Emergency contraception for individuals weighing 80kg or greater: a randomized trial of 30 mg ulipristal acetate and 1.5 mg or 3.0 mg levonorgestrel

Abstract

Objective:

To compare the efficacy of oral emergency contraception (EC) regimens used within 72 hours of unprotected intercourse in individuals weighing ≥80 kg.

Study Design:

We enrolled healthy women aged 18–40 years with a weight of at least 80 kg requesting EC in a multi-center, single-blind, randomized study of levonorgestrel 1.5 mg (LNG1X) and 3.0 mg (LNG2X) and ulipristal acetate 30 mg (UPA) (enrollment goal 1200). Key eligibility requirements included regular cycles, unprotected intercourse within 72 hours of enrollment, no use of hormonal contraception since last menstrual period, a negative urine pregnancy test (UPT), and willingness to abstain from intercourse until next menses. Study staff directly observed EC ingestion. To assess our primary outcome of incidence of pregnancy, participants completed home UPTs; if no menses by 2-weeks post-treatment, or a positive UPT, they returned for an in person visit with quantitative serum human chorionic gonadotropin (hCG) and ultrasound.

Results:

We enrolled and randomized 532; 44 were not dosed or not evaluable for primary end-point, leaving an analyzable sample of 488 (173 LNG1X, 158 LNG2X, 157 UPA) with similar demographics between groups [mean age 29.6 years (5.74), BMI 37.09 kg/m² (6.95)]. Five pregnancies occurred during the study (LNG1X n = 1, LNG2X n = 1, UPA n = 3); none occurred during the highest at-risk window relative to estimated day of ovulation (day of ovulation and the 3 days prior). We closed the study prior to achieving our enrollment goal because the low pregnancy rate in all groups established futility based on an interim blinded analysis.

Conclusion:

Although slow enrollment limited our study power, we found no differences in pregnancy rates between oral EC regimens among women weighing 80 kg or more. Our results are not able to refute or support differences between the three treatment arms.

Implications:

Women weighing 80 kg or more experienced no differences in pregnancy rates between oral EC regimens but due to several significant study limitations including sample size and the lack of a study population at high risk of pregnancy, our results are not able to determine if differences in treatment effectiveness exist.

1. Introduction

Approximately 45% of all pregnancies in the United States are unintended [1]. Emergency contraception (EC) provides a second chance for individuals to prevent unintended pregnancy after unprotected intercourse [1]. Early studies evaluating oral EC suggest a reduction in pregnancy risk of up to 81–90% [2–5]. However, existing evidence raises concerns of decreased efficacy of oral EC methods in individuals with a body mass index (BMI) of ≥30kg/m² and essentially no efficacy for individuals taking levonorgestrel-containing EC weighing 80 kg or more [6].

Glasier, et al. re-analyzed data from two large randomized control trials to identify risk factors for EC failure [6]. Individuals with a BMI of ≥30kg/m² using levonorgestrel-based EC (LNG EC) had a greater risk of pregnancy compared to those with a BMI <25kg/m² (OR 4.41, 95% CI 2.05–9.44). Notably, these authors determined that the decline in efficacy relates to absolute weight; LNG-based EC appeared to have a ceiling of efficacy at 70 kg and no efficacy for women 80 kg and above. A reanalysis of four World Health Organization EC trial also demonstrated a differential risk of LNG EC failure for individuals weighing over 80 kg but overall, rates of pregnancy were low in this study [7].

We hypothesize that oral LNG EC effectiveness is reliant on achieving a rapid peak serum LNG concentration at a critical time point prior to the LH surge [2,8]. Studies have demonstrated that obesity significantly alters LNG pharmacokinetics [9,10], which may explain the increased failure rates seen in those with a BMI > 25kg/m² and a weight of ≥70 kg. Following the standard LNG 1.5 mg dose, individuals with a BMI of ≥ 30kg/m² achieved peak LNG concentrations approximately 50% lower than those obtained from women with a BMI <25kg/m²; doubling the LNG EC dose to 3 mg corrected the PK differential [10]. Although the study did not include a pharmacodynamic assessment, these results led to some changes in clinical recommendations. UK clinical guidelines suggest that women of higher weight should consider taking LNG 3.0 mg for EC rather than 1.5 mg [11]. LNG EC products in the USA and Europe do not contain a warning that the products may be less effective in individuals of higher weight or BMI.

Obesity is common in the US population.[12] Therefore, a need exists to develop contraceptive methods that are equally effective regardless of weight. We designed this clinical trial to compare the efficacy of single and double dose LNG-EC and Ulipristal Acetate (UPA)-EC in individuals weighing ≥80 kg used within 72 hours of unprotected intercourse.

