Design of epidemiological studies measuring genital and plasma HIV-1 outcomes: lessons from a randomised controlled trial.