It is a very well known fact that good many drugs have done very good in small PhII trials and then failed miserably in large PhIII trials.
Furthermore, IMHO, DNDN's trial in question has been designed and executed in a way leaving plenty of room for data manipulation and misinterpretation.