Sponsors get the answer they want
Before we get going, we need to establish one thing beyond any doubt: industry-funded trials are more likely to produce a positive, flattering result than independently funded trials. This is our core premise, and you’re about to read a very short chapter, because this is one of the most well-documented phenomena in the growing field of ‘research about research’. It has also become much easier to study in recent years, because the rules on declaring industry funding have become a little clearer.
We can begin with some recent work: in 2010, three researchers from Harvard and Toronto found all the trials looking at five major classes of drug – antidepressants, ulcer drugs and so on – then measured two key features: were they positive, and were they funded by industry?1 They found over five hundred trials in total: 85 per cent of the industry-funded studies were positive, but only 50 per cent of the government-funded trials were. That’s a very significant difference.
In 2007, researchers looked at every published trial that set out to explore the benefit of a statin.2 These are cholesterol-lowering drugs which reduce your risk of having a heart attack, they are prescribed in very large quantities, and they will loom large in this book. This study found 192 trials in total, either comparing one statin against another, or comparing a statin against a different kind of treatment. Once the researchers controlled for other factors (we’ll delve into what this means later), they found that industry-funded trials were twenty times more likely to give results favouring the test drug. Again, that’s a very big difference.
We’ll do one more. In 2006, researchers looked into every trial of psychiatric drugs in four academic journals over a ten-year period, finding 542 trial outcomes in total. Industry sponsors got favourable outcomes for their own drug 78 per cent of the time, while independently funded trials only gave a positive result in 48 per cent of cases. If you were a competing drug put up against the sponsor’s drug in a trial, you were in for a pretty rough ride: you would only win a measly 28 per cent of the time.3
These are dismal, frightening results, but they come from individual studies. When there has been lots of research in a field, it’s always possible that someone – like me, for example – could cherry-pick the results, and give a partial view. I could, in essence, be doing exactly what I accuse the pharmaceutical industry of doing, and only telling you about the studies that support my case, while hiding the reassuring ones from you.
To guard against this risk, researchers invented the systematic review. We’ll explore this in more detail soon (p.here), since it’s at the core of modern medicine, but in essence a systematic review is simple: instead of just mooching through the research literature, consciously or unconsciously picking out papers here and there that support your pre-existing beliefs, you take a scientific, systematic approach to the very process of looking for scientific evidence, ensuring that your evidence is as complete and representative as possible of all the research that has ever been done.
Systematic reviews are very, very onerous. In 2003, by coincidence, two were published, both looking specifically at the question we’re interested in. They took all the studies ever published that looked at whether industry funding is associated with pro-industry results. Each took a slightly different approach to finding research papers, and both found that industry-funded trials were, overall, about four times more likely to report positive results.4 A further review in 2007 looked at the new studies that had been published in the four years after these two earlier reviews: it found twenty more pieces of work, and all but two showed that industry-sponsored trials were more likely to report flattering results.5
I am setting out this evidence at length because I want to be absolutely clear that there is no doubt on the issue. Industry-sponsored trials give favourable results, and that is not my opinion, or a hunch from the occasional passing study. This is a very well-documented problem, and it has been researched extensively, without anybody stepping out to take effective action, as we shall see.
There is one last study I’d like to tell you about. It turns out that this pattern of industry-funded trials being vastly more likely to give positive results persists even when you move away from published academic papers, and look instead at trial reports from academic conferences, where data often appears for the first time (in fact, as we shall see, sometimes trial results only appear at an academic conference, with very little information on how the study was conducted).
Fries and Krishnan studied all the research abstracts presented at the 2001 American College of Rheumatology meetings which reported any kind of trial, and acknowledged industry sponsorship, in order to find out what proportion had results that favoured the sponsor’s drug. There is a small punchline coming, and to understand it we need to cover a little of what an academic paper looks like. In general, the results section is extensive: the raw numbers are given for each outcome, and for each possible causal factor, but not just as raw figures. The ‘ranges’ are given, subgroups are perhaps explored, statistical tests are conducted, and each detail of the result is described in table form, and in shorter narrative form in the text, explaining the most important results. This lengthy process is usually spread over several pages.
In Fries and Krishnan  this level of detail was unnecessary. The results section is a single, simple, and – I like to imagine – fairly passive-aggressive sentence:
The results from every RCT (45 out of 45) favored the drug of the sponsor.
