Impact of Polymer Source Variations on Hydrogel Structure and Product Performance in Dexamethasone-Loaded Ophthalmic Inserts

Abstract

Localized drug delivery can enhance therapeutic efficacy while minimizing systemic side effects, making sustained-release ophthalmic inserts an attractive alternative to traditional eye drops. Such inserts offer improved patient compliance through prolonged therapeutic effects and a reduced need for frequent administration. This study focuses on dexamethasone-containing ophthalmic inserts. These inserts utilize a key excipient, polyethylene glycol (PEG), which forms a hydrogel upon contact with tear fluid. Developing generic equivalents of PEG-based inserts is challenging due to difficulties in characterizing inactive ingredients and the absence of standardized physicochemical characterization methods to demonstrate similarity. To address this gap, a suite of analytical approaches was applied to both PEG precursor materials sourced from different vendors and manufactured inserts. ¹H-NMR, FTIR, MALDI, and SEC revealed variations in end-group functionalization, impurity content, and molecular weight distribution of the excipient. These differences led to changes in the finished insert network properties such as porosity, pore size and structure, gel mechanical strength, and crystallinity, which were corroborated by X-ray microscopy, AI-based image analysis, thermal, mechanical, and density measurements. In vitro release testing revealed distinct drug release profiles across formulations, with swelling rate correlated to release rate (i.e., faster release with rapid swelling). The use of non-micronized and micronized dexamethasone also contributed to release profile differences. Through comprehensive characterization of these PEG-based dexamethasone inserts, correlations between polymer quality, hydrogel microstructure, and release kinetics were established. The study highlights how excipient differences can alter product performance, emphasizing the importance of thorough analysis in developing generic equivalents of complex drug products.

Graphical Abstract

1. Introduction

Ophthalmic drug delivery faces numerous challenges that impact the efficacy and safety of therapeutic interventions in the eye. The unique anatomical and physiological features of the eye, such as the presence of protective barriers like the cornea and conjunctiva, limited tear volume, and rapid drainage mechanisms, pose significant challenges to effective drug delivery (Urtti 2006, Gote, Sikder et al. 2019, Macoon and Chauhan 2021). The rapid drainage mechanism and complex structure of the ocular tissues demands precise targeting and controlled release of pharmaceutical agents to ensure optimal therapeutic concentrations at the intended site of action (du Toit, Pillay et al. 2011, Agrahari, Mandal et al. 2016). Additionally, patient compliance is a critical factor for traditional eye drops along with low bioavailability due to drainage and washout effects (Urtti 2006, Al-Kinani, Zidan et al. 2018, Fang, Yang et al. 2021). Addressing these challenges requires innovative strategies, including novel localized sustained-release formulations.

Localized drug delivery offers the potential to enhance therapeutic efficacy while minimizing systemic side effects. By providing sustained release and controlled delivery, these systems optimize drug concentrations at the site of action, improving therapeutic outcomes and patient compliance. Innovative technologies, including sustained-release implants, nanoparticles, and hydrogel-based formulations, have shown promise in achieving localized drug delivery for various ocular conditions (Robinson 1980, Kumari, Sharma et al. 2010, Costello, Liu et al. 2023, Costello, Liu et al. 2023, Xie, Lin et al. 2023, Jian 2025, Santos, Delgado et al. 2025).

Polymeric inserts have emerged as a promising strategy for localized ophthalmic drug delivery, and Dextenza^™, a dexamethasone ophthalmic insert, serves as an example in this domain. Ophthalmic inserts can be inserted in the lower lacrimal punctum and into the canaliculus, acting as a punctum plug which slowly releases an API, such as dexamethasone. Dexamethasone is indicated for the treatment of ocular inflammation and pain following ophthalmic surgery, and for the treatment of ocular itching associated with allergic conjunctivitis (FDA). Ophthalmic inserts such as Dextenza offer several advantages over traditional eye drops, including enhanced patient compliance due to its prolonged therapeutic effect and the elimination of the need for frequent administration. According to Dextenza’s drug product labeling (FDA), 4-arm 20K polyethylene glycol N-hydroxysuccinimidyl glutarate (PEG-SG) is used as the hydrogel forming polymer, which undergoes an irreversible crosslinking reaction with trilysine acetate (TLA). The resulting crosslinked polymer matrix provides sustained and controlled release of the drug, minimizing systemic exposure and side effects. Notably, PEG-SG is a novel excipient in ophthalmic inserts, and the in-depth evaluation of the potential impact of quality variability of this inactive ingredient on product performance would be beneficial for product understanding for this class of drugs.

Despite the notable promise of polymeric implants in localized ophthalmic drug delivery, the development of generic equivalents faces substantial challenges related to inactive ingredient variability and manufacturing processes. The complex interplay of these factors may impact the safety and efficacy of ophthalmic implants, as has been explored for other PEG-based ophthalmic products.(Wang, Williamson et al. 2018, Fea, Novarese et al. 2022, Murakami, Hoshi et al. 2022) Ensuring comparable drug product performance requires careful attention throughout formulation and manufacturing process development, adequate controls, and thorough product characterization and testing, often necessitating a comprehensive understanding of the intricacies of polymeric materials and use of analytical methods with improved capabilities for drug product attribute assessment. Additionally, achieving similarity in drug release, swelling, and biodegradation characteristics is critical for ensuring product performance similarity of generic ophthalmic inserts containing poorly water-soluble API. Successfully navigating these challenges is needed to expand access to cost-effective alternatives while maintaining safety and efficacy.

This study was undertaken to evaluate the variations in functional excipient, PEG-SG, and its impact on dexamethasone ophthalmic inserts, to support the understanding and characterization of product performance. This research also represents a first-ever attempt to promote comprehensive physicochemical characterization of PEG-SG containing products, emphasizing the potential impact of inactive ingredient quality on overall performance.

