Pharmacokinetics of levonorgestrel and etonogestrel in rat or minipig following intravenous, subcutaneous, or intradermal administration

Abstract The objective of these studies was to determine the pharmacokinetics of levonorgestrel and etonogestrel in Sprague-Dawley rat or Göttingen minipig following various administration routes. Four sequential crossover studies were conducted: Study 1 administered levonorgestrel 30 µg intravenously and intradermally in four minipigs; Study 2 administered levonorgestrel 30 µg intravenously in 12 rats; Study 3 administered levonorgestrel 60 µg intravenously and subcutaneously in 12 rats; and Study 4 administered etonogestrel 30 µg intravenously in 12 rats. Samples were quantified using liquid chromatography-tandem mass spectrometry and pharmacokinetic parameters were estimated via noncompartmental analysis. C max and AUCinf for etonogestrel and levonorgestrel were similar following 30 µg intravenous bolus in rats, suggesting comparable pharmacokinetics. Levonorgestrel exposure was dose-proportional in rats, based on two-fold higher AUCinf following levonorgestrel 60 versus 30 µg. The bioavailability of intradermal and subcutaneous levonorgestrel was 97.7% (Study 1) and 90.3% (Study 3), respectively. The minipig levonorgestrel clearance was 21.5 L/hr, which was about 20-fold higher than both the rat levonorgestrel (range: 0.985–1.45 L/hr) and etonogestrel clearance (range: 0.803–0.968 L/hr). These studies contribute to the gap in knowledge of nonclinical levonorgestrel and etonogestrel pharmacokinetics, which is necessary for the ongoing development of long-acting reversible contraceptives.


Introduction
Long-acting reversible contraceptives (LARC) are among the most effective methods of birth control, which consist of intrauterine devices, long-acting injectables like depot medroxyprogesterone acetate, and rod implants of levonorgestrel (LNG) and etonogestrel (ENG) (Mansour et al. 2010). Developing innovative alternatives and increasing accessibility to these contraceptives remains a global health priority, especially in addressing the issue of unintended pregnancies in low-and middle-income countries (Sully et al. 2020). The main challenge of LARC development, however, is predicting human pharmacokinetics (PK) based on the observed nonclinical PK. More specifically, predicting the clinical release rate based on nonclinical data is an essential step in determining whether to move forward with a LARC formulation. The release rate can be calculated using either basic PK principles or mathematical deconvolution of the drug absorption profile (Margolskee et al. 2016), but in either case the intravenous (IV) PK is necessary. A review of publicly available literature suggests that there is limited and outdated information on nonclinical PK of synthetic progestins, especially IV PK. Our studies aim to determine the IV PK of LNG and ENG in Sprague-Dawley rat or G€ ottingen minipig and estimate the bioavailability of LNG administered subcutaneously (SC) and intradermally. The results from our studies narrow the gap in the literature on nonclinical LNG and ENG PK and provide valuable IV PK necessary to estimate LNG and ENG release rate from LARC formulations in development.

Materials and methods
Four nonclinical sequential crossover studies were conducted to assess PK following LNG or ENG in rats and LNG in minipig administered via various routes.
predefined phase, with a 14-or 15-day washout period between phases. LNG 30 mg was administered as an IV bolus injection with a one-minute push in Phase 1; as a 24-hr IV infusion in Phase 2; and as two intradermal injections (at 0.15 mL per site) in Phase 3. Four female G€ ottingen minipigs were enrolled for this study, five months of age at baseline with a mean body weight (± standard deviation (SD)) of 11.5 ± 0.4 kg for Phase 1 (IV bolus), 13.0 ± 0.5 kg for Phase 2 (24-hr IV infusion), and 14.2 ± 0.5 kg for Phase 3 (Intradermal). Animals were allowed to acclimate for 32 days prior to initial dose administration.

