Transitioning from a lab-scale PLGA microparticle formulation to pilot-scale manufacturing

Abstract

The complexity of scale-up manufacturing of PLGA microparticles creates a significant challenge when transitioning from benchtop-scale formulation development into larger clinical scale batches. Minor changes in the initial formulation composition (e.g., PLGA molecular weight, solvent type, and drug concentration) and processing parameters (e.g., extraction kinetics and drying condition) during scale-up production can result in significantly different performance of the prepared microparticles. The objectives of the present study were to highlight the in vitro and in vivo performance of a candidate benchtop-scale batch created with a rotor-stator mixer, transitioned into an in-line manufacturing process at ~15× scale of a long-acting naltrexone formulation. Physicochemical properties (such as drug loading, residual benzyl alcohol content, and morphology) as well as the in vitro release characteristics of the prepared naltrexone microparticles between the benchtop-scale and in-line process pilot-scale were determined. The pharmacokinetics of the naltrexone microspheres were investigated using the rat model. The results demonstrate that while the morphologies of the particles were different from a visual assessment and slight differences were observed in the in vitro release profiles, the in vivo pharmacokinetics illustrate similar kinetics. Our study shows that scale-up production having the same drug release kinetics can be made by controlling the formulation and processing parameters.

1. Introduction

The inherent complexity of formulating and manufacturing poly (lactide-co-glycolide) (PLGA) microparticle-based products under GMP conditions is demonstrated by the lack of generic versions approved by the Food and Drug Administration (FDA). Zoladex® and Lupron Depot® for treating prostate cancer were approved in 1989, and since then, approximately 20 PLGA-based long-acting products have been approved [1]. A generic product needs to be both qualitatively (Q1) and quantitatively (Q2) similar, although Q1/Q2 sameness does not imply similar performance – in vitro and in vivo [2]. Minor changes in formulation and manufacturing process have demonstrated significant differences in the final composition of the microparticles and performance [2–4]. To further complicate matters, the in vitro testing method, commonly used to discriminate between formulations, also influences the in vitro release kinetics [2,4].

Scalability is crucial for developing long-acting injectable microparticle formulations, although the available literature is relatively scant on this topic. The scale-up of high shear devices such as rotor-stator mixers is based on a practical approach rather than theoretical [5]. Challenges may arise when moving from a benchtop or lab-scale process, typically milligram quantities, into a much larger scale, even simply 10×. Furthermore, extraction kinetic differences may need to be accounted for between lab/formulation design sizes and larger batches. Also, transferring the seed emulsion into an extraction tank is not often considered at small scales. Significant differences in particle size and flow kinetics may be encountered when scaling rotor dimensions and tank sizes change. Flow regimes in the extraction tank may differ depending on the volume, geometry, and impeller shape [5]. Finally, when considering scaling nano-sized PLGA-based formulations, additional complications may arise due to the mechanistic change in particle properties based upon their respective sizes.

Microfluidics is often mentioned as a means of scaling microparticle manufacturing due to such characteristics as scale-independent, reproducible, and fast and easy to control [6]. Microfluidic processes typically result in very narrow/monodisperse particle size distributions [7] that may be advantageous and disadvantageous. Particle size is a property known to influence drug release rates due to surface area and the respective diffusion path lengths [8]. The distribution of sizes in a single dose can potentially aid in maintaining zero-order kinetics over a longer duration due to the distribution [9]. Microfluidics offers significant control over potential microparticle properties, although narrow particle size distributions/monodispersity are typically found based upon the microparticle formation mechanism [10]. Second, these systems often require very low concentrations due to viscosity limitations, ultimately impacting encapsulation efficiencies significantly; an encapsulation efficiency for risperidone of 19.3% was found in a 1% PLGA 0.5% risperidone in dichloromethane (DCM) solution [11]. Extremely low concentration of discontinuous phases may result in longer extraction times leading to greater drug loss. Additional studies with a greater emphasis towards formulation tunability may need to be examined to determine the true potential of microfluidics as both a scalable and viable method for large-scale microparticle manufacturing relative to traditional methods.

