Impact of Injection Technique on Microsphere Distribution During Transarterial Radioembolization in a Successively Bifurcating In Vitro Model

Abstract

Purpose

This study aimed to experimentally investigate how injection technique affects the distribution of microspheres during transarterial radioembolization in a successively bifurcating in vitro model.

Methods and Materials

A symmetrical phantom, bifurcating 3 times into 8 outlets, was incorporated into a flow circuit. A blood-mimicking fluid was pumped through the phantom using a physiological representative waveform. A microcatheter was placed into the lumen of the phantom, and holmium-165 microspheres were administered with a conventional administration device and a newly designed controlled administration device, containing a rotating syringe to keep the microspheres in suspension during administration. Two clinicians performed manual injections to establish clinically relevant injection rates. Then, different injection profiles were tested using syringe pumps: pulsed vs continuous injections (24 mL/min), and reduced continuous injection rates (10 and 5 mL/min). Microspheres were collected at each outlet and their distribution over the 8 outlets was analyzed.

Results

Continuous high injection rates led to more homogeneous radial distributions of microspheres over the right side of the phantom (outlet 5-8 received 16.5%-23.1% of the microspheres per outlet) compared with the clinically standard used pulsed injections (outlet 5-8 received 11.3%-40.1% of the microspheres per outlet). In contrast, reduced continuous injection rates led to more selective distributions (outlet 5-8 received 2.5%-68.8% of the microspheres at 10 mL/min and 1.0%-80.0% at 5 mL/min).

Conclusions

Injection technique strongly influences the distribution of microspheres. During high continuous injections, more mixing between microspheres and blood-mimicking fluid was observed. This led to more uniform radial microsphere distributions, creating a more predictive setting for transarterial radioembolization.

Introduction

Transarterial radioembolization (TARE) is an established treatment option for primary and secondary liver cancer.^1,2 During TARE, radioactive microspheres are injected into the hepatic artery via a microcatheter. The technique relies on the unique dual blood supply of the liver; tumor tissue is predominantly supplied by the hepatic artery, whereas normal liver parenchyma is mostly supplied by the portal vein. The hepatic arterial blood flow transports the injected microspheres primarily to the tumors, where they lodge in the arterioles and emit a high radiation dose to the tumor cells.

During the TARE workflow, cone beam computed tomography (CBCT) contrast injections and scout dose injections are used to predict microsphere deposition. CBCT contrast injections are performed continuously via a contrast media injector using high rates (up to 120 mL/min),³ whereas technetium-99m-macroaggregated albumin (^99mTc-MAA) scout doses are injected manually from a syringe at lower, mostly unknown rates, 1 to 2 weeks before the treatment. In contrast, TARE microspheres are injected manually using pulsed bolus injections (0.1 mL per pulse)⁴ with a conventional administration device (AD). The conventional AD consists of a V-vial setup that enabled pulsed injection of microspheres via overpressure (Fig. EA1, EA). The CBCT contrast agent, scout doses, and actual microsphere delivery, differ in physical properties and therefore may respond differently to injection dynamics. Furthermore, the variation in injection profiles may also contribute to discrepancies in predictive accuracy.⁵

The impact of injection rate on microsphere distribution has already been shown in vitro and in silico.^6,7 Caine et al⁶ found that higher injection rates improve mixing of microspheres with blood, resulting in a distribution that more closely aligns with the overall blood flow. Basciano et al⁷ showed that microspheres follow fluid stream lines within the vessel lumen and that proper mixing between the microsphere fluid stream lines and the blood only occurs during peak systole. This concept, however, has not yet been evaluated using the conventional AD currently employed in clinical practice.

Optimizing microsphere-blood mixing may benefit lobar or whole-liver TARE treatments by ensuring even distribution to downstream branches. In superselective TARE, guiding microspheres along specific streamlines may be preferable. However, in clinical practice, different injection rates and profiles are applied, which may contribute to differences in TARE outcomes. Therefore, a better understanding of the relation between injection technique and microsphere distribution could be beneficial to improve microsphere deposition predictions, with the ultimate goal of improving TARE outcomes.

The aim of this study was to investigate currently used injection rates and profiles in clinical practice and how the injection rate and profile affects the distribution of microspheres in a successively bifurcating in vitro model.

