@Science1000 has already answered the bit about the ratio of active to placebo patients in the trial, but another important part of the trial is the randomisation. From what I've read (and this is normal in scientific experiment design) they've done blocks of randomisation, so the blocks are each readout point, and inside those the ratio is the same as the overall trial.
For example:
First 90 patients should be 45:45 if the trial is 1:1 (which we think it is)
This also makes data analysis easier... you shouldn't really have a readout point if the readout is on trial data that isn't representative of the larger trial, because then your trial design is "changing" slightly.
I tried to think of an analogy and the best I can do at this hour is this - It's kind of like if you're making pancakes and your recipe says 3 cups flour and 6 cups of milk, but then half your friends show up early and demand pancakes, so you just mix 1.5 cups of flour and 3 cups of milk, and start cooking pancakes,and then if the late friends come in time you put the rest of the ingredients in and keep going.
OK... so over to the other question on things that could make the Covid trial come back as a fail, along with the risk I personally prescribe to each (but DYOR on these, google phase 3 trials for more info):
- Ineffective product: I give this about a 2 out of 100 chance. It's possible the original 12 patients were some kind of crazy statistical anomaly or that there was a massive human bias in recruiting them, but it's unlikely. The evidence suggests that the cells work, the result was statistically overwhelming.
- Badly designed trial: Also a 1 out of 100 chance. The trial is exactly what the FDA was asking for in the recent ODAC meeting
- External factors (eg. standard of general care increases enough to make the cells statistically insignificant): Really unlikely for the endpoint of the trial, as deaths in developed countries are still happening at a high rate. Possible reason we don't get overwhelming efficacy though. 1 out of 100 for trial, 50 out of 100 for preventing overwhelming efficacy at first readout (I'd accept criticism that I'm being too pessimistic on this one for first readout, but these are my scores, and that's how I see it at present)
- Bad choice of efficacy indicators: Definitely an outside risk, as it's "all mortality" so hopefully no cured patients get hit by a car on their way home from the hospital... it is a common point of failure for phase 3 trials (as we know too well), in this case the choice seems sensible, risk is low... 5/100 maybe?
- Something goes wrong with the trial (eg. trial is compromised or there's defective product): This would be really out of character and the team has successfully run other trials with good process. No risk here.
- Trial can't recruit the desired cohort of 300: Really doesn't seem like this will be an issue looking at infection rates. Risk < 1/100
That's all I have, but let me know if I've forgotten something, or if you see a risk that I've missed. I think the biggest barrier for this product is not efficacy, but production and cost of treatment. It's going to get pumped hard if there is overwhelming efficacy but there may be something cheaper that comes along that could deal with most of the covid patients (like a vaccine or antiviral or plasma therapy) though there's nothing too close yet... the oxford vaccine trial is probably the one I have my eye on most closely. I think there's still be a role to play for MSB even if we do get a vaccine, though, and the general ARDS utility is always going to be there. Exciting couple of months coming up!
(20min delay)