The goal of this manuscript is to highlight the preliminary scale-up of a 2-month naltrexone microsphere formulation from a lab-scale to a pilot-scale and the resultant impact on the respective in vitro and in vivo release kinetics. Sharifi et al. demonstrated significant formulation effects in this in-line process at a smaller scale, including but not limited to increasing PLGA concentration, resulting in increased efficiency [12]. Although, as commonly found in other studies, an increase in drug loading also shortened release duration [13]. Based upon this study, a mid-sized molecular weight PLGA with a lactide:glycolide (L:G) ratio of 85:15 was determined to provide the best balance between yield in the target size range, 25–150 μm, drug loading, and release. Careful control of the formulation and processing parameters provides a reliable means to scale-up production for further clinical development.

2. Materials and methods

Ester end-capped 85:15 PLGA from LACTEL® (Birmingham, Alabama) and naltrexone-free base anhydrous SpecGx, L.L.C. (St. Louis, MO) were used for the study. Dichloromethane (DCM), benzyl alcohol (BA), acetonitrile, methanol, potassium phosphate monobasic, and sodium azide were purchased from Fisher Scientific (Fair Lawn, NJ). Emprove® Essential 40–88 (Poly(vinyl alcohol) (PVA)) was purchased from Millipore Sigma (Darmstadt, Germany). Sodium L-ascorbate and phosphate-buffered saline with 0.05% Tween® 20, pH 7.4 (PBST) were purchased from Sigma Aldrich (St. Louis, MO).

2.1. Preparation of naltrexone microparticles

The organic phase consisted of PLGA 85:15 (Lactel B6006–2) and naltrexone-free base dissolved in DCM and BA. For the benchtop scale, the organic phase was mixed in a 20 mL scintillation vial – a 20% (w/v) PLGA in DCM and 46% naltrexone in BA were prepared and mixed with a target 1.67 g batch size and 40% drug loading. The naltrexone mixing was kept to <15 min to minimize the degradation of PLGA by naltrexone [12,14]. Then, 10 mL of the continuous phase was added to the top of the organic phase and homogenized at 5000 RPM for 60 s with an IKA T25 homogenizer with S25N - 10G generator (IKA Works, Inc. Wilmington, NC). The mixture was then transferred into 760 mL of water and stirred at 4 °C for 8 h. Microparticles were then collected with a 25 μm sieve. The product retained by the sieve was de-watered for 15 min at 22 °C and vacuum dried for about 16 h. The microparticles were then suspended and washed in 400 mL of a 25% (v/v) ethanol solution for 8 h at 22 °C to remove the emulsifying agent (PVA) and any residual solvents from the microparticles. The washed microparticles were then collected on a 25 μm sieve and vacuum dried for 48 h.

For the pilot-scale batches, an in-line mixing assembly was fit onto a Silverson L5M-A homogenizer to convert it into an in-line homogenizer. The continuous phase, 1% PVA in ultrapure water, and discontinuous phase, PLGA, and NTX dissolved in DCM, and BA with the same formulation composition were pumped through a Cole-Parmer gear pump and syringe pump at 300 mL/min and 100 mL/min respectively with a target batch size of 25 g. These flow rates were chosen to keep the volume ratio of the continuous and discontinuous phases similar to the lab batch scale, while also providing pilot-scale manufacturing quantities in a relatively short period of time (<60 s). The two phases were pumped into the homogenizer via a tube-in-tube design to minimize premixing before homogenization. Homogenization was performed at 1800 RPM with a medium emulsor screen. The phases were subsequently transferred directly into the extraction vessel loaded with 15.2 L of water at 4 °C and hardened for 4 h. The hardened microparticles were collected and de-watered on a 25 μm sieve mesh. Microparticles were further hardened overnight in a vacuum oven for 18 h and washed with 4 L of 25% ethanolic solution at 22 °C for 8 h. Postextraction, microparticles were collected between 25 and 150 μm, dewatered, and dried under vacuum for 48 h.

2.2. Characterization of naltrexone microparticles

2.2.1. High-performance liquid chromatography

The quantification of naltrexone was performed with an Agilent 1260 HPLC system with a UV absorbance detector set to 210 nm. The HPLC had the following conditions: Mobile Phase: 65:35 methanol:potassium phosphate buffer, pH 6.6; flow rate: 1.0 mL/min; autosampler temperature: room temperature; column temperature: 30C; detection: 210 nm (UV); total run time: 7 min; injection volume: 2.5 μL (drug loading) 10 μL (in vitro release); column: Zorbax SB-C18 150 × 4.6 mm, 5 μm; approximate retention time of naltrexone: 4.8 min.