Methods and Materials

Part 1: investigation of pulsed injection methods with the conventional AD

In this first part of the study, which was executed without the use of a phantom, the injection rate and profile required to achieve proper agitation of microspheres in the conventional AD were determined. This was done by asking a nuclear medicine physician and an interventional radiologist who routinely perform TARE to replicate the procedure. Both clinicians followed the same procedural steps. The number of microspheres delivered from the conventional AD to the catheter tip, in combination with the instructions for use (IFU) for QuiremSpheres,⁴ served as a foundation for the investigated injection rates and profiles with the successively bifurcating phantom in the second part of this study.

In practice, holmium microspheres are administered through a centrally placed glass V-vial within the conventional AD. This vial is connected to a delivery set, consisting of tube lines A and B (Fig. EA1, EA). Manual injections of 0.9% saline solution via tube line A and B are performed in a pulsed manner (0.1 mL per pulse) to administer microspheres; see IFU for Quiremspheres.⁴

For the experiments, a 130 cm 2.7 Fr. microcatheter (Progreat, Terumo Europe) was attached to the distal part of tube line A of the delivery set. V-vials were filled with 300 mg nonradioactive holmium-165 loaded poly(L-lactic acid) microspheres (Quirem Medical B.V.) with a density of 1.4 g/mL⁸ and a mean size of 30 µm (±5 µm)⁹ and 2 mL phosphate-buffered pluronic solution.

Five injection methods were tested (Table 1), with each experiment repeated 3 times. The nuclear medicine physician and interventional radiologist independently determined the fluid volumes for microsphere agitation (2.5 mL) and flushing (4 mL and 5 mL, respectively), reflecting their typical approach during a TARE procedure. Using the injection practices of the nuclear medicine physician and interventional radiologist as a reference, we designed 3 additional syringe pump settings aimed at closely replicating their manual injection techniques. Microspheres and saline solution reaching the catheter tip were collected in separate centrifugal collection tubes per injection repetition. To determine the number of microspheres in the solution, the solution was washed 3 times with 30 mL distilled water after settling of the microspheres, using intermittent centrifuge steps (10 minutes, 1500 rpm) and removal of the supernatant. After evaporating the remaining fluid, the microsphere weight was determined,¹⁰ and compared for the different tested injection methods.

Table 1.

Experimental design investigating various pulsed injection methods with the conventional administration device

	Protocol
Injection method	Administration (tube line B)	Flush (tube line A)	Repetitions
Manual injection (nuclear medicine physician)	2.5 mL (pulses of $\sim$0.1 mL, $\sim$25 ×)	5 mL (pulses of $\sim$0.1 mL, $\sim$50 ×)	5
Manual injection (interventional radiologist)	2.5 mL (pulses of $\sim$0.1 mL, $\sim$25 ×)	5 mL (pulses of $\sim$0.1 mL, $\sim$50 ×)	5
Programmable syringe pump injection	2.5 mL (pulses of 0.1 mL in 0.6 s, 0.2 s pause, 25 ×)	5 mL (pulses of 0.1 mL in 0.6 s, 0.2 s pause, 50 ×)	5
Programmable syringe pump injection	2.5 mL (pulses of 0.1 mL in 0.25 s, 0.2 s pause, 25 ×)	5 mL (pulses of 0.1 mL in 0.25 s, 0.2 s pause, 50 ×)	5
Programmable syringe pump injection	2.5 mL (pulses of 0.1 mL in 0.125 s, 0.2 s pause, 25 ×)	5 mL (pulses of 0.1 mL in 0.125 s, 0.2 s pause, 50 ×)	5

Part 2: microsphere distribution in a successively bifurcating phantom for different injection rates

