by Burch, Druin
This time the Medical Research Council was persuaded. A randomised, double-blind, placebo-controlled trial was set up and, from 1970, began enrolling patients. Archie Cochrane, still convinced that cardiologists deserved constructive mockery, taunted them that their expensive coronary care units might be made redundant by something as old and cheap as aspirin.
John Crofton had found his involvement in the MRC streptomycin investigations less than inspiring – trials, he concluded, were important but ‘not intellectually challenging’. Elwood was the other way around. Having been bored as a junior doctor, he found the world of trials altogether different. ‘I found it so intellectually intoxicating I used to walk home reading a textbook and often I would get so entranced I’d walk into a lamp-post.’ (Jeffreys, 2004)
On a Saturday in 1972, Elwood was in his office. The phone rang. It was a call from an American pharmacologist named Herschel Jick, working in Boston. Jick was involved in a study looking at what drugs patients were taking during the few days before coming into hospital. It was a trawl to see if they could pick up associations, good or bad – and, like all trawls, it had a high likelihood of fishing out something meaningless. Jick’s trawl involved forty different diseases and around sixty different drugs. Even if every one of those drugs was an identical placebo, that meant you could expect to find some relationships that looked meaningful when they weren’t.
However, the association that Jick’s team discovered did not sound at all meaningless to Elwood. Heart attack patients were taking a third less aspirin than others. If Jick’s finding was not chance, it meant that people on aspirin were coming into hospital with heart attacks much less than expected. That implied one of two explanations. Either aspirin was preventing heart attacks, or it was making them so much worse that patients never made it to the hospital alive.
The use of speculative studies that go looking for associations is in turning up hypotheses. You find a link between admission rates and aspirin and you theorise to yourself that it is not just there by chance. Then you set up an interventional study to figure out whether it really is cause and effect. You give some people aspirin and see what happens. You carry out a study, in other words, exactly like the one that Elwood and O’Brien had laboriously set up. Jick’s team wanted Elwood to dissolve their trial, to end it prematurely just in case they were killing people with aspirin. ‘Look,’ they said, ‘we have to know if aspirin is beneficial or harmful.’
When it comes to designing a trial, or understanding a treatment, certain outcomes matter more than others. The ones that ultimately count are called ‘hard’ outcomes. Others, often used because they are easier to count and collect, act as proxies. They are ‘soft’ outcomes. For aspirin and heart attacks, Elwood’s team were using the hardest outcome of all – death. That was what they felt really mattered; did the drug help people survive?
They could have chosen softer outcomes. Not everyone who has a heart attack dies from it, for example. If you count heart attacks instead of deaths then your trial is quicker and easier and cheaper. There are simply more heart attacks than deaths. Elwood and O’Brien were concerned that, since aspirin was a pain killer, it might stop people from noticing a small heart attack. (Even without drugs, some heart attacks are curiously painless.) In other words, they felt that the number of observed heart attacks would be a poor proxy for figuring out the actual effects of aspirin. The soft outcome was not reliable enough. The drawback of their trial’s high quality was that it needed to recruit large numbers of people, enough that a sufficient number would die. A difference in deaths between two groups – or enough observed deaths to show no difference at all – was needed to determine what aspirin actually did.
Elwood pointed out to Jick that the MRC trial of aspirin had so far recorded only seventeen deaths. If aspirin were a miracle drug, with a massive effect, those seventeen might be enough to reveal the truth about it – if all seventeen had been taking the placebo, say, or if all of them had been on the real drug. Aspirin was not new, though, and despite Craven’s suggestion that it stopped every single heart attack and stroke, no one believed that its effects could possibly be that overwhelming. So seventeen deaths were too few to make the trial meaningful. What the American team were asking did not really make sense. Stopping a trial early means you get results sooner, but it also increases your chances of them being useless or misleading. In order to go along with the American request, Elwood’s and O’Brien’s trial needed to remove the double-blinding that kept secret who was on aspirin and who on placebo. That meant compromising their trial.
Oddly, and after some debate, Elwood and Cochrane agreed with the American request. Eleven of the patients who had died turned out to be taking placebos, six aspirin. As predicted, it was meaningless. Aspirin neither killed everyone nor saved everyone. Those numbers were too small and similar to be taken as saying anything else.
Reassured that their patients were not all dropping dead as the result of their aspirin, Elwood and his colleagues correctly concluded that their trial efforts needed to be started up again. They did so, reporting in 1974 on their subsequent recruitment of 1,239 men with recent heart attacks. Giving half of them aspirin and half a placebo, they got results that were suggestive of the drug’s helping to keep people alive. But they were not conclusive.
There is no absolute guarantee that any result is not the work of chance. The more unlikely it is, though, the more confident you can be. The chance that Elwood’s trial was due to luck, rather than an effect from aspirin, was more than one in twenty, and therefore over the conventional cut-off for statistical significance.
