Over the past decade, increasing attention has focused on emergency contraceptive pills as an important means to reduce rates of unintended pregnancy and abortion. Because of the potential public health benefit as well as the safety and simplicity of the method, prominent medical and public health organizations have supported efforts to maximize access to it, including a recent application to the United States Food and Drug Administration to allow distribution of emergency contraceptive pills over the counter.1 Concerns have been raised by activists, providers, and women themselves, however, that easy availability of emergency contraceptive pills could undermine use of more effective contraceptive methods, particularly condoms. Decreased contraceptive and condom use could raise rates of both pregnancy and sexually transmitted infections. Method substitution has been demonstrated in other situations in which multiple contraceptive options were promoted.2,3
We designed this study to investigate these concerns. Our trial compared two approaches for providing emergency contraceptive pills. In the “standard access” approach, we informed women about how to obtain emergency contraceptive pills when needed, at usual charges. The “increased access” approach made taking the pills as effortless as possible: we gave women two packages of emergency contraceptive pills at admission and proactively provided them with free replacements after each package was used or lost. The aim of the trial was to determine how easier access would affect rates of pregnancy and sexually transmitted infection.
PARTICIPANTS AND METHODS
We conducted the trial in Nevada and North Carolina between October 2002 and June 2005. The protocol was approved by the institutional review boards of University of California at San Francisco, which managed the Nevada site, and Family Health International. All participants signed informed consent forms before data collection began. The study was monitored by a Data and Safety Monitoring Board, which reviewed aggregate interim outcome data and data pertaining to trial conduct. We adhered to CONSORT guidelines in the design and reporting of this study.4
We recruited sexually active women, aged 14–24 years, who did not desire pregnancy. We excluded women who were using or planned to use longer-term contraceptive methods (sterilization, intrauterine device, or hormone injections, implants, patch, or vaginal ring) and women who had been pregnant within the past 6 weeks or were breastfeeding. At the admission visit, we interviewed each volunteer, and she completed a self-administered computerized questionnaire and submitted urine and self-collected vaginal specimens. We tested the urine for pregnancy and sent the vaginal specimen to the Chlamydia Laboratory at Indiana University for gonorrhea, chlamydia, and trichomonas testing using polymerase chain reaction assays.5 We then assigned the volunteer to either the increased access group or the standard access group by opening the next in a consecutively numbered set of sealed opaque envelopes containing random assignments. The randomization scheme was stratified by site and used randomly permuted blocks with sizes of 4, 6, and 8 generated by computer at Family Health International before the start of enrollment at each site. We gave participants assigned to the increased access group two free packages of emergency contraceptive pills (Plan B, Barr Pharmaceuticals Inc, Pomona, NY) to take home. If a clinician with prescribing authority was not available at enrollment, we sent packages to the participant as soon as possible by mail. We advised participants in the standard access group how to obtain emergency contraceptive pills from the study site if needed. We counseled participants in both groups to take emergency contraceptive pills as a single dose of 1.5 mg levonorgestrel as soon as possible after unprotected intercourse, but we gave them no other special instructions.
We asked each participant to return to the clinic at 6 and 12 months after admission. Data collection procedures at follow-up visits were similar to those at admission. We reviewed available medical charts for relevant interim events. If a participant could not complete a visit in person, we mailed a pregnancy test kit and vaginal swab to her. We asked her to perform the pregnancy test herself and to report results by telephone along with other data and to mail the vaginal specimen directly to the study laboratory. At approximately 2, 4, 8, and 10 months after admission, we sent each participant by mail or e-mail a short survey about contraception use in the last 2 weeks. We did not tell participants in advance about the planned timing of these surveys.
We asked participants in both groups to notify the study site every time they used emergency contraceptive pills. We provided those in the increased access group with a replacement package for each package used or lost. At each follow-up visit, we questioned each increased access participant about the number of unused packages in her possession, and we gave additional packages to those who had fewer than two. The goal was to ensure that each increased access participant had two unused packages available at all times.
