Impact of drug loading on release from levonorgestrel intrauterine systems

Abstract

Levonorgestrel intrauterine systems (LNG-IUSs) are polydimethylsiloxane (PDMS) based non-biodegradable complex drug-device combination products providing efficacy for up to several years based on the strength. A large amount of LNG (e.g., 52 mg in Mirena and Liletta) must be loaded in the LNG-IUS products to maintain the long-acting effect even though LNG is a potent hormone. However, the high amount of LNG not only poses the potential risk of dose dumping, but also leads to drug waste due to incomplete drug utilization close to the end of usage. It has been unclear whether the duration of usage of these products should be extended for full drug utilization or products with lower drug loading should be developed. Therefore, it is critical to understand the impact of strength (or drug loading) on drug release from LNG-IUSs. In the current study, drug reservoirs with a broad range of drug loading (from 0.5% w/w to 50% w/w) were prepared and assembled into LNG-IUSs. Different accelerated release conditions were used to perform release testing of LNG-IUSs with different drug loading. 5% to 10% variation in excipient of the LNG-IUSs did not significantly alter the drug release profiles of the LNG-IUSs. The release rate of LNG-IUSs is inversely proportional to their drug loading at high drug loading (10% w/w, 25% w/w and 50% w/w). Drug release was incomplete for LNG-IUS with low drug loading (2.5% w/w and 1% w/w) and no drug release could be detected for the LNG-IUS with 0.5% w/w drug loading. In addition, the burst effect of the LNG-IUSs with different drug loading was investigated. This is the first research report covering ultra-long duration (more than four years) of real-time drug release from LNG-IUSs with different drug loading (0.5%−50% w/w). The amount of excipient (PDMS) used in the reservoir of LNG-IUSs was determined to be not a critical quality parameter in the formulation design since LNG-IUSs (50% w/w drug loading) with up to 10% variation in excipient did not show significant differences in their release profiles. The drug release kinetics/mechanism remained the same for LNG-IUSs with drug loading ranging from 1% to 50%. In addition, the accelerated release testing methods were confirmed to be representative of the real-time release profiles and this can give confidence in extending the duration of usage for these products provided that the device remains physically intact (no tearing or damage in the outer membrane) and the release rate is within the therapeutic window. It is recommended to perform both real-time and accelerated release testing simultaneously for LNGIUSs to understand the burst effect as well as the complete release characteristics. Lastly, drug/polymer interaction may play a role when designing LNGIUS formulations with low drug loading (<5% w/w) since drug/polymer interaction is significant when only a small amount of drug present.

1. Introduction

Long-acting intrauterine systems (IUSs) such as Mirena, are hormone-releasing intrauterine devices (IUDs) intended to release the hormone (i.e., levonorgestrel (LNG)) over a period of several years. There are four FDA approved LNG-IUSs (i.e., Mirena, Liletta, Skyla and Kyleena) and all of them are used for contraception (Fanse et al., 2022). In addition, Mirena, the first FDA approved LNG-IUS for contraception up to 8 years, is also indicated for the treatment of heavy menstrual bleeding up to 6 years for women who choose to use IUS as their method of contraception (Mirena prescribing information, 2022). LNG-IUSs are a complex drug-device combination product, composed of a T-shaped polyethylene frame with a polyethylene retrieval thread, a hollow cylindrical drug reservoir containing non-biodegradable polydimethylsiloxane (PDMS) polymer and the drug LNG, together with a PDMS based outer membrane (or sheath) used for controlling the drug release rate. The physical structure of the device can be found elsewhere (Mirena prescribing information, 2022; Bao et al., 2020; Bao et al., 2019). The complex physical structure of LNG-IUSs, make their formulation design, physicochemical characterization, as well as their in vitro and in vivo performance challenging.

LNG-IUS products are PDMS based controlled drug delivery systems which can deliver the potent hormone LNG locally in the uterus for several years at a near constant release rate with a gradual decline over time. To sustain drug release for several years, a large amount of drug substance must be incorporated in the drug reservoir of LNG-IUSs, which may pose an increased risk of dose dumping at the site of action. In view of this, it is critical to understand the appropriate strength or drug loading of LNG-IUS products. The first U.S. FDA approved LNG-IUS product Mirena has been on the market for more than two decades; it was originally approved (in 2000) for 5-year use for contraception as well as the treatment of heavy menstrual bleeding. It has been reported that the drug release ~60% at the end of 5-year use in vivo (Chi, 1991; Luukkainen et al., 1986). Another LNG-IUS product Liletta, originally approved (in 2015) for 3-year use and then extended to 5-year use (in 2018 (HCPlive.com, 2022) for contraception, showed ~50% drug release at the end of the 5-year use (Creinin et al., 2016). Both products have 52 mg dose strength (~50% w/w drug loading). Namely, the utilization of the drug in the formulation is onlŷ 50% and the other half of the dose (~26 mg) is wasted. This may also cause environmental pollution if not disposed properly (Unfpa, 2022). Most recently, both Mirena and Liletta obtained US FDA approval of new supplemental new drug applications to extend the duration of use. In 2019, Liletta was approved for 6-year usage. In 2021, Mirena was approved for 7-year use and in Aug 2022 the US FDA approved Mirena to prevent pregnancy for up to 8 years (Mirena prescribing information, 2022; Formulary Watch, 2022), which makes drug utilization in the device more efficient (~80% release of the full strength).

