Slight addendum to my above post:

If the null hypothesis is true, you should expect 5% of the studies to have p<0.05.