After randomization, we provided no unsolicited counseling about contraception unless a participant requested emergency contraceptive pills both four or more times total and more than once in any single month. However, we did not deny emergency contraceptive pills to such women. We notified participants who had positive pregnancy or sexually transmitted infection tests and referred them for care.
The target enrollment was 1,490 women. We selected this number to allow at least an 80% chance of showing with 95% confidence that the relative risk (RR) of infection, comparing the increased access approach with the standard access approach, was no more than 1.8. In our calculation, we assumed that the risk in the increased access group would be 6%, the increased access approach did not change that risk, and at most, 20% of the anticipated follow-up person-time would be missing. This sample size also provided at least 83% power to reject the null hypothesis of no difference in pregnancy rates between groups at the .05 significance level if the true RR were at least 1.6 and the pregnancy rates in both groups were at least 10%. Note that the null hypothesis for the sexually transmitted infection outcome was that the increased access approach would result in a higher infection rate than the standard access approach (with a true RR of at least 1.8), whereas for the pregnancy outcome the null hypothesis was that the risk is the same in the two groups.
Primary analyses included all enrolled participants. We analyzed each participant in the group to which she was assigned. For pregnancy analyses, we estimated the date of fertilization of each pregnancy using last menstrual period and ultrasound results, if available, and we excluded women pregnant at admission. Women not known to have become pregnant contributed time to the analysis through the later of 10 days before the last negative pregnancy test or 5 days after the last menstrual period. We tested the null hypothesis of no difference in pregnancy rates between the two treatment groups through 365 days after admission using a log-rank test, stratified by clinic. We defined incident sexually transmitted infections as gonorrhea, chlamydia, or trichomonas detected by the study laboratory or by a confirmed positive test performed elsewhere. For sexually transmitted infection analyses, we considered time in study to start at the earlier of the date of the first negative study test or, if the first study test showed a positive result for any of the three sexually transmitted infections, the date of single-dose treatment recommended by the Centers for Disease Control and Prevention. We assigned each subsequent infection to an interval between the date that the participant was last known to be uninfected and the date of the sexually transmitted infection diagnosis. We used a parametric Weibull proportional hazards model, which accounts for interval-censored data, to obtain an estimate of the RR of sexually transmitted infection along with a 95% upper confidence bound.6 We used proportional hazards models in secondary analyses of both primary outcomes to adjust for potentially important baseline covariates. We analyzed dates and times of coitus and emergency contraceptive pill use as reported by participants, except that we excluded all emergency contraceptive pill uses recorded as more than 2 days before or more than 7 days after sex (n=10) from analyses. We compared the median number of emergency contraceptive pill uses per participant and median delay between sex and emergency contraceptive pill use between treatment groups using median regression,7 implemented via the QUANTREG procedure in SAS 9.1 (SAS Institute Inc, Cary, NC). We considered P<.05 to be statistically significant throughout the analysis.
Between October 2002 and May 2004, 1,490 women enrolled in the study (Table 1). We did not keep records of women who were excluded or refused to participate. All participants met all admission criteria. The median age was 20 years. Many more participants intended to use hormonal contraceptives after admission than had been using these or other highly effective methods in the month before admission. Participants reported a median of four coital acts in the prior 14 days, including a median of two without condoms. Six percent had had an sexually transmitted infection in the past year, 39% had had more than one sexual partner in the previous 6 months, more than 25% were in a sexual relationship of less than 1 month duration, and 30% had partners who were probably or definitely not monogamous. The only notable difference between groups was that a higher proportion in the increased access group had a sexually transmitted infection at baseline. Participants in Nevada were slightly younger than those in North Carolina and were more likely to be Hispanic, white, and using nonhormonal methods of contraception.
Implementation of the intended protocol was successful in both groups throughout the study, with few exceptions. All increased access group participants were given or sent at least two free study emergency contraceptive pill packages within 8 days after admission to the study. Only 146 increased access participants (20%) experienced any time without emergency contraceptive pills after having used all previously dispensed packages; the mean total delay until resupply among all women was 7.2 days, less than 2% of the full expected year of follow-up. No increased access participants paid for study emergency contraceptive pill packages. One standard access participant was mistakenly given two study emergency contraceptive pill packages, which were both retrieved from her one week later.