2. Materials and Methods

2.1. Materials

Micronized dexamethasone and non-micronized dexamethasone were purchased from Spectrum Chemical Mfg. Corp. (Gardena, CA, USA) and Thermo Scientific (Waltham, MA, USA), respectively. Sodium phosphate dibasic and sodium phosphate monobasic were purchased from Sigma-Aldrich, St. Louis, MO, USA. Trilysine acetate (TLA) was purchased from Bachem Americas, Inc., Torrance, CA, USA. Polyethylene glycol - Succinimidyl Glutarate (PEG-SG) with varying arm number (4 and 8) and molecular weights (20K and 40K) were purchased from anonymized vendors; the 4-arm 20K PEG-SG was sourced from three vendors and is designated as ‘V1’, ‘V2’, ‘V3’, the 8-arm 20K PEG-SG is designated as ‘8a’, and the 4-arm 40K PEG-SG is designated ‘40k’.

2.2. Fabrication of Inserts

Inserts were made by mixing of hydrogel components. The insoluble dexamethasone drug substance is first suspended in a deionized water, and upon mixing the reactive excipient components, becomes entrapped within the hydrogel network as the aqueous solution-phase crosslinking reaction proceeds. The resulting hydrogel-drug suspension is casted, allowed to suitably cure, then dried in an incubator at 37 °C under low vacuum, and finally manually cut with a tube cutter and placed in sealed containers for desiccated storage.

2.3. Insert Characterization with optical and X-ray microscopy and AI Analysis

Initial examination of the commercial insert and in-house fabricated inserts was done by Keyence VHX digital microscope (Keyence Corporation, Osaka, Japan). Inserts were imaged in singlet via X-ray microscopy (XRM) and characterized with the I2S^™ (DigiM Solutions LLC) artificial intelligence (AI)-based image analysis software, to quantify and compare the pore and active pharmaceutical ingredient (API) size distribution, and API spatial distributions within inserts. Images were collected with the Zeiss Xradia Versa 520 or 620 instrument. The material phases of the insert samples were classified into pores, API, polymer and buffer salts. The volume fractions, size distributions of API and pores, and spatial distributions of the API particles were quantified. Further methodological details regarding the XRM and AI-based image analyses can be found in (Nagapudi, Zhu et al. 2021).

2.4. Proton Nuclear Magnetic Resonance Spectroscopy (¹H-NMR)

¹H-NMR (Bruker 850 MHz with room-temperature QXI probe) was used as a method for the identification of polymer end-group functionalization via integration comparison of the NHS and glutarate ester signal peaks relative to an internal trimethylsilylpropanoic acid (TMSP) standard, the primary peak arising from ethylene oxide groups, and the pentaerythritol core methylenes. More specifically, the integration ratio of the pentaerythritol core methylenes (labeled a in Figure 1) to the ester-adjacent methyl peaks (labeled c, e, and f in Figure 1) confirmed and quantified the presence of the chemical groups resulting from progressive stages in the synthesis, with the ratio of the NHS peak (labeled b in Figure 1) to the pentaerythritol peak additionally verifying presence of the NHS endcap. Free (unconjugated) NHS also features nearly the same chemical shift as the conjugated NHS, so the ratio of this peak to the others was used to quantify the residual free NHS impurity content. Samples were dissolved in D₂O or CD₃OD at 20 mg/mL.

2.5. Fourier-transform Infrared Spectroscopy (FTIR)

IR spectroscopy (Bruker Alpha Platinum-ATR FTIR) was used as a complementary method for the identification of end-group functionalization via the comparison of the 1738 cm⁻¹ peak, corresponding to the asymmetric stretch of C=O bonds in carbonyl-containing glutarate ester crosslinks and terminal NHS groups, relative to vector-normalized C-H peaks within the sample (1500–1225 cm⁻¹).

2.6. Size Exclusion Chromatography (SEC)

A size exclusion chromatography (SEC) instrument with quadrupole detection [UV-Vis, differential refractive index (dRI), multiangle light scattering (MALS), and viscometry] method was developed to determine the polymer molecular weight distribution and branching number. This was compared to the common practice of SEC with external column calibration. Two external calibration curves were performed using a set of linear PEG polymers (Polymer Standards Service-Agilent) with a peak molecular weight (M_p) range of (217.0, 44.0, 18.6, 11.4, 3.4, 2.1, and 0.6 kDa), and one using a set of protein standards (ThermoScientific and Sigma-Aldrich): bovine serum albumin (dimer and monomer), myoglobin, and cytochrome C (132, 66.4, 17, and 12 kDa, respectively). Polymer samples were prepared at a concentration of 10 mg/mL in 18.2 MΩ·cm ultrapure water, and then diluted to 5 mg/mL in a 1x phosphate buffered saline (PBS) solution. Due to precipitation of the V3 sample, it was initially filtered through a 0.45 µm Nylon filter prior to diluting into PBS. For SEC analysis, a TSKgel G3000PWxl column was used with 1x PBS as a mobile phase, at a flow rate of 0.5 mL/min, and an injection volume of 20 µL. Samples were run in triplicate and analyzed using ASTRA 8.2 software. Molecular weight determination with MALS-RI consisted of using a dn/dc value of 0.14, Zimm model with 1^st order fit degree, and detectors 4–16 selected.

2.7. Dynamic Light Scattering (DLS)

Batch-mode DLS experiments were performed on a Zetasizer Nano instrument (Malvern Panalytical). Samples were prepared using the same procedure as for SEC preparation which is described above. A ZEN0040 cuvette and backscatter angle of 173° was used for all experiments. Three different aliquots from each sample were measured to acquire an n=3 with appropriate standard deviation.

