Estimating emergency contraception efficacy with levonorgestrel and copper intrauterine devices

Abstract

Objectives:

This study aimed to determine observed vs expected pregnancy risk among levonorgestrel 52 mg and copper T380A intrauterine device (IUD) emergency contraception (EC) users.

Study design:

This is a secondary analysis of participants seeking EC randomly assigned to a levonorgestrel 52 mg or copper T380A IUD. Participants had at least one episode of unprotected intercourse (UPI) in the 5 days preceding enrollment and reported all episodes in the preceding 5 days. We report the proportion of pregnancies prevented of those expected at 1 month following IUD insertion using an established approach that assigns pregnancy risk by menstrual cycle day of intercourse. We calculated proportion of pregnancies prevented using both the day of most recent UPI and all reported UPIs within 5 days of presentation.

Results:

One pregnancy occurred among the 312 participants who received the levonorgestrel 52 mg IUD, with 137 (43.9%) reporting multiple episodes of UPI. No pregnancies occurred among the 318 participants who received the copper IUD, of whom 142 (44.6%) reported multiple episodes. In the levonorgestrel 52 mg IUD group, 14.8 pregnancies were expected using the most recent episode of UPI, and 22.0 pregnancies were expected using all episodes within 5 days of presentation. In the copper IUD group, we calculated 15.0 and 23.1 expected pregnancies, respectively. Levonorgestrel 52 mg IUDs prevented 93.2% to 95.7% of expected pregnancies, and copper IUDs prevented 100% of expected pregnancies.

Conclusions:

The results of this analysis provide additional data quantifying the pregnancy risk reduction for both the levonorgestrel 52 mg IUD and copper IUD in a realistic population of EC users.

Implications:

The levonorgestrel 52 mg IUD for emergency contraception prevented 93% to 96% of expected pregnancies using an established pregnancy risk method, while copper IUD users experienced 100% pregnancy prevention.

1. Introduction

Emergency contraception (EC) includes any method of contraception used to prevent pregnancy within days following an episode of unprotected sexual intercourse. Methods approved by the United States Food and Drug Administration include oral levonorgestrel and oral ulipristal acetate. Additionally, evidence supports off-label use of the copper T380A intrauterine device (IUD) for EC [1].

A growing body of evidence supports off-label use of the levonorgestrel 52 mg IUD for EC. A randomized controlled trial found the levonorgestrel 52 mg IUD provided noninferior pregnancy risk to the copper IUD for EC when placed within 5 days of unprotected intercourse [2]. Subsequently, the Society of Family Planning [3] and Planned Parenthood Federation of America [4] have recommended that providers offer the levonorgestrel 52 mg IUD as an EC option, though others have expressed a need for more data before routine use [5–7]. Analysis clarifying the efficacy of the levonorgestrel 52 mg IUD will allow more direct comparisons to other EC options.

EC trials frequently report efficacy as percent of pregnancies prevented using methods to estimate expected numbers of pregnancies. Conception only occurs with intercourse on the 5 days before and the day of ovulation, with daily pregnancy rates ranging from 0.04 to 0.29 in that 6-day window [8]. However, the day of ovulation cannot be reliably predicted based on the start of the last menses, which complicates pregnancy risk prediction in trials of EC that only collect menstrual cycle dating. Thus, methods have been developed that calculate the probability of pregnancy for intercourse on any given menstrual cycle day, which are commonly used in this context [9,10]. This analysis applies this method to report the proportion of pregnancies prevented among participants in the Randomized controlled trial Assessing Pregnancy for Intrauterine Devices for Emergency Contraception (RAPID-EC) study [2].

2. Materials and methods

This is a secondary analysis of the RAPID-EC study, a 1:1 randomized controlled noninferiority study comparing the efficacy of levonorgestrel 52 mg IUD to copper T380A IUD for EC. The original manuscript addresses the trial’s methodologic details [2]. The institutional review board at the University of Utah approved the trial protocol.

