|
Reproductive Health |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
The morning included five presentations on recent or ongoing barrier contraceptive effectiveness trials, summarized below. Comparing Polyurethane and Latex Condoms Comparing Non-latex and Latex Condoms Lea's Shield and FemCap Trials Efficacy Trial of Spermicidal Agents Comparing Polyurethane and Latex Condoms-- Terri Walsh, California Family Health Council The California Family Health Council conducted a comparative contraceptive efficacy study, with 805 couples randomly assigned to use polyurethane or latex male condoms. Couples used the study condom exclusively for six months but were also informed about emergency contraception (EC). Couples used EC only 33 times in the study; they were kept in the study results. To incorporate EC use into reported pregnancy rates, we added an estimated probability of pregnancy that would have occurred had EC not been used. Corrections for even this small number of uses increased the reported typical use pregnancy rates slightly in almost every category of measurement. The study, conducted from 1994 to 1996 with funding from the National Institutes of Health, measured performance (breakage and slippage), efficacy, usage and acceptability. Less than 5 percent were lost to follow-up, although only 65 percent of the couples completed six months in the study. The major reasons for dropout were personal, such as breaking up with a study partner or moving. Participants were recruited through advertisements in local newspapers, entertainment guides and radio spots. The study population was representative of monogamous couples who use condoms for contraception. The average age was 27, with 15 years of education. About two-thirds were white, 16 percent were Hispanic, 6 percent were African American and 12 percent came from other ethnic groups. About two-thirds were married or living together. The average length of the relationship was 3.4 years. Most participants were experienced condom users. Only 8 percent of men and 3 percent of women were inexperienced (i.e., 10 or fewer uses prior to study). About 3 percent of the women had used EC before entering the study. When designing the study, we faced an ethical dilemma. Study subjects are able to detect condom failure (breaks or slipoffs) and could use EC to reduce their risk of pregnancy. Should information about EC be withheld? Should subjects be asked to refrain from using EC? The answer to both questions was "No." So the question was: How would access to EC be incorporated into the study? We wanted to provide sufficient information about EC so the women could use it, but we wanted to distinguish between the study requirements and having access to EC. We wanted to create a neutral environment that neither discouraged nor promoted EC use, so that participants could decide what was best for them and would fully report what they did. Information and access to EC were provided during the study in the following way. We distributed a handout on EC use at the enrollment visit and offered to answer questions. The handout explained when EC might be used, risk of pregnancy relative to time in the cycle, side effects and how to contact us if they decided to use it. When asked about EC, we arranged for participants to receive EC at several clinics and private practices, and we paid for its use. We did not determine the protocols used regarding EC use. We did not suggest or recommend EC use, nor did we try to dissuade participants from using EC. Participants were not discontinued from the study after EC use. Information was gathered primarily through diaries kept by the couples. When couples requested information about EC use, we recorded as much information as possible. The diaries contained easy-to-use, pictorial icons for condom use, unprotected intercourse, condom breaks, slipoffs and discomfort. About 90 percent of the couples submitted diaries. Participants had abundant occasions for taking advantage of EC. In the study, there were 37,360 acts of intercourse using the study condoms. But participants did not use a condom in 2,755 unprotected intercourses outside of menses. Also, 418 condoms broke, and 181 slipped off. So, what happened on all of those occasions when pregnancy could occur? To our amazement, not much. Participants used EC only 33 times: two for the 2,755 unprotected intercourses, 21 for the 418 breaks, and 10 for the 181 slipoffs. In the analysis of pregnancy rates, we adjusted for use of EC. We started by estimating the ovulation date for each cycle of EC use. We calculated the average cycle length based on either study data if available or by history. We assumed that ovulation occurred 14 days prior to the expected menses. Then, we assigned a probability of conception based on the number of days before, on, or after ovulation that the exposure occurred, using estimates reported by Wilcox et al. Arriving at this estimate of probability was difficult. The variation in each woman's cycle affected our estimates, along with the variability in the luteal phase (although assumed to be 14 days, it can range from 10 to 16 days). In figuring the earliest and latest dates each woman might have ovulated, we arrived at ranges from six to 20 days, and even longer in a few instances. Hence, we felt very uncomfortable about assigning a specific probability of conception for each EC use. For this reason, we calculated the average probability of conception among all 33 EC uses in the study. We then reduced this number by the probability of miscarriage within the first six weeks of pregnancy, where we would not have a clinical pregnancy. As Table 1 shows, we calculated the average probability of conception as 0.1, with a factor of 0.76 to adjust for early miscarriages. Together, this is an adjustment of 0.076 for the probability of pregnancy with EC use. Thus, in the latex group, where there were eight EC uses, the estimated number of pregnancies that would have occurred without EC use was 0.6. The estimated added pregnancies for the polyurethane group was 1.2, for a total estimated added pregnancies to the study of 1.8. Put another way, without the availability of EC use during the study, there would have been an estimated additional 1.8 pregnancies.