2.8. Matrix Assisted Laser Desorption/Ionization Time of Flight (MALDI-TOF)

MALDI-TOF was used to measure molecular weight of the precursor PEG polymer materials, on a Bruker ultrafleXtreme MALDI-TOF/TOF, and data was processed using the FlexAnalysis software. The matrix consisted of 20 mg/mL 2,5-dihydrobenzoic acid matrix (30:70 acetonitrile:water ratio with 0.1% trifluoroacetic acid) and 1 mM sodium chloride as the cationization agent. The precursor polymers were prepared at 0.15 mg/mL in methanol, mixed with the matrix at a 1:1 volume ratio (20 µL of each), and then centrifuged for 6 s. After centrifugation, roughly 0.75 µL of sample deposited on a MTP 384 Target plate ground steel BC (Bruker) and left to air dry. For analysis, LP-5-50 method using linear mode, 50–60 % laser power, detector gain of 4.3, and medium detection. All spectra were baseline subtracted.

2.9. Calculation of Theoretical Chemical Crosslinks

Theoretical Chemical Crosslinks were calculated by taking the simple average of the number averaged molecular weights M_n from MALDI-TOF and SEC-MALS and normalizing relative to this value for the V1 polymer. The number of functional groups as determined by ¹H-NMR on a molar basis were then divided by the normalized M_n values to yield the number of theoretical chemical crosslinks per mass basis.

2.10. Differential Scanning Calorimetry (DSC)

A differential scanning calorimeter (TA Instruments Discovery DSC 2500) was used to generate thermograms for the precursor polymers, the dexamethasone API at different stages of the manufacturing process, and the finished insert materials. Approximately 5 mg of sample were used for each run. A two-cycle DSC method was used with cooling to −80 °C, and ramp to 100 °C at a rate of 10 °C/min, with isothermal holding period of 5 minutes at these temperatures.

2.11. Thermogravimetric Analysis (TGA)

Thermogravimetric analysis (TA Instruments Discovery TGA 5500) was used to assess thermal stability, degradation profile, and moisture content of the polymers, API, and crosslinked insert materials. The TGA methods used consisted of an isothermal holding period of 5 minutes, a temperature ramp of 10 °C/min to 600 °C, and a final 5-minute isothermal step, and were under either air or nitrogen gas flow at 50 mL/min.

2.12. Dynamic Mechanical Analysis (DMA)

A dynamic mechanical analyzer (TA Instruments DMA Q800) was used to perform mechanical testing on the dried and hydrated insert materials, using the film tension clamp accessory. Stress relaxation tests were performed at a preset amplitude of 20 μm with force tracking of 125% and preload force of 0.01 N, with an equilibration step at 22 °C, followed by an isothermal period of 10 min. Stress relaxation tests under temperature ramp were also carried out with preset amplitude of 10 μm and force tracking of 125% while increasing temperature at a rate of 2 °C/min until 80 °C. Strain-to-yield studies were performed through a displacement ramp method which first ramped the temperature up to and equilibrated at 80 °C, with an isothermal hold period of 0.5 min thereafter, and then stretched 5 mm long samples at a displacement rate of 2160 μm/min until failure. Inserts were also stretched using this method to a desired strain % for use in comparing swelling behavior vs strain, where upon reaching the target strain % the heated sample was then cooled at a rate of 15 °C/min until 22 °C, then were allowed to equilibrate for 10 min under nitrogen cooling gas control, after which the furnace was opened and samples were held at room temperature for a minimum of 2 h. DMA testing in the hydrated state was also carried out.

2.13. Rheology

Rheological studies were performed using a TA Instruments Discovery HR-3 Hybrid Rheometer with a hard anodized aluminum 40 mm, 1.0° cone plate on an aluminum Peltier plate stage, with a truncation gap of 25 μm. Flow Peak Hold experiments were used to examine viscosity of varying solid weight % solutions of precursor material, these were undertaken at 25 °C with a shear rate of 10 s⁻¹. Oscillation-Time experiments monitored the hydrogel curing kinetics by continuous oscillation under direct strain of 5%, frequency 1 Hz, temperature kept at 25 °C, for a first phase of duration 600 min. A second phase at a set torque of 9000 μN.m, frequency of 1 Hz, and temperature of 25 °C followed to examine the yield behavior of the materials. Oscillation Frequency sweeps were performed at both 1% and 5% strains from 100 Hz to 0.1 Hz at 25 °C. Oscillatory Amplitude sweeps to determine yield point were performed at a frequency of 1 Hz with stress ranging from 1.0E-8 MPa to 0.01 MPa or strain from 0.1% to 1000%.

Buffering conditions of 0x (0M phosphate), 1x (0.01M phosphate), 4x (0.04M phosphate) were used for the pH variation studies.

2.14. Material Swelling (Water Uptake)

Swelling studies of the in-house fabricated inserts were conducted using optical dimension measurement techniques. In-house inserts of known initial dimensions were immersed in 1x PBS, and dimensional changes were monitored over time. A Keyence VHX digital microscope (Keyence Corporation, Osaka, Japan) was used to measure changes in length and diameter at specified time intervals. The same methodology was used to determine swelling characteristics for samples subjected to the DMA stretching process to various strain % levels.

2.15. In Vitro Release Testing (IVRT)

A real-time IVRT was conducted for all in-house fabricated inserts. Samples were placed into polypropylene tubes containing 50 mL of deionized water in an incubator at 37 °C without shaking. One milliliter samples were withdrawn at 10 min, 30 min, 1 h, 2 h, 4 h, 24 h, and daily until day 8, with fresh media added to replenish the volume. Before taking the sample, the tubes were inverted gently to ensure uniform drug concentration within the tube. The release medium was completely renewed on day 7. The drug concentration in the withdrawn samples was determined by HPLC analysis following the Dexamethasone USP monograph (USP-NF).

3. Results

3.1. Precursor Polymer Characterization

3.1.1. ¹H-NMR

Variations in end group functionalization and impurity content among the PEG-SG precursor materials sourced from different vendors were observed as determined by ¹H NMR spectroscopy (Figure 1).

