who's responsibility is it?

Women of child bearing potential weren't always allowed to participate in clinical trials of new drugs.  That has only been permitted and encouraged within the last 20 years.  The reasoning for including such women is obvious because it is important that drug evaluations include young women as well as young men.  It is not conceivable that a new medication could be approved without including data from women 18-50 or so. However by allowing women of childbearing potential to be included, the FDA is essentially exposing pregnant women to new drugs with unknown mutagenic or teratogenic effects.

For this reason women of childbearing potential are required to use an effective form on contraception for the duration of participation in a trial (and often for a period of time before and/or after completion as well).  This requirement is explicitly outlined in study specific informed consent document that all participants must sign before participating in a trial.  To me, this should serve as documentation enough that a subject is aware of the unknown risks of a new drug to a fetus, but if a site reports a pregnancy one of the first things asked by the sponsor is: was the subject practicing an approved method of contraception and was she educated on the necessity of it?

I don't think that any investigator would willingly expose a woman wanting to become pregnant to a new drug because two lives could be at risk, but a lot of times the sponsor takes their requests of documentation too far. For example I have heard of studies where the site must contact participants on a regular schedule during the study to ask if they are still using contraception and if they are pregnant.  This to me is overkill.  It should be sufficient for a subject to sign consent (maybe initial a place saying they agree to use adequate contraception, for good measure) and then the site to document the type of birth control used.  Outpatient studies are supposed to be exactly what they say they are and subjects should be permitted a least some bit of freedom during a trial.  What ever happened to trusting a subject when she says that she agrees to use contraception and not get pregnant during a study?

As long as sites go in to the appropriate amount of depth in the informed consent discussion and make females aware of the requirements to participate, then the signature of subjects should be sufficient documentation of the necessity of using birth control during a trial.  That signature is good enough to prove a subject reviewed the trial duration, obligations, drug side effects, and randomization scheme, so why can't it suffice for the need for adequate contraception?