HUNDREDS of years ago, if you had a pain, a cough or a fever, an apothecary might prescribe you a tincture or – joy – a restorative course of leeches. Thankfully, medicine has come a long way since then. It is by no means perfect, but hospitals, drugs and healthcare have made our days inestimably more comfortable.
Much of this is thanks to that bastion of science, the clinical trial, which tests whether a medicine or treatment is safe and effective. Evidence from such trials is considered the gold standard, and over the years it has helped us sort the quackery from the cures. It might be surprising to hear, then, that a growing number of doctors think the way we test medicines needs an overhaul.
Advertisement
For all their strengths, clinical trials often take years to deliver a verdict. This drawback was exposed during the covid-19 pandemic, when we desperately needed treatments for a new disease. Doctors were forced to use quicker methods of assessment, and at this juncture, it seems they paid off. “We were able to achieve in weeks what would have otherwise taken years,” says epidemiologist .
If we can get robust answers about medicines in a faster way than standard clinical trials can, surely we are ethically obliged to do so. Some say yes: helping more people more quickly must be a good thing. Others worry that rushing medicines into use has got us into trouble before. Whether there really are speedier, more reliable ways of doing clinical trials is rapidly becoming one of the most critical questions in medicine.
The origins of the clinical trial can be traced back to a major problem faced by British sailors in the middle of the 18th century. Thousands of them were dying of scurvy, a condition that causes spontaneous bleeding, pain in the limbs and ultimately death. A physician called James Lind suspected that scurvy could be cured by changes to sailors’ diets and, in 1747, he set up a small but pioneering test of the hypothesis. He divided 12 men with scurvy into six groups and augmented their diets in different ways. The ones who got extra fresh fruit fared much better than the others – and the rest is history.
Lind got the right result. But his test was missing two elements – apart from scale – that we now consider crucial to a clinical trial. First, is the control group. In modern clinical trials, people receiving a new drug or treatment usually aren’t compared against those getting other, similar treatments, but against a group receiving a placebo, a dummy pill. The point is to have the control group treated in what seems like the same way, so that the difference between this group and those given a new treatment can be understood clearly.
The second important facet is randomisation. When you divide people up, you can accidentally introduce differences between the groups without realising it. We know, for instance, that doctors sometimes subconsciously choose healthier individuals to receive a new drug, which can make it look as if an ineffective treatment works.
These two key ideas were first put into practice in a 1947 trial involving the newly discovered antibiotic streptomycin. The US doctors who ran the trial used a system that ensured neither the doctors nor the patients knew who was receiving the drug and who was getting the placebo until the end. This format coalesced into what we now call a randomised controlled trial, or RCT. Today, trials proceed through phases I, II and III, which test drugs’ safety and efficacy on progressively larger groups of people.
“It often takes upwards of 10 years to chaperone a drug from invention to approval”
The system works. But even before the pandemic hit, it was clear that it came at a high price. Between 2015 and 2016, , much of which was paid for by pharmaceutical companies, which claim this contributes to high drug prices. Then there is the time it takes to chaperone a drug from invention through trials to approval, which is often upwards of 10 years. Trials can also be costly to the volunteers themselves. There is often a 50:50 chance that someone involved in a trial will get only a placebo. Perhaps that explains why participation is woefully low – .
Doctors have been bothered by all this for a long time. In the early 1980s, Robert Bartlett, a surgeon at the University of Michigan, wanted to test whether extra-corporeal membrane oxygenation (ECMO) machines could be used to do the work of the heart and lungs in premature babies. Until then, the machine had only been used in adults. Bartlett used an unorthodox trial design that borrowed from game theory. The first infant enrolled would receive the standard kind of ventilation at the time, which involved a continuous positive airway pressure, or CPAP, machine. If it survived, the next would receive the same standard care. If it died, the second enrollee would be hooked up to an ECMO. The first baby died. The second received ECMO and survived. So did the third, the fourth, and all the rest up until the 12th baby. Given that nearly all similar premature babies had died without ECMO, .
Bartlett’s peers thought the trial was too small, however, and the design too unorthodox, to accept ECMO as the new standard of care. According to clinical statistician David Robertson at the University of Cambridge, the reception of Bartlett’s work put others off trying out new kinds of clinical trial for years to come. ECMO wasn’t accepted until 1996, when a group of UK scientists assessed the machine using a full-blown RCT that involved 185 babies. As in Bartlett’s trial, . For , Canada, though, it was a pyrrhic victory. As a father, he says, all he can think about are the babies who didn’t make it. “Did all those kids really have to die just to prove a point?”
In the 1990s, more doctors started to gradually experiment with adaptations of the clinical trial. This produced two basic kinds of novel trial that each address different shortcomings of traditional RCTs (see “Trials on trial”).
The first big problem with conventional trials is that they build the same research infrastructure for each and every drug tested. Everything from the recruitment of administrative staff to bed space and the statistical models are organised from scratch for each trial. That is like building a new stadium for every football game, says statistician Munya Dimairo at the Sheffield Clinical Trials Research Unit, UK.
To get around this, clinicians now sometimes use what is called a platform trial. Here, scientists create an infrastructure that can be used again and again. What’s more, researchers can share a single control group across multiple trials of different drugs, as well as test therapies head-to-head against each other. This kind of trial doesn’t ask “is one particular drug effective?”, but “which drug is the best at treating a certain condition?”.