Figure 1.

a) Chemical structure for the PEG-SG polymers, with PEO repeat unit per arm (n) varying by polymer type, and number of pentaerythritol cores m = 1 for the 4-arm polymers and m = 3 for the 8-arm polymer. b) ¹H NMR spectrum of different PEG-SG polymers illustrating differences in functionalization among them. V1, V2 and V3 indicate 4-arm, 20K PEG-SG from different vendors. 8a and 40k indicate 8-arm, 20K PEG-SG and 4-arm, 40K PEG-SG. c) inset region highlighting impurity content, particularly prevalent in V3. d) Summary table of % functionalization according to polymer type, and corresponding theoretical crosslinking sites available based on nominal branching (4 or 8).

The overall percentage of functionalized arms and the resulting theoretical crosslinks on a molar basis differed substantially (Figure 1d). For example, the 8-arm sample (8a) showed the highest functionalization (96%) and thus the greatest capacity for crosslinking (7.68), whereas the 4-arm 40K material (40k) displayed a lower functionalization (76%) and fewer crosslinks per unit mass (1.52).

Among the 4-arm, 20K PEG-SG samples (V1, V2, V3), V3 exhibited the lowest functionalization (76.5%) and thus potentially lower reactivity when forming hydrogels. These discrepancies could likely originate from differences in raw material purity, reaction efficiency in raw material manufacturing, or storage-related factors such as partial hydrolysis of NHS end groups. From a formulation standpoint, such variations are highly relevant, as the extent of functionalization and number of arms directly influence hydrogel crosslink density, mechanical properties, and swelling behavior in the final ophthalmic insert.

3.1.2. Fourier-transform infrared spectroscopy

Notable differences were observed in the carbonyl stretching region near 1738 cm⁻¹ using FTIR. The inset in Figure 2 focuses on the 1720–1760 cm⁻¹ window, where the NHS-ester and glutarate groups typically exhibit strong absorption bands. Variation in the intensity and shape of this peak reflects the overall level of carbonyl-containing end groups and thus the extent of functionalization in each raw material.

Figure 2.

FTIR spectrum of different PEG-SG polymers showing difference in end-group functionalization via the comparison of the 1738 cm⁻¹ peak.

The normalized peak intensity around 1738 cm⁻¹ helps distinguish subtle discrepancies in carbonyl content also observed in the ¹H NMR data. For instance, polymer 8a exhibited the highest carbonyl peak, consistent with the greater number of arms (eight) per unit mass and greater degree of functionalization indicated by NMR (96.0% functionalized). Conversely, polymer 40k displayed a weaker carbonyl band, in line with a lower calculated functionalization level and fewer crosslinking sites per unit mass. Meanwhile, V1 and V2 exhibited little relative difference. V3 showed an atypically high carbonyl signal after normalization, partly because of unreacted NHS groups and potential impurities.

In summary, V1 and V2 are quantitatively similar (94.74% vs. 94.44%) with respect to overall carbonyl content, normalized to 100% for 8a. The 40k sample displayed 80.38% of expected carbonyl content, while V3 showed a spectrum skewed by free NHS, yielding a 124.25% relative carbonyl content.

3.1.3. Molecular weight determination by MALDI-TOF and SEC-MALS

Molecular weight determination of the PEG-SG precursors, V1, V2, V3 (all 4-arm, nominally 20 kDa), 8a (8-arm, 20 kDa), and 40k (4-arm, 40 kDa), was initially performed by MALDI-TOF (Figure 3). All three 4-arm, 20 kDa samples displayed principal peaks near the expected 20–21 kDa range. Molecular weights for V1 and V3 were comparatively similar, while V2 was slightly lower. In contrast, the sample 8a (8-arm, ~20 kDa) exhibited slightly higher m/z values than V1–V3, consistent with its additional functional groups contributing to a larger overall molecular mass. The sample 40k showed the most pronounced mass range, centered near 40–41 kDa, and displayed higher polydispersity.

Figure 3.

Molecular weight determination of different PEG-SG polymers by MALDI-TOF.

To confirm the MALDI-TOF observations, the precursor polymers were analyzed by SEC-MALS. Figure 4a overlays the 90° light-scattering (LS) response (primary y-axis) with the computed molecular weight (secondary y-axis) across elution time. V1 and V3 exhibited narrow, single peaks centered at approximately 13.5 minutes, with molecular weight (number averaged, M_n) values in the 22 kDa range. V2 eluted slightly later, with M_n around 20 kDa (Figure 4a), and smaller hydrodynamic radius R_h,v (Figure 4b). The 40K polymer eluted at a lower retention time (12–13 min) than the 20K polymers, reflecting larger M_n and hydrodynamic volume. Sample 8a showed a broader and slightly later elution peak than V1–V3, consistent with its additional branching and hydrodynamic contraction, and suggesting heightened polydispersity. To support the assessment of crosslinking density across all inserts, branching analysis was completed using results from SEC (Table S1). Similar branching ratios (g’) and number of branch units per molecule were observed for V1-V3. The 8a sample showed much smaller g’ value, suggesting a more branched or contracted architecture. Radii for the polymers were determined orthogonally via DLS (Figure 4c), which also revealed multimodal distributions indicating larger species, likely caused by aggregation at high concentration or possible impurities.

Figure 4.

Chromatograms of SEC-MALS analysis showing a) molecular weight ranges; b) R_h,v ranges determined by online viscometry. Figure inset shows a second peak in the light scattering data that is in low concentration which could be related to larger impurities; and c) Size distributions for the precursor polymers via DLS.

Differences in polymer molecular weight distributions measured by MALDI and SEC-MALS in combination with end group functionalization from NMR and FTIR, yielded theoretical crosslinks per unit mass for the polymer precursors which could be used as a key metric for assessing quantitative similarity, shown in Table 1. The degree of crosslinking in hydrogel systems is intrinsically linked to the molecular weight of the polymer precursors. In general, lower molecular weight polymers result in a higher density of functional end groups per unit mass, which can lead to a higher crosslinking density under the same crosslinking conditions. Conversely, higher molecular weight polymers have fewer reactive ends per unit mass, resulting in a more loosely chemically crosslinked network.

