Estimating the efficacy of medical abortion

Comparisons of the efficacy of different regimens of medical abortion are difficult because of the widely varying protocols (even for testing identical regimens), divergent definitions of success and failure, and lack of a standard method of analysis. In this article we review the current efficacy literature on medical abortion, highlighting some of the most important differences in the way that efficacy has been analyzed. We then propose a standard conceptual approach and the accompanying statistical methods for analyzing clinical trials of medical abortion and to explain how clinical investigators can implement this approach. Our review reveals that research on the efficacy of medical abortion has closely followed the conceptual model used for analysis of surgical abortion. The problem, however, is that, whereas surgical abortion is a discrete event occurring in the space of a few minutes or less, medical abortion is a process typically lasting from several days to several weeks. In this process, two events may occur that are not possible with surgical abortion. First, the woman can opt out of the process before a fair determination of efficacy can be made. Second, the process of medical abortion allows time for surgical interventions that may be convenient for the clinician but not strictly necessary from a medical perspective. Another difference from surgical abortions is that, for medical abortions, different medical abortion protocols specify different waiting periods, giving the drugs less time to work in some studies than in others before a determination of efficacy is made. We argue that, when analyzing efficacy of medical abortion, researchers should abandon their close reliance on the analogy to surgical abortion. In fact, medical abortion is more appropriately analyzed by life table procedures developed for the study of another fertility regulation technology; contraception. As with medical abortion, a woman initiating use of a contraceptive method can change her mind after some period of exposure and opt out. Also, as with medical abortion, a contraceptive can fail, usually with the risk of failure depending heavily on whether or not the woman follows the protocol for that method precisely. Finally, as with medical abortion, medical conditions may arise that necessitate discontinuing use of the contraceptive method. In both cases, these medical conditions are sometimes open to interpretation or subject to the skill, judgment, or experience of the clinician involved. The appropriate information to collect for a multiple decrement life table analysis of medical abortion includes data on compliance with the protocol, timing of the event of interest (abortion) when it is observable, and, because we argue that these should be regarded as events of interest, a typology of any surgical interventions that are conducted during the woman's participation in the study.