There's a lot of misinformation about how trials work in the announcement thread... here's a few things that people have said:
"This announcement is good because it means they are seeing some results and want Mesoblast to continue!"Not the case, it just means they didn't stop the trial for overwhelming efficacy ("we can already tell this works") or overwhelming futility ("based on what we've already seen there's no way this works!") it could continue with no observed effect, or even a slightly worse effect than the control group, as long as there is still a path to efficacy. It's likely that we are on the way to a good result, but the continue recommendation doesn't MEAN anything about the results other than it wasn't amazingly good or horribly bad. It's not normal to have either result at a 30% read out, so this announcement is bland. People in the other thread like
@DocMcstuffins and
@Techinvestor have tried their hand at explaining this to people but they're having limited success.
"This means we are certain to get overwhelming efficacy at the 45% readout"Based on the tables in the first post, which are the best approximation I have so far for what could happen (until we do some Bayesian stuff at least), depending on your assumptions about control group mortality it did look likely that there wouldn't be overwhelming efficacy at the 30% mark, but much more chance of achieving it at the 45% mark. This is no guarantee though, and if the control group mortality is actually around 30-40% and the treated group mortality is 15% then at p < 0.001 there's about a 50-50 chance of significance in a t-test. I would be cautiously optimistic about the 45% readout, but quite optimistic about the 60% readout if the cells perform as per the trial. In general, performance in blinded trials is lower, so if you gave me $20 to bet on a result I'd put it on the 60% readout based on available information, or maybe even efficacy at the end of trial. If control group mortality is 50% or higher then there's a good shot for overwhelming efficacy at the 45% readout. (Original tables reproduced below for 30% mortality and 50% mortality). Note: at 85% survival as per the first trial, we'd expect 10.2 people to die in the treated group at the 45% readout, and 13.5 people to die in the 60% readout... so it becomes much easier to get a result as the power of the trial increases. Fully acknowledge that this is only a guide and Bayesian stats would have different requirements here, but it shows how significance changes with power in a nice way.
Original table (assuming 30% control mortality):Summary table (assuming 50% control mortality):"I do not see this announcement as implying "overwhelming efficiency" has not been reached. It may actually of been achieved but they need more readouts to give greater confidence these results were not due to something such as contamination, mistakes, or fraud."No, it absolutely means overwhelming efficacy hasn't been reached at this readout. I think people misunderstand how hard it would be to achieve overwhelming efficacy at a 30% readout. Even getting it at 60% would be amazing.
Hope this helps clear things up, but it's understandable people are trying to make sense of this and asking all kinds of questions. I thought we would see a bit of a drop after this announcement. Clearly the share market is smarter than I gave it credit for.
(20min delay)