2.2.2. Drug loading and benzyl alcohol content

Approximately 5 mg of microparticles were accurately weighed, dissolved in 5 mL of acetonitrile, and diluted with the mobile phase. 2.5 μL was then injected with the same HPLC conditions as the in vitro release samples.

2.2.3. In vitro drug release

20 mL of pH 7.4 phosphate-buffered saline with 0.05% Tween 20 and 0.0625% (w/v) sodium ascorbate and approximately 5 mg of microparticles were placed in a stoppered 50 mL Erlenmeyer flask and placed in a 37.0 °C (±0.3 °C) glycerol baths at 30 RPM in a shaking incubator. 1 mL aliquots were taken at various time points and replaced with fresh release medium. Naltrexone content in the buffer was analyzed via HPLC.

2.2.4. Imaging

The morphology of various lots was characterized with a Tescan Vega3 scanning electron microscope. For the internal morphology assessment, microparticles were placed in a − 90 ^°C freezer under sealed conditions, sectioned with a razor blade, and subsequently placed under vacuum in a desiccator to equilibrate to room temperature to avoid any water-induced changes. Microparticles were then mounted onto carbon taped aluminum stubs and sputter-coated with a gold-palladium mixture under vacuum in the presence of argon.

2.2.5. Polymer molecular weight

Samples were dissolved in acetone, filtered through a 0.22 μm PTFE filter, and collected into an HPLC auto-sampler vial for injection. The samples were analyzed using GPC with a quadruple detector (4D). The GPC-4D system consisted of an Agilent 1260 Infinity II HPLC connected to Dawn Heleos II (MALLS) coupled to Dynapro Nanostar DLS via optical cable, Optilab T-rEX (RI detector), and Viscostar III viscometer operated by Astra 7 software. GPC analysis was performed by injecting 50.0 μL of ~2.5 mg/mL polymer solution. The separation was performed with a linear gradient column (Tosoh Bioscience LLC, TSKgel GMHHR-L, 7.8 mm × 30 cm) at 0.6 mL/min flow of acetone with a 60-min run time.

2.2.6. Powder X-ray diffraction

Powder diffraction (XRD) data were collected on a Panalytical Empyrean X-ray diffractometer equipped with Bragg-Brentano HD optics, a sealed tube copper X-ray source (λ = 1.54178 Å), soller slits on both the incident and receiving optics sides, and a PixCel3D Medipix detector. Samples were packed in metal sample cups with a 16 mm wide and 2 mm deep sample area. Anti-scatter slits, divergence slits, and masks were chosen based on sample area and starting θ angle. Data were collected between 4 and 35° in 2θ using the Panalytical Data Collector software.

Samples for stability were prepared (~100 mg) and filled into individual vials under an N₂ purge for each stability time point. Samples were stored in an incubator at either 25 or 40 °C and subsequently analyzed.

2.2.7. Thermal analysis

A Perkin Elmer DSC 7 differential scanning calorimeter was used for thermal analysis. Samples (~5 mg) were analyzed in hermetically sealed aluminum pans under a dry argon purge at 50 mL/min. Indium was used for temperature and heat of fusion calibration (ΔH_f). Samples were heated at 40 °C/min to temperatures approximately 40 °C above the glass transition (T_g).

2.2.8. Particle size distribution

The particle size distribution was measured using a CILAS 1190 particle size analyzer (Madison, WI). Approximately 50 mg of microspheres were dispersed in 1.5 mL of a 0.1% Tween 80 aqueous solution and subsequently analyzed. Each sample was measured in triplicate.