2. Materials and Methods

2.1. Study participants

The National Institute of Child Health and Human Development (NICHD) Female Contraceptive Clinical Trials Network conducted this phase IIb, multi-center, single-blind, randomized study from June 2017 to April 2022. A total of 17 US sites participated (Oregon Health & Science University, Portland, OR; Eastern Virginia Medical School, Norfolk, VA; Johns Hopkins University, Baltimore, MD; University of Pennsylvania, Philadelphia, PA; University of California, San Francisco, CA; University of California Davis, Sacramento, CA; University of Hawaii, Honolulu, HI; Essential Health Access, Los Angeles, CA; University of Colorado, Aurora, CO; University of Utah, Salt Lake City, UT; University of Cincinnati, Cincinnati, OH; New York University, New York, NY; Women & Infants Hospital of Rhode Island, Providence, RI; Columbia University, New York, NY; Planned Parenthood League of Massachusetts, Boston, MA; University of Pittsburgh, PA). Advarra (Columbia, MD), a central national institutional review board (IRB), approved the protocol with local IRBs ceding full authority or performing secondary reviews for local regulatory requirements. All participants completed written informed consent prior to any study procedures. The study used many central and local recruitment activities including the use of several trial recruitment platforms, IRB approved flyers, provider referrals, radio advertisements, social media, and newsletters. Participants received compensation.

We recruited healthy women aged 18–40 years old, weighing 80kg or more, at risk for pregnancy and requesting EC within 72 of unprotected intercourse. We defined ‘at pregnancy risk’ as regular menstrual cycles with non-use, incorrect use, or failure of a contraceptive method (see Supplemental material for full inclusion/exclusion criteria). Consenting enrolled participants also accepted the risk that oral EC is less effective than a copper intrauterine device (IUD) and agreed to abstain from additional acts of unprotected intercourse and not to initiate a hormonal contraceptive method or intrauterine device until the end of the study.

2.2. Procedures

Participants completed an in-person screening visit within 72 hours of their unprotected act of intercourse to obtain written consent, complete screening activities, and collect baseline demographic and health information. The screening and enrollment visit typically occurred in immediate succession (on the same day) to ensure administration of study drug dosing within 72 hours of unprotected intercourse. At enrollment, study staff measured and confirmed weight eligibility and collected a blood and urine sample. Study sites stored serum samples at −80C, and shipped frozen samples to a central laboratory (ONPRC Endocrine Core Laboratory, Beaverton, OR) at the end of the study where they underwent batched analysis for progesterone (P4), luteinizing hormone (LH), estradiol (E2), follicle stimulating hormone (FSH), and serum hCG as previously described.[13] After performing a urine pregnancy test (UPT) and confirming negative results, study staff entered the participant in a central database maintained by the data coordinating center and received a treatment allocation from a computer-generated randomization schema using permuted blocks (1:1:1 allocation ratio). Participants were masked to treatment arm and received either one LNG 1.5mg tablet plus a placebo tablet (LNG 1.5mg arm), two LNG 1.5mg tablets and no placebo tablet (LNG 3mg arm), or one UPA 30mg tablet plus a placebo tablet (UPA 30mg arm). The placebo tablets were all the same no matter the treatment arm but not necessarily identical to the active drug. Study staff limited close inspection of the study tablets by participants and directly observed ingestion of study drug. Participant masking was maintained until after completion of all data collection and entry.

Study staff also trained and provided participants with a daily paper diary to report any vaginal bleeding or spotting, sexual activity (including whether additional contraceptive method was used), and any symptoms (nausea, vomiting, fatigue, etc.). They also supplied two urine pregnancy tests and instructed participants to perform a UPT 5–7 days and 12–14 days after the next expected menses and to report these results to the study staff during scheduled phone contact follow-up. If a urine test was positive, then the participant returned to the site as soon as possible for a serum hCG and a pelvic ultrasound. The participant could exit the study if both UPTs were negative and menses had returned; if they experienced no menses, then an additional follow up occurred 19–21 days post-expected menses for a serum hCG.

