Impact of Injection-Based Delivery Parameters on Local Distribution Volume of Ethyl-Cellulose Ethanol Gel in Tissue and Tissue Mimicking Phantoms

Abstract

Objective:

Local drug delivery aims to minimize systemic toxicity by preventing off-target effects; however, injection parameters influencing depot formation of injectable gels have yet to be thoroughly studied. We explored the effects of needle characteristics, injection depth, rate, volume, and polymer concentration on gel ethanol distribution in tissue and phantoms.

Methods:

The polymer ethyl cellulose (EC) was added to ethanol to form an injectable gel to ablate cervical precancer and cancer. Tissue mimicking phantoms composed of 1% agarose dissolved in deionized water were used to establish overall trends between various injection parameters and the resulting gel distribution. Additional experiments were performed in excised swine cervices with a CT-imageable injectate formulation, which enabled visualization of distribution without tissue sectioning.

Results:

Needle type and injection rate had minimal impact on gel distribution, while needle depths ≥13 mm yielded significantly larger distributions. Needle gauge and EC concentration impacted injection pressure with maximum gel distribution achieved when pressure was 70–250 kPa. Injection volumes ≤3 mL of 6% EC-ethanol minimized fluid leakage away from the injection site. Results guided the development of a speculum-compatible hand-held injector to deliver gel ethanol into the cervix.

Conclusion:

Needle depth, gauge, and polymer concentration are critical to consider when delivering injectable gels.

Significance:

This study addressed key questions related to the impact of injection-based parameters on gel distribution at a scale relevant to human applications including: 1) how best to deliver EC-ethanol into the cervix and 2) general insights about injection protocols relevant to delivery of injectable gels in tissue.

Introduction

Local injection of drugs is an increasing area of interest for biomedical researchers. Localized delivery reduces systemic toxicity by preventing off-target effects of the agent on non-target cells[1, 2]. Local delivery can achieve enhanced efficacy of drugs at smaller doses, which alleviates many side effects observed with systemic delivery[1, 2]. There are more than 95 ongoing clinical trials investigating local intratumoral injection of different therapeutics, including chemotherapy and immunotherapy (clinicaltrials.gov)[3–5]. To achieve safe and efficient intratumoral drug delivery, it is essential to understand the role of injection-based delivery parameters on drug distribution. There are more than 25 clinical trials testing various injection techniques, including investigating the impacts of needle size and injection rate[6]. However, uniform drug distribution and retention of therapeutics within tumors remains a challenge [7].

One approach to improve the uniformity of drug distribution and retention in tumors is to load drugs within injectable hydrogel delivery systems, which undergo a phase change upon injection into tissue [8]. For example, in situ-gelling hydrogels are injected in liquid form and undergo a solution to gel transition in tissue [8]. Many groups are interested in developing in situ-gelling hydrogels[8]; examples include: an injectable hydrogel for spatiotemporal control of anti-vascular endothelial growth factor delivery[9], a temperature-responsive injectable hydrogel for intratumoral administration of doxorubicin [10], and injectable biomaterial scaffolds for cancer immunotherapy [11]. However, the influence of injection parameters on the resulting distribution of injectable gels have yet to be thoroughly studied and characterized[7]. While positive pressure driven infusion into tissue is not new, gel formation has a significant impact upon distribution volume and molecular mass transport in the tissue [8, 12], adding complexity not thoroughly studied in prior work[13–15]. Upon injection, the tissue deforms, forming a cavity. The phase change propagates from the external surface of this cavity inwards [16]. This cavity grows as more fluid is injected, with a portion of the injectate fluid permeating into the porous tissue [16]. If the pressure at the boundary of this cavity exceeds the critical pressure of the tissue (i.e. the pressure at which fracture occurs), it can lead to leakage away from the injection site through cracks that are formed in the tissue (i.e. crack formation) [16]. Similarly, leakage may occur in regions of the tissue that were damaged during needle insertion, such as those abutting the walls of the needle (i.e. backflow up to the tissue surface [16]. All these phenomena occur simultaneously, resulting in a complex process. Here we sought to identify the dominant delivery parameters that govern the resulting distribution volume of phase-transitioning therapeutic injections – specifically, a gelated form of ethanol ablation.