Six outpatient clinical sites in Utah enrolled RAPID-EC trial participants from August 2016 through December 2019. We recruited potential participants from individuals already presenting in person for EC. Study participants included individuals aged 18 to 35 years who reported at least one episode of unprotected intercourse in the 5 days preceding enrollment, had regular menstrual cycles occurring every 21 to 35 days, and a known last menstrual period ( ± 3 days). We excluded potential participants with a positive point-of-care urine pregnancy test at presentation (n = 3). We obtained written consent from all participants. Research staff randomized participants to receive either a levonorgestrel 52 mg IUD or copper T380A IUD. The trial’s primary efficacy outcome was pregnancy identified by a positive urine pregnancy test at 1 month following insertion. For the 48 participants who did not report this outcome, we obtained pregnancy outcome information based on responses to 1, 3, and 6-month follow-up surveys, review of electronic health data from the system that provided the care at enrollment and any visits participants reported at other sites.

At their enrollment visit, participants completed a questionnaire identifying the timing of their most recent episode of unprotected intercourse and the reason why they needed EC (e.g., “I did not use a contraceptive method” or “I missed pills, contraceptive patch, vaginal ring”). Additionally, participants used a calendar to identify all episodes of intercourse in the 14 days preceding enrollment and checked episodes when no method of birth control was used or when they used a method that might put them at risk for pregnancy (e.g., broken condom, missed pills). We excluded participants who did not report intercourse timing from this analysis. Using these data, we assigned cycle day of intercourse from one through 35 for each episode of unprotected intercourse within 5 days of IUD placement. We collected and stored data using the secure Research Electronic Data Capture platform [11,12].

Expected pregnancy rates were calculated using a modification of the algorithm derived from data initially collected in the North Carolina Early Pregnancy Study, which followed a cohort of participants seeking pregnancy. In this cohort study, 221 participants from 1982 to 1985 provided daily urine samples to measure metabolites of estradiol (estrone 3-glucuronide) and progesterone (pregnanediol 3-glucuronide). An algorithm based on the ratio of these metabolites identified the day of ovulation in 625 menstrual cycles with complete data. From these, a highly sensitive and specific urine assay for human chorionic gonadotropin identified 192 pregnancies. Conception only occurred on the 5 days before ovulation and the day of ovulation, with conception probabilities ranging from 0.04 to 0.29 [9,10]. However, day of ovulation is not reliably predicted based on the start of the last menses as shown in EC trials in which pregnancies occurred in participants who reported no intercourse inside the estimated fertile window [13,14]. In 2001, Wilcox et al. [9] published an analysis of the North Carolina Early Pregnancy cohort data to provide a probability of pregnancy for a single act of unprotected intercourse that can be used with menstrual dating in EC trials based on a single episode of unprotected intercourse from cycle day 1 to 40. In 2015, Li et al. [10] updated these probabilities to incorporate the nonindependent relationship between frequency of intercourse and ovulation, which results in an increased estimated probability of pregnancy overall.

For both the levonorgestrel 52 mg and copper IUD study groups, we calculated the number of expected pregnancies using conception probabilities per episode of unprotected intercourse based on the cycle day of intercourse using the probabilities in participants with regular cycles published by Li et al. [10]. Detailed descriptions estimating pregnancy risk for each cycle day are provided in the original manuscript. These methods have been used previously in trials evaluating the efficacy of EC [13,15–17]. For our primary analysis, we calculated the range of expected pregnancies per group with two approaches assuming the probabilities of pregnancy risk are known exactly for each cycle day. First, we included only the most recent episode of unprotected intercourse. Then we included all reported episodes in the 5 days preceding enrollment, assuming an independent contribution to pregnancy risk by every individual act of intercourse. For example, consider a participant who reports intercourse 2 days and 4 days before enrollment, which corresponds to days 17 and 15 of their menstrual cycle. This participant would contribute one times the predicted pregnancy risk of intercourse on cycle day 17 (1 * 0.064) to the first method of analysis and one minus the product of one minus the pregnancy risk on cycle day 17 and one minus the pregnancy risk on cycle day 15 (1 – [1 – 0.064] * [1 – 0.082]) to the overall expected number of pregnancies in the second. From here, we calculated overall pregnancy risk by summing the predicted pregnancy risk of all participants in each IUD group. The percentage of pregnancies prevented was calculated as the number of expected pregnancies minus actual pregnancies divided by expected pregnancies for each study arm. We calculated risk difference between levonorgestrel 52 mg IUD and copper IUD groups and reported 95% CIs. In addition, we completed three sensitivity analyses. The first restricted the analysis to only the participants who reported a single act of unprotected intercourse within 5 days of IUD placement. The second included only those reporting no method, incorrect use of the rhythm method, or failed use of the withdrawal method at the time of most recent unprotected intercourse in order to estimate pregnancy risk among those at highest risk of pregnancy and exclude participants in the trial who reported recent hormonal contraception or condom use. We also included six of nine participants who selected ‘other’ when their descriptions indicated that they did not use a method. The third included only participants who met criteria for both prior sensitivity analyses. All analysis was performed using STATA v18.0 (StataCorp, LLC, College Station, TX).

