> we're more likely to be mislead than enlightened by interpreting the results of small trials

Potentially yes, but when you consider anecdotal evidence to be trials with sample size 1, this is still a large improvement.