Previously we studied the local injection of ethanol as a potential ablative agent to treat precancerous and cancerous lesions in the cervix. Ethanol ablation has previously been used to treat hepatocellular carcinomas[17], cardiomyopathies[18], parathyroid tumors[19], and pancreatic tumors[20]; however, its efficacy is limited by early leakage of ethanol from the tumor injection site, potentially causing off-target damage and risk of adverse effects[21]. To address this issue, we added the polymer ethyl cellulose (EC) to the injectate, which undergoes a phase change, forming a crosslinked gel-like material upon exposure to aqueous environments (i.e. tissue). This acts to reduce leakage and enable sustained local ablation near the site of injection. Therapeutic efficacy of EC-ethanol was initially evaluated in chemically induced oral squamous cell carcinomas in the hamster cheek pouch[22]. Results showed 3% EC-ethanol injections significantly stunted tumor growth compared to pure ethanol injections over a 1-week period. We also found that slowing the injection rate from a manual injection (~100 mL/hr) to 10 mL/hr further reduced leakage from the tumor. To study the impact of larger injection volumes, we next injected excised swine liver tissue with an EC-ethanol-fluorescein solution to enable visualization of distribution volumes after tissue sectioning[23]. We found that increasing EC concentrations from 3% to 6% further enhanced distribution volume, and it reduced fluid leakage away from the injection site. Considering these key results, we next studied how to best deliver EC-ethanol into the specific anatomy of the cervix to enable treatment of cervical dysplasia and cancer[24]. Specifically, we injected 6% EC-ethanol mixed with fluorescein into excised swine cervices (because its anatomy has many similarities to that of the human cervix). For shallow injections (<8 mm into tissue), we found that keeping injection volumes low (≤500 μL) and pointing the needle bevel radially outward from the endocervical canal minimized fluid leakage away from the injection site.

While our initial studies show promise for using EC-ethanol to treat precancerous and cancerous lesions, key questions remain regarding the impacts of various delivery parameters on distribution volume to deliver EC-ethanol into the cervix to adequately cover precancerous lesions and deeper early invasive disease. High-grade cervical pre-cancer, which is still confined to the epithelium, varies in extent from a mean depth of 2 mm up to a maximum of 5 mm and a mean breadth of 7.3 mm up to a maximum of 20.5 mm[25]. Early invasive disease (stage 1A), which has invaded the basement membrane, can extend between 3 to 5 mm deeper into the stroma (i.e up to 10 mm below the surface of the cervix)[26]. Here, we performed a series of experiments in which the impact of needle type, gauge, injection depth, injection rate, injection volume, and EC concentration were evaluated in tissue mimicking phantoms and cervical tissue using novel imaging methodologies to visualize distribution volumes. We sought to maximize injectate distributions by minimizing two primary forms of fluid leakage away from the injection site: 1) backflow, which is retrograde flow up the needle pathway to the tissue surface, and 2) crack formation, which is the formation of cracks in tissue into which injectate flows. Our motivation behind using the agarose phantoms was to utilize a biophysically relevant in vitro model as much as possible before working with tissue to reduce the number of animals that need to be sacrificed. Agarose phantoms provide high throughput characterization of the injection parameter space and low-cost visualization of injectate volume without the need for a contrast agent. Owing to its tunable mechanical properties, biocompatibility, and controllable degradation, agarose has been widely used as a tissue mimicking phantom to optimize infusion protocols[27, 28]. In a previous study, we found that 1% agarose most closely matched the mechanical properties of liver tissue[16]. Likewise, 1% agarose gels have been used to mimic brain tissue and inform clinical brain infusion protocols[29]. For subsequent tissue experiments, we developed a new methodology to reliably visualize ethanol distribution within intact swine cervices, as their anatomy is similar to that of humans[30]. Specifically, we developed a CT-imageable version of our EC-ethanol injectate via adding iohexol and imaged samples with a micro-computed tomography (micro-CT) system. This methodology enabled us to reconstruct and quantify the 3D injectate distribution while leaving the tissues intact.

Methods

Overview of experiments.

