Ann: FDAs OTAT Agrees to Primary Endpoint for Back Pain, page-353

p = 0.05 (usually the minimum standard set by FDA, TGA etc for clinically significant results) meaning there is only a 5% probability that the positive treatment results versus placebo would have occurred by chance. In other words if you ran the trial with same patients and conditions statistically speaking the treatment would show similar positive treatment results versus placebo 95 times out of 100. This is usually the most uncertainty the FDA will accept.

Alternatively, p values above 0.05 are usually labeled not clinically significant. For example a p value of 0.10 means there is a 10% probability that the treatment results happened by chance (too much uncertainty for FDA to accept).

Conversely, a p value less than 0.05 increases confidence further that the positive treatment results are due to the treatment and not by chance alone. For instance, a p value of 0.001 is highly significant and statistically means there was only a 0.1% chance (or 1 in 1000 chance) that the treatment effect occured by chance alone. In other words if you ran the trial again 1,000 times under the same conditions 999 out of 1,000 times you should see a similar treatment effect and only 1 out of 1,000 not showing this treatment effect. This is a very good p value and can be totally relied on by the FDA that the treatment is the reason for the trial results - in other words you can be extremely confident that the result (lets say decrease in mortality) is due to the treatment as opposed to differences between the treatment group and placebo group creating a skewed result.

In short, the lower the p value the better (and it must be 0.05 or lower to get the attention of the FDA to consider the positive trial results as clinically significant).

On the other point, technically speaking p values do not give us the same probability that the next trial will show similar results because you are running different patients under slightly different conditions. However, having said that, if the next trials replicate the same conditions as much as possible (which MSB will aim for) or even better target more responsive patients (e.g. diabetics with high crp or CLBP less than 6 months) then it is very likely that the trial results will be replicated (or improved) if the former trial had a low p value. Essentially you are running a similar trial again, so if the original trial p value was 0.01 for instance, then there is an excellent chance similar results will carry into the next trial (i.e. a similar trial run again with the p value of 0.01 should in theory not differ much from having a 99 times out of 100 possibility of achieving the original result from the original trial).

If you take nothing away from this apart from a p value <0.05 is good and if you can replicate the first trial as best as possible in the confirmatory phase III trial then there is a very good chance that similar trial results will prevail (and if the original trial p value way say 0.001 then you can bet your bottom dollar that you will achieve similar trial results as the first trial).

Hope this helps, albeit a confusing concept. Good luck.