Table 1.

Theoretical Chemical Crosslinks for the Polymer Precursors

Precursor identity:	V1	V2	V3	8a	40k
Theoretical chemical crosslinks (mass basis):	3.52	3.63	3.03	7.12	1.64

3.2. Insert Manufacturing

Inserts were fabricated through a sequence of steps that were applied uniformly to all samples, each containing a different precursor polymer type. Upon weighing out the appropriate amounts of each material, the reactive components PEG-SG and TLA were dissolved separately into water. Buffer components were separately dissolved, and dexamethasone API was mixed into the PEG-containing solution to form a slurry, prior to mixing the reactive components together. Upon mixing all components, the crosslinking reaction proceeded quickly, taking several minutes to reach a highly crosslinked and gel-like state.

Strong pH-dependence of the crosslinking reaction was evident by tracking the rheological modulus crossover in response to buffer changes, indicating control over reaction pH is imperative, Figure 5a. A comparison of the plateau storage modulus G’ measurements for materials composed from V1 and 8a showed ~13% difference, which accounts for different contributions of chemical crosslinking and molecular entanglement between the polymer types, shown in Figure 5b.

Figure 5.

a) Rheology showing curing kinetics at different phosphate buffer strengths (0x = 0 mM, pH 6.57; 1x = 10 mM, 7.35; 4x = 40 mM, pH 7.22) for the V1 gelation system. b) curing profiles for 4-arm (V1) vs 8-arm (8a) species

The 8-armed PEG system reaches its gel point more quickly than the 4-armed variants owing to the relative abundance (2x) of crosslinking sites that undergo the amidation reaction with TLA, which is present at a 1:1 molar functional group (amine : NHS) ratio. Thus, a higher density of crosslinks quickly forms relative to the mass of pre-gelled mixture which is fixed across sample types, imparting gel-like properties at a faster rate.

3.3. Insert Characterization

3.3.1. XRM and AI Analysis:

Dry inserts were characterized by X-ray microscopy (XRM), except for V2 which was omitted for this phase of the study for cost reasons. Segmentation successfully identified the different material phases, with the volume fractions of the pores, polymer, API, and salts are shown in Table 2.

Table 2.

Volume fractions for each phase in the inserts, as determined by XRM imaging.

Volume Fractions (%)	V1	V3	8a	40k	Commercial Insert	V1-L*
Pores	2.8	11.6	15.9	40.9	8.5	0.9
Polymer	50.3	45.7	42.0	30.7	42.3	49.3
API	46.8	42.6	42.0	28.3	44.9	48.9
Salts	<0.1	<0.1	<0.1	<0.1	4.3	0.9

^*V1-L indicates the insert was compositionally identical to V1 with the exception that it contained larger (“L”) non-micronized dexamethasone instead of micronized dexamethasone.

Varying pore size distributions were observed among inserts made from polymers of different vendors, molecular weights, and arm numbers. Increasing either arm number (e.g., 8a) or molecular weight (e.g., 40k) increased the porosity. Differences in vendor also led to different porosity behaviors.

The rank ordering of insert porosity was the following:

V1-L < V1 < Commercial Insert < V3 < 8a < 40k

And for average pore size the order was the following:

V1-L < Commercial Insert < V1 < V3 < 8a < 40k

Interestingly, examining the micrographs displayed in Figure 6, the trend in insert porosity also appeared to match the inserts’ visual characteristics, where frequency/density of internal pores can be visually gauged due to the translucent nature of the materials.

Figure 6.

X-ray Micrograph in grayscale and the corresponding AI segmentation overlay for all insert samples (top), and volume-weighted pore size distribution within the different insert types (bottom). The commercial insert is here denoted as ‘C.I.’

Higher porosity in V3 compared to V1 may be due to differences in end-group functionalization and impurities. Increases in branching (e.g., 8a) or molecular weight (e.g., 40k) also yield higher porosity and pore size. Notably, the observed rise in porosity for 8a, where crosslink density should theoretically decrease porosity, may reflect incomplete mixing or air bubbles. It’s posited that the differing dynamic viscosities which develop during the curing of the reactant mixtures contribute to the solutions’ ability to entrap air bubbles during the mixing process, which likely leads to the observed disparities in porosity and pore size. Other factors, such as interchain associations and crystallinity, can also influence the dried material’s structure. The discrepancies in salt volume % determined for the commercial and in-house inserts may be attributed to known deviations in the processing conditions for these samples which differentially alter salt nucleation. Despite having identical composition, this distinction highlights the importance of manufacturing processes on end physicochemical properties.

XRM was also used to examine size distribution of the dexamethasone API in the inserts in the dry state, displayed in Figure 7. The V1-L insert was made with non-micronized API and exhibited larger API particle size distribution than others. Samples fabricated using the same micronized dexamethasone raw material (V1, V3, 8a, and 40k) showed relatively comparable dexamethasone size distributions within the finished inserts (D₅₀ of 9–10 µm), Table 3. Notably, the 40k sample showed a significantly lower overall volume fraction of API (28.3 % vs. 42–47 % for other inserts). The lower API fraction highlights potential differences in how the larger molecular-weight polymer network entraps or distributes dexamethasone particles, possibly leading to more open or porous network regions in the dried state. By contrast, V1-L, which was formulated using non-micronized dexamethasone, exhibited a higher volume fraction of 48.9% and a distinctly larger particle size distribution (D₅₀ of 13 µm and D₉₀ of 18 µm).

Figure 7.

Differential size distribution of API within the insert materials and the commercial insert, denoted ‘C.I.’, determined by XRM imaging.

Table 3.

