Hydrogel-Forming Microneedle Arrays for Sustained and Controlled Ocular Drug Delivery

Age-related macular degeneration (AMD) and diabetic retinopathy are among the leading causes of permanent vision loss in developed countries [1–4]. A complete cure to these chronic eye diseases is not currently available. Instead, treatments can only reduce progression of the disease. The main challenge of treating these posterior eye diseases lies in the delicacy of the eye tissue and the many ocular barriers that may prevent access to the virtuous humor [5–7]. Topical or oral drug delivery fails to quickly and accurately localize the drug to desired areas, and thus, requires administering higher doses, leading to waste, high toxicity, or additional side effects [8,9]. To localize the drug delivery and overcome the scleral barriers in the eye, intravitreal injections are commonly used to deliver drugs to the vitreous humor. For example, bevacizumab is most commonly used drug for treating AMD via intravitreal injections every 4–6 weeks. However, intravitreal injections cause pain and are associated with a number of complications [10–13].

Microneedles (MNs), which provide a minimally invasive method for drug delivery, are being considered as an alternative solution to intravitreal injections [14–16]. Also, MNs have been shown to deliver drugs in a more sustained and controlled manner than intravitreal injections. Thakur et el. used a hollow MN to deliver an implant to the scleral tissue for sustained delivery for up to 24 h [17]. Hydrogel-forming MNs showed a more sustained drug release than using solid, coated, or hollow MNs. For example, chitosan MNs were used to deliver protein drugs in a relatively steady rate for up to 8 days [18]. Another study using chitosan MNs showed a prolonged ovalbumin exposure at the insertion site for up to 14 days with most of drug released during the 7 days [19].

The area available for trans-scleral drug delivery is limited to the front portion of the posterior segment of the eye to avoid damaging the sensitive retina. Also, the inaccessibility toward the posterior segment of the eye needs to be considered due to the confined orbital space [20]. Therefore, it is important not only to carefully design the number of MNs and their layout in the MN array to deliver the targeted dosage effectively in a sustained manner but also to minimize the pain and potential tissue damages. A few researchers have investigated the relationship between design factors of MNs and drug release kinetics. Olatunji et al. investigated the effect of MN geometry and the number of MNs on drug release with a focus on compressive strain and its influence on the skin permeability of drugs [21]. Davidson et al. studied the effect of the number of MNs on the drug release in coated MNs [22]. Al-Qallaf et al., demonstrated a mathematical optimization model factoring in the number of MNs in an array and its relationship to drug delivery [23,24]. This model predicted the release rate based on the number of MNs and the geometry for both solid and hollow MNs, but the model is not suitable for swellable or hydrogel MNs. There still is a lack of quantified relationship between the number of MNs in an array and the drug release kinetics for swelling hydrogel MNs. Previous studies on MN have not shown a sustained drug delivery over 4 weeks, which is similar to the interval between subsequent intravitreal injections.

In this study, the drug release kinetics of swelling MN arrays with varying numbers of MNs is studied. Immunoglobin G1 (IgG1), which has a similar molecular weight and structure to bevacizumab, is used as the drug in this study. Drug release over a 4-week period is measured to assess the sustainability of the new MN design. The amount of drug release during each individual week is also assessed. A drug release model is then established for each MN design to allow for prediction of the drug release profile as well as optimization of the number of MNs to be used given a critical therapeutic index.

Figure 1 shows the designs of the MN arrays (M8, M12, and M16). Each array has two rows of MNs while M8 has four columns, M12 has six columns, and M16 has eight columns. Table 1 lists the dimensions of the three different designs. These MN designs were created in a computer-aided design software, SolidWorks.

The fabrication process of the MN arrays was detailed in a prior study by the authors [25]. In short, the master molds, as the geometry shown in Fig. 1, were additively manufactured using a digital light processing-based 3D printer (B9 Core 530, B9Creations, Rapid City, SD). Next, a soft fabrication mold for each design of MN arrays was created by casting an elastomer SYLGARD 184 (Dow Corning) over the master mold following the procedure proposed by Bediz et al. [26] The ratio used in this research is 1:10 for curing agent to elastomer. A centrifuge (NuWind NU-C200R, NuAire, Plymouth, MN) was used to mix the elastomer with the curing agent for 5 min at 2300 rpm.

Polyvinyl alcohol (PVA) hydrogel (M_w 146,000–186,000, 99% hydrolysis; Sigma Aldrich, St. Louis, MO) was used to fabricate the MNs at the concentration of 16% w/w PVA to water. The mixture of PVA and water was stirred at 90 °C until the PVA dissolved in the water. Then, the PVA solution was poured into the fabrication mold, vacuumed, and centrifuged. Finally, the molded PVA was subjected to seven freeze-thaw cycles. Each cycle consists of freezing for 8 h at –20 °C and then thawing for 5 h at 25 °C [27,28].