The two groups contributed equal amounts of data. In the increased access and standard access groups, respectively, 709 (95%) and 703 (94%) had a final contact at 365 days after admission or later. In each group, 94% had known pregnancy status at 355 days after admission, and 93% had known sexually transmitted infection status at 365 days after admission (Fig. 1). Approximately 93% and 95% of the total possible person-time was ascertained for the pregnancy and sexually transmitted infection analyses, respectively, which was substantially higher than planned for in the power calculations. Over the entire study, 3,552 study emergency contraceptive pill packages were dispensed to the increased access group; the median number dispensed was four, and the maximum was 33. Most of the packages dispensed were either used by the participant or retained by her at the end of follow-up (Table 2). Less than 1% of the emergency contraceptive pills used by the increased access group were obtained outside the study. Women in the increased access group used emergency contraceptive pills substantially more often than women in the standard access group (Table 3). The median numbers of emergency contraceptive pill uses per participant in the two groups, respectively, were 2 and 0 (P<.01). In the standard access and increased access groups, respectively, 103 and 128 emergency contraceptive pill uses occurred within 1 day after admission. These immediate uses constituted 28% of total use in the standard access group, a much higher proportion than in the increased access group (6%). Emergency contraceptive pill users in the increased access group used the emergency contraceptive pills significantly sooner after sex (Table 2): the median delay was 12 hours in the increased access group and 36 hours in the standard access group (P<.01). Emergency contraceptive pill use patterns were similar at the two study sites. Participants in the standard access group reported having paid for 80% of the emergency contraceptive pills they used; the median charge was $15 in Nevada and $40 in North Carolina (overall range $1 to $60).
The incidence of pregnancy was similar in the two groups (Table 4) (hazard ratio 0.95, 95% confidence interval 0.68–1.33, log-rank P=.78). No interaction was observed between treatment group and study site. Adjustment for potentially important baseline covariates (previous pregnancy, black race, Hispanic ethnicity, marital status, and use of highly effective birth control methods in the month before enrollment) did not change this finding. However, the adjusted analysis suggested that women who had previously been pregnant were significantly more likely to have a study pregnancy and that women who used a highly effective birth control method before enrollment were significantly less likely to become pregnant. Also, pregnancies were much less common in North Carolina than in Nevada (6.8% and 11.3% of women contributing data, respectively). In the increased access and standard access groups, respectively, five and four participants had two pregnancies in the year after admission (the first pregnancy of one standard access participant was determined to have been fertilized before admission), and one woman in the increased access group had three pregnancies.
Our data provide significant evidence that, in the target population, the risk of the combined sexually transmitted infection outcome using the increased access approach is not substantially higher than the risk with the standard access approach. In fact, the observed risk was lower in the increased access group (hazard ratio 0.91, 90% confidence interval 0.66–1.26). Adjustment for site, youth, black race, positive sexually transmitted infection test at admission, sexually transmitted infection in the year preceding enrollment, and multiple partners in the 6 months preceding enrollment did not substantially affect this conclusion. Women with a sexually transmitted infection at admission, black women, and women with multiple partners in the previous year had a significantly higher risk of infection than women without those characteristics. Study site and age were not significantly related to sexually transmitted infection risk. No effect of treatment group on sexually transmitted infection rates was noted in subgroups defined by age category or study site. No significant differences between groups were observed in rates of any of the three sexually transmitted infections individually.
Participants’ coital activity and use of contraception, as reported 5–7 and 12–14 months after enrollment, did not differ significantly by group (Table 5), except for use of emergency contraceptive pills, which was much more common in the increased access group (P<.01, χ2). Behaviors reported at these two follow-up visits changed little compared with behaviors reported at enrollment: the proportion of women having sex decreased slightly, as did the proportion of sexually active women who used no contraception.
