For 60 years, clinical trials have provided the gold standard of evidence for showing whether new treatments work and whether they are safe before they are rolled out on a large scale. Trials are used to evaluate not just drugs, but a whole range of non-pharmacological treatments, including exercise programmes, behaviour change therapies, complementary and alternative treatments, and public health interventions. Many of the treatments investigated nowadays are “complex interventions”, or treatments with a number of different elements.
The basic design of the clinical trial – known as parallel groups – has been so good that it has remained essentially unchanged for decades. These trials involve one group of people being given an experimental treatment and another a “control” such as a placebo or a standard treatment. These two groups are then followed up and assessed over a period of time to fully measure an effect.
The study often described as the first modern clinical trial was conducted by the UK Medical Research Council in 1948, and investigated whether a drug called streptomycin could be used as a treatment for tuberculosis – a disease that was killing large numbers of people around the world. The control in this example was “routine care” – or bed rest. The researchers found noticeably fewer deaths after six months in the patients treated with streptomycin. Disappointingly, these patients eventually developed resistance to the drug, which meant their health still deteriorated, albeit more slowly. In the end the solution turned out to be a combination of streptomycin with another drug, but the clinical trial as a mechanism for evaluating treatments had proven a success.
Since then, variations on the parallel groups design have been developed. The most common is the crossover design in which one group of people is given the control treatment and assessed before being given the experimental treatment and another assessment. The process is reversed for the second group, who is first given the experimental treatment and then the control.
Setting up a clinical trial is a lot of work – it takes a huge amount of design, organisation, money and ethical oversight. And trials can of course go wrong. In 2006, six healthy male volunteers were treated for catastrophic reactions and organ failure after a clinical trial into a drug called TGN1412 went wrong. A subsequent review found that the company managing the trial was unclear about what dose was safe to start testing on humans and was wrong to test all participants at the same time.
This was an extreme case, and was an example of a Phase I trial – a trial conducted with a relatively small number of healthy volunteers, at a point in the development of a new drug that hasn’t previously been given to humans. Nevertheless, it provides a stark illustration of the ethical obligations of triallists towards their participants.
Things have moved on since then, and developments to trial design such as the crossover have an important advantage in that they require fewer people to take part in the research. This is beneficial because all clinical trials must be carried out according to strict ethical principles – one of the key principles being to have enough trial participants to provide a reliable answer to the research question, but no more than is necessary so less people are exposed to risk.
Trials involving fewer people (but still able to give reliable results) are therefore a good thing, and they may also be cheaper and quicker to run. This does not just save pharmaceutical companies money – it helps public and charitable funders of research as well: the UK National Institute for Health Research, which now funds many trials of treatments and therapies that are subsequently offered by the National Health Service, reckons it spends up to £5,000 per participant on running a trial, which can involve hundreds or even thousands of participants.
But crossover trials can’t always be done. They assume that when you withdraw one treatment, people go back to being a blank canvas, which then allows you to evaluate another treatment. But often, treatments have lasting effects; even when you withdraw them, people don’t go back to being exactly as they were.
Our own work is in Phase III trials – large-scale trials intended to provide evidence for the effectiveness of a new treatment – and in particular “pragmatic” trials, which compare treatments as they will actually be delivered by healthcare providers, and where the control is routine care. We thought we could design a simple, new way of carrying out these kinds of trials, which not only reduced the number of people taking part but didn’t assume (as the crossover design does) that people revert back to neutral when you withdraw the experimental treatment. We’ve called it the dog-leg trial.
Dog-leg and stepped wedge
The design, which we published in the International Journal of Epidemiology, took its inspiration from another variant of the parallel groups trial design called a stepped wedge. In a stepped wedge trial all participants end up getting the experimental treatment, but the treatment is staggered over a (sometimes large) number of groups and time intervals and can involve a heavy burden of repeated assessments.
We took the schedule of assessments in a stepped wedge design and attempted to thin it out to its essential core. What we ended up with was three groups of participants: one which is simply assessed after the new treatment (“A” for after), another which is assessed both before and after the new treatment (“BA”), and a third group which is not treated at all and is assessed once (“B”). The timing of the assessments in the three groups is crucial – in particular making sure that the “A” in one group lines up with the “B” in the next. This creates a crooked, snaking arrangement as to when assessments are made – which gave us the name dog-leg, not, as one reviewer asked, because of any literal link to canine anatomy.
When researchers plan a phase III trial, they use a mathematical calculation known as a “power calculation” to determine the number of people they need. Crucially, in many situations a dog-leg trial needs fewer people and fewer resources to achieve the same statistical power as a parallel groups trial to detect whether a treatment is effective. So this is the strange yet brilliant thing: it uses more groups but requires fewer people. It shouldn’t work but it does – when you do the maths you find that the dog-leg method is a surprisingly efficient way to run a trial.
Because the dog-leg design requires a patient group with a relatively stable condition (it must be possible to delay the delivery of the treatment to participants recruited into the trial) it will be particularly useful for research looking into ongoing chronic diseases. The ultimate aim is to make the research and development stage more efficient so that the findings can get out and benefit patients and the public.