2.1. Clinical trials registration

Clinicaltrials.gov registration number: NCT02175030.

3. Results

In this secondary analysis, we included 312 levonorgestrel 52 mg IUD and 318 copper T380A IUD users who received the assigned intervention and had timing of unprotected sexual intercourse and 1-month pregnancy data available (Figure 1). These represent 95.4% of the 327 levonorgestrel IUD users enrolled and 97.0% of the 328 copper IUD users enrolled. Participant characteristics have been previously reported [2]. One pregnancy occurred in the levonorgestrel 52 mg IUD group (1/312, 0.3%, 95% CI, 0.01–1.8) and no pregnancies in the copper IUD group (0/318, 0%, 95% CI 0–1.2) [2]. In Table 1, we report the most recent and all reported episodes of unprotected sexual intercourse within 5 days of enrollment by cycle day.

Figure 1.

Flow diagram of RAPID-EC study participants enrollment, randomization, and analysis. Utah, 2016–2019*. RAPID-EC: Randomized Controlled Trial Assessing Pregnancy for Intrauterine Devices as Emergency Contraception, IUD: intrauterine device, UPI: unprotected intercourse, EC: emergency contraception. *Turok DK, Gero A, Simmons RG, et al. Levonorgestrel vs. Copper Intrauterine Devices for Emergency Contraception. N Engl J Med. 2021;384[4]:335–344. doi:10.1056/NEJMoa2022141.

Table 1.

Cycle day of episodes of unprotected sexual intercourse in the 5 days preceding IUD placement reported by RAPID-EC trial participants, Utah, 2016–2019^a

	Levonorgestrel 52 mg IUD		Copper T380A IUD
Cycle day	Most recent	All episodes	Most recent	All episodes
1	4	4	8	1
2	4	6	1	2
3	5	6	3	4
4	9	16	4	10
5	3	7	9	12
6	8	16	9	17
7	21	29	18	29
8	14	16	15	23
9	13	22	11	19
10	11	29	14	23
11	23	33	11	23
12	12	19	12	22
13	14	22	21	32
14	19	27	23	32
15	15	22	12	19
16	12	17	20	25
17	12	19	13	23
18	14	22	12	20
19	19	25	11	16
20	10	18	13	24
21	12	14	10	15
22	11	16	10	19
23	8	13	9	13
24	11	13	11	15
25	12	12	5	13
26	1	3	8	11
27	4	4	7	12
28	5	7	9	10
29	2	5	3	7
30	4	7	5	6
31	2	7	0	5
32	3	4	0	2
33	0	1	1	2
34	1	2	1	4
35	2	3	1	2

IUD, intrauterine device, RAPID-EC: Randomized Controlled Trial Assessing Pregnancy for Intrauterine Devices as Emergency Contraception,

^aTurok DK, Gero A, Simmons RG, et al. Levonorgestrel vs. Copper Intrauterine Devices for Emergency Contraception. N Engl J Med. 2021;384[4]:335–344. doi:10.1056/NEJMoa2022141.

Table 2 presents the number of expected pregnancies and the proportion of pregnancies prevented. In the levonorgestrel 52 mg IUD group, we expected 14.8 pregnancies using the most recent episode of unprotected intercourse and 22.0 pregnancies using all episodes of unprotected intercourse within 5 days of enrollment. The proportion of pregnancies prevented ranged from 93.2% to 95.7%. In the copper IUD group, we expected 15.0 pregnancies using the most recent episode and 23.1 pregnancies using all episodes, respectively, with 100% of pregnancies prevented. Risk difference did not differ significantly between the levonorgestrel 52 mg IUD and copper IUD groups (risk difference 6.8% [p = 0.50] with analysis including only the most recent episode of intercourse and 4.5% [p = 0.49] in the analysis including all episodes within 5 days).

Table 2.