2.3. Statistical Methods

Given the uncertainty regarding estimates of failure between groups to base a power analysis, we planned a descriptive analysis of pregnancy outcomes in this exploratory study. We defined pregnancy as a positive UPT (or serum hCG) or by confirmation of pregnancy by ultrasound or both. We assumed a sample size of 1200 (400/group) would provide relatively precise estimates of contraceptive failure of the three regimens (95% CI within 1–2% of a point estimate of 3%) but would not likely have sufficient power to detect differences between groups. The a priori analysis plan included the following cohorts: all randomized participants and efficacy evaluable participants. We also planned an interim analysis once approximately 200 individuals per arm completed follow-up. We stopped the study due to futility in achieving the enrollment goal of 1200 given low and slow enrollment at all study sites. We performed the interim analysis following enrollment of 532 participants. We described participant characteristics for each study group using means and medians for continuous variables. We used SAS version 9.4 (SAS Institute, Cary, North Carolina) for all analyses.

3. Results

We stopped the study after enrollment of 532 individuals. Of these, we excluded 44 not evaluable for the primary outcome to obtain an efficacy-evaluable group of 486 (173 LNG1X, 158 LNG2X, 157 UPA,) (Figure 1). Study participants were similar demographically with a mean age 29.6 years (5.74), BMI 37.09 kg/m² (6.95) (Table 1). Many participants identified as Black or African American (282/532, 53%) (Table 1).

Figure 1.

CONSORT flow diagram. *Urine pregnancy test. LNG = levonorgestrel; UPA = ulipristal acetate.

Table. 1.

Characteristics of enrolled study participants

Parameter	LNG 1.5 mg	LNG 3.O mg	UPA 3O mg
Category/Statistics	(N = 181)	(N = 175)	(N = 176)
Age (y)
n	181	175	176
Mean (SD)	3O.1 (5.92)	28.8 (5.73)	29.8 (5.49)
Median	3O	29	3O
Min, Max	18, 4O	18, 4O	18, 4O
Body mass index (kg/m2) at baseline
n	181	175	176
Mean (SD)	37.O6 (7.O67)	37.11 (6.841)	37.11 (6.987)
Median	35.2	36.4	35.8
Min, Max	25.2, 72.7	24.4, 6O.3	25.8, 63.6
Weight (lbs)
n	181	175	176
Mean	224.5	222.7	223.5
Median	214	213.5	21O
Min, Max	177.4, 399.4	176.5, 385	175, 388
Race, n (%)
American Indian or Alaska native	3 (1.7)	5 (2.9)	3 (1.7)
Asian	2 (1.1)	1 (O.6)	3 (1.7)
Black or African American	93 (51.4)	1OO (57.1)	89 (5O.6)
Native Hawaiian or other Pacific Islander	1 (O.6)	1 (O.6)	3 (1.7)
White	56 (3O.9)	51 (29.1)	51 (29.O)
Other	13 (7.2)	9 (5.1)	17 (9.7)
Multiple	13 (7.2)	8 (4.6)	1O (5.7)
Ethnicity, n (%)
n	181	175	176
Hispanic or Latino	34 (18.8)	33 (18.9)	43 (24.4)
Not Hispanic or Latino	143 (79.O)	141 (8O.6)	132 (75.O)
Missing	4 (2.2)	1 (O.6)	1 (O.6)
Parameter	LNG 1.5 mg	LNG 3.O mg	UPA 3O mg
Category/Statistics	(N = 181)	(N = 175)	(N = 176)
Highest level of education completed, n (%)
n	181	175	176
Less than high school	9 (5.O)	6 (3.4)	11 (6.3)
High school or equivalent	56 (3O.9)	51 (29.1)	47 (26.7)
Some college, no degree	49 (27.1)	51 (29.1)	56 (31.8)
2-y college, associate’s	32 (17.7)	23 (13.1)	34 (19.3)
4-y college, bachelor’s	29 (16.O)	35 (2O.O)	24 (13.6)
Master’s or equivalent	6 (3.3)	9 (5.1)	4 (2.3)
Doctorate or equivalent	O	O	O
Marital status, n (%)
n	181	175	176
Never married	134 (74.O)	147 (84.O)	136 (77.3)
Married	2O (11.O)	2O (11.4)	19 (1O.8)
Separated	13 (7.2)	1 (O.6)	6 (3.4)
Divorced	14 (7.7)	7 (4.O)	15 (8.5)
Widowed	O	O	O
Nulligravid, n (%)
n	181	175	176
Yes	46 (25.4)	59 (33.7)	47 (26.7)
No	135 (74.6)	114 (65.1)	128 (72.7)
Missing	O	2 (1.1)	1 (O.6)

LNG, levonorgestrel; UPA, ulipristal acetate.

