Trial design
which is calculated based on some specific type one error rate or some specific power.” A type one error rate denotes the chance that the null hypothesis is rejected when it’s true, whereas the power is the likelihood that an effect is detected by the study. “So, if the statistical calculation demonstrates that we need 1,000 patients to achieve a certain power and to control the type one error rate to be below 0.05, for example, then we cannot change this number during the trial,” adds Zhou. That’s the traditional or conventional fixed design framework. “But using adaptive designs we can change some of the data, like the dose of the drug, or adding new drugs when they become available, or changing the sample size in the middle of the trial,” says Zhou. He specialises in Bayesian statistics, an innovative approach that turns the classical statistical model on its head. “The classical or frequentist paradigm is based on the viewpoint that the parameters are fixed quantities and the data is random,” Zhou says. “So under the frequentist framework, we view the treatment effect as something that is fixed – we don’t get to observe it, so we have to enrol patients and see what the random data looks like. But the Bayesian paradigm takes another approach: it views parameters as random and data as fixed, because we always observe the data.” This, he says, has one very important advantage: we can now make probability statements about the random parameters. The Bayesian approach, however, is only one among many adaptive methods. In fact, as James Wason, professor of biostatistics at Newcastle University, explains, the idea of doing something adaptive has been around for as long as the randomised clinical trial itself. “There’s an approach called adaptive randomisation”, Wason explains, “where you use patient outcome data. Instead of having a 50/50 allocation of the control and experimental treatment, if one of the treatments is doing better, you could favour that treatment for future allocation and allocate more patients in that arm.” That idea is from the 1930s. In fact, WR Thompson’s paper on adaptive randomisation procedures predates the MRC’s streptomycin trial by 15 years. This, as Wason explains, is common with statistical methods, where “the theoretical literature comes a long time before they’re used in practice”. Today, some of the key models of adaptive designs include platform designs, in which, Wason explains, “you can add in new arms as they become available, rather than starting a trial from scratch”, and umbrella designs, where “the treatment a patient gets is typically more tailored to their condition”. One question remains: If they have been around (at least in theory) for almost a century, then why have adaptive clinical trials only gained ground in recent years? “People are always willing to make changes to clinical
Clinical Trials Insight /
www.worldpharmaceuticals.net
trials for ethical reasons”, Zhou says, “but one of the main reasons adaptive trials weren’t so prevalent before has to do with the limits of computational power to better understand the operating characteristics of the designs”, he explains. “If we allow these changes to happen in the middle of the trial, then what is the property of the design? Nowadays, this can be simulated using computers, but these kinds of techniques weren’t available 30 years ago.”
The first recorded randomised controlled trial involved the tuberculosis treatment streptomycin, derived from the streptomyces griseus bacteria found in soil.
“People are always willing to make changes to clinical trials for ethical reasons, but one of the main reasons adaptive trials weren’t so prevalent before has to do with the limits of computational power to better understand the operating characteristics of the designs.”
Tianjian Zhou
Both Wason and Zhou note that the onset of Covid- 19 also marked a watershed moment in the upturn of adaptive trial designs. “With Covid”, Wason says, “certain types of adaptive designs had a big impact. The RECOVERY Trial, for example, used the platform approach and I think that really raised awareness of adaptive designs.” What is more, Zhou notes, “many ongoing clinical trials had to be stopped early, not because of evidence of efficacy or futility, but because of the interruption of Covid”. As in so many fields, the crisis of Covid shone a light on the feasibility of adaptive designs.
In addition to technological changes and the unprecedented onset of a pandemic, advancements in biological science have also changed the stakes for adaptive trials in recent decades. “In cancer,
35
AjayTvm/
Shutterstock.com
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53