API Particle Volume Fraction and D₁₀-D₉₀ values

Insert Type	Volume Fraction (%)	D10 (µm)	D50 (µm)	D90 (µm)
V1	46.8	6.4	9.3	13.5
V3	42.6	6.6	8.9	11.2
8a	42.0	7.7	10.0	12.2
40k	28.3	7.7	10.1	12.5
C.I.	44.9	3.8	8.9	12.3
V1-L	48.9	9.1	13.4	17.6

3.3.2. DSC

DSC findings confirm the hypotheses that the crosslinking and internal structure of the finished insert materials directly follow the functionalization % in the precursors, and thus control of the precursor properties is critical.

DSC thermograms of the polymer precursors (Figure 8a) and the corresponding crosslinked inserts (Figure 8b) are provided below. In general, the precursors exhibited melting endotherms with onset temperatures around 40–50 °C and peak melting transitions near 50–60 °C, reflecting the crystalline domains of PEG. The 40K polymer showed higher melting enthalpies and slightly shifted melting peaks, consistent with an increased polymer mass per molecule and more extensive crystallinity. Conversely, sample 8a displayed the opposite effects. Differences in crystallinity were corroborated with X-Ray Diffraction studies on the precursor powder materials (Figure S1). Once crosslinked into the insert form, the melting transitions (Figure 8b) shifted to slightly lower onset temperatures and lower enthalpies, indicating partial disruption of crystallite formation by the covalent crosslinks, and altered chain mobility within the hydrogel matrix. Figure 8c is a representative schematic illustrating how differences in branching (8a vs. 4a) and molecular weight (40K vs 20K) led to distinct network configurations, ultimately influencing the thermal behavior and mechanical properties of the final inserts.

Figure 8.

DSC thermograms for a) polymer precursors and b) inserts to showing differences in onset and peak temperatures as well as heats of melting indicating potential differences in crystallinity and crosslinking. c) Idealized 2D depiction of the polymer network structure for inserts composed of different precursor types, showing representative differences in polymer chain arrangement and interactions, as well as crosslink density.

3.3.3. TGA

As NHS-ester bonds tend to hydrolyze in storage if exposed to water, TGA experiments for the present study evaluated water content in the precursors (Figure 9a) and finished inserts (Figure 9b). In addition, the insert moisture content has an effect on rigidity and brittleness, which increases difficulty in handling and testing of the material. Additionally, TGA studies provide needed information regarding thermal stability of the polymers, where alternative manufacturing techniques (e.g., hot-melt extrusion) may expose the materials to higher temperature and potential degradation (Figure S2).

Figure 9.

Thermogravimetric curves under N₂ atmosphere for a) polymer precursor powders and b) insert samples. Insets show determination of residual moisture content.

3.3.4. Mechanical Properties

Dynamic mechanical analysis on the dried inserts is a useful tool to understand key critical properties such as resilience to excessive bending or clamping forces during administration, as well as a surrogate technique that is predictive of performance characteristics like swelling and drug release. The tensile mechanical properties of the inserts correspond to polymer functionalization and material crosslinking.

The stress-strain response curves (Figure 10a) revealed distinct differences in the mechanical resilience of the samples across different formulations and processing conditions. The elastic modulus (Figure 10b) highlights variations in stiffness among the formulations. For instance, V2 inserts exhibited one of the highest moduli, suggesting enhanced crosslinking contributing to its rigidity. On the other hand, the significant reduction in modulus for V3 and 40k may reflect insufficient crosslinking or structural integrity, likely impacting mechanical performance and potential application. The 8a insert demonstrated only a marginal increase in modulus compared to V1 and V2, consistent with the slight increase in storage modulus for 8a as measured via rheology. This finding points to the complex interplay of arm length, branching, geometric constraints, and chemical/physical crosslinking which underlies these materials, which have been studied in detail for other branched polymer materials.(Kida, Tanaka et al. 2019, Giuntoli, Hansoge et al. 2020, Giuntoli 2021, Nanok, Khanom et al. 2023) The different yield points exhibited by the different insert materials indicate their potential processability during manufacturing and durability during application.

Figure 10.

a) Stress-strain response curve with visual guides to the 1^st and 2^nd LVRs, in pink and blue, respectively, depicting the slope of stress/strain = E’. b) Elastic moduli (Stress/Strain) for the 2^nd LVR region (blue) of the inserts in measured in a).

3.3.5. Physical Properties

Polymer source variation did not show any significant (p>0.05) impact on the dry insert diameter Figure 11a. Polymer source variation did show significant (p<0.05) impact on the insert density Figure 11b, indicating that insert density can be a useful measurement along with mechanical properties as a surrogate for effective material crosslinking.

Figure 11.

a) Dry insert diameter mean diamonds scatter plot and Tukey-Kramer plot for the inserts made without any stretching. b) Leverage plot for the Insert Type variable, and Tukey HSD of least squared means indicate that a significant difference is found on insert density between V3 and the other two 4-arm 20k polymers V1 and V2.

3.3.6. Material Swelling

Swelling of the in-house fabricated inserts showed different swelling rates for inserts made with similar polymers (arm number and molecular weight) from different vendors (Figure 12). Inserts fabricated with 4-arm PEG with 20K molecular weight showed similar swelling rates for V1 and V2 but swelling rate was higher for V3. This difference can be attributed to V3’s lower overall functionalization (per ¹H NMR) and higher impurity content, both of which reduced its crosslink density and created a more open network that admits water more readily. Inserts made with 4 arm PEG with 40K molecular weight and 8 arm PEG with 20K molecular weight showed slower swelling rates compared to inserts made with 4-arm PEG from different vendors. The lower crosslink density in 40K leads to a more expansive swollen network at equilibrium, while the kinetics of the swelling are hampered early on by its enhanced polymer crystallinity, as corroborated by DSC melting behavior. The contracted architecture of the 8a material, with its higher theoretical crosslink density, produces considerably limited swelling kinetics and final swelling state compared with the other samples.

Figure 12.