2.2.9. Residual solvent determination

Residual dichloromethane and ethanol were determined using a Shimadzu GC-2010 Plus with HS-10 autosampler using He as carrier gas. Approximately 25 mg of microparticles were weighed and dissolved in 5 mL dimethylacetamide (DMAC). This solution was diluted 5× with MQ water and subsequently crimped and sealed. A calibration curve was prepared with dichloromethane and ethanol in DMAC. A ZB-624 column (30 m, 0.32 mm ID with 1.80 μm film thickness) was used. The initial column temperature was maintained at 40 °C for 5 min, ramped to 150 °C at 10 °C/min, and held at 150 °C for 1 min. The FID detector temperature was 250 °C. Calibration curves for dichloromethane and ethanol were prepared in the same DMAC: MQ water ratio as the samples.

2.2.10. In vivo pharmacokinetic study

The Purdue University Institutional Animal Care and Use Committee approved all animal procedures. Sprague-Dawley rats from Envigo (Indianapolis, IN.) were used for the study and were acclimated for one week before the study. Each formulation was given to four rats injected subcutaneously in the scapular region at 50 mg/kg in an aqueous-based vehicle composed of 0.9% sodium chloride, 0.02% Tween 20, and 0.5% sodium carboxymethylcellulose. The animals were observed for overt toxicity and any existing test site abnormalities during the study, including redness, swelling, bleeding, discharge, and bruising at the injection site. Body weights were taken and recorded at administration and various blood draw time points for ~100 days. Rats were anesthetized and bled (approximately 250 μL) via the tail or submandibular vein. Blood was collected in labeled potassium ethylenediaminetetraacetic acid tubes. Blood was centrifuged for 10 min at 5000 rpm at 4 °C. Before analysis, the plasma fraction was transferred to labeled 1 mL plastic tubes and stored at −80 °C prior to analysis.

Naltrexone was analyzed by liquid chromatography mass spectrometry. A stable labeled deuterated analog of naltrexone was used as an internal standard and for quantitation. D3-Naltrexone was purchased from Sigma Aldrich (St. Louis, MO). All stock solutions were prepared using 100% methanol and stored at −20 °C when not in use. The plasma calibration curve was 1–50 ng/mL final concentration. Plasma samples were stored at −80 °C until analysis. The plasma was thawed and 0.1 mL aliquoted into a tube for naltrexone extraction. 5 ng/mL of d3-naltrexone was added to each sample before extraction. Each sample was extracted with a 5× volume of methyl tert-butyl ether (MtBE). After vortexing for 10 min, the samples were centrifuged at 13,000 rpm for 10 min. The supernatant was collected, transferred to a new tube, and dried using a rotary evaporation device. The samples were subsequently reconstituted in 0.1 mL of 5% acetonitrile +0.1% formic acid prior to LC/MS/MS analysis. The analysis was done with an Agilent 1260 Infinity II liquid chromatography system coupled to an Agilent 6470 QQQ mass spectrometer (Santa Clara, CA). Reverse-phase chromatography used a Water’s T3 column (2.1 × 50 mm, 3.5 μm) for separation (Santa Clara, CA). Buffer A consisted of water +0.1% formic acid, and buffer B was acetonitrile +0.1% formic acid. The linear LC gradient was as follows: time 0 min, 0% B; time 1 min, 0% B; time 10 min, 95% B; time 10.5 min 95% B; time 11 min, 0% B; time 15 min, 0% B. The flow rate was 0.3 mL/ min with a total run time of 10 min. Multiple reaction monitoring was used to analyze each compound (Table 1). Positive polarity electrospray ionization was used with the following source conditions: gas temperature 330 °C, gas flow 8 L/min, nebulizer pressure 45 psi, sheath gas temperature 250 °C, sheath gas flow 7 L/min, capillary voltage 4000 V, nozzle voltage 1000 V, and an electron multiplier voltage of +400. Data were processed using Agilent Masshunter Quantitative analysis software (V.B.08).

Table 1.

Comparison of naltrexone loading, encapsulation efficiency, and residual benzyl alcohol contents (%).

Lactel Lot#	Scale	NTX loading (%)	EE (%)	Residual BA (%)
A18–091	Benchtop	34.7 (0.8)	86.8	0.73 (0.01)
A18–091	Pilot	35.9 (1.0)	89.9	1.22 (0.03)*
A17–028	Pilot	36.2 (0.1)	90.6	1.12 (0.03)*
1613-52-01	Pilot	36.4 (0.7)	91.1	1.11 (0.02)*

^*Statistically significant difference relative to the benchtop formulation.