* There were a total of 33 EC uses, but only 24 occurred in the cycles contributing to the pregnancy analysis. The next challenge was to incorporate this calculation into the typical use and consistent use pregnancy rates for the entire study. Table 2 shows pregnancy rates for the entire study before and after incorporating the risk associated with EC use. To obtain the adjusted rates, we added the 0.076 adjustment for probability of pregnancy with EC use into the appropriate time interval within the life table calculations. With the overlap of the 95 percent confidence interval, none of the differences in pregnancy rates for the two condom groups were statistically significant. Table 2. Impact on Pregnancy Rates
It is interesting to note that the typical use rate for the latex condom adjusted for EC use was the same as the unadjusted rate. This is due to two factors: There were few EC uses in the latex group, and the denominators were different for these two rates. We excluded all data subsequent to EC use when calculating the unadjusted rate, but used all data in the adjusted rate. If the excluded cycles had a lower risk of pregnancy than cycles contributed by women who had never used EC, the unadjusted rates would include a larger proportion of higher risk cycles than the adjusted rates. In summary, we observed three major findings regarding EC use in a barriers trial. First, participants rarely requested EC appointments when merely provided with an information sheet on EC. Second, participants were more likely to use EC following a condom failure than following unprotected sex. And third, difficulties in determining the day of ovulation of women who used EC make it difficult to determine what the EC correction factor should be.
Comparing Non-latex and Latex Condoms-- Dr. TMM Farley, UNDP/UNFPA/WHO/World Bank Special Programme of Research, Development and Research Training in Human Reproduction, World Health Organization (WHO) Study Background In early 1996, we conducted a pilot project for a randomized study comparing contraceptive effectiveness of non-latex and latex condoms. The project sought to establish whether acceptability and slippage/breakage studies in selected volunteers could translate to differences in pregnancy rates between the different types of condoms. If this was the case, then equivalence in slippage and breakage rates could be used as a surrogate for equivalence in actual clinical performance in family planning clients. This would reduce the requirement for large scale, difficult and costly contraceptive effectiveness studies whenever a new condom material or innovative condom shape was developed. The study populations had to represent as closely as possible normal family planning clients who elected to use condoms as their method of contraception. Recruitment was through family planning clinics that provided services to a sufficient pool of condom users annually. Volunteers would remain in the study up to a maximum of 12 months, or until they no longer wished to participate, chose to use another contraceptive method, or chose to use only non-study condoms. Since pregnancy was the primary study endpoint, women were the study unit, though it was expected that the majority of volunteers would be in a stable relationship with a single partner. The pilot study was conducted in three locations with diverse social and economic situations: Chengdu (China), Sagamu (Nigeria) and Manchester (United Kingdom). What About Emergency Contraception? Investigators in Manchester raised the question of EC. Their family planning clinic served mainly a university population, and EC was routinely promoted and provided as part of normal clinic services. They felt that recruiting study volunteers who were unaware of EC was impossible, and that it was not ethical to deny EC if a volunteer requested it following a condom break, slip or nonuse. The Manchester site was participating in a WHO study comparing the Yuzpe and levonorgestrel EC regimens, so awareness of EC may have been higher among clients there than in other parts of the United Kingdom. We felt that EC use had to be addressed for any site in the United Kingdom and possibly other countries too. Even if EC had not been promoted in other participating countries in 1996, we anticipated that the situation might change during the study. U.K. practice at the time of the study was to provide EC through family planning clinics or general practitioners (GPs). While we expected the majority of EC requests would come directly to the study clinic, it was possible that a volunteer might request EC from another source, such as her GP or another family planning clinic, if she were away for a weekend or on vacation. Our study instruments therefore had to record EC requests at the clinic or from other sources. We decided to gather this information through the scheduled follow-up interviews, consulting the daily diary of menses, acts of intercourse and condom use, if available. Calculating an EC-Adjusted