In the second part of the study, it was investigated how microspheres distribute in a successively bifurcating phantom using different injection rates and profiles. A symmetrical in vitro phantom (Fig. EA2, EA) was integrated in a flow circuit. A detailed illustration of the phantom setup is available in previously published work.¹⁰ The combination of a continuous and pulsatile pump generated a physiological representative flow profile of the right hepatic artery, inspired on the work of Aramburu et al¹¹ The resulting flow passed through the symmetrical phantom, with its outlets directed into an open fluid collection reservoir. The outlets did not include additional resistances. The blood-mimicking fluid (BMF) used in the setup was a mixture of water, glycerol (Glycerol waterfree s.g.1,26, Boom B.V.) and urea (99.3+% purity, Thermo Fisher Scientific Inc), with respective weight percentages of 55.8%, 34.2%, and 10%, yielding in a density of 1112 kg/m³ and viscosity of 3.5 cP.¹²

Upstream of the phantom, a straight-shaped 110 cm 5 Fr. guiding catheter (Radifocus Glidecath, Terumo Europe) was introduced in the system via an MBA Hemostasis Valve (Merit Medical Systems Inc). The distance from the guiding catheter to the first bifurcation in the phantom was 10 cm. The 130 cm 2.7 Fr. microcatheter was inserted via the guiding catheter and positioned in the phantom, 2 cm before the first bifurcation (Fig. EA2, EA).

Outflow calibration

To check whether the flow to all outlets was equal, a calibration run was performed in which the BMF running through the phantom was collected in collection tubes at each outlet for 3 minutes, without injection of microspheres, but with the guiding catheter and microcatheter in place.

Microsphere administration

For the microsphere administration, 2 different ADs were used. First, the conventional AD with a centrally placed V-vial from which the suspended microspheres are flushed toward the catheter (Fig. EA1, EA). Second, a newly designed controlled AD, containing a rotating syringe to keep microspheres in a continuous suspension during administration. This controlled AD was recently tested in a clinical study.¹³ Both ADs were operated by 2 syringe pumps: one to load microspheres in the tube line and microcatheter (PHD ULTRA, Harvard Apparatus, Inc, ± 0.25% accuracy, ± 0.05% reproducibility)¹⁴ and one to flush the tube line (NE-4000 Multi-Phaser, New Era Pump Systems Inc, ± 1% accuracy, ± 0.1% reproducibility).¹⁵ Following the IFU with 0.1 mL per pulse,⁴ a workable pattern is injection in 0.25 s (resulting in an injection rate of 24 mL/min per pulse), followed by a 0.2 s pause. This approach generates sufficient pressure to adequately agitate microspheres within the vial for the conventional AD.

V-vials of the conventional AD were filled with 300 mg holmium-165 loaded microspheres and 2 mL phosphate-buffered pluronic solution. The protocol included 5 repetitions of intermittent administration of 2.5 mL saline solution via tube line B for the microsphere suspension and 5 mL saline solution via tube line A for the flushing. The controlled AD was also filled with 300 mg microspheres, 2 mL 0.1% phosphate-buffered pluronic solution and 10.5 mL saline solution (total 12.5 mL), to create a homogeneous suspension of 24 mg/mL. The protocol included 5 repetitions of intermittent administration of 2.5 mL from the controlled AD and 5 mL saline solution from the flushing syringe. For the controlled AD, similar injection rates as for the conventional AD were tested (24 mL/min), as well as lower injection rates (10 and 5 mL/min). An overview of the tested settings is available in Table 2. Each experiment was repeated 5 times.

Table 2.

Experimental design investigating the microsphere distribution in a successively bifurcating phantom for various injection rates and profiles

			Protocol
Administration device	Injection rate (mL/min)	Profile	Administration	Flush	Repetitions
Conventional device	24	Pulsed	2.5 mL (pulses of $\sim$0.1 mL in 0.25 s, 0.2 s pause, 25 ×)	5 mL (pulses of $\sim$0.1 mL in 0.25 s, 0.2 s pause, 50 ×)	5
Conventional device	24	Continuous	2.5 mL in 6.25 s	5 mL in 12.5 s	5
Controlled device	24	Continuous	2.5 mL in 6.25 s	5 mL in 12.5 s	5
Controlled device	10	Continuous	2.5 mL in 15 s	5 mL in 30 s	5
Controlled device	5	Continuous	2.5 mL in 30 s	5 mL in 60 s	3

Qualitative analysis of catheter outflow patterns

To monitor the microsphere flow from the catheter tip into the vessel, blue dye (Dr Oetker) was added to the injection solution (30 mg/mL). The outflow pattern at the catheter tip was recorded using both a top-view and side-view camera position, operating at 30 frames/s and 3840 × 2160 image resolution (Logitech BRIO webcam, Logitech), resulting in approximately 135 pixels over the vessel lumen and thus a pixel size of 0.028 mm.

