Look at the standard error

There has been a huge amount of speculation about what happened with the last trial in the USA (batch variation in the USA, contamination of the vehicle with active product, swapping of the US control and active ingredient data, incompetence by the USA CRO, even conspiracy theories like "big pharma has interfered with the trial").

All of these theories have overlooked the most obvious answer which is the difference between the different cohorts and countries is all just due to measurement error and noise. If we look at the data presented by BOT in their last investor presentation we can see that they have included the standard error (SE) values (these are the inverted T's under each bar). These are quite large relatively to the measured reductions - this is just a way of showing that there was a large amount of variability within each cohort.



I measured off the SE values from this figure and created a new figure which makes this variability more obvious. The red regions in my figure below are the approximate ranges that we should expect the results to land if we ran the trial say 1000 times (it is a little more complex than this, but this is a close enough approximation for our purposes). I have drawn a line across (in pink) where you can see that all the values (except 1) fall within what we would expect if the differences are due to random scatter. The only exception is the 5% BID US cohort which shows less reduction than the vehicle control (i.e. it worked worse).



What does all this mean? The null hypothesis in this trial was that BTX 1503 has no effect on inflammatory lesions. Since the results fall within what we would expect by measurement error and random noise we can't reject the null hypothesis. At this point in time the trial results (as presented by BOT) suggest that BTX 1503 is doing nothing at all.

I should add that this conclusion is based on the data BOT has released to the market which may not be correct (for example they may have shown incorrect SE values on the figures presented), but if so this is nearly as bad. If I owned BOT I would be asking management some hard questions at the AGM.