Pregnancy Rate Because of potential use of EC during the study, we made the rate of EC requests an important additional study endpoint. Reasons for requesting and using EC could completely mask any differences between study groups. For example, a higher rate of no condom use or of condom breaks in one study arm may not translate into differences in pregnancy rates if EC were used. Similarly, even if there were no differences in the rates of no condom use or slips and breaks, differences in the rates of EC requests may indicate that the reliability of the study condoms were perceived differently by the volunteers. Methods for calculating the pregnancy rates in the absence of EC use are readily available. The number of pregnancies (n) occurring in m months of observation follows a Poisson distribution with E(n) = V(n). Thus, the estimated standard of error of the unadjusted pregnancy rate r = n/m is given by In order to incorporate EC use into the analysis of the primary endpoint (pregnancy), we defined a new statistic, the "EC-adjusted pregnancy rate." For each EC request, we added a fraction of a pregnancy corresponding to the probability of conception on the day of the cycle when the act of intercourse occurred. This provided an estimate of the number of pregnancies that would have occurred if EC had not been used, and gave us a method to adjust for possible differences in the rates of EC use according to study group. The EC-adjusted number of pregnancies We also wanted to calculate a standard error and a confidence interval (CI) for the EC-adjusted pregnancy rate. We calculated the standard error and CI treating the r' as if it were a Poisson random variable, but this may not be appropriate. Since the pi are fixed weights, n' is a weighted sum of Poisson variables and does not follow a Poisson distribution. To calculate the EC-adjusted pregnancy rate, we also had to determine which conception probabilities to use. A number of authors have provided estimates of conception probabilities according to the day of the cycle, but when we developed the analysis strategy, no consensus on the best rates had been established. We elected to use the smoothed rates of Dixon et al. Revised conception probability estimates computed by formally combining data from all available studies have recently been published by Trussell et al. For a randomized comparative study such as ours, the exact choice of conception probabilities would make little difference. In contrast, the exact choice of probabilities would have more of an impact on studies that provided an unbiased estimate of the true, or absolute, effectiveness rate, or studies that compared condom effectiveness rates in different groups, with EC available to some but not other groups. Pilot Study Results The pilot study involved 90 women followed for a maximum of three months. A total of two pregnancies and five EC requests were recorded in 253.0 woman-months. The five EC requests resulted from reports of two condom splits, one break, one perceived leak, and one intercourse in which no condom was used. The five acts of intercourse occurred on cycle days 4, 8, 13, 16 and 21, or days -10, -6, -1, +2 and +7 relative to the estimated day of ovulation. Table 3 shows the conception probabilities compared to the estimated day of ovulation. The Dixon conception probabilities are: 17 percent on day -1, 5 percent on day +2, and zero on the other days on which intercourse occurred. This results in an estimated 2.22 EC-adjusted pregnancies (95 percent CI: 0.3 - 7.7) or 10.5 EC-adjusted pregnancies per 100 woman-years (95 percent CI: 1.6 - 36.5). Calculations that used the revised conception probabilities from Trussell et al. yielded 2.3 EC-adjusted pregnancies, corresponding to 10.9 EC-adjusted pregnancies per 100 woman-years. Table 3. Conception Probabilities and EC Requests by Day of Cycle
Anticipated Problems for Main Phase of Study With the limited number of pregnancies and EC requests observed in the pilot study, little further analysis was possible. However, in the main phase of the study, we plan to recruit 3,000 women and observe 25,000 months of condom use. Currently, EC is only readily available in family planning services in two centers that would each recruit 300 women during the main study. However, EC may be introduced in a third center of similar size. If the rate of EC use in the pilot study is repeated (five requests in 71 woman-months), we would anticipate approximately 500 EC requests. This amount of EC requests would result in a number of problems that will need to be resolved:
Future Plans for Main Phase of Study The main phase of the study is currently on hold pending funds and availability of suitable condoms. All study instruments have been revised following the pilot study and the main phase could be launched at short notice. We have additionally modified the study to collect detailed slippage and breakage information on all volunteers. Since one of the objectives of the study is to demonstrate whether differences in slippage and breakage rates predict differences in pregnancy rates, it is essential that the slippage and breakage information be obtained in the same study population. We intend to collect details of the first five condoms used at the start of the study and the first five condoms used after the six-month follow-up visit -- a double-nested slippage and breakage study. We have not finally decided which condoms to include. Lea's Shield and FemCap Trials-- Dr. Christine Mauck, Contraceptive Research and Development Program (CONRAD) In the two efficacy studies I will discuss, participants were told to use only the study method and implicitly agreed not to use EC. In the studies, four cases of EC use were reported among more than 1,100 women. Data from these four women were included in the final results because the most appropriate way to handle such data had not been established. This dilemma is in part what prompted this meeting, since FHI was providing considerable assistance to us in data analysis. Of the coital acts, about 7 percent involved incorrect method use or no contraception, situations where women might want to use EC. Both of the trials were designed to calculate typical use pregnancy probabilities for the barrier method alone. Both studies were randomized and fairly large. The Lea's Shield study, with 300 women, compared efficacy among women using this device with and without spermicides. The FemCap trial, with about 840 women, compared efficacy of the FemCap device to a diaphragm. Lea's Shield and FemCap are mechanical barrier devices for use by women, similar in some respects to cervical caps and diaphragms, but each with distinctive design features. Is it ethical to ask women to agree not to use EC? If so, what are the key elements of an informed consent process if women will be asked not to use EC? We, in essence, did ask women not to use EC, although we did not go into any detail about it. In the Lea's Shield trial, the consent form read: "I must be willing to use Lea's Shield as my only form of contraception for six months." The FemCap form read: "I will use [the assigned method] as my only means of contraception." The possible use of EC was not specifically addressed in either trial's consent process. The trials were designed and initiated in the early 1990s, before EC was widely known or available. As awareness and availability increases, I think EC should be specifically addressed. In the two trials, four women reported using EC -- one in the Lea's Shield study and three in the FemCap trial. Some detail on each of these four can provide a flavor of what can actually happen in clinical trials. The woman in the Lea's Shield trial had sex 28 times in 40 days during the study. All uses were "perfect," meaning the device was used alone and correctly according to the woman's coital log, except for unprotected sex for the 23rd act. This act was on day 14 of the cycle, which would normally be near the day of ovulation except that her previous cycle was 42 days, not 28. She started EC the day after the unprotected sex. Three weeks later, she had a negative pregnancy test. She was kept in the analysis, without any adjustment for a probability of pregnancy from the EC use. This might be handled differently in the future. In the FemCap study, subject #644 had sex 102 times in 216 days with 83 perfect uses, 16 other method uses (condom), two uses with the device upside down, and unprotected sex on the 15th act with no EC use. One of the times she used a condom (31st act), she also used EC the same day. It is not clear why she used condoms so much or why she used EC this once. The 31st act occurred on day 15 of her cycle, which might have worried her. (Her previous cycle was 30 days.) Twelve days after taking EC, she had a negative pregnancy test. She completed the study and her data were kept in the analysis. Again, this might be handled differently today. Subject #2105 in the FemCap study had sex four times in eight days. We have no coital logs after the fourth act. All four uses were perfect, but the device was found to have sperm in it on the fourth act. (She wore it to the clinic, per protocol, and it was checked for sperm after the fit was found to be suboptimal.) She started EC use on the day of this act, which was day 19 of her cycle; her previous cycle was 30 days. Twelve days later, she had a negative pregnancy test. She discontinued the study 54 days later because of inability to fit the device. She probably should have been discontinued at the time of her EC use. She was kept in the treated population, which was used to calculate pregnancy probabilities, but not in the comparison group with the diaphragm. She was excluded from this group not because of the EC use but because of the fitting issue. Subject # 2116 in the FemCap study had sex 88 times in 190 days. Of these, 81 uses were perfect, five were other method uses (condom), one involved unprotected sex (no EC use), and one involved removing the device