In generic product development, it is important to confirm if minor variation (<5%) in the excipient amount (drug loading) will impact the drug release characteristics of the finished products. Generally, 5% variation in excipient amount for compositionally (Q1/Q2) equivalent formulations is acceptable (FDA ANDA Submissions, 2022). However, it is unclear if the 5% variation in excipient amount also applies to LNGIUSs since there has been no previous reports on this aspect. In addition, for a monolithic system, drug loading can have a significant impact on drug release mechanisms and characteristics. Formulations with higher drug loading have more porosity and connectivity when releasing the drug while those with lower drug loading have less connectivity of pores and channels (Baker, 1987; Barrett et al., 2018; Andersson et al., 2004). The difference in the connectivity of the pores throughout monolithic drug reservoirs may lead to different drug release patterns. Accordingly, it is important to understand if the drug loading in LNGIUSs has any effect on their release mechanisms and hence their release characteristics. Another important aspect is the potential for drug molecules to interact with the PDMS polymer. It has been reported that such drug interaction lead to low recovery of drug in a vaginal ring device (Murphy et al., 2016). Based on previous reports on LNG-IUSs with 50% w/w drug loading, no drug/polymer interaction issues were observed (Bao et al., 2018). However, it is important to investigate potential drug/polymer interaction for LNG-IUSs with different drug loading and any impact on the drug release characteristics.

Although research on LNG-IUSs can be traced back to the 1980 s, the manufacturing process, formulation design, release testing methods and mechanisms of drug release from LNG-IUSs are still not well understood. Knowledge has been gained from recent reports (Bao et al., 2020; Bao et al., 2019; Bao et al., 2018) on compositionally equivalent LNG-IUSs (using Mirena as the reference product) in the aspects of formulation preparation, physicochemical characterization, in vitro release testing (both accelerated and real-time) as well as potential release mechanisms. However, these previous reports were all based on compositionally equivalent LNG-IUSs, where the drug loading (50% w/w) and the PDMS polymer (liquid silicone rubber MED4840 from Nusil^®) of the formulations were not altered. Accordingly, no knowledge was acquired on the impact of formulation factors such as material (PDMS) attributes and drug amount (drug loading) on the performance of LNG-IUSs. The most recent research (Fanse et al., 2021; Fanse et al., 2022) explored the impact of material attributes such as crosslinking on the physicochemical properties of LNG-IUSs. The present study investigates the effect of drug loading on the release characteristics of the finished LNG-IUSs over a range (0.5% w/w to 50% w/w) of drug loading. In addition, in vitro release testing of representative LNG-IUS formulations is conducted under real-time and accelerated release conditions to understand the potential impact of the release conditions on the release characteristics (release kinetics as well as burst effect) of LNG-IUSs with different drug loading. This work will enable the use of suitable accelerated release methods for quality control and formulation optimization of IUSs. In addition, it may facilitate bioequivalence establishment of IUS products (with different drug loading) through an in vitro approach. Correlations between the in vitro drug release rate constant (release percentage) and the drug loading will be investigated, in an effort to help predict drug release rates from IUSs with different drug loading.

2. Materials and methods

2.1. Materials

Levonorgestrel with particle sizes of 16 μm (D90) and 80 μm (D90) were purchased from Tecoland Corporation (Irvine, CA, USA) and Cayman Chemical (Ann Arbor, MI, USA), respectively. PDMS tubing (outer membrane or rate-controlling membrane, Silastic^®, cat.no. 508–007) was purchased from Dow Corning (Midland, MI, USA). Liquid silicone rubber (MED-4840 Part A and Part B) was purchased from NuSil^™ (Carpinteria, CA, USA). Sodium chloride and sodium dodecyl sulfate (SDS) were purchased from Sigma-Aldrich (St. Louis, MO, USA). Tetrahydrofuran (THF) and tert-Butanol (TBA, certified) were purchased from Fisher Scientific (Fair Lawn, NJ, USA). Unless otherwise specified, all materials were of analytical grade.