Delivery parameters were systematically varied across experiments in tissue mimicking phantoms and tissue, as summarized in Table 1. In Experiment 1, we evaluated if food dye was suitable for visualization of distributions in phantoms, as had been described previously by Morhard et. al[22]. We mixed varying amounts of dye with EC-ethanol and found that the amount of dye significantly altered the observed distribution volumes, indicating that added dye contributed to experimental error (see Supplementary Fig. 1). Thus, instead of relying on the dye for visualization, the EC gel alone dictated the distribution landscape, as the white hue is readily visible against the translucent tissue phantom background. In Experiment 2, the effects of different needle types (beveled tip, blunt tip, and Whitacre spinal needles) were evaluated in phantoms while other variables were held constant. Specifically, 27G needles were initially selected as 27G Potocky needles have been used previously for cervical injections and smaller needle diameters have been shown to reduce injection pain[31]. Needle length was selected as 1.3 cm (1/2 inch) because it is a standard length for 27G needles, and a needle depth of 13 mm was selected because it is the full length of the needle. The selections of 6% EC-ethanol, an injection volume of 300 μL (i.e. between 100–500 μL), and a rate of 30 mL/hr (i.e. between 10–100 mL/hr) were guided by our previous results described above[24]. Shallower needle depths ranging from 4–13 mm were explored because we are interested in treating precancerous and cancerous lesions extending from 2–10 mm below the cervix surface. Our previous results indicated that dimpling (i.e. tissue deformation around the needle tip) led to 2–3 mm lower actual needle depths compared to programmed needle depths[24], which is why to achieve 2–10 mm actual needle depths, we selected 4–13 mm programmed needle depths. In Experiment 4, we assessed the effect of needle gauge on distribution volume and assessed the relationships between needle gauge, injectate viscosity (i.e., EC concentrations of 6, 9, and 12%), and injection pressure in phantoms. We explored higher EC concentrations up to 12%, which have been shown to increase depot volume[32]. However, due to its viscosity, 12% EC-ethanol is difficult to inject through a 27G needle; thus, we increased our range of gauges from 27 to 22G as lower gauge needles (e.g. 22G and 23G) are more commonly used in the clinic for a variety of applications. In order to compare 27, 25, 23, and 22G needles to each other, we sought to keep the needle length constant and therefore selected a length of 2.5 cm (1 inch), which was a standard length for all 4 gauges. In Experiment 5, injection volumes of EC-ethanol were administered separately to assess the distribution volume achieved within ex vivo swine cervical tissue. Here we varied the volume from 300 to 3000 μL to determine the upper limit of what volumes could be injected into the cervix before substantial fluid leakage occurred. Results from experiments 2–5 guided the development of a speculum-compatible hand-held injector to deliver ethanol into the cervix. The injector was designed around a 22G 8.9 cm (3.5 inch) long needle connected to a needle extender, which is typically used to administer lidocaine through the speculum into the cervix prior to a LEEP procedure. The injector could be operated manually or automatically via a programmable syringe pump. Experiment 6 outlines a comparison of manual and automated injection methods in which the impacts of injection rate and volume on distribution was observed.

Table 1.

Overview of injection parameters evaluated in each experiment and summarized main result.

Exp	Sample	Needle Gauge; Type; and Length	Needle Depth (mm)	Injection Rate (mL/hr)	EC content (%w/v)	Injection Volume (μL)	Contrast Agent	Main Result
1	Phantom, n=5	27G; bevel; 1.3 cm	13	30	6%	300	Blue dye (drops of)	Adding dye to EC-ethanol increases error (see Supplementary Fig. 1)
2	Phantom, n=5	27G; blunt, bevel, Whitacre; 1.3 cm	13	30	6%	300	None	No significant differences in distribution between bevel needle and other needle types
3	Phantom, n=5	27G; bevel; 1.3 cm	4, 8, 10, 13	10, 30	6%	300	None	Deeper needle depths (≥13mm) significantly increased distribution
4	Phantom, n=5	22G, 23G, 25G, 27G; bevel; 2.5 cm	13	10	6%, 9%, 12%	300	None	Lower gauges (22G) generate larger distributions for higher viscosity solutions (12%) while higher gauges (27G) generate larger distributions for lower viscosity solutions (6%)
5	Swine cervical tissue (Ex vivo), n=4–5	27G, bevel; 1.3 cm	13	10	6%	300, 500, 750, 1000, 1500, 2000, 3000	Iohexol (40 mg/mL)	Significant leakage was observed when injected volumes ≥ 3000 μL
6	Swine cervical tissue (Ex vivo), n=6	22G; bevel; 8.9 cm	13	10, manual	6%	500,1000	Iohexol (40 mg/mL)	Automatic and manual injections showed no significant differences within volume conditions

Phantom preparation:

Tissue mimicking phantoms were used to initially assess trends in the distribution of EC-ethanol achieved with different injection parameters. Phantoms consisted of a 1% agarose powder (UltraPure Agarose, Invitrogen, Carlsbad, CA) dissolved in deionized water (10 grams of agarose per 1000 milliliters of deionized water). The agarose and water were mixed in a flask on a heated stir plate set at 200°C until the solution becomes clear (Fig. 1A (I)). The mixture was then poured into 30 mL polystyrene drams, which were covered with snap caps and placed into a refrigerator at 4°C for 24 hours prior to experiments.

