As a result of the growing availability and acceptance of emergency contraception (EC), new issues need to be considered when designing a barrier contraceptive effectiveness trial. The assumption behind these new issues is a general objective to determine how well a barrier method works.
Before the EC era, the primary objective of new barrier method effectiveness studies was to measure the typical use pregnancy probability for a new barrier method when used alone. Another possible objective was to compare this measure to the typical use pregnancy probability of an existing barrier method. For example, trials of the cervical cap versus the diaphragm, the sponge versus the diaphragm, and the Reality female condom all assessed what I would like to call the method specific effectiveness. The pregnancy probability estimates from these studies are best interpreted as the probability of becoming pregnant for women who plan to use the barrier method alone to prevent pregnancy. In the EC era, we may still be interested in determining the method specific effectiveness.
We now might also be interested in measuring the combined method effectiveness of a barrier method with EC backup. This can be estimated by the typical use pregnancy probability for women who plan to use the barrier method with EC backup if needed, for example, after a malfunctioning of a device or after an episode of nonuse. This typical use pregnancy probability would be interpreted as the cumulative probability of becoming pregnant among women who plan to use the barrier method with EC backup if indicated.
Ideally, we would estimate precisely both the method specific effectiveness and combined method effectiveness for a variety of potential users of a new barrier method. We cannot assume that the availability of EC will reduce the risk of pregnancy by the same amount for different barrier methods. Two barrier methods that are equally effective when used alone may have different levels of effectiveness when used with EC as a backup. This may be the case because problems may be more easily recognized for one method than another, and consistency of using the two methods may vary as well. For example, a new barrier method that has a better method specific effectiveness than the standard latex male condom might have a worse combined method effectiveness than the standard latex male condom.
The first question raised by the availability of EC is whether measures of both the method specific and the combined method effectiveness are needed to evaluate sufficiently a new barrier method. And if both are needed, is one more important than the other? Some may think that the answer to these questions should vary by method. For example, it may become standard practice to provide all acceptors of barrier methods that must be obtained by prescription with a prescription for EC as well. If this is the case, then combined method effectiveness may be of greater interest for a prescription-only barrier method. However, for a method that women will be able to obtain over the counter, there may be more reason to measure method specific effectiveness. Regardless of which measurement is the primary objective for a study, design and implementation issues arise that we did not have to deal with before the EC era.
Several issues arise when the primary objective is to assess the method specific effectiveness of a new barrier method. The first issue is whether it is ethical to ask women enrolling in a barrier method study to agree not to use EC. Historically, women enrolling in barrier method trials have agreed not to use other methods of contraception besides the assigned barrier method, even though use of an additional method -- for example, use of a male condom in a diaphragm trial -- would provide extra protection against pregnancy.
I'm not aware that the ethics of asking women to rely solely on their assigned method was ever questioned prior to the EC era. This new ethical concern may be due to the fact that EC gives a woman a chance to reduce her risk of pregnancy after a method malfunctions or after nonuse.
Assuming it is considered ethical to conduct a trial where women are asked not to use EC, then we must decide what information about EC should be provided during the informed consent process. We need to make sure that women understand what they are giving up by agreeing to rely on the barrier method alone without EC backup.
Once women have given fully informed consent not to use EC and have enrolled in a method specific effectiveness trial, issues related to EC still arise. Even if a woman has agreed not to use EC during the trial, she may ask for EC during the trial, for example, because the condom she was assigned to use broke. What should she be told? What if a woman reports that the condom broke yesterday, but she does not request EC? Should clinic staff be specifically instructed not to bring up EC in this case in a method specific effectiveness study? Should the counseling be the same if the woman reports an episode of nonuse but doesn't request EC? Certainly all investigators in a multicenter study should be asked to approach these situations in the same manner.
Even if EC is not supposed to be used in a study, some women will likely use it anyway. What data on EC use should be collected in a method specific effectiveness trial? And if EC use is reported, how should it be handled in the primary analysis? One option is to analyze just the data as is, that is, not do anything special because EC was used. The resulting typical use pregnancy probabilities would be interpreted as the cumulative risk of pregnancy among women who intend not to use EC.
