Using Pharmacokinetic Results to Investigate Treatment Compliance As a Potential Contributor for Contraceptive Failure

Pearl Index Study With Levonorgestrel-releasing Intravaginal Rings

R. Nave; J. Höchel; U. Mellinger; A. Kohnke; J. Elliesen; H. Schmitz


Hum Reprod. 2020;35(11):2515-2523. 

In This Article


This multicenter, open-label, uncontrolled Phase 3 study was performed in the USA and in Japan to investigate the contraceptive efficacy and safety of LNG (40 μg/day) delivered via a new IVR. In total, 1471 generally healthy women aged 18–35 years and in need of contraception were assigned to treatment. The study was prematurely terminated after approximately one-third of the planned exposure (about 4 months out of the 1 year planned) due to the high number of pregnancies. The early study termination resulted in a relative high rate of subjects not attending the follow-up visit likely due to loss of motivation or lack of understanding of the necessity. By the time the study was terminated, 30 on-treatment pregnancies had been reported. This resulted in a PI of 6.1 of relevant exposure. Since 28 of these 30 pregnancies occurred in the USA, the data suggest a marked regional difference: the PI was higher in the US study population (8.2, relevant exposure of 341.7 woman-years, 1166 participants), while in the Japanese study population, the LNG IVR demonstrated good contraceptive efficacy with a PI of 1.3. However, the results from the Japanese subgroup are based on a relatively small number of participants (305 participants) with on average 6.5 treatment cycles resulting in 154.0 woman-years of relevant exposure.

The PK of LNG was assessed in subgroups both in the USA and in Japan. In Japan, the median plasma concentration of LNG (331 ng/l) after 6 months of treatment was as expected and in line with earlier results with the same type of IVR delivering 40 μg of LNG per day. A median concentration of 347 ng/l at the end of the wearing period was reported in that Phase 1 study in Japanese participants (Nave et al., 2018a). The PK data obtained in the current PI study indicated, however, a markedly lower exposure of LNG in the US participants with median concentration of 173 ng/l after 6 months (see Table III).

This observation in the US study population is in contrast to all previously collected PK data after placement of IVRs releasing daily 40 μg LNG, which all showed trough concentrations (Ctrough) >300 ng/l. Previous Phase 1 studies mainly conducted in Europe characterized well the PK of LNG delivered from an IVR (Nave, 2019). It was demonstrated that the LNG concentration during an IVR wearing period was stable during the 28-day period. Geometric mean concentration of Ctrough (Day 28), and Cav over a period of 28 days for a daily release of 40 μg LNG were 384 and 355 ng/l, respectively (Schultze-Mosgau et al.,2016). Similar PK parameters were obtained in another Phase 1 study with 338 ng/l at the end of the wearing period (Nave et al., 2018b). It is concluded from these findings that a single PK sample prior to IVR replacement is representative for the LNG exposure for the entire intended treatment period. The LNG monotherapy IVR was investigated prior to the PI study in a Phase 2b study resulting in geometric Ctrough values at the end of the wearing period of 329 ng/l (Nave et al., 2019). In summary, for IVRs releasing daily 40 μg LNG, Ctrough in steady state of >300 ng/l was expected based on Phase 1 and Phase 2b study experience (Nave, 2019). Therefore, the relatively low LNG plasma concentration observed in the US study population is surprising and potential reasons are discussed below.

Covariates like age and body weight affecting the PK of LNG were investigated using population PK approaches (Reinecke et al., 2017, 2018). The identified covariate effect was that total LNG exposure decreased with increasing body weight (with about 1.1%/kg). Based on the estimated covariate effect, the magnitude of the observed difference in the current study between LNG exposure in the USA (91 ng/l) and Japan (263 ng/l, see Table III) study populations cannot be explained by the difference in body weight (mean difference: 25 kg, see Table II). In addition, no differences in the LNG PK are known for different races or ethnicities. Furthermore, the SHBG concentrations at baseline and the decrease under treatment were similar in both regions and overall as expected (Supplementary Figure S1).

With respect to PK analysis, the variability in data should always be considered. In a global Phase 2b study performed in more than 100 study sites, the coefficient of variation (%CV) for LNG was in the range of 46.5–64.1% and was overall comparable with previous Phase 1 studies investigating the PK of the IVR (Nave, 2019; Nave et al., 2019). The observed CV% in this PI study was, however, considerably higher, and in particular in the USA (see Table III, e.g. 267% at Month 6 for all PK participants). This is mainly driven by the LLOQ data. In 17 out of 108 samples in the USA and in 2 out of 57 samples in Japan, the LNG concentration at steady state was below the LLOQ (10 ng/l), which is not expected with continuous use of the IVR. Based on an elimination half-life of about 24 h for LNG (Fotherby, 1995; Nave et al., 2018b) it would take approximately 5 days without wearing the IVR for a plasma LNG concentration of 300 ng/l to decrease to levels below the LLOQ. In addition, in 5 (out of 108 samples) in the USA, the LNG concentration at steady state was below 50 ng/l, which is assumed to be an unexpectedly low concentration. Thus, the lower LNG exposure in the US participants cannot be explained by factors related to the PK of LNG, but rather suggests that PK samples were taken days after IVR removal (i.e. non-compliance). This was partially observed for the EOT samples, where PK samples were taken at the visit although the IVR had been removed much earlier, i.e. when the participant was informed about the termination of the study (true EOT before EOT visit; even though the participant was advised to continue to wear the IVR until EOT visit). This is also supported by the observation that for most participants with an LNG concentration below the LLOQ in one sample, the concentration in the other sample taken at a different visit was in the normal range (see Figure 1).