Fig. 1:

Overview of experimental methodology, including (I) sample preparation, (II) injection, and (III) imaging + analysis for (A) tissue mimicking phantoms and (B) ex vivo swine cervical tissue.

Tissue preparation:

Fresh swine reproductive tracts supplied by Animal Biotech Industries (Doylestown PA) were dissected to remove the uterus and uterine horns. Then the location of the cervix within the lower reproductive tract was determined by locating the interdigitating pads along the tract; tissue below the pads was removed. The cervix was then cross-sectioned into 1-inch-long slices (Fig. 1B (I)) and stored in 10% PBS prior to experiments, which were conducted within 6 hours of receiving the tracts.

Injection procedure:

Samples were injected with EC-ethanol using the injection setup described in Fig. 1A and 1B (II). The injectate was loaded into a 3 mL syringe (Becton, Dickinson and Company, Franklin Lakes, NJ), and the syringe was placed in a NE-300 Just Infusion^™ Syringe Pump (New Era Pump Systems, Farmingdale, NY) programmed to 8.66 mm diameter to be compatible with the 3 mL syringes. The Syringe Pump was programmed for various injection volumes and injection rates depending on the experiment. The syringe was loaded with EC-ethanol, placed in the syringe pump, and a needle was attached to the syringe. Samples were placed on their side, level with the syringe pump for the injection. The needle bevel was orientated upwards for all injections, then the needle was carefully inserted into the sample to a specified depth and the injection was performed. The needle was left in place for 5 minutes post-injection to ensure that the full volume of EC-ethanol was delivered. For tissue experiments, the needle was inserted horizontally into tissue in the middle of the cervical wall, parallel to the endocervical canal for automatic injections or vertically into tissue for manual injections as shown in Fig. 1B (II). In Experiment 4, the injectate-driving pressure was monitored with a Honeywell 3.3V pressure sensor connected to the system via a hose barb tube splitter (see Supplementary Fig. 2A). Because tubing was used in this setup, the elastic expansion of the tubing was accounted for as described previously[23]. Specifically, the tubing compliance was determined by plugging the end while injecting fluid and measuring the pressure (see Supplementary Fig. 2B). The volume of unpressurized tubing was approximately 2000 μL, and the tubing expanded with a compliance of 1.4 to 2 μL/kPa (depending on the EC%), which was used to account for the effect of tubing expansion on distribution area in experiment 4 (see Supplementary Fig. 2C).

EC-ethanol formulation and contrast agent for visualization:

Mixtures of ethyl cellulose (Sigma Aldrich, St. Louis, MO) and ethanol (200 proof, Koptec, King of Prussia, PA) were created using a hotplate stirrer (VWR Hotplate/Stirrer) at room temperature. The ethyl cellulose and ethanol mixture (EC-ethanol) was continuously mixed until the EC was fully dissolved. EC concentrations of ≥6% (EC to ethanol, %weight: volume) were used in all experiments, per previous studies that showed demonstrating significant increases in distribution volume for ≥6% EC[23].

For tissue experiments, iohexol (Omnipaque, GE Healthcare, Chicago IL) was added to EC-ethanol to provide CT image contrast between the injectate and surrounding cervical tissue. This enabled evaluation of the 3-dimensional (3D) distribution volume achieved in intact tissue without the need for tissue sectioning. Iohexol was selected as a CT contrast agent for its high solubility in ethanol[33]. To determine the concentration of iohexol suitable for imaging, various amounts of stock iohexol were added to 6% EC-ethanol. The resulting concentrations of iohexol were 8, 30, 40, and 50 mg/mL. The EC-ethanol solution was placed on a stir plate and iohexol was added and stirred until fully mixed. The EC-ethanol-iohexol solutions were injected into phantoms, imaged with the micro-CT, and the resulting image contrast between EC-ethanol-iohexol and the surrounding phantom was then quantified (procedures are described below). Results indicated a concentration of 40 mg/mL of iohexol maximized the signal to background ratio (see Supplementary Fig. 3); thus, 40 mg/mL was used for all tissue experiments.

