Abstract
Purpose:
In a benchwork particle counting analytical evaluation, the number and type of particles in intravitreal injection formulations of three different agents against vascular endothelial growth factor were investigated.
Methods:
Commercially available ready-to-use aflibercept and brolucizumab glass syringes, vials containing bevacizumab (off-label use in ophthalmology), and repackaged ready-to-use plastic syringes containing bevacizumab were tested without filtration. Total visible, subvisible, and nanoparticles numbers and size distributions were quantified using light obscuration, flow imaging, resonant mass measurement (RMM), tunable resistive pulse sensing, and dynamic light scattering.
Results:
Repackaged bevacizumab showed overall low particle numbers, aflibercept showed high numbers of micrometer sized particles but low nanoparticle numbers, brolucizumab showed low to moderate numbers of micrometer sized particles but high nanoparticle numbers. RMM measurements identified particles in the nanometer range as either proteinaceous or silicon oil; the nature of the other particles was not further evaluated.
Conclusions:
Repackaged bevacizumab shows no inferior particle quality compared to ready-to-use products. It is relevant to study nanoparticle load of the products as the micrometer-sized particle numbers do not in all cases correlate to nanoparticle counts. Particularly for the high concentration product Beovu (brolucizumab), high nanoparticle numbers were found despite low numbers of micrometer sized particles. Silicone oil droplets did not account for high particle numbers as the measured numbers were low.
Translational Relevance:
Different side effects are registered in different frequencies with different intravitreal anti-VEGF-drugs and syringes, which are applied by injection by small 30G needles through the sclera directly to the intravitreal cavity. The study of nanoparticles and silicone oil droplets may be able to contribute to narrowing down the causes.
All products implemented in the study were sent to the place of measurement under cooling conditions and stored at 4°C after arrival. Avastin (10 syringes, 3.75 mg bevacizumab/150 µL) was repackaged by Asklepios Klinik Nord, Hamburg into 1 mL BD Plastipak Luer-Lock syringes (Becton, Dickinson and Company, Franklin Lakes, NJ) and in total kept for a maximum of five days. Avastin was provided in glass vials (4 mL, 25 mg/mL). Eylea (40 mg aflibercept/mL) and Beovu (120 mg brolucizumab/mL) were both obtained as prefilled syringes from the manufacturer with 90 µL and 165 µL each, respectively.
Depending on the small number and volume of the samples, different pools had to be formed, as indicated in
Table 1.
Table 1. Sample Names, Concentration, Primary Packaging, and Number of Pools and Aliquots Formed at LMU Munich
Table 1. Sample Names, Concentration, Primary Packaging, and Number of Pools and Aliquots Formed at LMU Munich
None of the products of the study were filtrated before pools were formed or aliquots were aspired. To form two pools from repackaged Avastin (Avastin 1 and 2), five syringes each were discharged into two 15 mL Greiner polypropylene (PP) tubes.
One pool was formed by discharging eight syringes of Eylea into one 15 ml Greiner PP tube.
Two pools of Beovu (Beovu 1 and 2) were prepared discharging four syringes each into two 15 mL Greiner PP tubes (Greiner tubes from Sigma-Aldrich Inc., Darmstadt, Germany).
From the comparably large Avastin vial, aliquots of 0.8 ml each were aspired with a sterile 18G × 1 ½ needle into four sterile 1 mL Luer-Lock Tip syringes (both Becton, Dickinson and Company). Each syringe was discharged into a 15 mL Greiner PP tube to form four samples (Avastin vials 1–4).
Due to the small volume of the available products, the products had then to be diluted to allow all analytical measurements. All products were diluted 1:20 with the corresponding product buffer (placebo) solutions. The corresponding buffer solutions of the marketed products were used to avoid any incompatibility with other diluents was taken from official documents.
2,9,10 There were no gas bubbles inside the evaluated syringes.
All measurements were performed on calibrated equipment: FlowCAM and PAMAS were calibrated with Duke Standards and Count-Cal Particle Count Controls (NIST Traceable Size Standards 2, 10, and 25 µm) (all Thermo Fisher Scientific, Fremont, Waltham, MA). Different measurement techniques were used to assess particle sizes and numbers. To detect subvisible and visible particles, light obscuration was used, in principle following USP guidelines. As orthogonal technique, flow imaging was applied, which is able to detect translucent particles in the subvisible size range better than light obscuration. To measure particles in the nanometer range, resonant mass measurements and tunable resistive pulse sensing were applied. RMM also allowed the differentiation of protein and silicon oil particles by their difference in densities.