Study design and treatments
Studies 2 and 3 were prospective, two-treatment crossover studies in which the rat received two treatments sequentially, with at least a 14-day washout period. For Study 2, LNG 30 mg was administered as a 24-hr IV infusion in Phase 1 and as an IV bolus injection with a one-minute push in Phase 2. For Study 3, LNG 60 mg was administered as a 1-hr IV infusion in Phase 1 and as an SC injection in Phase 2. Female Sprague-Dawley rats were enrolled for these studies (N ¼ 12 for each study). For Study 2, rats were 14 weeks of age at baseline with a mean body weight (± SD) of 286 ± 10 g for Phase 1 (24-hr IV infusion) and 305 ± 12 g for Phase 2 (IV bolus). For Study 3, rats were 18 weeks of age at baseline with a mean body weight (± SD) of 247 ± 23 g for Phase 1 (1-hr IV infusion) and 251 ± 17 g for Phase 2 (SC injection). Animals were allowed to acclimate for 12 days in Study 2 and three days in Study 3 prior to initial dose administration.

Study design and treatments
Study 4 was a prospective, two-treatment crossover study in which rats received two treatments sequentially, with a 14day washout period. ENG 30 mg was administered as a 24-hr IV infusion in Phase 1 and as an IV bolus injection with a one-minute push in Phase 2. 12 Female Sprague-Dawley rats were enrolled for this study, 14 weeks of age at baseline with mean body weight (± SD) of 270 ± 11 g for Phase 1 (24-hr IV infusion) and 290 ± 12 g for Phase 2 (IV bolus). Animals were allowed to acclimate for five days prior to initial dose administration.

Monitoring, dose administration, and sampling
The animals were weighed at the time of animal selection and on each day of dose administration. Cageside observations for general health and appearance were performed once daily. Detailed observations were performed on all available animals 0 (predose) and at 1, 4, and 24 hr post dose following IV bolus and at 4, 26, and 48 hr following the start of the 24-hr IV infusion. The IV bolus and infusion doses were administered via an FVC. Approximately 0.5 mL of blood was collected via the jugular cannula or jugular venipuncture. The sampling schedule was staggered so that samples were taken from a subset of four animals at four or five time points. Samples were collected at 0 (predose), 1, 2, 4, 8, 12, 24 (end of infusion), 25,26,28,30,33,36,40, and 48 hr post-dose for 24-hr IV infusion and at 0 (predose), 0.1, 0. 25, 0.5, 1, 2, 4, 8, 10, 12, 24, 36, and 48 hr post-dose for IV bolus.

Animal environment
Animals in all studies were selected for use based on overall health, body weight, and cannula patency. Animals were individually housed, except in Study 3, in which they were group housed (up to three animals per cage). Animals were provided the appropriate certified diet (PMI, Inc. for minipig and Envigo RMS, Inc. for rat) and fresh water daily. The temperature was maintained at 20-26 C and there was a 12-hr light/12-hr dark cycle. These studies were not considered to be within the scope of the Good Laboratory Practice Regulations. Animals were maintained in compliance with the Animal Welfare Act Regulations (9 CFR 3).

Dose preparation
All doses were prepared by diluting stock solutions of LNG or ENG with 10% hydroxypropyl ß-cyclodextrin (HP-ß-CD) in normal saline and then stirring until the solution was clear and colourless. Contact and solution stability experiments (Velesco Pharmaceutical Services, Wixom, MI) showed that there was minimal or no loss of potency of LNG or ENG in the vehicle (syringes, syringe infusion kits, and IV bag infusion kits) for ! 24 hr. All tested concentrations of LNG remained stable at 20 ± 2 C for five days and ENG remained stable at 2-8 C for one month.

