by Ben Goldacre
When the committee charged with approving the drug cast their votes, they went 11–3 in favour.
It’s hard to know what to make of this process, since the vote went against not only the surrogate outcome data, but also the evidence from very large trials showing no benefit on real-world outcomes or survival. But we are all human, and it is hard to reject a drug when you’re faced with moving life-and-death testimony. One FDA scientist told John Abraham during his field work: ‘[Patient testimonials] definitely have an influence over advisory committees. That’s what Iressa proves.’ Several of these patients had been funded to attend the FDA advisory committee meeting by AstraZeneca. We can only wonder if individuals who had not been successfully treated with Iressa would have been flown across the country to speak their personal truth. Perhaps not. Perhaps they might be dead.
The FDA could have rejected the view of its expert committee, and that might have been wise. Not only was there no evidence of benefit: there were reports from Japan of fatal pneumonia associated with Iressa, affecting 2 per cent of patients, a third of whom died within a fortnight. But the FDA approved the drug all the same. AstraZeneca was compelled to conduct a further 1,700-patient study, which again found no benefit over placebo. Iressa stayed on the market. Another treatment appeared, and this one was effective in third-line treatment of non-small-cell lung cancer. Iressa stayed on the market.
The FDA did send out a letter saying that no new patients should be started on Iressa, but drugs that are on the market get used by doctors, often quite haphazardly, driven by marketing, habit, familiarity, rumour and lack of clear current information. Iressa continued to be prescribed for new patients. And still it stayed on the market.
We can see from the percentages in surveys that post-marketing trials requested by regulators are often neglected; and cynical doctors will often tell you that ineffective drugs are commonly marketed. But midodrine and Iressa are, I think, two cases that really put flesh on those bones. Accelerated approval is not used to get urgent drugs to market for emergency use and rapid assessment. Follow-up studies are not done. These accelerated approval programmes are a smokescreen.
The impact on innovation
As we have seen, drugs regulators don’t require that new drugs are particularly good, or an improvement on what came before; they don’t even require that drugs are particularly effective. This has interesting consequences on the market more broadly, because it means that the incentives for producing new drugs that improve patients’ lives are less intense. One thing is clear from all the stories in this book: drug companies respond rationally to incentives, and when those incentives are unhelpful, so are drug companies.
To explore whether new drugs represent any kind of forward movement in a field, we would have to examine all those approved over a period of time. This is exactly what a recent paper by some Italian researchers achieved.30 They took every drug acting on the central nervous system that had been approved since the first day the European regulator began approving drugs, and looked to see whether they represented any degree of innovation.
As you would expect by now, they found several serious problems with the data given to support these drugs’ applications. All the approved drugs had only been shown to be better than a placebo. Important information was missing from the trial reports: for example, there was no clear data about the number of people dropping out of each trial, which is important information, as it helps to show whether a drug is intolerable because of side effects. Then there were serious problems in the design of the trials. The majority (seventy-five out of eighty-three) were very brief. They were also small: not one of the submitted studies had enough participants to accurately detect a difference between a currently available treatment and a new one, on the rare occasions when they tried to do so.
The researchers concluded that the problem was straightforward: if the rules don’t require a company to show a new drug’s superiority to current treatments, they are unlikely to develop better medicines.
This is well demonstrated by the phenomenon of ‘me-too’ drugs. If you think back to the previous chapter, you will remember that developing a completely new molecule, with a completely new mechanism of action in the body, is a very risky and difficult business. Because of that, once a company has an established drug on the market, others will often try to produce their own version of that drug: so there are a great many antidepressants around from the class known as ‘selective serotonin reuptake inhibitors’, or SSRIs, for example. Developing a drug like this is much more of a safe bet.
Often these me-too drugs don’t represent a significant therapeutic benefit, so many people regard them as wasteful, an unnecessary use of development money, potentially exposing trial participants to unnecessary harm for individual companies’ commercial gain rather than medical advancement. I’m not entirely sure this is correct: among a class of drugs, one may be better than the others, or have fewer idiosyncratic side effects, so in that sense these copycats can be useful, sometimes. On the other hand, we have no way of knowing what amazing new drugs might have been created if we incentivised companies by insisting that they demonstrate superiority. These are not easy counterfactuals to unpick, and I never feel fully satisfied with the economists’ models of the impact on innovation, on either side.
However, we can see, by tracing the life of these me-too drugs, that the market does not work entirely as we might wish it to for us, the people who collectively pay for health services.31 For example, you might expect that multiple competing drugs in the same market would bring prices down, but an economic analysis, using Swedish data, showed that drugs regarded by the FDA as showing no advantage over the existing options enter the market at the same price. Another followed the price of an ulcer drug called cimetidine, and found that it became more expensive when ranitidine, from the same class, came on the market; and both drugs’ prices kept on rising when the competitors famotidine and nizatidine came out.