Volumetric swelling ratio of in-house fabricated inserts with time. a) Early timepoints till 6 hours and b) completion of the study at day 12

The swelling study results are in agreement with the precursor polymer characterization data, as functionalization corresponds to theoretical crosslink density and crosslinking density post-curing determines the swelling rate.

3.3.7. Strain Memory Effect on Swelling

Another important quality attribute to examine was the stretching of the inserts and the ensuing shape memory effects upon swelling. The commercial dexamethasone insert, which was designed for intracanalicular insertion, swells predominantly in the radial direction to ensure secure lodging beneath the punctum, but not in the axial direction so that the insert would undesirably protrude, risking dislodgement in the outward direction. Thus, the inserts were measured in the radial and axial dimensions upon swelling in 1x PBS, documented at the 1 h and 24 h timepoints, shown in Figure 13.

Figure 13.

Strain-swelling studies examining the differences in sample dimensional changes upon water uptake, at the 1 h and 24 h timepoints, for inserts that have been stretched in dry, heated conditions between 0% (no stretching) and up to 400% strain. a) cross-sectional area change at 1 h swelling. b) cross-sectional area changes at 24 h swelling. c) length change at 1 h swelling. d) length change at 24 h swelling

From these data, to achieve the desired physical dimension changes under swelling, a stretching process achieving 100% strain is necessary. At both 1 h and 24 h of swelling, the length change from dry to swollen state is negligible for all the inserts, centering around a swollen:dry ratio of 1. In the radial direction, swelling in slight excess of the ratio of the diameter of the dry insert to the diameter of the punctum is desired, so that the insert remains in place after swelling. The 100% strain-processed inserts are on the order of 0.6 mm wide when dry, and exhibit a cross-sectional area change of approximately 6 to 9-fold upon swelling, corresponding to a diameter change of about 2.5–3x. This seems like an appropriate change given the width dimensions of the human cadaver vertical canaliculus and lower punctum are about 1.6 ± 0.5 mm and approximately 0.5 mm, respectively, and considerable variability in healthy volunteers, with punctal diameters ranging from 0.41 ± 0.16 mm to 0.65 ± 0.15 mm.(Tyson 2020) This range of radial swelling is achieved for the vicinity of 50–150% strain during processing.

3.3.8. In Vitro Release Testing

The in vitro drug release profiles of the in-house fabricated inserts are shown in Figure 14. Differences in drug release were observed in inserts made from the 4-arm 20K PEG polymers (V1, V2, V3). The 8a and 40k inserts exhibited differences from the others in drug release as well. In addition, particle size of the dexamethasone also appeared to play an important role, with formulation containing non-micronized dexamethasone (i.e., V1-L) exhibiting slower release. Overall, the current IVRT method was able to discern minor differences in polymer characteristics and particle size via f2 similarity factor (Table S2). IVRT results were also in good agreement with the crosslinking density (depending on the % functionalization) and swelling study results. Therefore, both particle size distribution and IVRT data can be utilized as components of a comprehensive characterization approach to corroborate performance sameness across different PEG-based inserts.

Figure 14.

Dexamethasone release from in-house fabricated inserts. V1, V2, V3, 8a, and 40k inserts were made with micronized dexamethasone, while V1-L is made with 4-arm 20K PEG from the same vendor as V1 but contains non-micronized dexamethasone.

Different release models (e.g., zero-order, first-order, Higuchi, Log(t) and Korsmeyer-Peppas) were performed to fit the release profiles of the in-house fabricated inserts. Based on the IVRT data and release fitting parameters, the Korsmeyer-Peppas model showed the best goodness of fit for all inserts (R² value>0.99), (Table S3) confirming that the release mechanism follows a combination of classical diffusion theory and transient swelling phenomena, with a moving boundary of waterfront triggering solubilization and diffusion of drug outwards through the matrix.(Peppas and Sahlin 1989, Kosmidis, Argyrakis et al. 2003, Siepmann and Peppas 2011, Peppas and Narasimhan 2014, Heredia, Vizuete et al. 2022) The relative contributions of Fickian diffusion and swelling phenomena to the release mechanism can be extracted from the model, and are demonstrated to vary by polymer type in a manner consistent with the swelling and release data. Additionally, the release rate was found to be governed by the solubility limit of API in the release medium. Together, these two factors control drug release, leading to sustained released characteristics.

3.3.9. Flory-Rehner Mesh Size ξ calculations

Along with the swelling studies, hydrogel mesh size calculations were carried out using the method described in literature for similar (degradable crosslinked PEG star) polymer systems.(Canal and Peppas 1989, Lu and Anseth 2000, Zustiak and Leach 2010) Mesh size calculations in conjunction with the volumetric swelling and density data provide an approximation for the internal pore dimensions within the hydrogel network that relate to swelling in water over time and gradual degradation.

The calculated internal dimensions of the gel networks correlated well with the swelling data, with the same trends appearing in rank order at both the first timepoint and the rate of change between time points, as evident in Figure 15. A preliminary short-term degradation study indicated that polymer source and differences in functionalization and impurities has an impact on degradation rate, with V3 featuring some degradation, matching accelerated mesh size changes, within a 28-day period, while the other vendors remained unchanged.

Figure 15.

Mesh size calculations among swollen inserts made with different PEG-SG polymers after 1 hour, 1 day, and 12 days swelling + release timepoints in PBS bath.

The mesh size information provides context for the mechanism of API release from the insert. From examining API PSD measurements via XRM, it was determined that the particulates greatly exceeded the mesh dimensions by more than 10-fold early on and they continued to exceed mesh dimensions even as the inserts further swelled and began to degrade. The release mechanism of dexamethasone is thus linked to the gradual solubilization of dexamethasone API particles which are immobilized within the hydrogel, primarily by physical entrapment, until a sufficient reduction in size is paired with a 1.5–2x increase in the hydrogel mesh size later in the swelling process.