2.2.11. Statistical data analysis

All data are presented as means with standard error of the mean (SEM) or standard deviation (SD). Statistical analyses were performed using Prism 9.0 (GraphPad Software, La Jolla, CA) using a paired student t-test with a significance level of a p-value <0.05.

3. Results and discussion

Scalability is essential for developing novel PLGA-based products, as this product is susceptible to manufacturing changes. As typical with any formulation development program, small-scale bench-top batches were first prepared to determine the critical process parameters that influence the resultant product attributes. These benchtop batches prepared were <1 g of material; to place this quantity in perspective, this is less than a single dose of the final proposed resultant product. Minor changes in the manufacturing process, including (co-)solvent systems, rate of solvent extraction via temperature changes or addition of solvent to the extraction media, and post-treatment method, have all been demonstrated to influence the physicochemical properties of PLGA microparticles [15]. As previously shown, the performance of naltrexone-PLGA microparticles is sensitive to manufacturing and formulation modifications such as solvent system, solvent removal rate, temperature, discontinuous phase viscosity, and polymer properties such as molecular weight [3,13].

Fig. 1 illustrates the in vivo pharmacokinetic profile of the lab candidate formulation batch in rats dosed at 50 mg/kg. This formulation is being developed to extend the duration of naltrexone to improve retention during opioid use disorder treatment. Currently, Vivitrol® is the only FDA-approved long-acting version of naltrexone. Other products in development include the OLANI implant [15] and Naltrexone Pellet Implant (BICX104) [16]. The required attributes of the formulation under development were equal to or greater naltrexone loading than Vivitrol, minimizing and eliminating quantity of exposure during burst release, and a longer duration (preferably 2 months or greater) until the C_MEC is reached relative to Vivitrol. The lab-scale formulation also demonstrated a drug loading of 34.7% naltrexone, meeting the loading and duration of release criteria. This formulation was chosen as a lead candidate for scale-up and further testing.

Fig. 1.

Plasma concentration profile of a long-acting naltrexone benchtop formulation (rats, n = 4, SC, 50 mg/kg).

An in-line homogenization process, as highlighted in Sharifi et al., was chosen as the technique to scale the process [12]. This process was chosen to enable a continuous seed emulsion to be produced under identical conditions across the batch. Flow rates of 300 and 100 mL/min were chosen as this represents a similar volume ratio of the continuous to discontinuous phase to that of the bench-top formulation. In the bench-top process, the entire seed emulsion was poured into the extraction solution in a single instance. In the in-line method, the seed emulsion flows through a series of pipes into the extraction solution tank. The difference in the initial extraction kinetics, specifically when the seed emulsion meets the extraction tank, could potentially result in variability across the batch in terms of skin formation kinetics, ultimately resulting in physicochemical differences between the benchtop and pilot-scale. Although, an additional step and/or processing modification could be implemented whereby the seed emulsion can be combined with an extraction solution before entering the extraction tank or a tangential flow filtration during extraction to modify the extraction kinetics as the scale increases.

Fig. 2 illustrates the in vitro release kinetics between the batches produced at the pilot-scale and benchtop formulation candidate using a sample and separate method. Various methods have been utilized to characterize the in vitro release profiles of PLGA microparticles, such as sample and separate, dialysis, USP II, and USP IV methods [2–4,17,18]. While the USP IV has been recommended as a dissolution method for microparticles [19], financial and throughput limitations can hinder this method during formulation development and screening. The advantages of the sample and separate method are that it allows for a number of formulations to be screened and/or characterized simultaneously and cost-efficiently. The three pilot-scale batches prepared with the three different lots of polymer are virtually indistinguishable from each other in the sample and separate method. While there appears to be a slight increase in a release from Day 2 to 7 compared to the benchtop scale, the most notable differences between the two methods are found in the 2nd phase of release. PLGA microparticles release profiles are typically triphasic: an initial burst release, followed by a period with an about constant drug release rate, and a final rapid drug release phase leading to complete drug exhaust [20,21]. These three phases correspond to hydration of the microparticle and surface/near-surface drug release, swelling of the microparticle and drug release via constant diffusion, and finally, substantial swelling and degradation where the last and final amounts of the drug are released. This difference in the release could be due to several different factors, including but not limited to differences in drug loading, polymer molecular weight, particle porosity, and particle size. The drug loading of the pilot-scale batches is ~3.5–4.9% higher, and the residual benzyl alcohol content is ~60–75% (Table 1) higher. The residual dichloromethane and ethanol content was <10 and 30 ppm, respectively, for all batches – likely not contributing to any potential difference observed in the performance. In general, the particle size distributions between the benchtop scale and pilot-scale batches are similar (Table 2), although a few instances were observed where statistical differences were noted between d₁₀ and d₅₀ values.

