In a preponderance of the evidence test, liability must be more likely than not under the evidence.

Instead, judges, non-experts in statistics, except by chance, must find that the chances of injury due to the allegedly bad drug at least doubles the risk of injury or death.

No one seems to doubt that Merck mislead the FDA and that Vioxx has hurt or killed many people.  But because the studies do not meet the Havnar statistical test, there is no liability for harm cause.

5 times as many heart attacks is not enough.

Merck & Co., Inc. v. Garza, 347 S.W.2d 256 (Tex. 2011).