In the increased access and standard access groups, 246 (33%) and 280 (38%) of women, respectively, admitted at least once to having had unprotected sex without having used emergency contraception afterward. In both groups, the most common reasons cited for failure to use the emergency contraceptive pills were inconvenience and failure to appreciate risk of pregnancy. Participants in the increased access group used emergency contraceptive pills in 17 of the 74 total menstrual cycles (23%) in which pregnancy occurred; the corresponding figure for the standard access group was 2 of 74 cycles (3%). No serious adverse events related to the study occurred during the trial.
In our study, a strategy designed to enhance women’s ability to take emergency contraceptive pills when needed led to substantially increased emergency contraceptive pill use and greater promptness of use after unprotected coitus. This strategy had no effect on coital and contraceptive use patterns or on incidence of sexually transmitted infections. However, it did not have any apparent benefit in reducing pregnancy rates.
Recently published research on other programs to increase access to emergency contraceptive pills, including distribution in advance of need, direct provision by pharmacists, and over-the-counter marketing, has yielded findings consistent with ours.8–14 Although the evidence of absence of harm is reassuring, the failure to demonstrate a population-level contraceptive effect has been disappointing. One leading explanation has been that the ability of earlier studies to detect a benefit was limited by acknowledged flaws in study design and execution, such as low power, low baseline risk for pregnancy, adequate emergency contraception access in the comparison group, short follow-up, and crossovers between groups. Also, the interventions tested in many of the previous studies may have been intrinsically ineffectual: only one or two emergency contraceptive pill packages were provided, or participants were required to make special efforts to obtain replacements. Our trial largely avoided all of these weaknesses. In particular, the difference in amount of emergency contraceptive use between the groups in our study was substantial, greater than in previous studies. Thus, other explanations for the failure of increased emergency contraceptive use to translate into lower pregnancy rates must be considered.
One possibility is that emergency contraceptive pills are simply not highly efficacious. Published estimates suggest that after a single act of intercourse, the levonorgestrel regimen reduces pregnancy risk by 60–94%.15 However, the method used to derive these estimates is questionable because it did not take into account factors other than emergency contraceptive pill use that might have accounted for differences in pregnancy rates between women who did and those who did not use emergency contraceptive pills. More convincing evidence of efficacy comes from studies showing that emergency contraceptive pill treatment can increase the chance of anovulation and other physiologic events incompatible with pregnancy.16 Furthermore, two randomized trials showed that the levonorgestrel regimen is significantly more efficacious than an older regimen. Therefore even if the older regimen is no better than placebo, the levonorgestrel regimen logically must have some efficacy.17 But robust data on the specific level of efficacy are unavailable.
A second hypothesis is that any contraceptive benefit of the increased emergency contraceptive pill use may be counteracted by increased risk taking. However, like prior researchers,8–14,18 we found no gross differences between groups in reported coital behavior or use of regular contraception other than emergency contraceptive pills. Admittedly, these data are self-reported and impossible to verify, as are the data on emergency contraceptive pill use. However, consistent with one prior study,12 we also observed no difference in rates of sexually transmitted infections between groups, which is objective evidence of a lack of a clinically meaningful effect on condom use.
A third hypothesis is clearly supported by our data. More than one third of women in both study groups admitted to having had unprotected sex at least once without using emergency contraceptive pills afterward. This number is probably an underestimate because of poor recall, denial, and desire to please the researchers. Furthermore, as has been previously reported,8,10,11 most participants who became pregnant did not use emergency contraceptive pills in the menstrual cycle in which the pregnancy occurred. At least 146 increased access participants experienced some time in the study (7.2 days, on average) during which they did not have unused emergency contraceptive pills in their possession. Clearly, despite increased access, many risky coital acts remained “uncovered” by emergency contraception.
Our proactive intervention to keep increased access participants stocked with emergency contraceptive pills was expensive in terms of both labor and commodities. For this reason, it would probably not be feasible for widespread, long-term use outside a study. If emergency contraceptive pills are to achieve a measurable direct population-level impact on pregnancy rates, strategies to target high-risk women and high-risk coital acts may be needed. We plan to examine our data in more detail to evaluate possible explanations for our negative findings, which we hope will help to inform the development of such strategies.