This extreme finding has a very interesting side effect, for those interested in time-saving shortcuts. Since every industry-sponsored trial had a positive result, that’s all you’d need to know about a piece of work to predict its outcome: if it was funded by industry, you could know with absolute certainty that the trial found the drug was great.
How does this happen? How do industry-sponsored trials almost always manage to get a positive result? It is, as far as anyone can be certain, a combination of factors. Sometimes trials are flawed by design. You can compare your new drug with something you know to be rubbish – an existing drug at an inadequate dose, perhaps, or a placebo sugar pill that does almost nothing. You can choose your patients very carefully, so they are more likely to get better on your treatment. You can peek at the results halfway through, and stop your trial early if they look good (which is – for interesting reasons we shall discuss – statistical poison). And so on.
But before we get to these fascinating methodological twists and quirks, these nudges and bumps that stop a trial from being a fair test of whether a treatment works or not, there is something very much simpler at hand.
Sometimes drug companies conduct lots of trials, and when they see that the results are unflattering, they simply fail to publish them. This is not a new problem, and it’s not limited to medicine. In fact, this issue of negative results that go missing in action cuts into almost every corner of science. It distorts findings in fields as diverse as brain imaging and economics, it makes a mockery of all our efforts to exclude bias from our studies, and despite everything that regulators, drug companies and even some academics will tell you, it is a problem that has been left unfixed for decades.
In fact, it is so deep-rooted that even if we fixed it today – right now, for good, forever, without any flaws or loopholes in our legislation – that still wouldn’t help, because we would still be practising medicine, cheerfully making decisions about which treatment is best, on the basis of decades of medical evidence which is – as you’ve now seen – fundamentally distorted.
But there is a way ahead.
Why missing data matters
Reboxetine is a drug I myself have prescribed. Other drugs had done nothing for this particular patient, so we wanted to try something new. I’d read the trial data before I wrote the prescription, and found only well-designed, fair tests, with overwhelmingly positive results. Reboxetine was better than placebo, and as good as any other antidepressant in head-to-head comparisons. It’s approved for use by the Medicines and Healthcare products Regulatory Agency (the MHRA) in the UK, but wisely, the US FDA chose not to approve it. (This is no proof of the FDA being any smarter; there are plenty of drugs available in the US that the UK never approved.) Reboxetine was clearly a safe and and effective treatment. The patient and I discussed the evidence briefly, and agreed it was the right treatment to try next. I signed a prescription saying I wanted my patient to have this drug.
But we had both been misled. In October 2010 a group of researchers were finally able to bring together all the trials that had ever been conducted on reboxetine.6 Through a long process of investigation – searching in academic journals, but also arduously requesting data from the manufacturers and gathering documents from regulators – they were able to assemble all the data, both from trials that were published, and from those that had never appeared in academic papers.
When all this trial data was put together it produced a shocking picture. Seven trials had been conducted comparing reboxetine against placebo. Only one, conducted in 254 patients, had a neat, positive result, and that one was published in an academic journal, for doctors and researchers to read. But six more trials were conducted, in almost ten times as many patients. All of them showed that reboxetine was no better than a dummy sugar pill. None of these trials was published. I had no idea they existed.
It got worse. The trials comparing reboxetine against other drugs showed exactly the same picture: three small studies, 507 patients in total, showed that reboxetine was just as good as any other drug. They were all published. But 1,657 patients’ worth of data was left unpublished, and this unpublished data showed that patients on reboxetine did worse than those on other drugs. If all this wasn’t bad enough, there was also the side-effects data. The drug looked fine in the trials which appeared in the academic literature: but when we saw the unpublished studies, it turned out that patients were more likely to have side effects, more likely to drop out of taking the drug, and more likely to withdraw from the trial because of side effects, if they were taking reboxetine rather than one of its competitors.
If you’re ever in any doubt about whether the stories in this book make me angry – and I promise you, whatever happens, I will keep to the data, and strive to give a fair picture of everything we know – you need only look at this story. I did everything a doctor is supposed to do. I read all the papers, I critically appraised them, I understood them, I discussed them with the patient, and we made a decision together, based on the evidence. In the published data, reboxetine was a safe and effective drug. In reality, it was no better than a sugar pill, and worse, it does more harm than good. As a doctor I did something which, on the balance of all the evidence, harmed my patient, simply because unflattering data was left unpublished.