Expected number of pregnancies and proportion of pregnancies prevented by IUD group in the RAPID-EC trial^a using the Li-Wilcox^b method using the most recent and all episodes of unprotected intercourse within 5 days of IUD placement, Utah, 2016–2019

Included episodes	Levonorgestrel 52 mg IUD		Copper T380A IUD		Pregnancy risk difference between levonorgestrel 52 mg and copper IUD EC users (95% CI)
	Pregnancies expected/observed (n)	Proportion pregnancies prevented (%) (95% CI)	Pregnancies expected/observed (n)	Proportion pregnancies prevented (%) (95% CI)
Most recent episode of unprotected intercourse	14.8/1	93.2% (90.5, 95.9)	15.0/0	100% (98.8, 100)	6.8% (−6.5, 20.0)
All episodes of unprotected intercourse within 5 days of placement	22.0/1	95.7% (93.5, 97.9)	23.1/0	100% (98.8, 100)	4.5% (−4.4, 13.5)

IUD: intrauterine device, RAPID-EC: Randomized Controlled Trial Assessing Pregnancy for Intrauterine Devices as Emergency Contraception.

^aTurok DK, Gero A, Simmons RG, et al. Levonorgestrel vs. Copper Intrauterine Devices for Emergency Contraception. N Engl J Med. 2021;384[4]:335–344. doi:10.1056/NEJMoa2022141,

^bLi D, Wilcox AJ, Dunson DB. Benchmark Pregnancy Rates and the Assessment of Post-coital Contraceptives: An Update. Contraception. 2015;91[4]:344–349. doi:10.1016/j.contraception.2015.01.002.

The sensitivity analysis restricted to participants who reported a single act of unprotected intercourse within 5 days of IUD placement included 347 participants with 176 levonorgestrel IUD users (53.8% of included participants) and 171 (53.8%) copper IUD users. The single pregnancy occurred in this group with 8.1 pregnancies expected in the levonorgestrel 52 mg IUD group with 87.7% pregnancy risk reduction and 8.0 expected in the copper IUD group with 100% risk reduction. The sensitivity analysis including participants who used no method of contraception at last intercourse included 309 participants with 148 (47.4%) levonorgestrel IUD users and 161 (50.6%) copper IUD users. No pregnancies occurred in this group with 8.5 pregnancies expected in the levonorgestrel 52 mg IUD group and 10.8 expected in the copper IUD group. When restricting the analysis to participants who met both of these criteria, 229 participants with 112 (35.9%) levonorgestrel IUD users and 117 (36.8%) copper IUD users were included with 4.8 pregnancies expected in the levonorgestrel 52 mg IUD group and 5.3 expected in the copper IUD group.

4. Discussion

In this secondary analysis, both the levonorgestrel 52 mg IUD and copper T380A IUD prevented a high proportion of expected pregnancies. Established oral EC methods show 52% to 85% EC efficacy using Wilcox 1995 [16,17] or Wilcox 2021 [18] pregnancy probability estimates. A recent randomized controlled trial showed increased efficacy of oral levonorgestrel when combined with oral piroxicam, a cyclo-oxygenase inhibitor, with 95% of pregnancies prevented with the Wilcox model [15]. While methodologic differences between the RAPID-EC trial and other EC studies limit the ability to make a direct comparison to other methods, the 93% to 96% EC efficacy of the levonorgestrel 52 mg IUD and the 100% EC efficacy of the copper T380A IUD is reassuring that these methods are effective in preventing pregnancy in this population and compare favorably to established oral methods.

The participants in the original cohort that generated the Li-Wilcox method were seeking pregnancy. EC users in contrast are likely to have lower baseline pregnancy risk. While these methods have been used in several similar analyses [13,15–17], the original data were not designed to address this issue. Similarly, baseline pregnancy risk broadly complicates the assessment of EC efficacy and is a limitation in the direct comparison between studies. Specifically, in the RAPID-EC trial, baseline pregnancy risk likely differs compared to other EC trials due to inclusion of participants reporting more than one episode of unprotected intercourse in the current cycle and inclusion of participants who disclosed recent use of contraceptive methods. Regarding episodes of intercourse, EC users commonly have more than one episode of unprotected intercourse per cycle [19], and in the RAPID-EC trial, we included all eligible individuals who self-identified the risk of pregnancy regardless of the number of acts of intercourse they reported, supporting external validity. Thus, relative to a trial that excludes those with greater than one episode of unprotected intercourse [15,20,21], RAPID-EC participants may have had a greater baseline pregnancy risk. Our sensitivity analysis only including participants who report a single act of intercourse is supportive of our primary findings.

