Just as there are numerous adaptations that fall within the umbrella of adaptive designs, there are several different statistical methods that can lead to the construction of a synthetic control arm. Cytel’s ebook, “Demystifying synthetic control arms”, is an effort to explain common strategies for their construction. Download the ebook by clicking on the button.
While a randomized control trial is still the industry gold-standard for regulatory submissions, there might be occasions when such trials are either infeasible or unethical. They might be infeasible due to small sample sizes, or other resource constraints. There could also be pragmatic or ethical issues to offering standard of care in certain therapeutic areas. For example, patients might drop out of an oncology trial if enrolled into the control arm, or it might simply be unethical to give the standard of care to an ill patient making a single arm trial a necessity. When a single arm trial is implemented by necessity, a synthetic control arm can play an important role for augmenting data collected for submission.
Awareness of these varying practices might enable the non-specialist to determine under what circumstances a placebo-controlled RCT is most appropriate and under what conditions to use single arm trial combined with a synthetic control. The book might also provide insight into what kinds of methods to enquire about when thinking of obtaining real world data, and working with a quantitative strategist.
This book will introduce Propensity Score Weighting and Bayesian Dynamic Borrowing, two of the most common strategies for the development of synthetic control arms. Then it will go on to discuss data considerations, and offer some key considerations for those considering a synthetic control arm.