Quantitative analysis of catheter outflow patterns

During microsphere injection, BMF, injected saline solution, and microspheres were collected in collection tubes at each outlet. The dye concentration of the total collected fluid volume was measured using a spectrophotometer (Tecan Infinite 200 PRO, Tecan Trading AG). To determine the number of microspheres in the solution, a similar method was followed as described in part 1 of the study (including washing steps, centrifuge steps, and removal of the supernatant). After evaporating the remaining fluid, microsphere's weight was determined.

Statistical analysis

The Shapiro-Wilk test (n < 50) was used to determine whether the data were normally distributed.¹⁶ Mean and SDs were used unless indicated otherwise. The microsphere distribution was expressed as the percentage of total microspheres collected across all outlets, indicating how much of the total ended up in each specific outlet. For comparison of microsphere distributions, the Pearson’s correlation coefficient (r) was used, with 0 ≤ r < 0.1 none, 0.1 ≤ r < 0.3 poor, 0.3 ≤ r < 0.6 fair, 0.6 ≤ r < 0.8 moderate, 0.8 ≤ r < 1 very strong, r = 1 perfect correlation.¹⁷ The data were processed using SPSS statistical software package version 29 (SPSS Inc).

Results

Part 1: investigation of pulsed injection methods with the conventional AD

The manual injections performed by the nuclear medicine physician followed a similar pattern compared with the 24 mL/min syringe pump injection, whereas the manual injections performed by the interventional radiologist followed a similar pattern compared with the 48 mL/min syringe pump injection (Fig. 1). The manual injections by the 2 clinicians and the 24 mL/min syringe pump injection demonstrated high variability across the 3 repeated measurements, whereas the 10 and 48 mL/min syringe pump injections demonstrate relatively narrow ranges; see shaded bars in Fig. 1. None of the injection profiles led to a 100% administration of the present microspheres within the 5 injected repetitions (Y-axis in Fig. 1). Details on the residual number of microspheres in the V-vial and the tubing for the investigated injection profiles can be found in Supplemental file EB. The 24 mL/min injection rate was chosen as reference during the second part of this study.

Figure 1.

Accumulated percentage of injected microspheres for various pulsed injection methods performed with the conventional administration device. Percentages after each repetition with respect to the total number of microspheres present in the V-vial (300 mg) are provided. Median values from 3 repeated measurements are plotted, and ranges are indicated by shaded bars.

Part 2: microsphere distribution in a successively bifurcating phantom for different injection rates

The measured flow profile in the phantom and the radial catheter position during the experiments can be found in Supplemental file EA. Mean overall lumen blood flow rate was 2.7 mL/s, and mean peak systolic rate was 6.5 mL/s.

Outflow calibration

During the calibration run (3 minutes), the BMF volume collected across the 8 outlets ranged from 63.93 to 69.79 mL, resulting in a maximum deviation of 1.1% among the outlets. This consistency confirmed uniform outflow before microsphere injection.

Qualitative analysis of catheter outflow patterns

Comparing the injections performed with the conventional AD, a more selective injection stream line was observed for the pulsed injection profile (Fig. 2.1) with respect to the continuous injection profile (Fig. 2.2). This means that more of the blue dye ended up at both sides of the branched vessel for the continuous injection. The injections performed with the conventional AD and the controlled AD at similar rate (24 mL/min) showed very similar results (Fig. 2.2 and 2.3, respectively). The reduced continuous injection rates (Fig. 2.4 and 2.5), tested with the controlled AD, led to even more selective injection stream lines.

Quantitative analysis of catheter outflow patterns

The microsphere distribution over the 8 outlets for the 5 experimental settings can be found in Fig. 3. With the current catheter position, >80% of the microspheres ended up in outlets 5 till 8 for all tested injection profiles. For the conventional AD, a more selective microsphere distribution toward outlet 7 for the pulsed injection compared with the continuous injection was observed (40.1% and 23.1%, respectively), although the correlation between pulsed injection and continuous injection was still very strong (r = 0.830). Correlation between the 2 ADs at a continuous 24 mL/min injection was very strong (r = 0.988). Lastly, it can be observed that at lower microsphere injection rates (10 and 5 mL/min) more microspheres ended up in outlet 7 (68.8% and 80.0%, respectively).