Imaging and image analysis procedures:

In phantoms, the distribution volume was visualized through direct imaging of the gel, which turns white upon contact with aqueous environments and can be readily seen against the translucent phantom background (Fig. 1A (III)). To prepare phantoms for imaging, first the surface of each phantom was wiped to remove any excess fluid that collected at the surface in order to enable clean cuts to be performed. Then each phantom was removed from the dram, cut along the largest plane (parallel to the needle), and photographed with a digital camera (Apple iPhone 11), as illustrated in Fig. 1A (III). Images were analyzed using ImageJ Fiji (NIH, Bethesda MD). The thickness of the gel front was quantified by measuring the widest horizontal length between the outermost edge of the distribution and the outermost edge of the clear region formed inside the depot.

For tissue experiments, excised tissue was injected with the EC-ethanol-iohexol injectate and subsequently imaged using a micro-CT (Bruker SkyScan 1276, Billerica, MA). Samples were placed upright during imaging, with insertion site oriented upwards. Pre- and post-ablation micro-CT images were taken of each sample. The micro-CT images were taken using step-and-shoot projections at a pixel size of 81.072 μm and a field of view of 9.31 × 158.75 cm. Images were acquired in full rotation (360°) in a 504 × 336 matrix using 65 kV, 200 μA, 50 ms exposure time, 0.5 mm aluminum filter, with a 1.6° rotation step. Projection images taken by the micro-CT were reconstructed using NRecon (Bruker). Post-alignment correction was performed on all samples, and an appropriate ring correction was applied. Images were then exported as TIF files and loaded into 3Dslicer (Kitware, Clifton Park, NY) to quantify the volume of EC-ethanol within the tissue. Maximum entropy auto-thresholding was applied to data to segment EC-ethanol-iohexol[34]. Depending on variations in tissue structure, the threshold values for specific tissues were manually increased to capture the entirety of the injectate and remove any background noise if the automatic thresholding under-selected the injectate. Once the injectate was segmented, the total distribution volume was calculated in 3DSlicer Fig. 1B (III). To enable comparisons across groups, ratios of the total volume to the initial volume (i.e., the programmed injection volume) were determined.

Handheld injector:

Results from experiments 2–5 guided the development of a speculum-compatible hand-held injector to deliver ethanol into the cervix. The injector features a 22G, 8.9-cm-long needle connected to a 12.7-cm-long needle extender that was over-molded with a custom Polyurethane handle (Fig. 2). The longer needle was selected to facilitate access to the cervix through a speculum and is typically used to reach and inject the cervix with lidocaine prior to LEEP procedures. To control needle depth, an adjustable depth stop was incorporated, consisting of a sliding mechanism, that may be secured in place with a set screw. The distal end of the injector can be connected to either a syringe pump (for automated control of injection rate and volume) or a standard syringe (for manual control of injection rate and volume) via Luer-lock connections.

Fig. 2:

Side and top view of handheld speculum-compatible injector designed to deliver EC-ethanol into the human cervix.

To initially assess the repeatability of the injection depot achieved with the single needle injector, we injected two volumes (500 and 1000 μL) of EC-ethanol-iohexol into ex vivo swine cervices and imaged with the micro-CT system described previously. Specifically, the single needle injector was used to insert the needle 13 mm into the tissue and deliver 6% EC-ethanol-iohexol solution into the cervix through: 1) automated control: the injector was connected to a syringe that was placed into syringe pump to control injection rate (10 mL/hr) and volume (500 or 1000 μL), or 2) manual control: the injector was connected to a syringe through which injection volume (500 or 1000 μL) could be manually delivered.

Statistical analysis:

N=5 trials were performed for each parameter value for all the experiments unless otherwise indicated. Wilcoxon rank sum, Kruskal Wallis, and Dunn’s tests (non-parametric, two-tailed, alpha = 0.95) were performed in MATLAB (MathWorks, Natick, MA) to assess if there were significant differences between groups in Experiments 2–6. A significance level of p = 0.05 was considered to reject the null hypothesis for all analyses.