Fig. 2.

In vitro comparison of benchtop scale to pilot-scale batches.

Table 2.

Comparison of the d₁₀, d₅₀, and d₉₀ of the particle size distributions of the respective formulations.

Lactel Lot#	Scale	d10 (μm)	d50 (μm)	d90 (μm)
A18–091	Benchtop	38.5 (1.1)	69.1 (1.6)	106.4 (3.1)
A18–091	Pilot	35.0 (0.7)*	68.3 (2.3)	111.4 (4.1)
A17–028	Pilot	35.5 (1.3)	65.6 (1.5)	106.8 (3.9)
1613-52-01	Pilot	32.8 (1.2)*	62.3 (0.3)*	101.5 (1.3)

^*Statistically significant difference relative to the benchtop formulation.

Modifications in formulation and processing parameters have been well documented to produce many resultant differences in properties. Some commonly altered components are solvent systems [2], extraction temperature [22], emulsification type (i.e., magnetic stirring, vs. homogenization) [3], and PLGA supplier [23]. Trying to keep similar parameters across scales when increasing batch size has not typically been explored or documented. As previously discussed, the flow rates of the discontinuous and continuous phases in the in-line process were placed at a 3:1 flow ratio, keeping the volume ratio similar to that in the benchtop scale to create a similar seed emulsion process. Fig. 3 shows the morphology of microparticles prepared between the benchtop process and 3 in-line batches. It is readily apparent the benchtop process results in a combination of buckled and wrinkled particles, compared to the in-line process that resulted in predominately smooth particles.

Fig. 3.

SEM images of the benchtop batch (A) and three in-line pilot-scale batches 1613-52-01 (B) A17–028 (C), and A18–091 (D) (Scale bar = 200 μm).

Polymer chains and drug molecules accumulate near the skin during solvent removal. As this region grows, mechanical stress accumulates, and once this stress is released, resultant instability occurs, causing inward buckling [24]. Wrinkling on the surface occurs due to stress relaxation due to various interfacial instability caused by mechanical stress, thermal expansion, and swelling-shrinking [24]. While these types of morphologic differences are typically in a continuum, tiny changes in the process can largely influence the morphological properties of the microparticle. The kinetics of the skin formation process and overall extraction kinetics may be different between the two processes. The higher drug loads resultant from the in-line method may be due to faster skin formation in the extraction solution tank due to the greater initial concentration gradient during extraction, as the seed emulsion continuously flows into the tank over a time frame of ~60 s, where the initial flowing seed emulsion observes a greater concentration gradient than the final portion of the seed emulsion, rather than a single transfer step as performed in the benchtop scale. Differences are evident visually, and elevated drug loadings and residual benzyl alcohol content are observed in the in-line process. Furthermore, the high amount of residual benzyl alcohol found in the in-line formulations likely kept the polymer molecules more mobile and inhibited them from collapsing into a more compact structure; therefore, smooth particles were observed. The question now is whether the morphological differences result in an observable variability in the in vivo pharmacokinetic profiles or not.

Gel-permeation chromatography (GPC-4D) was used to determine the polymer molecular weights of the pilot-scale batches. Unfortunately, an insufficient amount of material was available to perform an accurate assessment of the benchtop scale formulation. Tables 3 and 4 illustrate the results of three separate PLGA lots characterized by Durect before and after manufacturing. The two GMP lots (A17–028 and A18–091) of PLGA exhibit minimal to insignificant differences in their number (M_n), weight (M_w), and z-average (M_z) molecular weights, whereas the 1613-52-01 (R&D lot) exhibits a lower initial M_n in the initial assessment. In the final formulation, no differences are observed between the A17–028 and A18–091 formulations, whereas the 1613-52-01 formulation shows slightly lower molecular weights. We hypothesize these differences are not significant enough to cause a difference in release, although the summation of the minor differences between formulations could lead to the observed variability.