2.2. Preparation of LNG-IUSs with different drug loadings

The in-house LNG-IUSs were prepared as previously described (Bao et al., 2019; Bao et al., 2018). In brief, there were two major steps: 1) preparation of the LNG-PDMS drug reservoir; and 2) fabrication of LNG-IUSs. Briefly, the LNG-PDMS drug reservoir was prepared using liquid silicone rubber (Part A: Part B = 1:1) mixed with different amounts of LNG depending on the drug loading. The LNG-PDMS mixture was transferred into a polycarbonate mold and cured to form a hollow cylindrical drug reservoir. The drug reservoirs were then cut into 100 mg pieces and enveloped onto plastic rods. Lastly, a 24 mm-long piece of outer membrane (pretreated with hexane) were symmetrically coated onto the drug reservoirs to obtain the LNG-IUSs. Residual solvent was removed under vacuum at room temperature. The finished LNG-IUSs are shown in Fig. 1. The LNG-IUSs with different drug loading are listed in Table 1.

Fig. 1.

Pictorial image of finished LNG-IUSs.

Table 1.

Product design parameters for manufacturing LNG-IUSs^*.

No.	LNG-IUSs with different drug loading	Comments
1	0.5% w/w DL	Drug loading ranging from 0.5% w/w to 50% w/w was used to investigate the impact on drug release characteristics. 50% w/w drug loading was used as a reference.
2	1% w/w DL
3	2.5% w/w DL-T*
4	2.5% w/w DL-C*
5	5% w/w DL
6	10% w/w DL
7	25% w/w DL
8	50% w/w DL
9	48.8% w/w DL	5% more in excipient amount
10	51.1% w/w DL	5% less in excipient amount
11	47.6% w/w DL	10% more in excipient amount
12	52.6% w/w DL	10% less in excipient amount

^*T and C represent the API obtained from Tecoland and Cayman, respectively. API from Tecoland was used in the formulations without T or C.

2.3. HPLC method

LNG concentration was analyzed using an Agilent 1290 Infinity HPLC system. A UV detector was used and set at 240 nm. A mixture of acetonitrile and water (60/40, v/v) was used as the mobile phase with a flow rate at 1.5 mL/min. A Phenomenex^® Kinetex^® C18 column (250 × 4.6 mm, 5 μm) was used for sample separation and the column temperature was set at 30 ^◦C. The injection volume was 50 μL. The data analysis was performed using Agilent OpenLAB CDS ChemStation.

2.4. Drug loading and content uniformity of the prepared drug reservoirs

The drug loading and content uniformity of the prepared LNG-PDMS drug reservoirs were determined using a solvent extraction method as described previously (Bao et al., 2020; Bao et al., 2018). In brief, the drug reservoirs were cut into approximately 8 mg pieces and extracted by submerging in THF for 2 days. The samples were then diluted and injected into an HPLC for analysis. To test the content uniformity, three samples from different regions of the reservoirs were obtained.

2.5. Morphological studies

The LNG and PDMS polymer distribution as well as the solid state of the LNG particles in the prepared drug reservoirs were visualized using a Keyence high resolution VHX-7000 digital microscope (Keyence Corporation, Itasca, IL, USA) with a polarized lens and an FEI Nova Nano-SEM 450 scanning electron microscope (FEI company, Hillsboro, OR, USA). The drug reservoirs with different drug loadings were laminated into thin slices. The samples were then mounted on carbon taped aluminum stubs and sputter coated with gold for 3 min at 6 mA with a deposition rate of 6 nm/min prior to imaging.

2.6. In vitro drug release testing of LNG-IUSs

In vitro drug release from the LNG-IUSs was carried out using a water shaker bath as previously described in the method section of (Bao et al., 2019). The LNG-IUSs were placed in PYREX^® glass bottles (Corning Inc., NY, USA) with screw caps. The real-time release testing of the LNG-IUSs was conducted at 37 ^◦C in normal saline with a rotation speed of 100 rpm. The rotation speed was setup initially to ensure the samples inside the bottles do not move around (the samples will move around when the rotation speed is above 120 rpm) and yet have reasonable release profiles. This condition has been proven to be very close to the in vivo situation (data not shown). Using the RLD Mirena, the in vitro release rate tested under this condition is close to the in vivo release rate (the release testing is still ongoing due to the extremely long release duration). We expect that rotation speed may have some impact on the release rate if not restricted by the instrumentation setup. The other constraint is time. It would take several years to investigate the impact of different conditions on the final release rate, which is extremely challenging. Based on previous studies (Bao et al., 2019), release from LNGIUSs can be accelerated via different conditions such as the addition of alcohols and surfactants (e.g., SDS) as well as via elevation of temperature. In this study, to fully understand the impact of accelerated conditions on release from the LNG-IUSs with different drug loading, six different conditions were investigated (Table 2). The sampling plans for all the tests are listed in Table 2. The release testing experiments were carried out under sink conditions and performed in triplicate.