The incidence of pregnancy for the efficacy-evaluable population was no different between groups. Five pregnancies occurred during the study [LNG1X n = 1; LNG2X n = 1; UPA n = 3, overall pregnancy rate 1.03% (95% CI 0.334, 2.38)] (Table 2). We also saw no meaningful differences between groups in the incidence of pregnancy when grouped according to the timing of unprotected sex relative to the estimated day of ovulation (Table 3). Incidence of pregnancy was too low to analyze differences between groups as it relates to the onset of treatment following unprotected sex exposure (0–24 hours, >24 to 48 hours, >48 to 72 hours) but we have included additional information regarding timing of coitus and study medication dosing in supplementary data materials. Two serious adverse events occurred: 1 participant in the LNG 1.5 mg group experienced an unrelated event while on study (diverticulitis) and 1 participant in the UPA 30 mg group experienced an ectopic pregnancy.

Table. 2.

Post-treatment pregnancies (efficacy-evaluable population)

Characteristic	LNG EC 1.5 mg (n -173)	LNG EC 3.0 mg (n-158)	UPA 30 mg (n-157)
Post-treatment pregnancies
Number pregnancies (%)	1 (0.58)	1 (0.63)	3 (1.91)
95% Confidence interval	0.02. 3.18	0.02. 3.48	0.4, 5.48

EC, emergency contraception; LNG. levonorgestrel; UPA, ulipristal acetate.

Table. 3.

Summary of coitus and conception events relative to the estimated day of ovulation (ITT)

	LNG 1.5 mg (N=173)		LNG 3.0 mg (N=158)		UPA 30 mg (N=157)		Total All groups
Coitus Day Relative to Estimated Day of Ovulation*	Number of Subjects	Number of Pregnancies	Number of Subjects	Number of Prcgnancics	Number of Number of Subjects Prcgnancics	Number of Subjects	Number of Prcgnancics
>= 7 Before	31	1	21	0	21	1	72	2
6 Before	12	0	8	0	10	0	30	0
5 Before	5	0	6	0	7	0	18	0
4 Before	12	0	6	0	10	1	28	1
3 Before	11	0	9	0	5	0	26	0
2 Before	7	0	10	1	9	0	26	1
1 Before	11	0	5	0	11	0	27	0
0 Before	4	0	10	0	7	0	21	0
1 After	4	0	6	0	5	0	15	0
2 After	9	0	8	0	4	0	21	0
3 After	8	0	6	0	6	0	20	0
>= 4 After	59	0	63	0	62	1	184	1

LNG, levonorgestrel; UPA. ulipristal acetate.

Gray shaded area is the highest risk of pregnancy.

^*Estimated day of ovulation = last menstrual period + average cycle length - 14.

Discussion

We found no differences in pregnancy rates between oral EC regimens in women weighing 80 kg or more but unfortunately, our study is under powered to determine if actual differences exist. Current clinical evidence suggests that LNG EC is ineffective in women weighing more than 80 kg. Results of clinical studies suggest that pharmacokinetic (PK) differences contribute to this disparity, and that doubling the LNG dose will correct the PK profile in obese women [10]. We designed this study in hopes of identifying whether this approach improves outcomes among women weighing >80 kg at risk for pregnancy seeking EC. Given the uncertainty surrounding efficacy of oral LNG EC in this group, we included UPA EC as a comparison group. We hypothesized that LNG2X (3mg) would have similar efficacy to UPA when dosed with 72 hours of unprotected intercourse, and that standard dose LNG (1.5mg) would not.

Our study has important limitations. First, we experienced slow recruitment and fell considerably short of our enrollment goal of 1200 women. Due to low enrollment, we lack sufficient statistical power for our planned analysis. However, the absence of any indication of a difference in effect size between groups in our interim analysis supported our decision to suspend the study. While an increase in sample size can improve the confidence surrounding the point estimate of an effect leading to statistical significance, we would not expect further enrollment of this same population to result in clinically important differences in the pregnancy rates between treatment groups.

This slow enrollment surprised us, as we engaged multiple sites within the established NICHD-funded Female Contraception Clinical Trials Network with expertise in clinical trials. These sites reflected considerable geographic and demographic variation, a decided strength of our study design. Enrollment challenges were global, and despite substantial effort and investments to increase enrollment using a variety of traditional and social media-based advertisement platforms, showed no trend toward improvement over the entire study period. Further, the COVID pandemic began during the study and exacerbated difficulties in enrollment. While enrollment efforts continued at some of the sites during the COVID emergency, both exposure to unprotected sex and access to health care may have been reduced in many areas, affecting both pregnancy risk and enrollment. We used historical data to estimate risk of failure of EC. Although an untreated control group would have provided the best estimate of failure, obvious ethical considerations prevent consideration of that approach. Strengths of our study include a prospective, randomized evaluation of oral EC effectiveness in the population of interest, enrollment of a diverse group of women, the inclusion of UPA as a positive control, a rigorous set of inclusion/exclusion criteria to identify women at risk, same day ability to screen and enroll participants at all sites, and high follow-up of enrolled participants.