early. When the device was removed early, she used EC three days later. This occurred on day 16 of her cycle; her previous cycle had been only 16 days but the cycle before that had been 30 days. She completed the study the day EC was prescribed. She had a negative pregnancy test 30 days later. She was also kept in the analysis. In the two studies, 7 percent or less of the women either used no method or used the method alone but incorrectly. These are the most likely reasons women would be interested in using EC. If a trial were incorporating information on EC use for a larger portion of women than we had, detailed coital logs might be useful. But in our studies, we determined that overly detailed logs would be difficult to use. Women filled out diary cards, and the study coordinators transcribed that information onto coital logs. We chose not to collect some data, such as the time of insertion and removal and the number of acts during a single insertion. This information would have been helpful but would probably have included many additional data problems, given the number of problems observed on the simpler coital logs.
Efficacy Trial of Spermicidal Agents-- Dr. Elizabeth Raymond, Family Health International (FHI) FHI is currently conducting a randomized trial testing five spermicidal products. The objectives of the trial are to measure and compare contraceptive effectiveness, safety, acceptability and product use. Eleven research centers in the United States are collaborating on the trial, which aims to enroll a total of 1,800 women. Recruitment began in June 1998. The study is scheduled to take three years, which includes 29 months of recruitment and seven months of follow-up for each woman. As of September 25, 1998, 125 participants had been enrolled, and so far, none had requested EC. The study is funded by the National Institutes of Health. Each participant in the trial will make four scheduled visits to the research center (a screening and admission visit and three follow-up visits) and three phone calls to the center between the follow-up visits. A highly sensitive pregnancy test will be administered at each visit and before each phone call. The women will keep daily coital and menstrual diaries. The data collected on the diaries include: whether bleeding occurred, whether it was the first day of menses, and for each coital act, the time, method used, and minutes between spermicide insertion and coitus. The study protocol includes specific policies regarding EC use. Each candidate must agree to use the assigned spermicide as her only method of contraception for 30 weeks, except for emergency contraceptive pills (ECPs). Women are not required to use ECPs at any time. ECPs will be provided free of charge to participants when needed and requested. ECPs will not be provided in advance, that is, upon admission to the study or during follow-up visits before any indication of need. The indications for ECP use are: no method used, incorrect use (e.g., spermicide was not inserted long enough before coitus), method did not function properly (e.g., the spermicide tablet was found not to have dissolved), or other incidents in which the investigator judges that EC use may be needed. If EC is needed, the investigators will provide either the Yuzpe regimen or a levonorgestrel-only regimen. The Yuzpe regimen is four active Nordette pills to be taken within 72 hours of unprotected intercourse, followed 12 hours later by a second dose. For the levonorgestrel regimen, 20 Ovrette pills are to be taken within 72 hours after unprotected intercourse, followed 12 hours later by a second dose. At admission, we will collect data on whether the woman thinks she will use EC if an indication arises. We will also collect information on ECP requests and provision, including the date, indication of need, prescription, date and time of coital act prompting need, and dosage provided. After each use of EC, similar data will be collected, including the date, time of use and side effects. The planned primary study analysis will estimate the "combined method effectiveness" of spermicide backed up by ECPs, if the woman chooses to use them. This analysis will ignore use of ECPs during the study. If a woman who uses ECPs becomes pregnant, she will be counted as having a pregnancy. If she does not, she will be counted as not having a pregnancy. A secondary analysis will be conducted to estimate the effectiveness of spermicide alone. This analysis will count each ECP use as a fraction of a pregnancy, similar to the method used by the California Family Health Council in their condom study. The fraction of pregnancy will be equal to the probability of clinically detectable pregnancy after unprotected intercourse on the cycle day of the act leading to the ECP use. Another secondary analysis will be conducted to estimate the effectiveness of spermicides alone. This analysis will exclude women who say at admission that they intend to use ECPs if they have a problem with the spermicide.
