Read on Twitter
1/n I was asked to give an industry statistician's view on A) below. Disclaimer: I do not know about the exact regulations (which might also be region-specific). What I offer is a 1st hand experience of what happens around a trial stopping early. A thread. https://twitter.com/PWGTennant/status/1325795962439217154
3/n I was the trial statistician for this trial: https://www.nejm.org/doi/full/10.1056/NEJMoa1614598
Planned efficacy interim after ~245 PFS events.
1) Data with sponsor, except for tmt assignment (=rando codes).
2) Rando codes with IxRS vendor.
3) Indep. Stat. Reporting Group (ISRG) coordinates.
Planned efficacy interim after ~245 PFS events.
1) Data with sponsor, except for tmt assignment (=rando codes).
2) Rando codes with IxRS vendor.
3) Indep. Stat. Reporting Group (ISRG) coordinates.
4/n If you'd like to know more about the setup have a look at Janet Wittes' slides which she presented at recent EFSPI regulatory statistics workshop: https://www.efspi.org/EFSPI/Events/Regulatory_Meetings/5th_efspi_workshop_on_regulatory_statistics.aspx?&WebsiteKey=0b28fec8-fcd6-4283-b360-0c689a893140&hkey=b26f2559-d1d2-4fb4-b837-8deb3b2759e4
5/n Sponsor sends clean data to ISRG & IxRS vendor sends rando codes. ISRG generates *unblinded* outputs and shares with iDMC. So, ISRG + iDMC approx know results. They meet, discuss, and submit recommendation to sponsor. Call this day 0.
6/n I got this call at 10:30pm. Not much sleep that night 
Initiate
1) submission of rando codes to sponsor,
2) generation of pre-programmed outputs,
3) get iDMC outputs.
Stats team starts reviewing results, then shares with broader project team. If you are efficient, day 5.
Initiate1) submission of rando codes to sponsor,
2) generation of pre-programmed outputs,
3) get iDMC outputs.
Stats team starts reviewing results, then shares with broader project team. If you are efficient, day 5.
7/n Now ISRG + iDMC + sponsor project team (stats, data mgmt, programmers, clinicians, regulatory colleagues, ...) know results. Start involving further folks at sponsor, such as business team, chain of command up to the top. Easily, this can be 100 people by now.
Say, day 10.
Say, day 10.
8/n *All* these people involved had to be prespecified and submitted to authorities - to give them means to check insider training. And AFAIK they actually do check.
Typically, a press release is then issued (say day 10), with topline results, so that everyone is on same page.
Typically, a press release is then issued (say day 10), with topline results, so that everyone is on same page.
9/n Q's remain:
1) Why not add precise numbers to press release? Typically you sit on results to break them at conference. I'd prefer to communicate them earlier in fact.
1) Why not add precise numbers to press release? Typically you sit on results to break them at conference. I'd prefer to communicate them earlier in fact.
10/n
2) Why not write paper along with press release? That would involve even more people (investigators, at least some of them!) and take more time, increasing the risk of insider trading.
2) Why not write paper along with press release? That would involve even more people (investigators, at least some of them!) and take more time, increasing the risk of insider trading.
11/n I hope this 1st hand account sheds some light on the process, and why it is in place. Happy to answer question, as much as I can!
Thanks @ADAlthousePhD for asking and providing a platform for explanation. Often, there are good reasons why things are as they are.
The end.
Thanks @ADAlthousePhD for asking and providing a platform for explanation. Often, there are good reasons why things are as they are.
The end.