Table 2.

Accelerated release conditions used in the in vitro release testing of LNG-IUSs with different drug loading.

No.	Accelerated release conditions	Sampling Plan
1	20% v/v THF, 80% v/v pH 7.4 PBS with 0.25% w/v SDS at 45 °C; 100 mL	1 mL samples were withdrawn at 2, 4, 6, 8 and 24 h eveiy day and fresh media was replenished after sampling; following sampling at 24 h, the media were drained and replenished with fresh media.
2	20% v/v TBA, 80% v/v pH 7.4 PBS with 0.25% w/v SDS at 45° C; 100 mL	1 mL samples were withdrawn oil days (1, 2, 3 and 4) in the first week, and replenished with fresh media. Following the day 7 sampling, all the media in the bottles were drained and replenished with fresh media. Thereafter, samples were withdrawn eveiy subsequent fourth and seventh day. The media were drained and replenished following sampling.
3	20% v/v TBA, 80% v/v pH 7.4 PBS with 0.25% w/v SDS at 65° C; 100 mL	Same as Condition 2
4	70% v/v TBA, 30% v/v water at 45 °C; 250 ml,	1 mL samples were withdrawn eveiyday up to 10 days and replenished with fresh media. Following the day 10 sampling, all the release media were drained and replenished. Thereafter, samples were withdrawn eveiy subsequent fourth and seventh day. The media were drained and replenished following sampling.
5	70% v/v TBA, 30% v/v water at 65 °C; 250 ml.	1 mL samples were withdrawn eveiyday up to 10 days and replenished with fresh media. Following the day 10 sampling, all the release media were drained and replenished. Thereafter, samples were withdrawn thrice a week (on alternate weekdays). The media were drained and replenished following sampling.
6	45% v/v TBA, 55% v/v water at 65 °C; 200 mL	1 mL samples were withdrawn eveiyday up to 10 days and replenished with fresh media. Following the day 10 sampling, all the release media were drained and replenished. Thereafter, samples were withdrawn thrice a week (on alternate weekdays). The media were drained and replenished following sampling on the seventh day of eveiy other week.

2.7. Statistical analysis

The data analysis and plotting were performed using OriginPro 2017 software (OriginLab Corporation, Northampton, MA, USA) and Excel 2016 (Microsoft Corporation, Redmond, WA, USA). All data were presented as average ± standard deviation (SD).

3. Results and discussion

3.1. Drug loading and content uniformity of the reservoir

All the prepared LNG-IUSs with higher drug loading achieved the target drug loading (over 90% drug recovery) with good content uniformity (less than5% of RSD) (Fig. 2). However, the LNG-IUSs with lower drug loading (less than 5% w/w) showed low drug recovery. For example, the LNG-IUSs with 0.5% w/w drug loading had less than 10% drug recovery. The low percentage of drug recovery for the LNG-IUSs with low drug loading may be due to drug/polymer interaction (or drug binding) during the manufacturing process (Murphy et al., 2016). The LNG particle size used in LNG-IUS preparation also played a role in the drug recovery. The particle size of both APIs obtained from Tecoland and Cayman were tested using polarized light microscopy with ImageJ. The results showed that the API particle sizes were 12 μm and 31 μm from Tecoland and Cayman, respectively. Tecoland LNG-IUSs with larger particle size (2.5% w/w DL-C, 31 μm) had higher drug recovery compared to LNG-IUSs prepared with smaller particle size (2.5% w/w DL-T, 12 μm). This may be explained by the smaller drug particles having a larger specific surface area, thus rendering a higher chance of drug/polymer interaction, and vice versa. The drug/polymer interaction effect was negligible for LNG-IUSs with high drug loading. However, this became a prominent issue when manufacturing LNG-IUSs with lower drug loading (less than 5% w/w) since the dose accuracy is more difficult to control at lower drug loading.

Fig. 2.

Drug recovery of LNG-IUSs with different drug loading (w/w) (mean ± SD, n = 3). * T represents the API from Tecoland Corporation with a particle size of ~12 μm and C represents the API from Cayman Chemical with a particle size of ~31 μm.