Figure 3.

Microspheres per outlet (mean from 5 repeated measurements) in percentage of the total of 8 outlets after microsphere administration with a conventional and controlled administration device. SDs are indicated by the black lines.

The correlation between microsphere distribution and injected volume via the catheter (Fig. 4) was very strong for all experimental settings (conventional AD, 24 mL/min, pulsed: r = 0.984; conventional AD, 24 mL/min, continuous: r = 0.994; controlled AD, 24 mL/min, continuous: r = 0.973; controlled AD, 10 mL/min, continuous: r = 0.980; controlled AD, 5 mL/min, continuous: r = 0.951).

Figure 4.

Microspheres per outlet and the injected volume via the catheter per outlet in percentages of the total of 8 outlets after microsphere administration with a conventional and controlled administration device.

Discussion

In the present study, the influence of injection technique on the microsphere distribution was investigated. Part 1 of this study showed that there is a variation in injection rates applied by clinicians. Despite the 2 distinct injection patterns from the clinicians (Fig. 1), the 24 mL/min injection rate, closely resembling the pattern of the nuclear medicine physician, was chosen as reference during the second part of this study. This injection rate was sufficient to achieve adequate agitation of microspheres in the V-vial and, additionally, higher injection rates (48 mL/min) resulted in large microsphere boluses during the first 2 injection repetitions (Fig. 1), which could potentially be unfavorable for tumor uptake.¹⁸ The most important finding of part 2 of the study was that the injection rate and the pulse of the injection impacts the microsphere distribution. This might indicate that the contrast CBCT and ^99mTc-MAA scout dose do not provide a proper prediction of actual microsphere injections, because these injection techniques differ from the conventional AD used for the actual TARE treatment. Previous studies comparing ^99mTc-MAA scout dose injections and therapeutic injections focused on variations in particle properties.^19,20 However, differences in injection technique may also contribute, suggesting that a holmium-166 scout dose, administered with the same AD as the therapeutic dose, could provide a more accurate prediction.²⁰

The injections performed with the conventional AD (24 mL/min pulsed vs 24 mL/min continuous) showed that pulsed injections lead to more selective microsphere distributions (Figure 2, Figure 3 [purple and green bar]). It might have been that during pulsed injections proper mixing of microspheres and BMF could not occur and that this prevented microspheres from following the main BMF flow, leading to inhomogeneous distributions over the outlets. A continuous injection, however, could create a more homogeneous distribution over the lumen. The more uniform radial distribution of microspheres might be favorable for lobar or whole-liver TARE treatments, because it provides a more predictive setting; it allows the microspheres to follow the blood flow. As hypervascular tumor sites receive a greater proportion of blood flow compared with normal liver parenchyma, it can potentially improve targeting toward hypervascular tumor sites.

Figure 2.

Outflow patterns at the catheter tip in the successively bifurcating phantom for different injection patterns (pulsatile vs continuous, 24-10-5 mL/min) after microsphere administration with a conventional and controlled administration device.

The continuous injections performed with the controlled AD (24-10-5 mL/min) showed that lower continuous injection rates result in even more selective microsphere distributions (Figure 2, Figure 3 [yellow, red, and blue bar]). For the conventional AD, a minimum injection rate of 24 mL/min was required, because lower rates impeded proper agitation of the microspheres in the vial. However, the homogeneous suspension of microspheres created and maintained using the controlled AD, allowed to apply and investigate lower injection rates. Injection rate was reduced until it matched the lumen blood flow rate. The mean lumen blood flow velocity was 26.5 cm/s (2.7 mL/s in a 3.6 mm vessel), comparable to the 5 mL/min experimental setting, which resulted in a 29.5 cm/s flow velocity (5 mL/min in a 0.6 mm catheter).

