Read on Twitter
(1/n) On disclosure, what a sponsor needs to know, how the choice of design has a huge impact on that, and how we might create a new system in the (far off?) future…..not a tweetorial…just my thoughts! Please comment. https://twitter.com/ADAlthousePhD/status/1325801566973079553
(2/n) I’m focused on ongoing trials and the potential disclosure of interim results of a continuing trial. For a trial ending (either at an interim or not) there are mandatory disclosures but we should also require timely dissemination of complete results.
(3/n) Interim results are also subject to disclosure. The most common are “flexible sample size” trials such as sample size re-estimation and group sequential designs. This is a sponsor choice, and there are implications to the choice.
(4/n) Sample size re-estimation typically looks at some early point and adjusts the maximal sample size..think “look at N=100 and decide a final sample size between 200 and 500”. This has disclosure implications since operations requires knowledge of sample size changes.
(5/n) If you are given the re-estimated sample size (and the protocol), you can back solve the observed interim treatment effect. This may be considered material information, but it’s also an FDA trial integrity concern that pervades multiple guidance documents.
(6/n) Section 7 of the FDA adaptive guidance is all about this.
https://www.fda.gov/media/78495/download
For a specific statement, see
https://www.fda.gov/media/139145/download (key passage highlighted below)
https://www.fda.gov/media/78495/download
For a specific statement, see
https://www.fda.gov/media/139145/download (key passage highlighted below)
(7/n) There are ways to limit this sample size re-estimation information leakage, but they require measures such as restricted protocols, etc. Thus you get some negatives to compensate for the potential efficiency of the flexible sample size.
(8/n) Group sequential designs look periodically and just ask “is this enough, or should we continue”. So think “at 200 we look, and either stop or continue…”. If you continue, this process continues at subsequent interims (200, 300, 400, etc.).
(9/n) Prior to the trial ending, the sponsor only knows “we didn’t stop”. You cannot back solve the treatment effect. I show an example with the vaccine trials in the attached thread…you get some information, but it is far more limited. https://twitter.com/KertViele/status/1307463136736354308
(10/n) Thus, a design choice can interact with disclosure requirements to protect trial integrity. (sidebar, my experience is that group sequential designs are as or more efficient than sample size re-estimation, but that may just be me)
(11/n) How might we address this in the future? There are tremendous efficiencies in multiple therapies being tested by third parties, for example in platform trials. https://www.nature.com/articles/s41573-019-0034-3
(12/n) These third parties can handle operational aspects internally, under confidentiality (they are now). Thus a sponsor could accurately say “we have no material information and are awaiting platform trial X to fully vet our potential therapy”.
(13/n) Note I am referring to a potential future world. Currently platform trials report similarly in time to other trials, and presumably will continue for the foreseeable future.
(14/14) Obviously sponsors can benefit from press releases, but this need not be zero-sum. Platform trials are very efficient, and a sponsor could effectively get a full readout of their therapy (at 50-75% of the cost) by the time they would otherwise just have a press release.