1. Tharp M. Petition filed on emergency contraception status. AAP News 2001;18:147.
2. Farr G, Acosta Castro LA, DiSantostefano R, Claassen E, Olguin F. Use of spermicide and impact of prophylactic condom use among sex workers in Santa Fe de Bogota, Colombia. Sex Transm Dis 1996;23:206–12.
3. Fontanet AL, Saba J, Chandelying V, Sakondhavat C, Bhiraleus P, Rugpao S, et al. Protection against sexually transmitted diseases by granting sex workers in Thailand the choice of using the male or female condom: results from a randomized controlled trial. AIDS 1998;12:1851–9.
4. Moher D, Schulz KF, Altman DG. The CONSORT statement: revised recommendations for improving the quality of reports of parallel-group randomized trials [in Chinese]. Zhongguo Zhong Xi Yi Jie He Za Zhi 2005;25:658–61.
5. Van Der Pol B, Kraft CS, Williams JA. Use of an adaptation of a commercially available PCR assay aimed at diagnosis of chlamydia and gonorrhea to detect Trichomonas vaginalis in urogenital specimens. J Clin Microbiol 2006;44:366–73.
6. Collett D. Modelling survival data in medical research. London: Chapman and Hall; 1994.
7. Yu K, Lu Z, Stander J. Quantile regression: applications and current research areas. J Royal Stat Soc: Series D 2003;52:331–50.
8. Glasier A, Baird D. The effects of self-administering emergency contraception. N Engl J Med 1998;339:1–4.
9. Jackson RA, Bimla Schwarz E, Freedman L, Darney P. Advance supply of emergency contraception. effect on use and usual contraception–a randomized trial. Obstet Gynecol 2003;102:8–16.
10. Hu X, Cheng L, Hua X, Glasier A. Advanced provision of emergency contraception to postnatal women in China makes no difference in abortion rates: a randomized controlled trial. Contraception 2005;72:111–6.
11. Lo SS, Fan SY, Ho PC, Glasier AF. Effect of advanced provision of emergency contraception on women’s contraceptive behaviour: a randomized controlled trial. Hum Reprod 2004;19:2404–10.
12. Raine TR, Harper CC, Rocca CH, Fischer R, Padian N, Klausner JD, et al. Direct access to emergency contraception through pharmacies and effect on unintended pregnancy and STIs: a randomized controlled trial. JAMA 2005;293:54–62.
13. Glasier A, Fairhurst K, Wyke S, Ziebland S, Seaman P, Walker J, et al. Advanced provision of emergency contraception does not reduce abortion rates. Contraception 2004;69:361–6.
14. Tyden T, Aneblom G, von Essen L, Haggstrom-Nordin E, Larsson M, Odlind V. No reduced number of abortions despite easily available emergency contraceptive pills. Studies of women’s knowledge, attitudes and experience of the method [in Swedish]. Lakartidningen 2002;99:4730–2, 4735.
15. Emergency oral contraception. ACOG Practice Bulletin No. 25. American College of Obstetricians and Gynecologists. Int J Gynaecol Obstet 2002;78:191–8.
16. Croxatto HB, Brache V, Pavez M, Cochon L, Forcelledo ML, Alvarez F, et al. Pituitary-ovarian function following the stan-dard levonorgestrel emergency contraceptive dose or a single 0.75-mg dose given on the days preceding ovulation. Contra-ception 2004;70:442–50.
17. Raymond E, Taylor D, Trussell J, Steiner MJ. Minimum effectiveness of the levonorgestrel regimen of emergency con-traception. Contraception 2004;69:79–81.
18. Gold MA, Wolford JE, Smith KA, Parker AM. The effects of advance provision of emergency contraception on adolescent women’s sexual and contraceptive behaviors. J Pediatr Adolesc Gynecol 2004;17:87–96.