Another option is to count women who use EC as having a fraction of a pregnancy, assuming no protection was provided by the barrier method on that day. The value of the fraction is set equal to the expected probability of pregnancy given the cycle day of intercourse that led to the EC use. This calculation involves making a number of assumptions, but it provides some estimate of what the pregnancy probability would have been had EC not been used. Simply excluding cycles of EC use from the analysis or stratifying the data by whether or not EC was actually used is not appropriate, as the background paper explains in more detail.
So, even if one decides to do a study in which women are asked not to use EC, one still has a lot of new issues to think about when developing a study protocol. Now I will turn to issues that arise when planning a study of combined method effectiveness.
A goal to measure combined method effectiveness implies that women would be asked at enrollment to agree to use EC whenever certain events occurred. If they do not use EC when indicated, their behavior would be considered incorrect method use. A question then arises: Should the trial exclude women who indicate that they would not use EC? One would need to exclude such women if the study objective is truly to assess the risk of pregnancy among women who consider the barrier method plus EC backup when indicated to be their primary method of contraception.
However, if the study enrolls women who do not expect to use EC even if it is indicated, the instructions for use should clearly describe EC as an option rather than as a standard backup. The typical use pregnancy probabilities estimated from a study where EC backup is considered optional might be thought of as a weighted average of the typical use pregnancy probabilities that would be obtained from a method specific effectiveness trial and from a combined method effectiveness trial.
Another question to consider is how to provide EC to the participants. At least three options are possible. First, the protocol might require that all women be provided an EC packet at admission and a replacement packet when their supply is used. Second, all women could be offered EC at admission -- they need not accept it -- and could be provided a packet any time they request one. Third, EC might be distributed only when a need arises. That is, women might be given the backup method only if they report nonuse or a malfunctioning of the barrier method.
If the study goal were to estimate the risk of pregnancy under optimal conditions for adherence, the first option would be the best approach -- unless possessing an EC packet changes one's consistency of barrier method use.
However, if the goal is to estimate the probability of pregnancy in normal-use settings, whether the method is available over the counter or only by prescription may determine whether providing EC in advance is the best. For a barrier method that will probably not be available over the counter, advance provision of EC during a trial may result in a study with too optimistic an estimate of pregnancy probability. That is, when outside the trial setting, women using this barrier method will need to make a special trip to a health-care provider to obtain EC backup.
A related question concerns the instructions that women should receive regarding when to use EC. Should they be instructed to use EC after every problem with a device or only after specific kinds of problems? Should they be instructed to use EC after every unprotected act or only if these events occur mid-cycle? Should women be asked to consult the investigator in all cases to help determine whether EC should be used? Instructions related to these questions need to be standardized across centers.
In a combined method effectiveness trial, the primary analysis is straightforward because EC is considered part of the method being evaluated. But there may be some interest in trying to assess the method specific effectiveness as a secondary objective of a combined method effectiveness study. There are at least two possible approaches.
One approach is to count each use of EC as a fraction of a pregnancy as discussed earlier. This approach will provide a measure of what the typical use pregnancy probability would have been had EC not been used.
A second approach could be applied if the trial includes women who at admission indicated that they would not use EC. Then we can estimate the typical use pregnancy probability separately for this subgroup. A typical use pregnancy probability estimate for this subgroup would be interpreted in the same way we would interpret a typical use pregnancy probability obtained from a method specific effectiveness trial.
It is important to note that these two approaches might lead to different typical use pregnancy probability estimates, for several reasons. The fractions of pregnancies needed for the first approach might come from an external population of women whose average fecundity is different from the study population. Also, women who do or do not plan to use EC may differ in coital activity, consistency of product use and average fecundity.
Finally, when the effectiveness of a barrier method is evaluated in a combined method effectiveness trial, how should the product label report effectiveness?
The last question on the agenda for this afternoon asks the group to reflect on what role the true efficacy trial that Markus Steiner described might have in the evaluation of new barrier contraceptives. Might studies such as the true efficacy study provide sufficient data to characterize a new barrier method?
return to table of contents