A second shortcoming of RCTs is that lots of participants are given a placebo. An approach that takes the edge off this is called an adaptive trial. You might think of these like a modern variation of Bartlett’s research. In these trials, patients aren’t randomised into groups, with each receiving different treatments. Instead, the treatment each person gets is determined by rules that are written before the trial. For example, if lots of patients have benefited from a particular drug, then the odds that a patient coming later in the trial will also be given that drug are increased. In practice, the allocation of treatment to a patient is determined by a computer algorithm, so the doctors running the trial remain blinded to who gets what.
A major gamble
Some trials combine both approaches, so that each arm in a platform trial tests in an adaptive fashion. One well-known trial that did things this way is the I-SPY2 trial. It was launched in 2010 by non-profit organisation the Quantum Leap Healthcare Collaborative to search for drugs that could prevent the progression of breast cancer, a common condition for which, at the time, there were few effective drugs. The trial tested six drugs simultaneously, using a single control group as the standard against which they were measured. Drugs that showed signs of efficacy would graduate for further evaluation and these would then become part of a new group of controls receiving the now-updated standard of care. I-SPY2 has enrolled more than 1400 participants and tested 16 drugs.
“The way we carry out clinical trials is like building a new stadium for every football game”
In a sense, I-SPY2 was a major gamble, says Scott Berry, a co-founder of Berry Consultants, who helped to set up the trial. At the time, the various agencies that approve drugs as safe for use were sceptical of adaptive trials and there was no guarantee that they would accept the evidence from I-SPY2 as valid.
But that has changed. In 2015, Janet Woodcock, who heads the US Food and Drug Administration (FDA), published an article in the New England Journal of Medicine that not only green-lighted novel designs, but actually encouraged them. She specifically cited I-SPY2 as an example of how adaptive and platform trials can successfully, quickly and safely identify promising therapies. This, says Ronenn Roubenoff at the pharmaceutical company Novartis, was a watershed moment. The FDA and the European Medicines Agency have now approved several drugs tested by the I-SPY2 programme.
Adaptive platform trials had been given a shot in the arm – and they were about to get another unexpected boost. In 2020, the covid-19 pandemic struck. Hospitals around the world were thrown into chaos and countries needed treatments, pronto.
One of the first and largest covid-19 platform trials was the UK’s RECOVERY trial, led by Landray and his University of Oxford colleague Peter Horby. Beginning in March 2020, RECOVERY aimed to test whether widely available drugs reduce the chances of someone dying from covid-19. The study has since enrolled nearly 50,000 participants and tested 10 drugs, three of which have proved effective. Similar kinds of trial have cropped up elsewhere, including efforts from the team behind I-SPY2 and the TOGETHER Trial in Canada and Brazil.
For Landray, the pandemic has shown that clinical research doesn’t have to stand apart from regular patient care. It can become an integral element of it in the form of adaptive trials. He points out that most medical practice isn’t actually based on evidence. Even in cardiology, which has particularly rigorous guidelines, only about a quarter of what doctors do has been supported by clinical trials. The biggest advantage of adaptive and platform trials, he believes, is their potential to shift clinical research away from large academic hubs and to smaller hospitals and surgeries. This would allow more physicians and patients to take part in studies.
Some doctors are nervous about these new kinds of trial. The statistical methods behind some of them can be arcane, which makes them difficult to replicate. If you don’t have a PhD in statistics, study methods can seem like a black box, says William Rosenberger at George Mason University in Virginia. And while adaptive trials get results more quickly, they require more planning work and a large support staff to run.
Platform and adaptive trials also aren’t suitable in all circumstances. For instance, if patients need to be followed for several years to determine a drug’s efficacy, researchers might not be able to conduct the interim analyses needed to alter randomisation. Then there are orphan diseases, conditions that affect a very small number of people, and for which there aren’t multiple new drugs available to test.
Still, spurred on by the pandemic, many doctors are now deciding to embrace adaptive trials. At Massachusetts General Hospital, neurologists Sabrina Paganoni and Merit Cudkowicz are looking for treatments for motor neurone disease, also known as amyotrophic lateral sclerosis, which robs patients of their ability to walk, talk, swallow and, ultimately, breathe.
With the HEALEY platform trial, which launched in 2020, they and their team are testing three drugs at once. With an adaptive platform design, they could get answers using only 480 participants instead of the 720 needed for a conventional clinical trial. Work on the first drugs is already wrapping up and two more are due to be added soon.
“We never could have gotten these answers with a standard design,” says Paganoni. “And patients are relieved that they are much more likely to receive an actual treatment.”
Trials for everything
There are all sorts of things that might help people live better lives that fall outside the sphere of medicine, such as providing sanitation, solar panels or textbooks. Global development organisations often use randomised controlled trials (RCTs) to evaluate what kind of interventions are the most effective. But traditional RCTs are slow and expensive, so charities are increasingly interested in using novel trial designs (see main story).
Luke Allen at the London School of Hygiene & Tropical Medicine is working on several such trials. Prescription glasses improve vision and enable children to learn better in school, but in parts of Asia and Africa, many people who are prescribed glasses never get them fitted. Along with several partners, including UK charity Peek Vision, the University of Botswana and the Kenyan Ministry of Health, Allen and his colleagues are designing adaptive trials that will test a range of interventions that could help. These include tweaking the nature of reminder messages and offering incentives such as money-off vouchers. Variations of these interventions will be tested all at once and algorithms will tilt the balance towards more successful strategies as the trials go on.
Trials of any stripe must get ethical approval from review boards before going ahead, which can take months. For adaptive trials, where speed is of the essence, this can be a major barrier. Allen says faster ethical review processes could help the world’s most disadvantaged communities benefit from adaptive trials. “They don’t have time to waste,” he says.
żěè¶ĚĘÓƵ audio
You can now listen to many articles – look for the headhones icon in our app