Sample analysis
In all studies, blood samples were immediately transferred to chilled cryoracks prior to centrifugation to obtain plasma. Centrifugation began within one hour of collection. Plasma was then maintained on dry ice prior to storage at approximately À70 C until analysed.
Separate stock solutions of LNG or ENG, prepared in methanol at 1.0 mg/mL in glass bottles and maintained at 2-8 C, were used for the calibration standards (CS) and for the validation samples (VS)/quality control (QC) samples. General spiking solutions were prepared by diluting the LNG or ENG stock solutions with methanol. CS and VS/QC samples were prepared by mixing 5 parts of respective spiking solutions with 95 parts of the control matrix (K 2 EDTA minipig or rat plasma obtained from commercial suppliers). The sample volume of 100 (for Studies 2-4) or 200 mL (for Study 1) was aliquoted into a 1.2 mL 96-well plate and mixed with 25 mL internal standard (10,000 pg/mL of LNG-D6 in water:methanol 50:50) and 75 (for Studies 2-4) or 175 mL (for Study 1) of water. The mixtures were vigorously shaken, vortexed, and centrifuged. The sample mixtures were then transferred to Biotage Isolute SLEþ 200 (for Studies 2-4) or 400 mg (for Study 1) plate and loaded to the sorbent using positive pressure. After about five minutes (min), samples were eluted into a 0.8 mL or 1.2 mL 96-well plate using 700 or 750 mL of ethyl acetate:hexane 25:75 and evaporated to dryness under nitrogen at 35 C. This elution and evaporation process was performed a second time in Study 1 only. The dried mixture was reconstituted with 50 mL of water:acetonitrile:formic acid 50:50:0.1 (Studies 3 and 4) or 75 mL of water:acetonitrile:formic acid 60:40:0.1 (Studies 1 and 2).
Samples were analysed on a Waters Acquity liquid chromatograph interfaced with a Thermo Scientific TSQ Vantage triple quadrupole mass spectrometer or on a Dionex Ultimate 3000 liquid chromatograph interfaced with a Thermo Scientific TSQ Quantiva triple quadrupole mass spectrometer, both with electrospray ionisation (AIT Bioscience LLC, Indianapolis, IN). Study 1 utilised both systems, Study 2 utilised the Waters Acquity system, and Studies 3 and 4 utilised the Dionex Ultimate system. Each extracted sample was injected onto a Waters BEH C18 column (2.1 Â 50 mm; 1.7 mm) equilibrated at 30 C, using a flow rate of 0.3 mL/min. 50 mL injection volume was used for the Waters Acquity system and 25 mL for the Dionex Ultimate system.
A gradient was used, with varying ratios of Mobile Phase A (MP A), ammonium fluoride (25 mM), and Mobile Phase B (MP B). MP B was a mix of acetronitrile:hexafluoroisopropanol 100:0.5 for Studies 1 and 2 and methanol:hexafluoroisopropanol 100:0.5 for Studies 3 and 4. The following gradient was used for Studies 1 and 2: From 0 to <1 min, a ratio of MP A:MP B 60:40 was used; 1 to <2.6 min, MP A:MP B 40:60; 2.6 min to <3.7 min, MP A:MP B 10:90; and 3.7-4 min, MP A:MP B 60:40. A slightly different gradient was used for Studies 3 and 4: From 0 to <3 min, a ratio of MP A:MP B 50:50 was used; 3 to <3.1 min, MP A:MP B 5:95; and 3.1-3.5 min, MP A:MP B 50:50. The observed retention time for LNG was 1.87 min for Studies 1 and 2 and 2.45 min for Study 3. The observed retention time for ENG was 2.52 min for Study 4. For Study 1, standard curves were linear over a concentration range of 10-10,000 pg/mL with an R 2 !0.99. For Studies 2-4, standard curves were linear over a concentration range of 25-25,000 pg/mL with an R 2 !0.98. The lower limit of quantitation (LLOQ) was < 10 or 20 pg/mL for Study 1 and < 25.0 pg/mL for Studies 2-4.

Pharmacokinetic analysis
The concentration-time data for each minipig were plotted and visually inspected. Pharmacokinetic parameters for each of the treatments were estimated using noncompartmental analysis (NCA) via Phoenix WinNonlin TM Version 8.1 (Certara USA, Princeton, NJ). The following parameters were measured or estimated, as appropriate for the treatment: peak concentration following dose (C max ), area-under-the-curve until the last measurable time point (AUC last ), area-under-thecurve extrapolated to infinity (AUC inf ), half-life (t 1/2 ), clearance or apparent clearance (CL or CL/F), and bioavailability (F). AUC calculations were estimated using the linear trapezoidal method except for Study 4, in which the linear/log trapezoidal method was used. Any concentrations below the LLOQ were treated as zero for descriptive statistics and pharmacokinetic analysis. Statistical analyses were limited to descriptive statistics such as mean and standard deviation, if applicable.

