Although the sample size is getting towards sufficient, there is no randomisation. The first fifty applicants are quite likely to be a nonrandom selection of leukaemia sufferers: the first volunteers may be people who are the most ill, willing to try anything, and most unlikely to repond to treatment.

Although the trials may have been set up double-blind originally, the experimenters decided the data didn't fit with their preconceptions, and manipulated the data so as the analysis came out with the answer they wanted. This defeats the object of the double-blind.