonstrate effectiveness Congress compromised safety. Here’s how. RCTs necessarily have a primary endpoint for purposes of calculating statistical significance. The assessment instruments and trial design hinge on getting the best possible data on the primary endpoint – which is always the demonstration of effectiveness. This focus on one endpoint means that other outcomes (safety issues) are side-lined.
As a result, if in the course of a study a safety problem happens at a greater rate on drug than on placebo, it will be poorly characterized, will be hidden under multiple codes and will almost always be dismissed as not statistically significant – even by FDA. Under oath, Andrew Witty will swear we had no evidence that our drug caused this problem, and he may not be lying in the conventional sense of that word.
The 1938 act FDA clearly took effectiveness into account. It used to be impossible to discuss safety without efficacy. For example, the early sulfonilamide drugs were dangerous but in view of their effectiveness in saving lives, there was little question but that they should be allowed on the market. A hypnotic shouldn’t cause birth defects or peripheral neuralgia and under the 1938 Act, thalidomide wasn’t licensed.
But under the 1962 Act, the claim that effectiveness has been demonstrated now means that some of the most problematic drugs in medicine, such as dopamine agonists like ropinirole that cause gambling, sex addiction and personality change, can be brought on the market for conditions like restless legs syndrome. Common sense doesn’t seem to apply any more.
As a means to promote safety, the effectiveness criterion fails. This might be acceptable if the criterion ensured we got effective treatments but comes close to lunacy if the drugs are not in fact effective, and is even more lunatic if doctors can be brainwashed into thinking they are effective.
Is the Effectiveness Criterion Unconstitutional?
In July 2012, GlaxoSmithKline paid a record $3 billion fine for off-label promotion of several of their drugs. A few months later and they might not have had to pay anything.
In December 2012, in the Caronia case, a Second Circuit court ruled that stopping a company representative from talking about off-label uses of the company’s drug was a breach of their First Amendment rights to free speech. There was consternation among supporters of regulation.
The case involved Alfred Caronia, then a sales representative for Orphan Medical, which sold Xyrem, a drug approved for narcolepsy. He had been caught on tape discussing the use of the drug as a treatment for insomnia, fibromyalgia and other conditions with a doctor who was a government informant. He was convicted in 2008 but appealed on the basis of the First Amendment.
Since 1962, companies have for the most part supposedly demonstrated drug A is effective for condition X in order to get a license. They are then restricted to promoting the use of A for X, while doctors can use A for anything. It is more rational for instance to use SSRIs for premature ejaculation than for depression, as SSRIs are extraordinarily efficacious for this. But if a company representative suggested such a use when his company had not applied for a license to market the drug for this purpose, the company could be penalized.
The ruling in Caronia stated:
“The government clearly prosecuted Caronia for his words — for his speech,” but “the government cannot prosecute pharmaceutical manufacturers and their representatives under the FDCA for speech promoting the lawful, off-label use of an FDA-approved drug.”
Judge Debra Ann Livingston disagreed, arguing that
“[this] calls into question the very foundations of our century-old system of drug regulation.”
In a statement to send a shiver up the spine of most “liberals”, the Pharmaceutical Manufacturers Association said:
“PhRMA believes that truthful and nonmisleading communication between biopharmaceutical companies and health care professionals is good for patients, because it facilitates the exchange of up-to-date and scientifically accurate information about new treatments.”
Whatever about politics and preferences, it’s intellectually difficult not to side with the judges and PhRMA here. As things stand, anyone on the planet can discuss off-label uses of most drugs except the company that has made them. Worse again, when it comes to other drugs it makes no sense to say a company could only talk about the use of an opiate analgesic for gall-bladder but not for lung cancer surgery. This is an artificial debate created by the effectiveness criterion.
Just as with drugs, Olive Oil has to meet a regulatory criterion to be let on the market. The criterion doesn’t specify use of Olive Oil as a salad dressing as opposed to its use as a cooking oil. If the criteria did specify this, once on the market, if it was discovered that Olive Oil was wonderful for massage purposes, it would clearly be an infringement of the right to free speech to stop a company representative talking about these benefits, especially of Extra Virgin Olive Oil.
The Pied Piper
So what had Michael Shepherd said in his lecture on The Placebo that reduced an American audience to silence?
At dinner after the symposium in New York, Lasagna told Shepherd the placebo had been completely eclipsed within American medicine. No-one had any idea what it meant – and no-one was likely to grasp the issues he had raised.
What Shepherd told his audience was that – bizarrely – the awareness most doctors have of placebos stems from controlled trials. But the RCT placebo has almost nothing to do with the Powerful Placebo of the 1950s. It’s a stochastic effect. It’s meaningless. In any particular trial no-one knows or can know whether it refers more to the biases of doctors or of patients or to the natural variability in the condition.
When it comes to treating an individual patient or evaluating what is likely to be going on, the fact that there is a placebo response in population studies of medical conditions tells us next nothing.
The placebo response in controlled studies looking for an effect on a structure or function of the body would be quite a different matter and is something that could be usefully explored for learning or other effects that might in their own right form a basis for other or additional therapies. Just as placebos in testing for analgesia can show profound conditioning and other effects, so too they could in testing for anxiolysis for SSRIs. In depression studies the role of the drug and placebo tell us nothing other than a regulator can possibly approve this drug. It’s almost anti-evidence based medicine.
Lasagna and Shepherd were at one on this. As Lasagna put it:
“Evidence Based Medicine has become synonymous with randomized placebo-controlled clinical trials even though such trials invariably fail to tell the physician what he or she wants to know which is which drug is best for Mr Jones or Ms Smith – not what happens to a non-existent average person”.
He went further, noting that all drug discovery in psychopharmacology had been made by men like Fritz Freyhan who had been so skeptical of clinical trials in 1956 (Empire of Humbug 1), rather than through trials:
“The days when a drug company would go to skilled and sophisticated psychiatrists and give them a supply of a new drug and ask them to try it on some different patients seem gone forever. Is this a cause for celebration or depression?”
When asked in 1996 about the developing “craze” for Evidence Based Medicine, Shepherd responded that if we went down this route:
“We would also, of course, never make any advances because all advances depend on guesses which we call hypotheses – most of which are wrong but some of which are right. In terms of the logic of the scientific process the evidence comes after the hypothesis. You must begin with a hypothesis and that is a guess.
“I knew Cochrane but [Evidence Based Medicine] is pushing the thing to, I think, an absurd extreme. Of course, it’s true that the whole discipline is cluttered up with procedures for which the evidence is meagre at best and there is a case for trying to make quite certain that that is minimised. If that’s what meant by it – certainly – but it eventually stiffles everything else”.
In fact where the placebo had once seemed a form of hypnosis, the focus on effectiveness and specificity has turned RCTs into a form of hypnosis. Doctors are now in thrall to the spell of the hypnotist from Oz, who’s RCTs can even make Snake Oil to be effective.
Hypnotized by the Piper’s tune, doctors and patients are lured to consume 6, 8 or 10 drugs at the same time – what could be wrong with taking drugs that work. Infants, elders are lined up and medicated. The only ones who may be saved are those who have been previously crippled or injured on whom the music has ceased to work. Few can see what is going on around them or the name of the theatre in which they are being duped – The Empire of Humbug.