In current clinical practice using the conventional AD, a certain degree of selectivity could be achieved by alternating between pulsed and continuous injection rates. Microsphere agitation via tube line B may be maintained in a pulsed manner, whereas the actual delivery of microspheres to the patient through tube line A can be alternated between pulsed and continuous modes, depending on the desired effect. Furthermore, a controlled AD as proposed in this article enables lower injection rates and could lead to selective injections. This might be of interest for superselective TARE, when unfavorable distributions are observed (ie, too much healthy liver tissue is targeted) whereas more selective microcatheter positions cannot be achieved. Lower injection rates can then be tested to increase the tumor-to-nontumor ratio. However, it is important to note that reducing the injection rate increases treatment time. For example, lowering the rate from 24 mL/min to 10 mL/min would extend a single injection repetition from 19 s to 45 s (10 mL/min injection rate) or 90 s (5 mL/min injection rate).

The conventional AD and controlled AD showed similar distributions (Figure 2, Figure 3 [green and yellow bar]) under comparable injection profiles (continuous, 24 mL/min). Although the number of microspheres delivered per repetition differed between the 2 methods, this did not appear to influence the overall distribution. Figure 1 illustrates that the conventional AD does not deliver microspheres consistently across injection repetitions, whereas the controlled AD maintains a constant suspension, resulting in a more uniform delivery per repetition. This suggests that the number of microspheres injected during each injection repetition does not affect their distribution. However, injecting substantially more microspheres might affect distributions because of, for example, increased particle-particle interactions, which lead to changes in the effective viscosity of the suspension.

The similar distribution observed between the 2 ADs pertains to macroscopic distribution, because this in vitro model does not account for tumor microvasculature. In vivo, microspheres are known to distribute heterogeneously.^21,22 Similarly, our results indicate that although a higher injection rate improved mixing, microspheres keep following specific blood flow streamlines, preventing uniform distribution across all downstream branches. Implications of these effects on a microscopic level should be further explored. Miller et al¹⁸ investigated Yttrium-90 glass microsphere deposition in tumors by comparing a bolus injection (36 mL/min, continuous) and a constant microsphere concentration injection (18 mL/min, continuous). They demonstrated greater distal deposition and more uniform tumor coverage when using the constant microsphere concentration injection. They suggest that this effect might be because of the lower injection rate, because slower injections cause less reduction in tumor blood flow compared with higher injection rates. However, both the injection rate and the number of microspheres delivered at a certain time point are different between the 2 described techniques and the influence of the latter is not described. For future studies on microsphere distribution in the tumor microvasculature, the controlled AD could be useful, because it allows control over both injection rate and the number of microspheres.

Finally, it is important to highlight that the microsphere distribution closely matched the injected volume via the catheter (Fig. 4) in all experimental settings (r > 0.951). This indicates that the microspheres follow the injected volume via the catheter, which in turn follows specific layers of the lumen blood flow.⁷ Differences between microsphere distribution and injected volume via the catheter were only seen at low injection rates, possibly because of dye diffusion into the BMF or settling of microspheres.

Our findings align with previous in vitro research,⁶ which demonstrated that higher injection rates enhance the mixing of microspheres and blood. However, despite the clinical recommendation for pulsed injection, prior studies have only examined continuous injection rates.^6,7,23, 24, 25, 26 Basciano et al⁷ found that proper mixing between microspheres and blood occurs only during peak systole, but aligning microsphere injection with the cardiac cycle is not achievable in the current clinical workflow. Furthermore, the influence of the cardiac cycle in our study will be minimal, because the injections lasted nearly 3 minutes, averaging the effect of injecting in various cardiac phases.