The drug release study was repeated three times using each of M16, M12, and M8 MN arrays along with three single injections. Fluorescein conjugated mouse IgG1 (No. 010-0202, Rockland Immunochemicals, Inc., Gilbertsville, PA), which has a similar molecular weight to bevacizumab, was used as the drug in this study. The IgG1 was diluted and loaded on each MN with 5 μg/10 μL of IgG1. Each single injection also consisted of 5 μg of IgG1. Figure 2 shows how the drug release test was conducted. The MNs were inserted through a parafilm and polyethylene/nylon sheet. Then, the base of MNs was wrapped with the parafilm and polyethylene/nylon to create an insulative environment in which only the MNs were in contact with the test solution. The test solution, a vitreous humor mimicking fluid, was prepared following the study by Fogli et al. [29]. The partially insulated MNs were attached to the lids of vials and then inserted into wells filled with the vitreous mimicking solution. For comparison, single injections were administered to inject the IgG1 into their perspective wells. A UV–visible spectrophotometer (Thermo Scientific Evolution 260 Bio-UV–Visible Spectrophotometer, Waltham, MA) was used to measure the amount of released drug at 485 nm wavelength [30]. Measurements were taken at the end of each week over a 4-week period.

The zero-order model is an ideal model for slow drug release by hydrogels [31]. To predict the drug release over time and to optimize the number of MNs for future studies, the zero-order model for each MN design was created following the formula given in Refs. [32] and [33].

Figure 3 shows a comparison between accumulated drug release from a single injection by a Gauge 31 needle, and MN arrays with 16 (M16), 12 (M12), and 8 (M8) needles, over a 4-week testing period. The intravitreal injection released nearly 100% of the total 5 μg IgG1 in the first week. M16 showed a higher release rate than the other two MN arrays and its release rate is steady in the first 3 weeks. By the end of week 3, nearly 90% of the drug was released. Both M12 and M8 showed a relative steady release rate over the entire 4-week period. By the end of week 3, M12 and M8 released nearly 70% and 50% of the drug, respectively. Both M16 and M12 released nearly the full amount of the drug loaded at the end of week 4, whereas only 80% of drug loaded on M8 was released at the end of the experiments, at a slower release rate than that of M12 and M16. All three MN arrays have a more uniform drug release profile than the single injection.

Figure 4 shows the amount of drug release during each week, 1–4, for the injection, M8, M12, and M16 treatments. At the end of week 1, the single injection released around 4.9 μg of the total 5-μg loaded drug amount of IgG1, while M16, M12, and M8 released 1.49, 1.21, and 0.86 μg, respectively. During week 2, all three MN arrays released a lesser amount of the drug than in week 1 (1.25, 0.65, and 0.61 μg for M16, M12, and M8, respectively). There is a significant difference between M16 and M12, as well as between M16 and M8. During week 3, all three MN arrays released a larger amount of the drug than in week 1 and week 2 (1.68, 1.47, and 1.05 μg for yM16, M12, and M8, respectively). No significant difference is found among different MNs. In the first 3 weeks, the amounts of drug release for all three MN arrays are proportional to the number of MNs in each array. During week 4, M16, in total, has released nearly all the drug, however it only released 0.39 μg during this time period. M12 and M8 continued releasing 1.39 and 1.24 μg of drug during week 4. There is also a significant difference between M16 and M12, as well as between M16 and M8.

Figure 5 shows a comparison of drug release week by week for each MN, as well as the remainder of drug after the 4-week period. Unexpectedly all three MN designs show significant differences among the 4 weeks. For M8 and M12, the differences are because of the lesser amount of drug release in week 2. It is believed that this can be contributed to phased drug released by the MN. In phase 1, the drug available for release in the needles is released. In phase 2, the drug in the base gradually transports to the needles which act as conduits, and thus causing a delay in drug release. The drug release rate then picks up during week 3. M16 has more conduits and needles than M8 and M12, and thus phased drug release is not as significant as that seen in M8 and M12. For M16, the difference is because of the low release in week 4 since most drugs have been released during the first 3 weeks. A further investigation can be conducted with more frequent measurements to clarify the effect of phased drug release. It is also worthwhile investigating on how drug transport within the MN array during drug release.

Figure 6 shows the drug release profiles of M8, M12, and M16, and their respective zero-order drug release models. Overall, the zero-order models predict the amount of drug release accurately. The regression zero-order models of M8, M12, and M16 are shown in Table 2. Q₀ is zero in all three models. Based on the zero-order model for M8, it is estimated that M8 can continue releasing the drug steadily for over 6 weeks. These models can be used to select the number of MNs needed to achieve a given critical therapeutic index.

In this study, MN arrays for sustained and controlled drug release are investigated. It is shown that sustained drug release for up to 4 weeks can be achieved by MN arrays. Sustained drug release using the proposed MN devices can be considered to be a subject of future studies. The drug release rate is proportional to the number of MNs in the array. The developed zero-order drug release models can accurately predict the amount of drug release, based on the given number of MNs. Knowing the relationship between the number of MNs in an array and the drug release rate aids in designing an optimal drug delivery device over a period of time. Moreover, the model can also be used to design a MN array with a targeted therapeutic index. This study demonstrates the feasibility of using MN arrays for sustained drug delivery for an extended period of time. It is expected that, with a slow but steady drug release, the treatment effect of chronic eye diseases such as AMD can be improved.

The authors acknowledge Dr. John McCloy and Jani Jesenovec for their assistance on the drug release tests.

• •

  U.S. Department of Defense - Congressionally Directed Medical Research Programs (W81XWH-18-1-0137) (Funder ID: 10.13039/100000005).