Twenty people is a pathetically small sample size. Medical trials need hundreds to be reliable. Small samples are unrepresentative. Imagine taking a random sample of four people from a well-stirred vat of people. One third of the time, you will pull out two girls or two boys rather than one of each. Even in a perfectly random sample, the small sample size means that the sample is terribly unrepresentative of the population at large at least one third of the time. And just consider all the other possible variables (age, ethnicity, etc.) that might also be important.

There was no proper placebo control (and therefore by definition, no double blind, as the experimenter is very likely to know which group isn't being given the pills).