The systemic LNG exposure data does not support the treatment compliance documented in the eDiaries. It turned out that the documentation of the wearing and replacement of the vaginal ring in the eDiary had been a challenge itself. Participants were supposed to document the date and time of each removal and insertion. There were, however, cases where these dates were either completely missing or incomplete so that the documentation suggested a prolonged wearing interval or wearing of two rings at the same time.

Participant characteristics can strongly influence the PI in clinical studies of hormonal contraceptives. It has been shown that in particular, Hispanic ethnicity, previous pregnancies and no previous use of hormonal contraceptives result in a higher PI (Gerlinger et al., 2014). Participant failures can be caused by non-compliance, significantly more Hispanic women, Black women and women who have previously been pregnant are likely to be in the group of non-compliant participants (Westhoff et al.,2012). Differences in compliance were obvious in another study, where Hispanic and Black women had fewer months of contraceptive coverage than White women (Borrero et al., 2013). Furthermore, it is known that the region also has some influence on the PI. The PIs derived from the USA are usually higher than those from Europe, e.g. Flexyess™ (Europe: 0.64 vs USA: 1.65) as discussed in literature (Rosenberg et al., 1995; Gerlinger et al., 2014). In this context, it should be noted that in our trial out of the 28 pregnancies in the USA: 16 were Black women, 7 were Hispanic women, 19 were parous women and 21 women had not used hormonal contraceptives (18 × barrier methods and 3 × none) as a contraceptive method at screening (Supplementary Table SII). All these observations are in line with reasons for contraceptive failure in the previously described publications.

The correct use of contraceptive methods directly affects the rate of pregnancy. Perfect use refers to the efficacy of the method when used correctly, while typical use refers to the efficacy during actual use (including inconsistent or incorrect use). Imperfect use has an impact on the pregnancy rates depending on how hard it is to use that method perfectly and on how unforgiving the contraceptive method is of imperfect use (Trussell et al., 2018). The tested LNG-IVR is a low-dose progestogen-only contraceptive method, which is very sensitive to incorrect use due to the fact that many women continue to ovulate while using this method. Even for vaginal rings that inhibit ovulation, the difference between 'perfect use' and 'typical use' was reported to be quite large, with a failure rate of 0.3% for perfect use and 7% for typical use. The main reason for these higher failure rates is incorrect use (Trussell et al.,2018). For example, the Nuvaring (releases 120 μg of etonogestrel and 15 μg of ethinylestradiol daily for 3 weeks with a 1-week ring-free period (NuvaRing, Prescribing Information) had a PI of 0.65 (95% CI 0.24; 1.41) from the NuvaRing Study in Europe. There was a total of six pregnancies: three of these six women appeared to have substantially violated compliance with Nuvaring in the cycle of conception (Roumen et al., 2001). In another PI study conducted in the USA, 21 pregnancies occurred, and 11 of these were attributable to non-compliance (Dieben et al., 2002). The overall PI (for both user and method failure) was 1.75 (95% CI 0.98; 2.89) for North America.

Contraceptive efficacy of LNG has been demonstrated following different routes of administration (e.g. orally, implants). The release rate and dose of LNG for the IVR was chosen to achieve an exposure similar to that of the approved low-dose progestin-only LNG pill (Norgeston®/Microlut®, 30 μg/day) and the LNG implant (Norplant II®/Jadelle®) after a wearing period of 2 years. For both products average plasma concentration around 300 ng/l are reported and therefore at least this threshold should safeguard contraception (Jadelle® Prescribing information; Norgeston® Prescribing Information; Reinecke et al., 2018). Exposure above this target concentration was reached in all previous studies using IVR delivering 40 μg LNG per day (Nave, 2019). The PK data from our study indicated that such a threshold seem to be reasonable considering the pregnancies and the low LNG concentrations reached under typical use conditions in the USA.

In conclusion, the PK evaluation in a study subpopulation indicated that the steady-state concentration of plasma LNG was markedly lower in the participants in the USA compared to Japan. This cannot be explained by intrinsic factors related to the PK of LNG in the two geographic groups, but rather suggests non-compliance in the US participants. The contraceptive efficacy of a POP-like contraceptive method is known to be very dependent on the user. Near perfect use would be a prerequisite for efficient contraception. The PK results suggest that non-compliance with the requirement of continuous wearing of the vaginal ring must have been one important reason for the lack of efficacy found in this Phase 3 study. This is an important finding in particular considering the small number of PK samples needed for the evaluation of the LNG exposure using IVRs. The non-compliance together with the choice of an LNG dose that is not inhibiting ovulation may explain the insufficient contraceptive efficacy of the LNG-IVR.