Particles in the micrometer range, were analyzed using the FlowCAM 8100 (Fluid Imaging Technologies, Inc., Scarborough, ME) with a 10× magnification cell (80 µm × 700 µm). The flow cell was rinsed, and cleanliness was verified with highly purified water (HPW, <100 particles/mL). Samples were measured in triplicates of 150 µL at a flow rate of 150 µL/min and a threshold of 1013. Results were evaluated with the software Visual Spreadsheet Version 4.7.6 (Fluid Imaging Technologies, Inc.).
An Archimedes system, equipped with a Hi-Q Micro Sensor (both Malvern Instruments, Malvern, UK) and the Archimedes software v1.20 was used for RMM analysis of particles up to 4 µm in size. The system was calibrated with polystyrene size standards of 0.994 µm specified diameter (Duke Standards; Thermo Fisher Scientific, Waltham, MA) and system cleanliness was verified.
Particle densities were set to 0.97 g/mL for positively buoyant particles (considered as silicone oil particles), and to 1.32 g/mL for negatively buoyant particles (considered as proteinaceous). Each sample was loaded for 40 seconds and the limit of detection was automatically determined by the instrument software. Minimum detectable particle sizes were approx. 485 nm for silicone oil particles and approx. 274 nm for protein particles.
Samples were analyzed in triplicates with a measurement time of 600 seconds as stop criterion, corresponding to an analyzed volume of ∼150 nL per replicate.
Data evaluation was performed with the LINK software platform v2.3.22.200619 (Lumetics, Nepean, ON, Canada).
Particles in the nanometer range (150–900 nm) were determined with TRPS using the qNano Gold (IZON, Nottingham, UK). For analysis, pore NP 300 (A57745, IZON) was chosen, a stretch of 47.01 mm and a pressure of around 15 mbar were applied. All samples, buffers, and the calibration particles CPC 400 (mean diameter 350 nm, 7.56 × 108 mg/mL; IZON) were spiked with a small amount of NaCl solution up to a final concentration of 140 mM of sodium chloride.
The lower fluid chamber was loaded with 80 µL of the respective buffer and the voltage was set to 0.34 V for the analysis of the Avastin vial and to 0.38 V for all remaining samples.
Twenty-five microliters of spiked sample or buffer were pipetted into the upper cell and measured in triplicates for 10 minutes. Results were calculated using the included software IZON CONTROL SUITE (IZON), with the focus on particle sizes of 150 and 300 nm.
To ensure no aggregation occurred due to spiking the samples, size and polydispersity (PDI) were counterchecked with dynamic light scattering for all samples used for TRPS measurements.
To determine size and PDI, 25 µL of each sample was pipetted in triplicates in a 384 well plate (Corning, Glandale, AZ). The plate was spun down at 1000 rpm for one minute at 20°C with the Heraeus Megafuge 40 centrifuge with a M20-well plate rotor (both Thermo Fisher Scientific), each well was sealed with 5 µL of silicone oil and centrifuged again. Then it was measured at 25°C with 10 acquisitions of five seconds for each well using the DynaPro DLS plate reader (Wyatt Technology Europe, Dernbach, Germany) and results were calculated using the included Dynamics V7.8 software.
Table 2 provides an overview of all particle numbers measured with FlowCAM and LO.
Table 2. Calculated, Cumulative Particle Numbers Per Milliliter (≥1, ≥10, and ≥25 µm) of Samples and Buffers, Measured with LO and FlowCAM
Table 2. Calculated, Cumulative Particle Numbers Per Milliliter (≥1, ≥10, and ≥25 µm) of Samples and Buffers, Measured with LO and FlowCAM
For both methods all the dilution buffers showed very low particle numbers and must not be considered further.
Looking at the four aliquots from the commercially available Avastin product vial, particle numbers and sizes varied noticeably for both LO and FlowCAM. Avastin vial aliquot 1 exhibited quite low particle counts, whereas vial aliquot 4 showed high numbers for small (≥1 µm) and medium (≥10 µm) particle sizes.
Repackaged bevacizumab syringes showed overall smaller particles numbers than the samples from the Avastin vial. The two pools showed almost identical particle numbers indicating a robust particle level.
Eylea syringes showed very high particle numbers in the small subvisible size range ≥ 1 µm in both applied methods. Particle numbers in the ≥10 and ≥25 µm range were measured low with LO and rather high with FlowCAM.