True Efficacy Study-- Dr. Markus Steiner, Family Health International (FHI) Participants in the conference received a copy of a manuscript summarizing this pilot study (Steiner et al.). A "true efficacy" study design is a new approach to measuring how well a barrier method really works to prevent pregnancy. The true efficacy approach may provide a better measure of how well a barrier method works than the measures obtained from a traditional contraceptive trial. We have successfully tested this approach in a small pilot study. We are all familiar with the traditional contraceptive trial of barrier methods. A cohort of women is enrolled for a prescribed period of time, usually six months or 12 months. Women are then randomized to use either a standard barrier or a new barrier. The main outcome is the relative difference in the pregnancy probability of the different methods at the end of the study. The absolute pregnancy probabilities are very difficult to interpret across studies because the design and analysis of studies have changed over the years. For example, today's studies include more sensitive and frequent pregnancy testing and hence are more likely to measure pregnancies that may have gone undetected in past studies. As a result, the absolute pregnancy probabilities may be higher than they would have been in past studies. However, even if all studies were designed and analyzed in exactly the same way, the absolute pregnancy probabilities would still be difficult to interpret because participants differ along factors that influence the risk of pregnancy. The first factor is the couple's ability to conceive. This factor will vary depending on the couple's age, history of STDs and other health-related variables. Even if a couple is very fecund, if they are having infrequent intercourse or are not having intercourse at the right time of the cycle, they will not get pregnant. So the second factor influencing the pregnancy probability is frequency and timing of intercourse. The third factor is the degree of compliance -- how correctly and consistently the method is used. An example of how participant characteristics can provide a misleading absolute pregnancy probability is a study from the 1970s. In this study, participants used N-9 condoms for two years and reported a 12-month typical-use rate of less than half a percent. A closer look at the men enrolled shows just how misleading the low absolute probability is. Half of the men were over 40 years old, and 20 percent were over age 50. The age of the female partners is not known, but we can assume they were similar in age and were no longer very fecund. Clearly the low pregnancy probability in this study does not demonstrate that N-9 condoms worked. Rather, the couples enrolled most probably had a low ability to conceive and were not having frequent intercourse. In addition, this study only counted pregnancies that ended in a live birth, further lowering the pregnancy probability. Even if the couples had not used any form of contraception, they may have had a similarly low pregnancy probability. The concept of how many women would have conceived had they not used any method is central to our new "true efficacy" approach. The proportionate reduction in pregnancy is a measure of effectiveness that takes this concept into account. We sometimes refer to this proportionate reduction as "the prevented fraction," which is shown below:
This fraction is the true measure of how well a barrier method prevents pregnancy if it is used correctly and consistently. An example from the EC literature shows how this fraction works. To estimate the effectiveness of EC to prevent expected pregnancies, the denominator in the fraction comes from historical control data. These data show that out of 100 women having one act of unprotected intercourse during the fertile part of the cycle, eight women will conceive. For the numerator, traditional Yuzpe trials have found that out of 100 women using EC within 72 hours of unprotected intercourse, two will conceive. Using the formula, we can then calculate the effectiveness of the Yuzpe regime as .75 or 75 percent:
For this EC calculation, it is important to note that we get the denominator from historical controls. Therefore, in using the formula, researchers have to assume that the ability to conceive is the same for the historical controls as it is for those using EC. This is a significant assumption. Ideally, we would like to have both the numerator and denominator come from the same population. The true efficacy approach seeks to accomplish this by avoiding the need for any historical controls or comparisons with previous studies. This approach uses other strategies to control for the factors that influence the risk of pregnancy: ability to conceive, frequency and timing of intercourse, and compliance. To estimate the participant's ability to conceive, we recruit women who want to conceive but are willing to postpone pregnancy by one cycle. This enables us ethically to randomize women to use no method of contraception. From the group randomized to use no method, we can estimate the whole