If you find that amazing, or outrageous, your journey is just beginning. Because nobody broke any law in that situation, reboxetine is still on the market, and the system that allowed all this to happen is still in play, for all drugs, in all countries in the world. Negative data goes missing, for all treatments, in all areas of science. The regulators and professional bodies we would reasonably expect to stamp out such practices have failed us.
In a few pages, we will walk through the literature that demonstrates all of this beyond any doubt, showing that ‘publication bias’ – the process whereby negative results go unpublished – is endemic throughout the whole of medicine and academia; and that regulators have failed to do anything about it, despite decades of data showing the size of the problem. But before we get to that research, I need you to feel its implications, so we need to think about why missing data matters.
Evidence is the only way we can possibly know if something works – or doesn’t work – in medicine. We proceed by testing things, as cautiously as we can, in head-to-head trials, and gathering together all of the evidence. This last step is crucial: if I withhold half the data from you, it’s very easy for me to convince you of something that isn’t true. If I toss a coin a hundred times, for example, but only tell you about the results when it lands heads-up, I can convince you that this is a two-headed coin. But that doesn’t mean I really do have a two-headed coin: it means I’m misleading you, and you’re a fool for letting me get away with it. This is exactly the situation we tolerate in medicine, and always have. Researchers are free to do as many trials as they wish, and then choose which ones to publish.
The repercussions of this go way beyond simply misleading doctors about the benefits and harms of interventions for patients, and way beyond trials. Medical research isn’t an abstract academic pursuit: it’s about people, so every time we fail to publish a piece of research we expose real, living people to unnecessary, avoidable suffering.
In March 2006, six volunteers arrived at a London hospital to take place in a trial. It was the first time a new drug called TGN1412 had ever been given to humans, and they were paid £2,000 each.7 Within an hour these six men developed headaches, muscle aches, and a feeling of unease. Then things got worse: high temperatures, restlessness, periods of forgetting who and where they were. Soon they were shivering, flushed, their pulses racing, their blood pressure falling. Then, a cliff: one went into respiratory failure, the oxygen levels in his blood falling rapidly as his lungs filled with fluid. Nobody knew why. Another dropped his blood pressure to just 65/40, stopped breathing properly, and was rushed to an intensive care unit, knocked out, intubated, mechanically ventilated. Within a day all six were disastrously unwell: fluid on their lungs, struggling to breathe, their kidneys failing, their blood clotting uncontrollably throughout their bodies, and their white blood cells disappearing. Doctors threw everything they could at them: steroids, antihistamines, immune-system receptor blockers. All six were ventilated on intensive care. They stopped producing urine; they were all put on dialysis; their blood was replaced, first slowly, then rapidly; they needed plasma, red cells, platelets. The fevers continued. One developed pneumonia. And then the blood stopped getting to their peripheries. Their fingers and toes went flushed, then brown, then black, and then began to rot and die. With heroic effort, all escaped, at least, with their lives.
The Department of Health convened an Expert Scientific Group to try to understand what had happened, and from this two concerns were raised.8 First: can we stop things like this from happening again? It’s plainly foolish, for example, to give a new experimental treatment to all six participants in a ‘first-in-man’ trial at the same time, if that treatment is a completely unknown quantity. New drugs should be given to participants in a staggered process, slowly, over a day. This idea received considerable attention from regulators and the media.
Less noted was a second concern: could we have foreseen this disaster? TGN1412 is a molecule that attaches to a receptor called CD28 on the white blood cells of the immune system. It was a new and experimental treatment, and it interfered with the immune system in ways that are poorly understood, and hard to model in animals (unlike, say, blood pressure, because immune systems are very variable between different species). But as the final report found, there was experience with a similar intervention: it had simply not been published. One researcher presented the inquiry with unpublished data on a study he had conducted in a single human subject a full ten years earlier, using an antibody that attached to the CD3, CD2 and CD28 receptors. The effects of this antibody had parallels with those of TGN1412, and the subject on whom it was tested had become unwell. But nobody could possibly have known that, because these results were never shared with the scientific community. They sat unpublished, unknown, when they could have helped save six men from a terrifying, destructive, avoidable ordeal.
