On analyzing therapeutic ineffectiveness reports