Our mathematicians (len, chasjac, others?) could give you a
better answer, but I think a sample size of 2,000 in a well
designed trial is actually quite large.

It's been years since I studied this stuff and I don't remember
enough, but there is a formula that can be used to relate sample
size and confidence in the conclusions drawn from the sample.  I
remember one rule of thumb in use many years ago was that samples
above 250 were considered very good.

The required sample size is also related to the study design.  If
you have tight control of all variables, you can use a smaller
sample.  If you don't have tight control, you need a larger
sample in order to drown out more of the possibly non-random
variations.

As you can imagine, running a trial is very expensive.  Sometimes
the drug itself is very expensive - especially given that the
inventor of the drug does not yet have a production line in place
to produce it and is working with small batches made by hand.
But even apart from the drug cost, and often much more than that,
there is the cost of recruiting the patients, counseling them,
administering the treatment, performing tests before and after
treatment, paying for time on expensive scanners or other
equipment, following up, correlating results, writing them up,
getting them reviewed, and so on.  So there is a great incentive
to keep sample sizes as low as possible and still get results in
which we have high confidence.

It can also be difficult and time consuming to recruit patients.
You need a lot of publicity to stir up interest.  You need to
recruit a lot of geographically disbursed study sites - and get
the interest and participation of the doctors in those sites.
Doctors don't always have time to take on new things, and
patients are often skeptical about being "lab rats".  Most
patients couldn't give a hoot about research.  Most want everyone
else to try the new drugs.  For themselves, they want the proven
cure.

The Phase II trial I was in was hoping for a sample of 30 men and
considered that 18 would be enough to get a minimally valid
result.  In the event, they were only able to recruit 11 men, in
spite of the fact that the trial was run in a large metropolitan
area (Bethesda, MD, in the environs of Washington DC) and that
all costs were paid for by the researchers.

The abiraterone trial can probably recruit men more easily
because they are looking for men who really have few or no other
options.

   Alan