3.2. Morphologies of the LNG-IUS drug reservoirs with different loadings

The morphologies of the prepared LNG-IUSs with different drug loading were characterized using high resolution light microscopy using a polarized lens (Fig. 3). A blank formulation (without API) was also characterized for comparison. The drug was shown to disperse uniformly in all the prepared drug reservoirs and remained in the crystalline state. LNG-IUSs with 0.5% w/w and 1% w/w drug loading showed comparatively less visible drug particles as compared with other IUSs with higher drug loading. This was consistent with the low drug recovery determined by the drug loading and content uniformity test. The background of formulations with a drug loading less than 0.5% w/w showed brighter and clearer boundaries of the particles, while the IUSs with a drug loading more than 10% w/w exhibited a darker or concretelike background. This was due to different degrees of transparency of the IUSs prepared with different drug loading. IUSs with less drug loading are slightly translucent, whereas IUSs with higher drug loading are opaque. To have a better morphological view of the IUSs with a drug loading higher than 10% w/w, the SEM images were captured (Fig. 4). The particle density of the IUSs were consistent with their individual drug loading. The higher the drug loading, the more abundant the drug particles in the PDMS matrix. In addition, the LNG particle size in all the reservoirs showed very similar size regardless of drug loading.

Fig. 3.

Polarized microscopic images of LNG-IUSs with different drug loading (scale bar: 100 μm; magnification: 500 fold).

Fig. 4.

SEM images of the prepared LNG-IUSs with 10% w/w, 25% w/w and 50% w/w drug loading (scale bar: 100 μm; magnification: 500 fold).

3.3. In vitro release testing for the LNG-IUSs with different drug loading

3.3.1. Drug release from LNG-IUSs with 5% and 10% variation in excipient amount

Based on the FDA guidance (FDA ANDA Submissions, 2022), the variation (or tolerance) of excipient amount in formulating Q1/Q2 equivalent formulations should be controlled to less than 5% (or with a concentration in the range of 95%−105%). To understand the effect of excipient variation on drug release from the finished formulations, LNG-IUSs with 50% w/w drug loading (similar to the commercial product Mirena), as well as formulations with 5% and 10% variation in the excipient amount (formulations 8–12 in Table 1) were prepared and release testing was conducted under accelerated conditions (Table 2, condition 2). The release profiles of these formulations with variations in excipient amount (5% and 10% difference) did not show significant differences in release rate, compared to the reference LNG-IUS with 50% w/w drug loading (Fig. 5), indicating the drug release rate is not significantly affected by up to 10% variation in the inactive ingredients for the LNG-IUS formulations.

Fig. 5.

Accelerated release profiles of the LNG-IUSs with 50% w/w drug loading and formulations with: (A) 5% variation in excipient amount; and (B) 10% variation in excipient amount. The release testing was performed in 20% v/v TBA and 80% v/v PBS containing 0.25% w/v SDS at 45 °C (condition 2) in a water shaker bath at 100 rpm (mean ± SD, n = 3).

3.3.2. Real-time release testing

The real-time release testing of the LNG-IUSs with 10% w/w, 25% w/w and 50% w/w drug loading was performed using a water shaker bath at 37 °C as described previously (Bao et al., 2019). The release testing duration was approximately 1400 days (~4 years). Due to the ultra-slow release rate of this system, drug release was not complete for all the LNG-IUSs, with approximately 60%, 40% and 20% release for the LNG-IUS formulations with 10% w/w, 25% w/w and 50% w/w drug loading, respectively (Fig. 6). It is worth noting that all the release profiles of the LNG-IUSs showed very consistent near-zero or first order release kinetics as reported previously (Bao et al., 2018), which means that release from LNG-IUSs with different drug loading showed very predictable release behavior over ultra-long durations (e.g., 4 years). In this sense, the drug release from LNG-IUS products is steady and well controlled with time. This gives confidence in extending the duration of usage for these products as long as the device remains intact (no tearing or damage in the outer membrane) and the release rate is within the therapeutic window.

Fig. 6.

Real-time release profiles of the LNG-IUSs with 10% w/w, 25% w/w and 50% w/w drug loading. Release testing was performed in 0.9% w/v saline at 37 ◦C in a water shaker bath rotated at 100 rpm (mean ± SD, n = 3).