Results
Figures 1-3 illustrate the concentration-time profiles from the studies: Figure 1 illustrates the LNG concentration-time in minipig for Study 1; Figure 2 the LNG concentration-time in rats for Studies 2 and 3; and Figure 3 the ENG concentration-time in rats for Study 4. There were several LLOQ values towards the last few timepoints for Study 1 and 4, which were treated as zero values, resulting in lower mean concentrations (Figure 1, Figure 3). The overall PK parameter results are summarised in Table 1. 24-hr IV infusions of LNG or ENG had lower C max values than the other routes of administration but maintained a steady concentration for the entire duration of the infusion. The IV bolus dose of LNG and ENG had the highest C max values, followed by slightly higher overall AUC inf compared to those following 24-hr IV infusions. The 1-hr IV infusion of LNG had similar PK parameter values to the SC administration of LNG in Study 3, as well as a similar concentration-time curve (Figure 2). Comparing the PK following 30 mg IV bolus of LNG and ENG in the rat (Studies 2 and 4), the C max and AUC inf for ENG were 6.67% and 25.5% higher, respectively, compared to those for LNG (C max : 91.1 vs. 85.4 ng/mL; AUC inf : 37.4 vs. 29.8 ngÁhr/mL for ENG and LNG, respectively). Comparing the PK following 30 and 60 mg of LNG in the rat (Studies 2 and 3), the AUC inf following 60 mg 1-hr IV infusion or SC injection was about twofold higher than that following 30 mg IV bolus (60.9 (1-hr IV infusion) or 55.0 (SC injection) vs. 29.8 ngÁhr/mL (IV bolus)). F of LNG was 97.7% following intradermal administration in minipig and 90.3% following SC injection in the rat. The LNG CL in minipig was 21.5 L/hr (1.87 L/hr/kg). The systemic CL of LNG and ENG in the rat following IV bolus and infusion were similar: LNG CL was 0.985-1.45 L/hr (3.31-5.07 L/hr/kg), and ENG CL was 0.803-0.968 L/hr (2.77-3.57 L/hr/kg).