The result of the aspirin trial persuaded some doctors to start prescribing it for those at risk of a heart attack. More usefully, it encouraged others to make their own attempts at similar trials. The larger the number of people studied, the more confidence you could have in the results. What was continuing to be clear was that aspirin was not a magic cure; the sort of impact that Craven reported simply did not exist. And the difficulty of proving moderate effects, ones that were often not apparent for some time after someone began swallowing a pill, were substantial.
The Medical Research Council, with Elwood and Cochrane, tried again. This time they recruited around 2,000 people to take either aspirin or a placebo. Again, the results were positive but inconclusive. ‘Well, gentlemen,’ said Richard Doll to the disappointed team, ‘the evidence may not be certain, but it is more convincing than that for most of the other drugs in the British Pharmacopoeia.’
Kerr White, a friend and colleague of Cochrane’s, recalled the two of them attending a conference in 1976 in New Zealand. Nervous of scaring his audience, a ‘staid group of white-coated clinicians’, White suggested to them that only 15 to 20 per cent of what doctors did for their patients was actually proven to work. ‘In mid-sentence Archie suddenly called out: “Kerr, you’re a damned liar, you know perfectly well that it isn’t more than 10%!”’ Both men were drawing their figures from a study of family doctors across Britain which had shown that only 9 per cent of all prescriptions written were for drugs whose effectiveness was properly established. That, after all the years of medical advances, was the extent to which doctors knew what they were doing.
Neither the two Medical Research Council trials nor another four done across the world showed a conclusive answer for aspirin. The last one reported in 1980, and after it appeared Cochrane and Elwood attempted to combine the results from all six. Finally, they thought, they had something persuasive. Altogether almost 11,000 people had been treated with aspirin as part of one of these randomised, double-blind, placebo-controlled trials. Put together, the numbers became convincing. If you started taking aspirin after having a heart attack, your chance of dying went down by around a quarter. Cochrane and Elwood calculated the odds of a genuinely useless pill giving such good results by chance. This time they came out at less than one in 10,000.
Aspirin decreased a patient’s chance of dying by a quarter in the year
after their heart attack. A big effect, but too small to be reliably apparent on the basis of a doctor’s clinical experience. Too small even to show up in trials involving 2,000-odd patients.
The benefits of aspirin are, in some ways, certain. If you give the pill to a large number of people, you save lives. It does a population good. On any given occasion, for any given patient, however, it can be useless. Swallow an aspirin after your heart attack, and there is a chance – a risk – that it will save your life. There is also a chance it will do nothing, and a chance it will make you bleed. Some of those bleeds go on to kill.
Before the introduction of aspirin, 88 per cent of people survived their heart attacks. By helping to prevent blood clots in coronary arteries, aspirin improved that number. Without aspirin the chances of dying were 12 per cent. Cut that by a quarter, and it went down to 9 per cent. In other words, if you give people a pill that improves their survival by a quarter, the effects you see depend very much on how likely they were to die in the beginning. Given that most people survive heart attacks, the impact of aspirin can seem minor. Rather than 88 per cent of people surviving a heart attack, 91 per cent do.
Aspirin drops your risk of dying by a quarter, your relative risk. Your overall chance of dying, however, goes from 12 per cent to 9 per cent – your absolute risk reduces by 3 per cent. Craven may have had some good ideas, but his results were not even close to being right. The only way he could have got the numbers he did was by introducing bias. That is not to say he consciously manipulated anything, only that it takes something as strict and careful as a randomised, double-blind, placebo-controlled trial for even the most humane and sincere people to avoid doing so. That 3 per cent is not enough for any doctor – even a cardiologist – to dependably notice in day-to-day clinical experience. Yet it matters. Heart attacks are common. Aspirin, properly used, saves 100,000 lives worldwide, every year: about 7,000 people per year in Britain, almost 30,000 a year in America.
* * *
1 The harmful effects of smoking are roughly equivalent to the combined good ones of every medical intervention developed since the war. Those who smoke, in other words, now have the same life expectancy as if they were non-smokers without access to any health care developed in the last half-century. Getting rid of smoking provides more benefit than being able to cure people of every possible type of cancer.
2 The word too easily means trials of widely varying reliability and power to uncover truths.
21 Large Trials and Grand Designs
BY 1980 THE need to investigate medicines reliably had become increasingly apparent. Those attracted to the effort were faced with two problems. One they knew about before they began: doctors were ignorant about the effects of most treatments. The other came as a surprise. Even when real evidence was placed in front of them, doctors sometimes took no notice of it. Habit was strong and clever explanations appealed more to clinicians than straightforward and partly statistical proofs. Researchers did not just need to change medical minds, they needed to win hearts.
It is difficult to shift fixed opinion. Two early randomised, double-blind, placebo-controlled trials were performed on homeopathy in the nineteenth century, the first in 1835, the second from 1879 to 1880. Both showed homeopathic medication did nothing different from an identically presented placebo. The British Medical Journal even drew doctors’ attention to the second of the two trials, lauding the homeopaths for their involvement, saying it was ‘highly creditable to those who ventured on an experiment involving so much peril to a favourite theory’. Neither trial, though, was understood. Failing to be convinced by either the results or the methods, believers in homeopathy went on prescribing remedies proven to be useless. Traditional doctors marched back to treatments that were even worse. The power and clarity of methodical testing were ignored.
