You could argue that dabigatran did "prove" that it reduced the number of deaths, since the p-value of 0.051 (at the higher dose) just narrowly misses statistical significance.