4. Discussion

Ensuring inactive ingredient sameness for PEG-based ophthalmic inserts can be challenging. In the 4-arm 20 kDa PEG systems that were studied, the critical hydrogel-forming excipient, although being nominally “the same”, can vary in end-group functionalization and molecular weight distribution. In this study, variations in end group functionalization and impurity content among the 4-arm PEG-SG precursor materials sourced from different vendors were determined by ¹H NMR spectroscopy and FTIR. Differences in polymer molecular weight distributions, as measured by MALDI and SEC-MALS, in combination with end group functionalization from NMR and FTIR, yielded theoretical crosslinks per unit mass, which was evaluated as a key metric for assessing qualitative similarity of polymeric precursors.

The polymer and insert characterization data indicated correlations between precursor functionalization and molecular weight distribution, crosslinking density in the cured polymer mesh network, hydrogel water uptake kinetics and swelling volume equilibrium, and ultimately the in vitro dexamethasone release profiles of the inserts. Higher polymer crosslinks per unit mass, governed by functionalization % and PEG MW distribution, yield higher crosslink density hydrogel materials that are less porous, which was corroborated by Flory-Rehner mesh size calculations. When the elastic crosslinked polymer network contacts aqueous media, the material absorbs the fluid, and the network starts to swell due to the thermodynamic compatibility of the polymer chains and water. The swelling force is counterbalanced by the retractive force induced by the cross-links of the network which can be interchain, intrachain, chemical and/or physical.(Peppas, Huang et al. 2000) Swelling equilibrium is reached when these two forces are equal, and thus both the swelling process and end state are governed by the physicochemical properties of the network constituents. Such a swelling phenomenon was consistent across all tested materials and was consistent with the network mesh size calculations that track internal dimensions of the gel material. In the present system, dexamethasone particulates on the order of 5–20 μm were entrapped within the mesh network during the mixing and curing manufacturing step, which exceeded the regular mesh size, even well into the swelling process. Thus, the dexamethasone release behavior was primarily governed by the diffusion of solubilized API, as opposed to diffusion of API aggregates through the mesh. The increased contact being made with water at higher swelling states accelerated the internal diffusion and eventual release from the insert surface.

Reverse-engineering or reconstituting the exact composition of crosslinked inserts can be difficult because the reaction is irreversible, meaning the individual components cannot be easily separated post-curing. In the present system, the amount of TLA on a mass basis was very small relative to the PEG and dexamethasone content, confounding any straightforward quantification approaches (e.g., elemental analysis). Some commercial inserts also incorporate additives such as visual aid dyes, covalently immobilized within the hydrogel, adding further complexity to composition-based comparisons. Consequently, there is growing emphasis on physicochemical (Q3) characterization to establish product equivalence when direct evidence of compositional sameness proves elusive.

Measuring the crosslinking efficiency and completion time within the cured insert device is also difficult. In lieu of this, mechanical and physical properties of the network material can act as surrogates to approximate crosslinking density and internal structure. These properties coupled with prior knowledge of the precursor identity or purity and manufacturing mixing conditions such as mixing intensity and pH provide a reliable means to evaluate the network crosslinking. X-ray microscopy also provides an additional avenue to characterize the internal structure of the insert materials, which while only performed in the dry state in this study, may prove useful to characterize differences in the swelling and degradation processes of hydrated inserts. Tensile or compression mechanical tests can offer additional insights into network properties that bear on clinical usability and device durability. Although tensile testing can be challenging for small inserts due to instrument limitations (e.g., sample-length requirements for DMA clamps), compression tests may be more relevant for predicting real-world performance and potential failure modes.

Thermal analysis, including DSC and TGA, demonstrated reproducible differences among the 4-arm, 20 kDa PEG-SG materials, though establishing direct links between thermal fingerprints and functional performance (e.g., drug release) requires further validation. Nevertheless, these data can serve as supportive evidence in demonstrating product similarity.

Swelling studies and IVRT revealed meaningful differences in the inserts made from various PEG-SG materials, distinguishing the impact of polymer source and dexamethasone particle size. Notably, inserts formulated with non-micronized dexamethasone exhibited slower release rates, indicating that within-insert particle size distribution was an important factor for product performance.

Swelling studies are also a necessary test for physicochemical sameness, with regard to the dimensional changes of the dry insert upon swelling when placed into the punctum. Minimal changes in length and suitably close changes in diameter compared to the commercial insert behavior, at both early and long timepoints, in appropriate media, may be beneficial in demonstrating product similarity. Similar long-term swelling and degradation effects can also be monitored, especially when compared against a reference product, to achieve control over the insert’s intended gradual resorption function and dynamic size.

Ultimately, our findings suggested that crosslinking density and API particle size were important factors for drug-release behavior in PEG-based ophthalmic inserts. Notably, crosslinking density and API particle size can be influenced by manufacturing processes, highlighting the need for robust, reproducible methods to control formulation quality and performance. This generalized methodological approach to product physical and chemical analysis could be useful when implemented for development and assessment of other hydrogel-based systems. Although, the release characteristics for amphiphilic and hydrophilic API may differ significantly from the system presented and would necessitate additional study to correlate excipient and implant properties with end product performance. By integrating a suite of physicochemical characterization tools, researchers and manufacturers can better navigate the complexities of ensuring excipient quality and maintaining consistent product performance across different polymer sources and processing conditions.

5. Conclusion

This study established a comprehensive framework of measurement methods to characterize PEG-based ophthalmic inserts, using raw materials sourced from multiple vendors as proof of concept. A series of physicochemical, thermal, and mechanical tests were applied to both intermediate and finished products, providing robust assessments of key quality attributes relevant to polymeric inserts. In addition, a preliminary in vitro release testing method was developed to explore how material attributes affect product performance. These findings enhance our ability to investigate ingredient and product sameness in PEG-based formulations and may assist in the development of cost-effective generic ophthalmic inserts, ultimately broadening patient access to innovative localized drug delivery options.