Is a three-level multilevel model appropriate for my research question?


For a study that is currently in the planning stage we are considering what kind of model would be most appropriate. We are interested in the interactive effect of a drug (vs. placebo) and hormone level (operationalized as point in time in the menstrual cycle, this is a factor with two levels) on the stress response. In the study, 80 women will be divided into two equally large groups of drug or placebo. Every participant is then tested at two points in time in the estrogen cycle. Our outcome variables are various indicators of the stress response to an experimental paradigm. So every participant is tested on two occasions (high and low phase of estrogen cycle) and will either be in the drug condition both times or the placebo condition both times.
My first intuition was that this would be a nested repeated-measures model, with measurement points nested within individuals and individuals nested within treatment groups.
However, there are two concerns I have regarding this conceptualization: First, there are only two treatment groups (drug or no drug), and therefore our level 3 groups would not be sampled from a theoretically infinite number of distinct groups. Second, our level 1 is not longitudinal in a traditional sense as we are not interested in time course per se, but rather in whether the drug's effect on the stress response varies according to phase of the estrogen cycle.
Is this best modeled via a three-level MLM? A cross-classified model?
Any insight you have would be much appreciated.