Beovu syringes showed medium particle levels, comparable to those measured for bevacizumab, falling between the repackaged syringes and the samples from the Avastin vials.
In all RMM measurements only low particle counts near the limit of quantification (3 × 10
5 particles/mL) were found. With that, calculated numbers for the undiluted products can only be considered as estimates. Particle numbers in the dilution buffers can be considered as negligible. The results for RMM nanoparticle quantification are presented in
Figure 1 and
Table 3.
Table 3. Calculated Results from RMM of All samples and Respective Buffers
Table 3. Calculated Results from RMM of All samples and Respective Buffers
For all the bevacizumab preparations, only very low nanoparticle numbers were recorded. The number of particles considered as silicone oil droplets was apparently in the same, very low order of magnitude as the protein particle numbers.
Interestingly, for Eylea and for the Beovu pools, comparably higher numbers of about 14 million counts/mL and approx. 29–32 million counts/mL were calculated in the size range between approx. 274 nm and approx. 1000 nm. In none of these cases did silicone oil droplets show up in relevant numbers.
Like for RMM, the overall particle counts were also very low; the particle rate during the measurements was below 100 particles/min.
The results for the calculated particle numbers at bin sizes of 300 nm are presented in
Figure 2 (results for 150 nm are similar, data not shown).
For the bevacizumab products, the nanoparticle numbers for the dilution buffers fell into the same range as for the diluted product samples, indicating that there is no relevant nanoparticle load present.
For Beovu, the situation resembles the same impression received from RMM, that is, having a higher nanoparticle number than for all the other products. The differences between the noise level represented by the buffers and the measured values for the Beovu product was not as large as for RMM. For Eylea, the values found with TRPS are very low, not different from the buffer.
Comparison of Aliquots from a Commercially Available Avastin Vial and Pools from Repacked Syringes from a Compounding Pharmacy
In the course of repackaging bevacizumab from vials into syringes for intravitreal application handling errors and stresses can be applied to the protein and the used syringes could potentially contain silicone oil that could be shed into the product during handling and intermediate storage.
3,25,36,37 In this study we cannot confirm previously found quality gaps for repackaged bevacizumab.
The results obtained from LO and FlowCAM show that the repackaged syringes exhibited a low particle count even compared to the samples from the vial regarding particles ≥ 1 µm.
The partly higher particulate burden in aliquots from the vial might be explained by the fact that this material was analyzed as provided, whereas in the compounding pharmacy, particles should have been removed by a filtration step in the course of the repackaging. All particle counts (≥10 µm) obtained by LO measurements would exceed limits set by the USP—counts for particles ≥ 25 and ≥ 50 µm were in line.
Avastin, both in vial and syringes, contains the lowest concentration of protein (25 mg/mL), Eylea a medium concentration (40 mg/mL) and Beovu the highest (120 mg/mL). Assuming an injection amount of 50 µL per eye, the overall dose of protein (in mg) applied thereby differs by a factor of about 3 to 5 between the products. When we try to correlate the particle numbers found with the protein concentration, we find that Eylea surprises with unexpectedly high numbers of particles in the range of 1 µm. For the larger particles of ≥10and ≥25 µm in Eylea, the dataset is not as clear, because the LO and the FlowCAM numbers differ. The low number for Eylea in LO and the high numbers for FlowCAM indicate that the aggregates formed are of a translucent and (for LO) hard to detect nature. Although being slightly higher in particles than bevacizumab, Beovu does not show a correlation of protein concentration and particle counts. Taking USP <789> into account, all three products would exceed the limit of particle counts set for particles ≥ 10 µm.
Nanoparticles provide a totally different picture. Here, Beovu shows relatively high numbers of nanoparticles, whereas the other products are low to very low. One must consider that the absolute values for Beovu are still low, but it is apparent that here the nanoparticle load correlated with the protein concentration. Eylea (with a medium high protein concentration of 40 mg/ml) is positioned in the middle, at least for one of the two methods applied.
Another observation is also remarkable, that is, that silicone oil nanoparticles play no relevant role in any of the products on the nanoparticulate level.
As mentioned in the introduction, Beovu is the ophthalmic product in the panel studied that contains the smallest therapeutic protein (brolucizumab exhibits a size of only 26 kDa), but with the highest concentration
12: It is therefore not possible that the nanoparticle counts were caused by high concentrations of the monomeric drug; it must be aggregates of any kind.