cohort's risk of pregnancy had they used no method. In other words, we now have a valid estimate of the denominator in our equation. To control for frequency and timing of intercourse, we ask participants to have intercourse on certain days of the cycle. In our pilot study, we gave women home ovulation kits. They were asked to abstain from vaginal intercourse from the beginning of the study until their day of ovulation, when they should have intercourse. They were also asked to abstain for five more days after ovulation. Since we enrolled participants during their menses, they had to abstain for about ten days, have intercourse once, and then abstain for five more days. Controlling for compliance may be the most difficult factor. Our pilot study enrolled women for only one cycle, so we think this short duration minimized non-compliance. Another strategy to minimize non-compliance was to have participants call an automated toll-free number once a day to answer a few questions about their coital activity, method use and menstrual bleeding. Of course, we cannot be absolutely certain that all participants were perfectly compliant with the protocol, but due to randomization, we think the level of compliance in the different arms was similar. The primary question with this approach was whether we could successfully complete such a study. In our pilot study, we enrolled 58 women at three sites. One site enrolled 25 women in five months. This level of recruitment is similar to the speed of recruitment during traditional barrier trials. We assigned women to use either the male condom, a spermicidal film or no method, when they had intercourse on their day of ovulation. We had a relatively low loss to follow-up with 54 women completing the study. We feel that this pilot study demonstrated the feasibility of this approach on a small scale. This small study does not have the precision to tell us the effectiveness of the two methods. Even so, it is useful to show how effectiveness would be calculated in a larger study of this design. The study found that the group using no method had a 12 percent pregnancy probability, while the spermicidal film group had an 11 percent pregnancy probability. None of the 19 women assigned to use condoms conceived. Using the formula shown above, we estimate that the effectiveness of the film is 8 percent while the effectiveness of the condom is 100 percent. Put another way, the film reduced the pregnancy probability from 12 percent to 11 percent, while the condom reduced it from 12 percent to 0 percent. We do not know how much this prevented fraction could be generalized to other cohort groups with different participant characteristics. However, we think the prevented fraction could certainly be generalized with more confidence than can the absolute pregnancy probabilities from a traditional trial. We have just received USAID funding for the full-scale study that we plan to initiate in the coming year. In this study we will use the same two approved products that we used in the pilot study: male condoms and spermicidal film (VCF). We hope to have a total of 500 women complete this study -- 200 women per group assigned to the two methods and 100 women in the no method group. Given our assumptions, this should provide sufficient power to detect a statistically significant difference between condoms and VCF. We estimated the size for the full study based on several assumptions. We decided to ask participants to have two acts of intercourse. The first act will be 17 days prior to the next expected menses while the second one will be on the day of ovulation. Based on the most recent review of probabilities for different days of the cycle (Trussell et al.), we estimate that 13.6 percent of the participants in the no method group will conceive during the study. We assume that 0.2 percent of the condom group and 4.0 percent of the VCF group will conceive. The condom probability comes from the most recent edition of Contraceptive Technology, while the VCF probability comes from FHI's recently completed international spermicide trial (Raymond et al.). The prevented fractions for condoms and VCF are assumed to be 98 percent and 70 percent respectively. The prevented fraction for condoms is close to what we measured in the pilot study (100 percent compared to 98 percent), while for VCF, the two prevented fractions are very different (8 percent and 70 percent). The outcome of most interest in the full study is to see how different the prevented fractions really are. With 500 participants completing the study, we will have 85 percent power to find a difference, given our assumptions. So what do we see as the future of this approach? Clearly the next step is to demonstrate that a full-scale study is possible to conduct. |
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Exact confidence intervals for Poisson variables can be computed from the Chi-square distribution (Johnson et al.).
, where ei is the number of EC requests corresponding to acts of intercourse on cycle day i, and pi is the probability of conception. The EC-adjusted pregnancy rate is r' = n'/m.