A limitation of this study is that no tumor behavior was simulated, such as increased flow typically observed at tumor-feeding outlets. The current focus was on the injection flow dynamics at the catheter tip before it reaches the tumor, isolating these effects from other variables such as varying outlet flows. A future step could be to focus on the behavior in a tumor model as microsphere distribution is highly influenced by fluid flow.²⁷ Such a model would also allow to study the embolization effect and capture changes in flow patterns as the tumor becomes saturated. Second, only 1 catheter tip location was investigated. The current catheter position (right lower quadrant) favored the outlets on the right side of the phantom. Variations in catheter position, both axial and radial, will lead to different distributions. Because the characteristics of the fluid stream exiting the catheter are primarily governed by the injection rate, which remains consistent across different catheter tip locations, the decision to investigate a single tip position is justified for the scope of this study. Nonetheless, future research would benefit from exploring multiple catheter tip locations to assess potential variations and broaden the applicability of the findings. Third, the silicone phantom had rigid vessel walls. In computer simulations performed by Childress and Kleinstreuer,²⁸ a flexible geometry was compared with rigid geometries. Differences were found especially during the systolic phase of the cardiac cycle. However, the time-averaged rigid geometry matched the microsphere distribution for the flexible geometry quite well and might therefore justify the use of rigid vessel walls, because an injection in the current study lasted almost 3 minutes. Moreover, the minor deviations observed during the calibration run (1.1%) were not corrected, because they did not affect the primary outcomes of this study. Although the findings of this study are likely transferable to other radioembolization products such as SIR-Spheres, owing to their similar size and density profiles and AD, extrapolation to glass microspheres (TheraSphere) is more limited. The substantially higher density, smaller diameter, lower particle count and difference in AD associated with TheraSphere may lead to distinct distribution characteristics that were not captured in this investigation. Future research could explore whether different microsphere types and ADs influence the results.

We recommend raising awareness among physicians that both the injection rate and pulsing pattern play a critical role in influencing microsphere distribution during TARE. In part 1 of this study, injections were performed by 2 clinicians and the high variability in these manual injections (Fig. 1) might indicate that it is difficult to perform consistent injections. However, the 24 mL/min syringe pump injection also showed high variability. This may be attributed to the lack of repeatability in the manual needle positioning in the V-vial of the conventional AD. For the 48 mL/min syringe pump injection, the high pressure generated in the vial ensures that most microspheres exit within the first few repetitions, making the needle position less critical. The percentage of injected microspheres for the 48 mL/min syringe pump injection and the manual injection performed by the interventional radiologist aligned closely with findings from Drescher et al²⁹ who investigated residual activity in the V-vial and delivery set for different injection protocols. It is however important to note that our study primarily focused on the injection profile and the agitation of microspheres in the V-vial, rather than complete emptying of the vial. For that, additional repetitions would have been required, especially at the lower injection rates.

For manufacturers, our recommendations would be to establish more standardized guidelines on administration practices, given the impact of injection rate on microsphere distribution. However, even if a standard injection rate would be used to administer microspheres, the blood flow in the hepatic artery is still highly variable and strongly dependent on hemodynamic alterations. In literature, ranges from 143 to 359 mL/min for healthy controls^30,31 and 220 to 813 mL/min for diseased livers^32,33 are described for the common hepatic artery. As a result, the ratio between injection rate and blood flow rate can be highly variable and it might be valuable to measure the blood flow rate in the common hepatic artery for each patient before TARE.

Conclusions

In conclusion, the injection technique had a major impact on the distribution of microspheres. High continuous injection rates (24 mL/min) resulted in better mixing between microspheres and blood-mimicking fluid and can therefore create a more homogeneous distribution over the outlets compared with pulsed injections and low continuous injections. This potentially offers a more predictive setting for microsphere deposition. Furthermore, because contrast CBCTs and scout dose injections are performed at different injection rates or using different injection profiles, their predictive values must be considered with caution. Acknowledging the impact of injection rate may contribute in reducing intra- and interprocedural differences in TARE and ultimately in improving TARE outcomes.

Disclosures

J. Frank W. Nijsen is co-founder of Quirem Medical that has been acquired by Terumo Europe NV in July 2020. J. Frank W. Nijsen has a scientific advisory role and is entitled to certain milestone payments from Terumo that are related to Quirem’s financial, operational and regulatory performance in the future. Furthermore, J. Frank W. Nijsen is inventor on the patents related to radioactive microspheres that are assigned to University Medical Center Utrecht Holding BV, Quirem Medical or BASF Corp. The activities of J. Frank W. Nijsen within Quirem Medical are approved and supported by the Board of Directors of the Radboudumc. Gerhard van Wolfswinkel is an employee of Quirem Medical (R&D) and provided support in the design of the controlled administration device (based on the input delivered by the Radboudumc) and commented on the manuscript. The other authors declare that they have no conflict of interest.