3.3.3. Accelerated release testing under different conditions

According to a previous report on the LNG-IUSs, several different accelerated conditions have been used to understand drug release from LNG-IUSs (Bao et al., 2019). To understand the impact of drug loading on release from LNG-IUSs, several conditions were investigated here (listed in Table 2). In a previous report (Bao et al., 2019), THF was shown to tremendously speed up drug release from IUSs, reducing test duration from many years to approximately one month to complete drug release from IUSs with 50% w/w drug loading. However, THF was not selected in further studies since it accelerated drug release from IUSs in an ultra-fast manner. In addition, THF solvent can be oxidated over time, leading to health concerns. TBA was selected to further investigate and optimize conditions for accelerated drug release from the LNG-IUSs. The release temperature (45 °C and 65 °C) and concentration of TBA (20% v/v, 45% v/v and 70% v/v) were considered to optimize the release testing conditions. It has been reported that higher temperature and/or higher concentrations of organic solvents may accelerate drug release to a large extend (Bao et al., 2019). The impact of temperature on the drug release profiles of LNG-IUSs have been investigated previously (Bao et al., 2019). The results demonstrated that the relationship between the temperature and the release rate (either first order or zero-order release model) followed the Arrhenius equation. The underlying reasons are that the solubility of the drug increases with elevation of temperature, which enhances the drug dissolution and accordingly drug diffusion through the matrix and the outer membrane of the LNG-IUSs. To develop appropriate accelerated release testing methods, several factors must be taken into consideration: (1) drug release kinetics and/or mechanism(s) should not be changed in the accelerated conditions compared to the real-time release conditions; (2) the duration of release testing should be practical such as 2–3 months instead of years; (3) simpler release media composition will make the experimental procedures straightforward and hence reduce human associated errors; and (4) sampling intervals should be sufficiently tight to cover the critical drug release percentages of the full release profiles.

3.3.4. Release media containing both surfactants and organic solvents

In a previous report (Bao et al., 2019), accelerated drug release was performed in release media containing low levels of organic solvents (i.e., 20% v/v) as well as surfactants (0.25% w/v SDS). Two of these conditions (Fig. 7) that have been shown to tremendously accelerate drug release from LNG-IUSs were utilized to investigate the release characteristics of the IUSs prepared with different drug loading⋅THF showed the strongest ability to accelerate drug release from for all different drug loading IUSs (Fig. 7A). However, THF is not suitable for long term use in release testing due to the above-mentioned reasons. Among all the hydroalcoholic media, it has been shown that alcohols with a larger number of carbons have a greater degree of drug release acceleration from LNG-IUSs. TBA hydroalcoholic media containing SDS showed a similar release trend as the THF release media, but with a slower release rate. It took approximately 4 months for complete drug release, using TBA hydroalcoholic media at 65 °C, from IUSs with 50% w/w drug loading. The methods showed good discriminatory ability among the IUSs with different drug loading. In addition, all the release profiles followed first order release kinetics.

Fig. 7.

Release profiles of the prepared LNG-IUSs with 10% w/w, 25% w/w, and 50% w/w loading under different conditions (mean ± SD, n = 3): (A) 20% v/v THF in 80% v/v pH 7.4 PBS with 0.25% w/v SDS at 45 ◦C (condition 1); and (B) 20% TBA in 80% v/v pH 7.4 PBS with 0.25% w/v SDS at 65 °C (condition 3).

3.3.5. Release media containing only water and TBA

To have a simpler composition of release media compared to the surfactant/organic release media, accelerated release testing of LNG-IUSs was investigated using water-TBA release media at different ratios (i.e., 45% and 70%, v/v) and elevated temperatures (i.e., 45 °C and 65 °C). As shown in Fig. 8, the release durations of the LNG-IUSs with 50% w/w loading in all the testing conditions were less than 6 months. Similar to the other accelerated release testing above, all the release profiles followed first order release kinetics with good discriminatory ability for the IUSs with different drug loading.

Fig. 8.

Release profiles of the prepared LNG-IUSs with 10% w/w, 25% w/w, and 50% w/w loading under different conditions (mean ± SD, n = 3): (A) 70% v/v TBA in water at 45 ◦C (condition 4); (B) 70% v/v TBA in water at 65 °C (condition 5); and (C) 45% v/v TBA in water at 65 °C (condition 6).

The release rates of all the release profiles under different release conditions were obtained using first-order modeling (Fig. 9A) and the relationship between the release rates and the drug loading was also investigated (Fig. 9B). The release rate (in terms of release percentage) was inversely proportionally to the drug loading for all accelerated conditions, following a similar trend to the real-time release profiles. The smaller the drug loading, it is understandable that it will take a shorter time to finish the drug release and vice versa. It is not clear whether the accelerated release conditions changed the release mechanism of the high drug loading IUSs (i.e., 50% w/w) since only approximately 20% of drug has released under the real-time conditions (to date), which is far from complete release. The release profiles of LNG-IUSs with low drug loading (i.e.,10% w/w) reached almost 60% drug release (to date), and therefore may be representative of the complete release profiles. The release profiles obtained from all the accelerated conditions were compared with the real-time release for IUSs with 10% w/w drug loading in order to determine whether correlations can be established (Fig. 10). With appropriate time scaling, all the accelerated release profiles matched well with the real-time release profiles, indicating that any of the developed accelerated conditions may be used to understand the overall real-time release characteristics, at least for the 10% w/w drug loading formulation.