In contrast to prior studies [6], we found no difference in the incidence of pregnancy between the treatment groups. Of equal importance, few pregnancies occurred overall in the study. Wilcox and colleagues used data from a prospective study of 221 women who were attempting to conceive to provide estimates of the likelihood of conception with a single act of intercourse [14]. They calculated a 3% probability of pregnancy with one completely random act of unprotected intercourse. As expected, risk varied based on cycle timing from negligible on days 1–3 to nearly 9% at midcycle. Given that prior studies have suggested no efficacy for LNG 1.5 mg in individuals with BMI >30 mg/k² [6], we expected to see pregnancy rates above 3%. The fact that we observed an overall pregnancy rate of under 2% in our study, with rates of below 1% in both the standard and double dose LNG groups is inconsistent with prior studies. The low rate of pregnancy in all groups suggested that other factors beyond treatment differences influenced our results, and that continued enrollment using our planned strategy was unlikely to change our conclusions. These results suggest differences in the study populations or compliance with protocol requirements that may have influenced baseline risk for pregnancy. In short, our study population appears to have been at low risk of pregnancy at baseline.

While we do not believe that our study has fully evaluated differences between oral EC treatments, our experience provides a cautionary lesson for future population-based EC studies. The sentinel studies that established the safety and efficacy of oral EC occurred in the setting of limited options for women experiencing contraceptive failure. In other words, participation in an EC study provided access to an otherwise unavailable therapy. Now, LNG EC is widely available for individuals of all ages without a prescription at a wide variety of commercial pharmacies. While no formal studies exist evaluating motivations to participate in an EC study, EC availability likely decreases the enthusiasm to participate in research to obtain treatment, even if that treatment is available at no cost. To facilitate enrollment, our study provided compensation for research participation. Compensation is intended to offset the time and effort required to participate, and not incentivize participation. However, as EC is widely available and treatment is time sensitive, we found compensation essential for recruitment and retention. Thus, participants in our study may have exaggerated their risk of pregnancy to qualify for the study in order to receive compensation. Many of the eligibility requirements for an EC study are based solely on participant reporting and cannot be proven; for example: was the last act of intercourse within 72 hours of presenting? When was the last menstrual cycle? When did you last use hormonal birth control or take an EC? We had research staff at various sites report incidents of individuals who did not qualify on initial screening for eligibility and then call back to answer screening questions differently or even offer to have unprotected sex so that they then could qualify. To compete against over-the-counter availability of EC, study sites also needed to accommodate urgent on-demand visits in order to provide EC within 72-hours of unprotected intercourse. These logistical challenges likely contributed to low enrollment.

Our study requirements likely influenced the discrepancy in expected pregnancies in the study population as compared to historical rates. Given that we enrolled only individuals > 80 kg with regular menstrual cycles, we do not believe our results reflect low fertility due to anovulation. We suspect that these low pregnancy rates reflect the actual risk of our enrolled population; Based on self-reported cycle timing, only about 20% of our participants appeared at risk of pregnancy (Table 3). Results of a separate analysis of baseline serum samples confirmed these cycle results and also revealed the presence of contraceptive steroids (an exclusion criterion) in many participants [ref companion paper].

Given that experienced contraceptive researchers supervised enrollment at all of our study sites, it is unlikely that future investigations using similar recruitment strategies will enroll a population with a greater pregnancy risk. This calls into question results of all recent EC studies that compare results to historical expected pregnancy risk. To adjust for this problem, future EC studies will need to consider much larger sample sizes, and our recent experience suggests that enrollment will prove difficult.

Although no studies have demonstrated that EC makes a difference on a population basis, the availability of EC makes an enormous difference to an individual at risk for unintended pregnancy following unprotected sex. Results from our study are inconclusive, and do not allow us to make evidence-based clinical recommendations. Given the limitations of EC research, we may never have definitive answers. Clinicians need to make the best decisions for patients based on limited data. While our results do not support superiority of double dose 3 mg LNG over 1.5 mg for EC, the approach appears safe. However, this higher dose increases costs and does not appear to provide an improvement in delay or inhibition of ovulation [15]. The potential benefits of UPA must weigh into this consideration. An informed discussion with patients will identify the most appropriate individual treatment approach.