I think it was the Millenium Women Study (US) that used the premarin and older subjects. It was good in the way it was random - controlled etc. However, they chose women who were past menopause and not having any symptoms. This was so that the symptoms did not interfere with results, but women without symptoms do not usually take hrt, and the average age was rather high.
The million women study (UK), I think, was a questionnaire study following women who came for breast screening during a certain period. The criticism of this study is the self selection nature of the people in the study - ie they came forward for screening and also agreed to do the questionnaire (about 70% did). Therefore, a higher proportion of those studied would have some reason to be concerned about developing breast cancer than the general population. You can't do random controls etc with this sort of investigation.
They were both good in that they managed to get a very large number of women, but there are compromises to be made in how studies this size are carried out and which ones are taken will affect the results in some way.