Fig. 9.

(A) First-order release rate constants of LNG-IUSs with different drug loading (10% w/w, 25% w/w and 50% w/w) under different release conditions (real-time and accelerated); and (B) linear regression of the release rate constant against the reciprocal of drug loading.

Fig. 10.

Correlations between release profiles obtained from different accelerated and real-time release conditions for LNG-IUSs with 10% w/w drug loading.

3.3.6. LNG-IUS formulations with 5% or less drug loading

The release profiles of the LNG-IUSs with a drug loading of 5% w/w or less are presented separately due to their abnormal release characteristics. For example, no drug was detected in the drug release testing (in condition 3) for the LNG-IUSs with 0.5% w/w drug loading (data not shown). For the IUSs with 1% w/w and 2.5% w/w drug loading, the release reached a plateau with incomplete release percentages, ~20% and 75% release for LNG-IUSs with 1% w/w and 2.5% w/w drug loading, respectively (Fig. 11). The final release percentage of IUSs with 5% w/w drug loading is close to that of the IUSs with 10% w/w drug loading, with approximately 90% drug release. The incomplete drug release from IUSs with low drug loading may be due to potential drug binding to the PDMS polymer (Murphy et al., 2016) and/or the inability of the drug to diffuse out of the matrix/membrane. These findings further corroborate with the hypothesis described in section 3.1, where LNG-IUSs with low drug loading showed significantly low drug recovery. Solid state characterization (such as solid-state NMR (Figs. S1 and S2) and FTIR (Fig.S3)) was utilized to identify any polymer drug interaction (refer to supplementary data). However, no peak shift or changes were observed, which may be due to the low detection limit of the current instrumentation.

Fig. 11.

(A) Release profiles of the prepared LNG-IUSs with lower drug loading (1% w/w, 2.5% w/w and 5% w/w) under different conditions (mean ± SD, n = 3) 20% v/v TBA in 80% v/v pH 7.4 PBS with 0.25% w/v SDS at 65 ◦C (condition 3); and (B) zoom-in view of the release profiles.

Previous reports have shown that monolithic drug delivery systems with different drug loading show different release mechanisms (Huang et al., 2006; Kim, 1998; Abbasnezhad et al., 2020; Wang et al., 2020). There has been no previous report on the impact of drug loading levels on the drug release characteristics of LNG-IUSs. It can be assumed that different drug loading levels might alter the release kinetics or mechanisms, considering LNG-IUSs with different drug loading (or concentration) have different distributions of drug particles in the polymer matrix. Intriguingly, other than incomplete release due to potential drug/polymer interaction, the release kinetics of the LNG-IUSs with low drug loading were similar to the LNG-IUSs with high drug loading (near zero-order or first order). This may be explained by the limiting step of release from LNG-IUSs being drug diffusion through the outer membrane of these complex reservoir-monolithic combinational drug delivery systems.

3.3.7. Burst effect

The burst effect refers to a significant amount of drug being rapidly released during a short period of time and this phenomenon occurs in many sustained and controlled drug delivery systems (Yoo and Won, 2020; Andhariya et al., 2019; Ficek and Peppas, 1993). Although the burst effect can be advantageous in some situations such as wound treatment and pulsatile release system design, the burst effect is undesirable in many cases and can be unpredictable. Burst release of drug may lead to toxicity, limited drug utilization, frequent dosing and shortened release profiles (Huang and Brazel, 2001). LNG-IUSs have been reported (Natavio et al., 2013; Wildemeersch and Andrade, 2010) to have an initial burst effect in clinical studies. However, this effect may not be reflected in all accelerated and real-time release profiles when plotting the release percentage against time. This is due to the strength of IUSs being ultra-high (52 mg LNG in IUS formulations with 50% w/w drug loading) compared to the daily release amount (e.g., 5–10 μg). Even if there is a high initial burst effect, the release profiles (release percentage vs time plot) will not show prominent burst release effects since the drug percentage release in the burst effect will be relatively small. The initial burst effect was observed in the real-time release profiles (Fig. 12A) for IUSs with high drug loading (i.e., 25% w/w and 50% w/w), but not in the IUSs with low drug loading (i.e., 10% w/w). The initial burst effect occurred immediately after the IUSs were placed in the release media (day 3 with detectable drug release). After approximately one month of release, the release rate tended to be constant. In contrast to the real-time release profiles, the accelerated release profiles showed an initial low release rate phase, followed by a peak rate and then a gradual linear rate decline phase (Fig. 12BCD). The accelerated release profiles do not reflect the initial burst release effect since the release rate is too high to pick up this initial phase.