That original researcher could not foresee the specific harm he contributed to, and it’s hard to blame him as an individual, because he operated in an academic culture where leaving data unpublished was regarded as completely normal. The same culture exists today. The final report on TGN1412 concluded that sharing the results of all first-in-man studies was essential: they should be published, every last one, as a matter of routine. But phase 1 trial results weren’t published then, and they’re still not published now. In 2009, for the first time, a study was published looking specifically at how many of these first-in-man trials get published, and how many remain hidden.9 They took all such trials approved by one ethics committee over a year. After four years, nine out of ten remained unpublished; after eight years, four out of five were still unpublished.
In medicine, as we shall see time and again, research is not abstract: it relates directly to life, death, suffering and pain. With every one of these unpublished studies we are potentially exposed, quite unnecessarily, to another TGN1412. Even a huge international news story, with horrific images of young men brandishing blackened feet and hands from hospital beds, wasn’t enough to get movement, because the issue of missing data is too complicated to fit in one sentence.
When we don’t share the results of basic research, such as a small first-in-man study, we expose people to unnecessary risks in the future. Was this an extreme case? Is the problem limited to early, experimental, new drugs, in small groups of trial participants? No.
In the 1980s, US doctors began giving anti-arrhythmic drugs to all patients who’d had a heart attack. This practice made perfect sense on paper: we knew that anti-arrhythmic drugs helped prevent abnormal heart rhythms; we also knew that people who’ve had a heart attack are quite likely to have abnormal heart rhythms; we also knew that often these went unnoticed, undiagnosed and untreated. Giving anti-arrhythmic drugs to everyone who’d had a heart attack was a simple, sensible preventive measure.
Unfortunately, it turned out that we were wrong. This prescribing practice, with the best of intentions, on the best of principles, actually killed people. And because heart attacks are very common, it killed them in very large numbers: well over 100,000 people died unnecessarily before it was realised that the fine balance between benefit and risk was completely different for patients without a proven abnormal heart rhythm.
Could anyone have predicted this? Sadly, yes, they could have. A trial in 1980 tested a new anti-arrhythmic drug, lorcainide, in a small number of men who’d had a heart attack – less than a hundred – to see if it was any use. Nine out of forty-eight men on lorcainide died, compared with one out of forty-seven on placebo. The drug was early in its development cycle, and not long after this study it was dropped for commercial reasons. Because it wasn’t on the market, nobody even thought to publish the trial. The researchers assumed it was an idiosyncrasy of their molecule, and gave it no further thought. If they had published, we would have been much more cautious about trying other anti-arrhythmic drugs on people with heart attacks, and the phenomenal death toll – over 100,000 people in their graves prematurely – might have been stopped sooner. More than a decade later, the researchers finally did publish their results, with a mea culpa, recognising the harm they had done by not sharing them earlier:
When we carried out our study in 1980, we thought that the increased death rate that occurred in the lorcainide group was an effect of chance. The development of lorcainide was abandoned for commercial reasons, and this study was therefore never published; it is now a good example of ‘publication bias’. The results described here might have provided an early warning of trouble ahead.10
As we shall shortly see, this problem of unpublished data is widespread throughout medicine, and indeed the whole of academia, even though the scale of the problem, and the harm it causes, have been documented beyond any doubt. We will see stories on basic cancer research, Tamiflu, cholesterol blockbusters, obesity drugs, antidepressants and more, with evidence that goes from the dawn of medicine to the present day, and data that is still being withheld, right now, as I write, on widely used drugs which many of you reading this book will have taken this morning. We will also see how regulators and academic bodies have repeatedly failed to address the problem.
Because researchers are free to bury any result they please, patients are exposed to harm on a staggering scale throughout the whole of medicine, from research to practice. Doctors can have no idea about the true effects of the treatments they give. Does this drug really work best, or have I simply been deprived of half the data? Nobody can tell. Is this expensive drug worth the money, or have the data simply been massaged? No one can tell. Will this drug kill patients? Is there any evidence that it’s dangerous? No one can tell.