Perhaps the clearest story is told by tracing the recent history of another class of drugs, the ‘proton pump inhibitors’ used to treat reflux and heartburn. These are common medical problems, so it’s a lucrative area, and omeprazole was one of the most lucrative drugs in the class. Around a decade ago it was making AstraZeneca $5 billion every year, which was about a third of its total revenue from all drugs. But it was about to come out of patent, and once generic drug manufacturers could make their own pills containing the same drug, the price would plummet, and the revenue would disappear. So AstraZeneca introduced something called a ‘me-again’ drug.
This is an interesting new twist on the original idea. Me-too drugs are entirely new molecules that work in a similar way to the old ones; with a me-again drug, the same molecule is relaunched onto the same market for the same disease, but with one clever difference.
Complex molecules like drugs can exist in right- and left-handed forms, called ‘enantiomers’. The chemical formula for each of the different molecules is the same, and you will find the same atoms in the same order attached to the same parts of the same rings. The only difference is that a particular twist in a chain goes one way in one enantiomer, and the other way in the other, in the same way that your left and right gloves are identical mirror-images, made of the same material, the same weight, and so on. But left- and right-handed versions of drugs can have slightly different properties. Maybe the molecule only fits neatly into the receptor where it exerts its influence if it’s the right-hand version. Maybe it only fits nicely into the jaws of the enzyme that will break it down if it’s the left-hand form. This will affect what it does in your body. Recently, with increasing frequency, drug companies have started to release ‘single enantiomer preparations’, where you only get the right-handed version, say, of an existing treatment in your pill. The companies claim this as a new drug, and so add a whole new patent lifetime to their profits.
This can be a good financial bet. It’s usually easy to get marketing authorisation, because the m
ixed form of the drug has already been licensed, and you have lots of trials showing that form of the drug to be better than nothing. The second job, of convincing people that the single enantiomer is better than the mixture, is down to your marketing department, and may not be subject to much detailed formal scrutiny from a regulator at all.
So if they can have genuinely different properties in the body, why do people often think it’s dubious (or ‘evergreening’) for a drug company to put out a version of an existing drug that is only the single enantiomer form? Firstly, these different properties are often slight, so all the concerns about me-too drugs are also applicable for me-again drugs. Then there’s the issue of timing: it’s striking that companies often put out a me-again drug towards the end of the patent life of the original. It’s also worth bearing in mind that there’s no free ride, as ever, and that treatments with beneficial effects can also have side effects. The right-handed version of fluoxetine (Prozac) seemed like a great idea: it had a longer half-life than the original mixture, and this raised the possibility of an anti-depressant tablet that could be taken once a week, rather than once a day; but it also turned out to cause something called ‘QT-prolongation’, a change in the electric patterns in the heart which is associated with things like an increased risk of sudden death. But lastly, most strikingly, alongside these possible extra risks, the new ‘single enantiomer’ pills often don’t seem to work much better than the mixed ones, despite being very much more expensive.
Let’s look at omeprazole, our heartburn drug. Come 2002, AstraZeneca knew it was about to lose $5 billion a year, a third of its revenue, which would be a disaster for its profits and its stock price. But in 2001 the company launched esomeprazole, and this was a great success: in fact, AstraZeneca still takes $5 billion a year from the drug today. In the US it’s huge, a top-three blockbuster. In the UK it takes £44 million a year, but the amount of esomeprazole we get for that large amount of money is tiny, because brand-new esomeprazole costs ten times as much as old-fashioned omeprazole.
Here’s the kicker: new esomeprazole, just the left-handed version of the molecule rather than a mixture of both forms, is really no better than the plain old mixture in omeprazole tablets. The evidence is mixed, but it’s clear that there is no dramatic difference between any of the various members of this class of drug, and certainly no special amazing, unique benefit from esomeprazole.32
So why do doctors prescribe it? This is the power of the marketing machine in medicine, as we will shortly see. The direct-to-consumer campaign in the US was vast: AstraZeneca spent $260 million on ads in 2003,33 and purplepill.com, its website to promote the drug, eventually pulled over a million visitors every quarter.34 Against this, there was a considerable backlash. Kaiser Permanente, the American medical insurance giant, kept esomeprazole off its list of prescribable drugs, after deciding it was pointlessly expensive. Thomas Scully, the head of Medicare and Medicaid, gave speeches explaining that the drug was a waste of money; but with no final control over what gets prescribed in the two organisations, he sat and watched as they spent $800 million on this vastly expensive drug every year. When I say Scully gave speeches, he said: ‘Any doctor that prescribes Nexium [the brand name for esomeprazole] should be ashamed of himself.’ AstraZeneca complained about him to the White House and on Capitol Hill. Scully says that he was put under pressure to ‘shut up’.35 He hasn’t.
Comparative effectiveness research
These are not stories that can leave anyone feeling actively impressed. But there is a more important problem behind our ethical concerns about how companies behave in situations like this: we have allowed ourselves to be left – as prescribers, as patients, and as the people paying for health care – without clear evidence comparing different treatments against each other. We have no idea which treatments are best, and by extension, we have no idea which are harmful. If you die from getting the third-best treatment, then you have died needlessly and avoidably, and have every right to be angry in your grave.