Table 3.

Molecular weights of PLGA as received.

PLGA	Lot #	Mn (kDa) (± SD)	Mw (kDa) (± SD)	Mz (kDa) (± SD)	MHS slope*
Lactel B6006—2P	A17–028	66.016 ± 0.377	78.198 ± 0.456	94.406 ± 1.107	0.648
	A18–091	66.038 ± 0.250	79.689 ± 0.114	96.626 ± 0.507	0.652
	1613-52-01 (R&D Only)	58.277 ± 0.279	74.102 ± 0.259	93.901 ± 0.738	0.637

^*M.H.S. slope: The slope a of the Mark-Houwink-Sakurada equation, [η] = KM^a.

The PXRD patterns of the benchtop scale batch and 3 in-line batches are shown in Fig. 4. As demonstrated by the diffraction peaks, a portion of the naltrexone is crystalline. The patterns are all similar without any clearly defined peaks present in any of them; therefore, they all likely consist of the same type(s) of polymorphic and solvated forms. The plots overlaid (Fig. 4B) clearly show the difference in the amorphous content between the different batches. Based on the microparticle size and naltrexone size present in the microparticles, coupled with similar drug loadings/polymer across the 4 batches, the benchtop batch likely has the most significant quantity of naltrexone crystallinity present based upon the greater area under the crystalline peaks and lower area under the amorphous halo. The 3 in-line batches appear to have a slightly lower percent crystallinity based on the abovementioned areas. As a difference in crystallinity is noted between the benchtop scale and the pilot-scale batches, a small-scale stability study was performed under storage temperatures of 25 and 40 °C to determine if a change in crystallinity can be observed as a function of temperature. Fig. 4C illustrates the patterns as a function of storage time for the A18–091 pilot-scale batch. A discernable difference between the samples is not readily noted, whereby the storage temperature or time frame was not sufficient to induce crystallization of the amorphous portion. The other samples can be found in the supplementary information. The glass transition was determined for each lot of polymer and formulation to further study the thermal properties of the system (Fig. 5). No statistical difference is noted between each polymer lot or between the benchtop and any pilot-scale formulation. A difference is noted between the raw polymer and formulation as expected due mainly to the decrease in molecular weight during processing, residual benzyl alcohol, and incorporation of naltrexone into the polymer matrix.

Fig. 4.

PXRD patterns of benchtop scale batch compared to in-line batches separately (A), overlaid (B), and patterns from stability study for batch A17–028 (C).

Fig. 5.

Heat flow as a function of temperature for the benchtop and 3 pilot-scale formulations and their respective T_g,onset.

*Statistically significant difference relative to the benchtop formulation.

Naltrexone exists in four different forms in the Vivitrol product, and the ratio of these forms influences the in vitro release rate according to US Patent 7,279,579 B2 [25]. Also, the percent crystallinity (AUC of amorphous halo subtraction → ratio of crystalline AUC to total AUC) was illustrated to impact the amount released on Day 14 in vitro and amount released on Day 10 in vivo. For the in vivo percent release at day 10, percent crystallinity of formulations of ~3.6, 9.0, and 17.1 provide ~79.2, 60.5, and 46.6% release (values obtained via digitization), whereby these percent crystallinity values correlate to a naltrexone percent crystallinity of ~10.7, 26.7, and 50.7. Standard deviations were not provided to determine the statistical importance of these values. Determining the impact of crystallinity on the release behavior of PLGA-based systems has not been thoroughly explored in sufficient detail. Nevertheless, this type of release behavior is common in amorphous dispersions for oral delivery, where the goal is obtaining an increase in release rate and/or supersaturation solubility to ultimately achieve an increase in bioavailability. In this formulation, a similar impact may exist and will be characterized in more detail in future studies. In addition to this, the impact of the environment (SC and IM) postinjection also needs to be further determined. Post injection, the formulation is exposed to both an increase in temperature and the presence of subcutaneous aqueous fluids. This combination could potentially lead to an increase in crystallization of the encapsulated naltrexone, ultimately mitigating some of the initial starting formulation properties. The stability experiments only probed the impact of temperature, whereas aqueous fluids could further increase the molecular mobility of both the polymer and drug. Finally, the variability often found in in vivo experiments could mitigate any minor changes in the performance observed in vitro.