Fig. 12.

Release profiles by plotting release rate (μg/day) against time (day) of IUSs with different drug loading (10% w/w, 25% w/w and 50% w/w) under the following release conditions: (A) real-time release; (B) accelerated release in 70% v/v TBA in water at 45 ◦C; (C) accelerated release in 70% v/v TBA in water at 65 °C; and (D) accelerated release in 45% v/v TBA in water at 65 °C.

The initial burst release from IUSs may vary based on the manufacturing process. The LNG-IUSs in the current study were manufactured through three major steps (Bao et al., 2018): (1) mixing PDMS prepolymers and LNG; (2) curing process to form hollow cylindrical drug reservoirs; and (3) placement of the sheaths or outer membranes. During step 3, the PDMS based outer membranes need to be swollen in hexane to enlarge the diameter of the tubing so that it can easily pass over the drug reservoir. The drug on the surface of the reservoir may be dissolved when contacting with the hexane swollen membrane. After vacuum drying, this dissolved drug may migrate to the surface of the outer membrane where it is readily available, leading to the burst effect. Other reasons for burst release include high release from the two ends of the device since the ends are not sealed (Fig. 1). The IUS with lower drug loading had an unnoticeable burst effect since the PDMS matrix is more predominant (Fig. 3) and the free drug on the surface of drug reservoir will be negligible.

3.3.8. Considerations of in vitro release testing method selection

The in vitro release testing of LNG-IUSs has been challenging due to the ultra-long duration of drug release (up to 8 years for Mirena). Development of appropriate release testing methods will benefit both formulation development and product quality control (QC) of the LNG-IUS products. Real-time release testing of LNG-IUSs requires tremendous and continuous expense and analytical method support for many years (which is impractical for QC purposes). Therefore, it is necessary to develop appropriate accelerated release testing methods that are representative of the real-time release. However, it appears that the accelerated release methods developed to date may not be able to pick up the first release phase (transient burst phase) of the LNG-IUSs. To mitigate this limitation of accelerated release testing methods, it is recommended that both real-time and accelerated release testing be developed and conducted. Both release testing methods (real-time and accelerated) should be conducted at the same time and run up to the time when the accelerated release is complete (within 3 months, at least 1 month). In this way, the initial real-time release data can be used to understand the burst effect and the accelerated release data can be used to evaluate the complete release profiles of the LNG-IUSs.

4. Conclusions

This is the first research report covering ultra-long duration (more than four years) of real-time drug release from LNG-IUSs with different drug loading (0.5%−50% w/w). The amount of excipient (PDMS) used in the reservoir of LNG-IUSs is not a critical quality parameter in the formulation design since LNG-IUSs (50% w/w drug loading) with up to 10% variation in excipient did not show significant difference in their release profiles. Although the release from LNG-IUSs with low drug loading (<5% w/w) was incomplete (due to drug/polymer interaction), it was confirmed that the drug release kinetics/mechanism of the LNG-IUSs with drug loading ranging from 1% to 50% remained the same. LNG-IUSs with different drug loadings (10%−50% w/w) demonstrated predictable controlled release behavior over ultra-long durations (e.g., 4 years). In addition, the accelerated release testing methods were confirmed to be representative to the real-time release profiles. This can give confidence in extending the duration of usage for these products provided that the device remains physically intact (no tearing or damage in the outer membrane) and the release rate is within the therapeutic window. Burst effects were observed in the real-time release profiles (release rate vs time) of LNG-IUSs with high drug loading, whereas the accelerated release profiles were unable to capture the initial burst. Therefore, it is recommended to perform both real-time and accelerated release testing simultaneously for LNG-IUSs to understand the burst effect as well as the complete release characteristics. In addition, attention needs to be paid to the drug/polymer interaction when designing LNG-IUS formulations with low drug loadings since drug/polymer interaction is significant when only a small amount of drug present.

Abbreviations:

LNG

levonorgestrel

IUD

intrauterine device

IUS

intrauterine system

Q1/Q2 equivalent

qualitatively and quantitatively equivalent

LNG-IUS

levonorgestrel intrauterine system

PDMS

polydimethylsiloxane

Data availability

No data was used for the research described in the article.