This would seem to be a simple matter: we need to do more trials after drugs come to market, comparing them against each other in head-to-head tests. Health care is a huge financial burden on all societies around the world, and in most countries outside the USA this burden is shouldered by the state. If we are so weak that we can’t force drug companies, through our regulators, to conduct meaningful trials, then surely it makes sense for governments to fund them? This seems especially sensible when you consider that the cost of irrational prescribing, in most cases, is vastly more expensive than the cost of research to prevent it.
A clear early illustration of this comes from the ALLHAT trial, which started in 1994 and cost $125 million. This project looked at high blood pressure, a condition that affects about a quarter of the adult population, with half of that number receiving medication for it. The trial compared chlorthalidone, a very cheap, old-fashioned blood-pressure pill, against amlodipine, a very expensive new one that was being very widely prescribed. We knew from trials comparing them head-to-head that the two drugs were equally good at controlling blood pressure, but these numbers aren’t what matters to patients: a trial was needed that gave some patients the old drug, and some patients the new drug, and then measured how many people had heart attacks, and died. When ALLHAT finally did this comparison, measuring the real-world outcome that matters to patients, it found – to everyone’s amazement – that the old drug was much better. The savings from this one project alone vastly outweigh the cost of the trial itself, even though it was a huge project: this study started in 1994, when I was still an undergraduate, and finished in 2002, when I had finished training (I tell you that, because later we will see how important and difficult it is for doctors to keep up).
So ‘comparative effectiveness research’, as this is called, is vitally important, but it has only recently been embraced. To give you an illustration of how slow and arduous this journey is: in 2008, shortly after being elected President, Barack Obama demonstrated to many academics and doctors that he had a clear understanding of the deep problems in health care, by committing to spend $1 billion on head-to-head trials of commonly used treatments, in order to find out which is best. In return he was derided by right-wing critics as ‘anti-industry’.
On this issue, since people with resources often defend the pharmaceutical companies from deserved criticism, it’s worth remembering one thing: health care really is one of those areas where we are all, in a very real sense, in it together. If you’re super-rich, in the top 0.2 per cent of the population, you can buy pretty much anything you want. But however rich you are, if you become sick you can’t innovate new medicines overnight, because that takes time, and more money than even you have. And you can’t know the true effects of the medicines we have today, because nobody does, if they’ve not been properly tested, and if some results go missing in action. The most expensive doctors in the world don’t know any better than anyone else, since any trained person can critically read the best systematic reviews on a given drug, what it will do to your life expectancy, and there is no hack, no workaround, for this broken system. Even if you are super-rich, even if you make $10 million a year, you are right here in it with the rest of us all.
So comparative effectiveness research is a vitally important field, for everyone, and in many cases the value of finding out what works best among the drugs we already have would hugely exceed the value of developing entirely new ones. This research is a very rational area in which to spend more money.36
Monitoring side effects
But effectiveness is only part of the story. Alongside issues of which drug is most effective, there is also the problem of safety. Like many doctors, I am constantly amazed by the enthusiasm with which doctors embrace prescribing new drugs. When a new medicine comes on the market, with no proven benefits over an existing one, doctors and patients are being offered a simple choice: do you want to use an old drug, a known quantity, for which we have many years of experience in m
onitoring side effects? Or do you want to take a completely new drug, with no demonstrated advantages, which may, for all we know, have some horrific idiosyncratic side effect quietly waiting to emerge?
I was taught at medical school that in this situation, a doctor should regard the rest of the medical profession as unpaid stunt doubles: let them make the risky mistakes on your behalf, sit back, watch, learn, and then come back out when it’s safe. In some respects you could argue that this is useful advice for life more generally. But how are side effects monitored?
Once a drug is approved, its safety needs to be assessed. This is a complex business, with – to be fair – genuine methodological challenges, and glaring, unnecessary holes. The flaws are driven by unnecessary secrecy, poor communication, and an institutional reluctance to take drugs off the market. To understand these, we need to understand the basics of the field known as ‘pharmacovigilance’.
It is important to recognise, before we even begin, that drugs will always come onto the market with unforeseen side effects. This is because you need data on lots of patients to spot rare side effects, but the trials that are used to get a drug approved are usually small, totalling between five hundred and 3,000 people. In fact, we can quantify how common a side effect must be in order to be detected in such a small number of people, by using a simple piece of maths called ‘the rule of three’. If five hundred patients are studied in pre-approval trials, that is only enough patients to spot the side effects which occur more frequently than one in every 166 people; if 3,000 patients are studied, that is still only enough to spot side effects which affect more than one in every 1,000 people. The overall rule here is easy to apply: if a side effect hasn’t yet occurred in n patients, then you can be 95 per cent confident that it will happen in fewer than one in 3/n patients (there’s a mathematical explanation of why this is true in the footnote below, if you want one, but it makes my head hurt*). You can also use the rule of three in real life: if three hundred of your parachutes have opened just fine, for example, then assuming no other knowledge, the chance of one not opening, and sending you to certain death, is at least less than one in a hundred. This may or may not be reassuring for you.