Discussion
LNG and ENG were patented in 1960 (Fischer and Ganellin 2006, p. 479) and 1972(Fischer and Ganellin 2006, respectively, and studied widely in humans over the span of 50-60 years. However, there is a clear gap in the literature for LNG and ENG PK in various animal species. There are a few studies that measured nonclinical PK for assay validation purposes or for experimental formulations (e.g. intravaginal ring), but dedicated IV PK studies of LNG and ENG in animals are sparse or outdated. The results from five studies that investigated LNG and ENG PK following IV route of administration are summarised in Table 2 (D€ usterberg et al. 1981;Nair et al. 1981;D€ usterberg and Beier 1984;Naqvi et al. 1984a;Geelen et al. 1993).
There are many caveats to the cited studies in Table 2 that prevent us from comparing our study results to those from the literature. First, not all the PK parameters typically derived from IV PK studies are reported and the   concentration-time profiles are generally not available. Second, the age, strain, and body weight of animals are all known to affect drug PK. From birth to adulthood, Sprague-Dawley rats and Camborough-29 pigs experience the increased expression of the two hepatic enzymes responsible for LNG metabolism, uridine diphosphate glucuronosyltransferase (UGT) and sulfotransferase (SULT) enzymes (Hu 2017;Xu et al. 2019). Age-and species-specific changes in hepatic enzyme production may consequently result in changes to overall CL. In addition, different strains of mice have shown different plasma concentrations of antibiotics like itraconazole, despite the same weight, dose, and route of administration (MacCallum and Odds 2002). Third, the assay methodology used results in different plasma drug concentration measurements. The radioimmunoassay (RIA) method is non-specific and detects both the drug and its metabolites, which overestimates plasma concentrations at lower concentrations and underestimates at higher concentrations (Xu et al. 2014). In addition to the methodology itself, different assay sensitivities can lead to differences in the characterisation of concentration-time curves, consequently affecting overall PK parameters. While RIA sensitivities are not mentioned in the D€ usterberg study (D€ usterberg et al. 1981), Naqvi et al. (1984a) and Nair et al. (1981) cite an RIA sensitivity of 10 pg per tube, or about 50 pg/mL, whereas liquid chromatography-tandem mass spectrometry (LC-MS/ MS) from our studies detected LNG and ENG concentrations as low as 10 pg/mL for the minipig and 25 pg/mL for the rat.
Higher LLOQ values will require more extrapolation past the last measurable time point and thus greater potential inaccuracies in estimating PK parameters. Finally, protein binding differences between animal species can affect the LNG and ENG PK. LNG and ENG are > 90% protein-bound in humans to albumin and sex hormone binding globulin (SHBG), with LNG demonstrating a higher affinity to SHBG compared to ENG (FDA 1996;FDA 2001). This difference in binding is evident by the change in total LNG CL observed with decreasing SHBG levels (Kuhnz et al. 1992). The literature suggests the presence of SHBG in rabbits, dogs, and monkeys, but not in rats and pigs (Renoir et al. 1980;Jenkins and Fotherby 1981). While there is currently no published literature to demonstrate the effects of SHBG on LNG and ENG PK, SHBG adds another layer of complexity to an interspecies and interstudy comparison between our studies and the cited studies in Table 2. Thus, the availability of nonclinical IV PK information does not necessarily translate to usability of said information unless the necessary assumptions are made to account for the limitations of the studies from literature.
The other two formulations tested in our studies are SC and intradermal injections. Six other studies that studied PK of LNG and ENG SC, intramuscular (IM), and intradermal (ID) in animal species are summarised in Table 3 (D€ usterberg 1984;D€ usterberg and Beier 1984;Naqvi et al. 1984b;Ananthula et al. 2015;He et al. 2018;. The extravascular studies in Table 3 have their own set of caveats that prevent a meaningful comparison between our results and those from the literature. The cited studies from the literature use a range of formulations, from sesame oil  Arithmetic mean ± standard deviation (SD) when SD was reported.
to long-acting solutions like poly(lactic-co-glycolic acid) (PLGA). In our studies, the main vehicle for LNG and ENG administration is 10% HP-ß-CD in normal saline, which was tested to be stable and compatible with the syringe kits (i.e. no loss of LNG or ENG potency). Differences in formulations can drastically affect the release rate of the drug, consequently affecting t 1/2 and overall exposure (e.g. C max and AUC). For example, Naqvi et al. (1984b) report a longer t 1/2 and higher exposures (after dose-normalization) following LNG 32 mg in sesame oil compared to LNG 60 mg in 10% HPß-CD from our studies. Ananthula et al. (2015) report a t 1/2 of 22.1 hr following LNG 1 mg in N-methyl-2-pyrrolidine (NMP) solution. Although not reported, He et al. (2018He et al. ( , 2020 utilise microparticles and PLGA solutions for ENG delivery, which will likely prolong the t 1/2 even