Nice WSJ article. The p value that the deaths were caused by torcetrapib was 0.007 – which is very statistically significant.
However, if there were only two fewer deaths, 80 instead of 82, the p value would have rose to 0.011. That would still be statistically significant under normal circumstances (signified by a p value < 0.05), but above Pfizer's target threshold of 0.01 for their study.
Bottom line: Those two extra deaths cost Pfizer $800 million.