This is a bizarre situation to arise in medicine, a discipline where everything is supposed to be based on evidence, and where everyday practice is bound up in medico-legal anxiety. In one of the most regulated corners of human conduct we’ve taken our eyes off the ball, and allowed the evidence driving practice to be polluted and distorted. It seems unimaginable. We will now see how deep this problem goes.
Why we summarise data
Missing data has been studied extensively in medicine. But before I lay out that evidence, we need to understand exactly why it matters, from a scientific perspective. And for that we need to understand systematic reviews and ‘meta-analysis’. Between them, these are two of the most powerful ideas in modern medicine. They are incredibly simple, but they were invented shockingly late.
When we want to find out if something works or not, we do a trial. This is a very simple process, and the first recorded attempt at some kind of trial was in the Bible (Daniel 1:12, if you’re interested). First, you need an unanswered question: for example, ‘Does giving steroids to a woman delivering a premature baby increase the chances of that baby surviving?’ Then you find some relevant participants, in this case, mothers about to deliver a premature baby. You’ll need a reasonable number of them, let’s say two hundred for this trial. Then you divide them into two groups at random, give the mothers in one group the current best treatment (whatever that is in your town), while the mothers in the other group get current best treatment plus some steroids. Finally, when all two hundred women have gone through your trial, you count up how many babies survived in each group.
This is a real-world question, and lots of trials were done on this topic, starting from 1972 onwards: two trials showed that steroids saved lives, but five showed no significant benefit. Now, you will often hear that doctors disagree when the evidence is mixed, and this is exactly that kind of situation. A doctor with a strong pre-existing belief that steroids work – perhaps preoccupied with some theoretical molecular mechanism, by which the drug might do something useful in the body – could come along and say: ‘Look at these two positive trials! Of course we must give steroids!’ A doctor with a strong prior intuition that steroids were rubbish might point at the five negative trials and say: ‘Overall the evidence shows no benefit. Why take a risk?’
Up until very recently, this was basically how medicine progressed. People would write long, languorous review articles – essays surveying the literature – in which they would cite the trial data they’d come across in a completely unsystematic fashion, often reflecting their own prejudices and values. Then, in the 1980s, people began to do something called a ‘systematic review’. This is a clear, systematic survey of the literature, with the intention of getting all the trial data you can possibly find on one topic, without being biased towards any particular set of findings. In a systematic review, you describe exactly how you looked for data: which databases you searched, which search engines and indexes you used, even what words you searched for. You pre-specify the kinds of studies that can be included in your review, and then you present everything you’ve found, including the papers you rejected, with an explanation of why. By doing this, you ensure that your methods are fully transparent, replicable and open to criticism, providing the reader with a clear and complete picture of the evidence. It may sound like a simple idea, but systematic reviews are extremely rare outside clinical medicine, and are quietly one of the most important and transgressive ideas of the past forty years.
When you’ve got all the trial data in one place, you can conduct something called a meta-analysis, where you bring all the results together in one giant spreadsheet, pool all the data and get one single, summary figure, the most accurate summary of all the data on one clinical question. The output of this is called a ‘blobbogram’, and you can see one on the following page, in the logo of the Cochrane Collaboration, a global, non-profit academic organisation that has been producing gold-standard reviews of evidence on important questions in medicine since the 1980s.
This blobbogram shows the results of all the trials done on giving steroids to help premature babies survive. Each horizontal line is a trial: if that line is further to the left, then the trial showed steroids were beneficial and saved lives. The central, vertical line is the ‘line of no effect’: and if the horizontal line of the trial touches the line of no effect, then that trial showed no statistically significant benefit. Some trials are represented by longer horizontal lines: these were smaller trials, with fewer participants, which means they are prone to more error, so the estimate of the benefit has more uncertainty, and therefore the horizontal line is longer. Finally, the diamond at the bottom shows the ‘summary effect’: this is the overall benefit of the intervention, pooling together the results of all the individual trials. These are much narrower than the lines for individual trials, because the estimate is much more accurate: it is summarising the effect of the drug in many more patients. On this blobbogram you can see – because the diamond is a long way from the line of no effect – that giving steroids is hugely beneficial. In fact, it reduces the chances of a premature baby dying by almost half.
Copyright © 2012 by Ben Goldacre