more. In terms of exposure, D€ usterberg (1984) reports a C max value of 6,300 ng/mL following administration of LNG 1 mg in a mix of benzyl benzoate and castor oil to Wistar rat, which is still 10-fold higher than the C max following LNG 60 mg in Sprague-Dawley rat (from Study 3) even after dose-normalization (6.30 ng/mL vs. 0.608 ng/mL). Another limitation with the cited studies is that the aforementioned publications neither report the F of their extravascular formulations nor the accompanying impulse data (i.e. IV bolus) in order to calculate relative F. In our studies, the same formulation is used for both the IV and extravascular routes, allowing us to calculate relative F from the SC and intradermal injection. Similar to the IV PK studies from the literature, there are multiple reasons that complicate a direct comparison between our study results to those from the extravascular PK studies in the literature, further highlighting the gap and need for nonclinical LNG and ENG PK. There are other nonclinical PK studies available but are excluded from this discussion due to differences in routes of administration (e.g. oral, nasal, and implant) and insufficient PK and study design information to be reported, among other reasons (Gabelnick and Hall 1987;Kuhnz and Beier 1994;Kuhnz et al. 1995;Shahiwala and Misra 2004;Ma et al. 2006;Ding et al. 2007). From our studies, the important findings are the comparability of LNG and ENG PK in rats, dose-proportionality of LNG exposure in rats, higher LNG CL observed in minipig than in rats, and F > 90% following LNG intradermal and SC injections. LNG and ENG PK following the same 30 mg IV bolus dose are comparable with one another, in which ENG C max and AUC inf values are 6.67% and 25.5% higher, respectively, compared to those for LNG. While there are notable differences between LNG and ENG, especially in terms of protein binding (FDA 1996;FDA 2001), this finding suggests a degree of interchangeability between the two synthetic progestins that demonstrate equivalent PK. The AUC inf following LNG 60 mg 1-hr IV infusion or SC injection is about two-fold higher than that following 30 mg IV bolus, suggesting dose-proportionality of LNG exposure and constant CL across doses. The estimated LNG CL in minipig is approximately 20-fold higher than that in rats (21.5 vs. 1.01 L/hr), but lower on a 'per weight' basis. While the exact reasoning for these differences in CL cannot be explained, this finding may be attributed to the number of LLOQ values observed with concentrations in minipig and suggests that a more sensitive assay should be used for minipigs following clinical doses of LNG. Finally, the high F values corroborate the high absorption of LNG and ENG when administered through the skin, both in animals and in humans (FDA 1996;FDA 2001).
One of the limitations of these studies was not using an assay that was sensitive enough to detect plasma concentrations <10 or 20 pg/mL for Study 1 with minipig. The LNG concentrations in minipig were lower than expected and thus not all PK parameters were within the acceptable criteria of the NCA to be reported. Another limitation not specifically related to the studies is the fact that we did not measure some of the physiological parameters associated with LNG and ENG, like oestradiol and SHBG levels. Comparing LNG and ENG concentrations with oestradiol and SHBG levels will allow better interpretation of the nonclinical PK results and subsequent translation to human PK.
The future direction of these studies would be to demonstrate the significance of nonclinical PK information in LARC development by summarising, implementing, and translating the study results to an in vivo release rate in women. Other directions include conducting further nonclinical PK studies in other species (e.g. rabbit and dog), comparing the accuracy and precision of the calculated in vivo release rates with the observed release rates, and exploring the applicability of this information in the context of currently accepted regulatory standards for calculating in vivo release rate from controlled-release or long-acting formulations.

Conclusions
Our studies set out to determine LNG and ENG IV PK as well as assess the bioavailability of LNG administered via extravascular routes in rats and minipig. LNG PK was comparable to ENG PK in rats, LNG exposure (i.e. AUC inf ) was shown to be dose-proportional in rats, LNG CL in minipig was higher than that of rat (but lower on a 'per weight' basis), and F of LNG in both species was high (> 90%). The findings from these studies will continue to narrow the gap in nonclinical LNG and ENG PK as well as support the ongoing development of innovative LARCs by providing IV PK data necessary to estimate in vivo release rates from controlled-release or long-acting formulations.

Disclosure statement
The authors report no conflict of interest.

Funding
Described studies and preparation of the manuscript were funded by the Bill & Melinda Gates Foundation contracted to Certara Inc. (INV-020041).

ORCID
Paul J. Ko http://orcid.org/0000-0003-3940-286X Table 3. LNG and ENG PK following SC/IM/ID injection in animal species from the previous studies a . Author  a Arithmetic mean ± standard deviation (SD) when applicable.