Here's an open question to the statistics mavens out there who...

Here's an open question to the statistics mavens out there who hold that the 2004 AOD9604 trial results to have been nothing more than a statistical fluke:

If the results were nothing but statistical "noise", and not just noise overlying real data, what's the statistical liklihood that the mean of each and every one of the five teated groups is on one side of the placebo mean? Wouldn't that be pretty unlikely if "noise" alone were driving the results?