So it seemed that doctors were merely repeating history when they took remarkably little notice of the combined aspirin trials. They did not trust them; statistical techniques were something they did not understand and did not like. The drug had been proven useful, but doctors were not using it.
‘Archie was immensely gratified’, said Elwood, ‘when, around this time, Richard Peto took up the matter of aspirin and presented a considerably elegant overview.’ What Peto, a statistician with medical interests, did was to take the data from the six trials of aspirin in heart disease, and put them together in a more statistically adept – and therefore more persuasive – manner. This helped, but only a little. The medical profession remained largely impervious.
Here is Dr Khan, writing in his Encyclopedia of Heart Diseases about the further development of aspirin as a treatment. ‘The timely 1983 study by Lewis et al. in the United States heralded a new era, and aspirin became widely known as a life-saving drug.’ He explains that the trial demonstrated tremendous benefit, dropping rates of heart attacks by a half in those with angina. (Angina is the shortened form of ‘angina pectoris’, the old and still current description of the pain people feel when their hearts are starved of blood; in the original Latin, it means a throttling pain about the chest: a cramp of the heart. Sometimes people say they feel as if they are about to die – as, indeed, some of them are.
The 1983 study that Khan refers to came after Richard Peto’s much more reliable synthesis of the aspirin trials showing that the drug extended lives. Peto’s overview involved very many more patients and it showed a change in a more important outcome – heart attacks matter; staying alive matters more. Yet cardiologists were unpersuaded by Peto, just as they were by Elwood and Cochrane before them. Khan is mistaken, too, in suggesting they paid much attention to Lewis’s study of 1983. Cardiologists thought very little of aspirin, and following their eminent lead the family doctors who might have prescribed aspirin were also unconvinced.
There is something charming about Khan’s retrospective optimism. He shows a wishful belief that cardiologists quickly understood the importance and relevance of the high-quality evidence that, for the first time, was being placed in front of them. He suggests that after ignoring the early conclusive evidence, they were open-minded enough to pay attention to the subsequent data. It is an amiable belief, but it is also wrong.
Cardiologists continued to be enchanted with their favourite theories, and indifferent to demonstrable facts. Aspirin was not new enough or powerful enough to appeal. They preferred warfarin, dicoumarol and other anticoagulants, drugs that you could really see working because they caused a lot of patients to bleed dramatically. Many died from their blood loss, which hammered home to the cardiologists the impression that their drugs were potent. And cardiologists, like all doctors, had a proud history of bleeding people to death while convincing them that it was for their own good. (Anticoagulants for heart attacks were later investigated with a large, randomised, double-blind, placebo-controlled trial. It found them to be useless.)
Backing up the cardiologists was the American Food and Drug Administration. At the tail end of 1980, the FDA refused permission for a drug company, Sterling, to start recommending aspirin on the basis of the new trial data. The company wanted to say that aspirin, in the wake of Peto’s overview, ‘has been shown to be effective in reducing the risk of death or re-infarction of patients who have suffered a myocardial infarction’, meaning a heart attack. The FDA, as unable to understand the evidence as most cardiologists, forbade Sterling from doing so. The importance of this was that it affected advertising. Aspirin was registered as a drug to treat fevers and pain. Without the FDA’s approval, companies were not allowed to market it for treating heart disease. Sterling was already taking a risk in trying to invest in aspirin, since the patent for it was long expired. Now the FDA was forbidding it from even advertising in order to try to familiarise doctors with the drug’s proven benefits. Doctors could still prescribe it for heart attack victims if they wished – they were under no compulsion to pay any attention to what the FDA thought – but few of them thought that it was of any use in such situations.
&
nbsp; In 1983, Sterling had another go. This time it got as far as persuading the FDA to formally listen to the case. Sterling flew over Peto and Elwood to present their data. The committee listened happily to the speaker who came first, a man who explained the theories behind why aspirin should work to prevent heart attacks.
Diarmuid Jeffreys’ Aspirin, a history of the drug, gives a wonderful account of the collision that followed, when the British doctors stepped up to speak. The committee distrusted the data from Elwood’s two MRC trials. And, more importantly, they were unconvinced by Richard Peto’s claim that the results from different trials could be usefully combined to give a robust conclusion.
Dressed in his customary brown corduroy jacket, with no tie and with longish blond hair falling over his collar, Peto cut an unusual figure in a room full of conservatively suited executives. He clearly anticipated that his ideas were going to be challenged and he was in no mood to be patronized by an American panel that he suspected held unflattering views of British science . . . His ad hoc style was more informal than the panel members were used to and his ironic tone seemed to suggest that if the Americans couldn’t see what he was on about, it was their failing not his.
Elwood recalled that the meeting collapsed. Peto, upset, was reduced to calling people fools and idiots. Sterling was denied permission to market aspirin to heart attack victims.