The mean plasma concentration-time profiles of the benchtop batch and the two pilot-scale batches (A18–091 and A17–028) are illustrated in Fig. 6. Lactel B1613-52-01 was obtained as an R&D batch and, therefore, not characterized in vivo. While it appears there is a significant burst peak in the Pilot Scale A18–091, a plasma concentration of ~73 ng/mL was determined for one of the rats in this group (9.03, 25.5, 72.9, and 7.280 ng/mL at t = 4 days), significantly skewing the profile (Fig. 6–A). This point was determined as an outlier and removed from the dataset during further analysis (Fig. 6–B). The steady-state release is observed with plasma concentrations of all 3 formulations ≥1.0 ng/mL at day 56 and slightly beyond (Fig. 6–C). Finally, a slight burst release is observed for the 2 in-line batches relative to the benchtop formulation (Fig. 6–D), which correlates to their respective in vitro release profiles. Furthermore, no statistical difference was noted in the AUC, C_max, or T_max between the benchtop and pilot-scale formulations (Table 5).

Fig. 6.

Plasma concentration vs. time profiles of benchtop formulation batch compared to the in-line batches at different zoom levels (A) – all raw data before outlier removal, (B) data after outlier removal, (C) 56-day highlight, and (D) 28-day highlight.

Table 5.

Pharmacokinetic parameters comparison of the 3 formulations tested in vivo.

Formulation	AUCd0–98 (ng•d/mL) ± SEM	Cmax (ng/mL) ± SEM	Tmax (day) ± SEM
Benchtop Scale A18–091	238.1 ± 21.0	9.98 ± 1.41	26.13 ± 11.36
Pilot Scale A18–091	224.1 ± 25.1	13.60 ± 4.12	9.25 ± 5.25
Pilot Scale A17–028	196.9 ± 13.0	11.49 ± 1.92	2.00 ± 0.68

^*Statistically significant difference relative to the benchtop formulation.

4. Conclusion

The impact of formulation and processing parameters on the resultant performance of microparticles is well known. Particle size/porosity, L:G ratio, molecular weight, and drug hydrophobicity/hydrophilicity are all properties that have fairly well-established correlations of their impact on performance. For transitioning these formulations towards clinical products, characterization of scaling a formulation and process and the resultant impact on the final properties has not been extensively documented. In this study, differences are noted in the visual appearance of the microparticles and in vitro release profiles of a small-scale benchtop process to an in-line homogenization process at an ~15× scale. However, the in vivo plasma concentration profiles of the pilot-scale are statistically similar across the AUC, T_max, and C_max to the benchtop formulation. Building upon this study, larger-scale batches need to be prepared to determine if the kinetic effects of skin formation and extraction are similar by simply increasing the run time of the homogenization process or if additional measures are required to produce microparticles with similar properties found in these two scales. Finally, tablets and capsules have multiple on/in-line parameters that can be monitored to ensure a manufacturing process remains within specification or brought back in (i.e., tablet weight, hardness, blend uniformity). To date, monitoring microparticle droplet size is the only potential in process control that may be utilized to monitor the formulation. Therefore, future development may need to focus on this aspect of microparticle development to enable tighter control of the process and resultant properties.

Table 4.

Molecular weights of pilot-scale batches.

PLGA	Lot #	Mn (kDa) (± SD)	Mw (kDa) (± SD)	Mz (kDa) (± SD)	MHS slope
Lactel B6006—2P	A17–028	43.470 ± 0.149	57.072 ± 0.181	75.053 ± 0.547	0.622
	A18–091	43.607 ± 0.174	57.292 ± 0.184	75.108 ± 0.540	0.630
	1613-52-01 (R&D Only)	42.329 ± 0.193	55.729 ± 0.200	73.463 ± 0.579	0.620