NorthWest Biotherapeutics Inc (NWBO)

[flipper44]

Pardon me for repeating this post. I just think it is highly relevant, because the DCVax-l phase III trial is an RCT with a (now) 2.3 to 1 ratio of treatment to control that was recently “augmented” by introducing concurrent/historical/synthetic control comparisons, probably because continuing the placebo randomization back in 2015 would have been unethical. In essence, getting the best from both control paradigms. (Aka: two heads are better tan one)

[External controls] are more common today. This 2019 article is comprehensive https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7218288/.

I found this excerpt helpful.

First, external data can be used to augment the precision of a concurrent control group in a randomized clinical trial (eg, using 4:1 randomization between treatment and control). Second, they can be used to create a stand-alone control group solely from external data. The latter is more common for rare diseases where single-arm trials predominate and randomizing to control is unethical, infeasible or highly inefficient.

Synthetic controls used to augment the precision of a concurrent control group generally have higher validity since they can be validated with the control arm in the performed RCT. Particularly control data from similar RCTs are valuable in this setting. Even if pertinent RCTs are not highly identical, much strength can still be gained.25 Where it is either unethical or infeasible to enroll patients to a control intervention, external data should be selected from the best possible source of data and synthetic controls can be used as a substitute for an absent control arm. In this setting (often a rare disease setting), pertinent RCTs are typically not available. There are many examples where various types of historical data, but increasingly data from prospectively recruiting patient registries as well as subsets from commercial real-world databases are being used. The scientific validity of these depend on the accuracy of the match. N-of-1 data, where each enrolled patient has at least several months of historical data on SOC, are ideal, but often not available. Data from similar medical centers from similar geographical regions to the target trial of interest can also improve validity. Of course, the more recent the higher validity, although a few months’ buffer should be allowed for full data curation processes (including quality assurance) to be finalized. Aggregate estimates from published cohort studies or case series may also be considered, either to validate primary sources of external data or as a substitute in the absence of individual-level patient-level data. Aggregate estimates from published studies nonetheless face limitations as detailed reporting on patient demographics, and clinical settings can be relatively sparse, and as such, similarity can be difficult to validate.