Conversely, multiple EC trials exclude patients who disclose recent use of a contraceptive method since use is likely to lower baseline pregnancy risk. Our sensitivity analyses identified that 48.1% of levonorgestrel 52 mg IUD users and 50.1% of copper T380A IUD users in this analysis did not report use of any contraceptive method at the time of most recent intercourse, and the results are supportive of high efficacy of both methods. It is worth noting as well that unreported use of hormonal methods of contraception complicates estimation of baseline pregnancy risk in any EC study. In a secondary analysis of a randomized controlled trial evaluating EC efficacy of oral ulipristal acetate vs oral levonorgestrel in individuals weighing 80 kg or greater [22], Jensen et al. observed that 22.5% of the participants had detectable levels of progestins at the time of presentation for EC and 10.6% had progestin levels above the minimal level for contraceptive effect [23]. The RAPID-EC trial did not conduct a similar assessment of exogenous progestin exposure though the sensitivity analysis in this manuscript focused on participants not reporting any method of contraception. Jensen et al. hypothesized that recruitment strategies and study reimbursement may have incentivized participants to misrepresent pregnancy risk. The drivers of this behavior are likely complex and incompletely understood. However, a major difference in recruitment strategies distinguishes these studies. In RAPID-EC, the only method of identifying potential participants was by presenting to a clinic requesting EC, whereas the study referenced by Jensen et al. recruited participants to clinics with external study advertisements. In addition, RAPID-EC participants who did not believe themselves to need EC could have also received the IUD they desired without cost at the same clinic without subjecting themselves to participant masked randomization.

A strength of this trial and the analysis is that participants determined their need for EC and had no incentive to over report recent unprotected intercourse. On the contrary, healthy behavior reporting bias would expect that people would under-report recent unprotected intercourse. In this study, participants reported this activity separately from their screening and after enrollment. Thus, by allowing people to provide the information without provider scrutiny, it may be more likely that each participant reported their exposure as it occurred. Additional strengths include the trial’s randomized design. Limitations include patient-reported timing of unprotected intercourse without validation, which is standard for EC studies.

A critique of the original trial [5] discusses the importance of identifying the mechanism of action for the levonorgestrel 52 mg IUD for EC, and this remains a limitation in this secondary analysis as well. We acknowledge that the mechanism of action is not fully understood and may include both contraceptive and contragestive mechanisms. This distinction is of importance to some researchers and clinicians. However, we lack data on whether what role this plays in EC user decision-making and if patient preferences regarding mechanism of action exist. Requiring certainty about the mechanism of action for the levonorgestrel 52 mg IUD for EC before its utilization interferes with access and autonomy for those EC users who prefer to have a levonorgestrel IUD. EC users express preference for the levonorgestrel 52 mg IUD over the copper IUD [24], and IUD users are less likely to become pregnant within a year after presentation when compared to oral levonorgestrel users [25]. Given evidence of patient preference, shared decision-making offers an ideal path forward. These conversations should include the fact that more data and certainty support the copper T380A IUD as the most effective option for preventing pregnancy. However, they may also discuss that the levonorgestrel 52 mg IUD has data supporting its efficacy and may be a preferred option for many seeking EC for reasons beyond efficacy alone. Some EC users and clinicians will use the constellation of data presented here, in the original trial publication [2] and another demonstrating no pregnancy risk in levonorgestrel 52 mg IUD EC users who resumed intercourse in the same cycle as placement [26] as sufficient data for levonorgestrel 52 mg IUD EC use. Others may decide these data are insufficient. The former group has clinical guidelines to support their selection [3,4] and the latter a Cochrane Review [6] and commentary [5,7] to support their refusal. All will benefit from data from an ongoing, multisite reproducibility study that will enroll 474 levonorgestrel 52 mg IUD EC users (NCT05444582) [27].

The results of this analysis provide additional data quantifying the pregnancy risk reduction for both the levonorgestrel 52 mg IUD and copper IUD in a realistic population of EC users. EC users desiring the greatest pregnancy prevention should be informed that the quantity and reproducibility of data support use of the copper IUD. EC users who strongly prefer a levonorgestrel 52 mg IUD and are willing to accept a lower level of certainty regarding efficacy can be reassured that this option is likely to be highly effective. Additional data from an ongoing reproducibility trial will further clarify the pregnancy risk for EC users selecting the levonorgestrel 52 mg IUD.