Results

Phantom experiments:

Tissue mimicking phantoms were injected with EC-ethanol under various conditions (Experiments 1–4, Table 1). First, common needle types were investigated, (Experiment 2), to assess their impact on distribution (Fig. 3). Common needle types included beveled needles (sharpened tip), blunt needles (unsharpened tip), and Whitacre spinal needles (tip shaped like a pencil with a hole on the side of the needle near the tip). All distributions in phantoms were coin shaped (see Supplementary Fig. 4). Because the thickness of the coin was difficult to quantify, we quantified distribution area from the side with the largest area only. No statistically significant differences in distribution area were observed between needle types. However, the blunt-tip injection had uncontrolled depot formation when viewed from the side, whereas both the Whitacre and bevel tip needle injections consistently led to depots with the largest area parallel to the direction the needle bevel was oriented. Thus, beveled needles were selected for subsequent experiments.

Fig. 3:

(A) Representative side-view images (surface of phantom located at the top of the image) in phantoms injected with different needle types (27G 1.3-cm-long needles, 13 mm needle insertion depth, 30 mL/hr injection rate, 6% EC-ethanol, 300 μL injection volume). The white trace highlights the borders of the distribution. A side view of the blunt-tip injection shows a slanted distribution, indicated by the white arrow. Scale bars = 5 mm. (B) The average distribution area was calculated for each condition (n=5 for each group). Error bars represent standard error. No statistically significance differences were observed between needle types.

Next, injection rate and needle depth were investigated, (Experiment 3), to assess their impact on distribution (Fig. 4). Distribution area significantly increased with insertion depth (Fig. 4), which may be attributed to a lower volume of EC-ethanol leaking out to the sample surface. In particular, a 13 mm insertion depth gave consistent circular distribution as observed in Fig. 4A, and therefore was used for subsequent experiments. At 10mm and 13 mm needle depths, 30 mL/hr had significantly higher distribution area than 10 mL/hr.

Fig. 4:

(A) Representative side-view images of distribution area in phantoms injected with different needle depths – 4, 8, 10, and 13 mm – and injection rates – 10 and 30 mL/hr (27G 1.3 cm-long beveled needles, 6% EC-ethanol, 300 μL injection volume). The white trace highlights the borders of the distribution. Scale bars = 5 mm. (B) The average distribution area was calculated for each condition (n=5 for each group). Error bars represent standard error. Statistical comparisons are indicated by asterisks on the graph (*p <0.05, **p < 0.005, ***p<0.0005).

Next, we assessed the relationship between EC concentration and needle gauge and their collective impact on both gel distribution area and injection pressure (Fig. 5). Fig. 5A shows gel distribution areas for each combination of EC concentrations and needle gauges used. The visual results seen in representative images in Fig. 5A reflect the quantitative results seen in Fig. 5B. We also recorded pressures at the top of the needle observed during the injections in experiment 4 (pressure curves are shown in Supplementary Fig. 5). The maximum pressure for each trial was determined and the average was calculated to give the average maximum pressure for each condition (Fig. 5C). Note that the maximum pressure of the sensor was ~465 kPa, and this value was observed in both 25G and 27G for 12% EC-ethanol. Three key trends emerged from this data set. First, EC concentration significantly increased distribution area when injected through 22 and 23G needles as seen in the first two columns of Fig. 5A. Second, large injection pressures above 250 kPa at the needle tip led to smaller distribution areas. Injection pressure increased with higher EC concentration and higher needle gauge. For example, when 12% EC-ethanol was injected through a 27G needle, distribution area significantly decreased (as seen in the last row of Figure 5A). Lastly, modest injection pressures between 70–250 kPa appeared to be beneficial as they were associated with significantly larger distribution areas. For example, the distribution area achieved when 6% EC-ethanol was injected through a 27G needle or when 9% was injected through a 25G needle was significantly larger than that achieved with lower needles gauges (as seen in the first and second rows of Fig. 5A and in Supplementary Figure 6).

Fig. 5:

(A) Representative images of phantoms injected with different needle gauges – 22, 23, 25, and 27G – and EC concentrations – 6, 9, and 12% EC-ethanol (2.5-cm-long beveled needles, 13 mm needle insertion depth, 10 mL/hr injection rate, 300 μL injection volume). The white trace from ImageJ highlights the borders of the distribution. Scale bars = 5 mm. (B) The average distribution area was calculated for each condition (n = 5). Statistical comparisons for the effect of needle gauge on distribution area are indicated by asterisks on the graph. (C) Statistical comparisons for the effect of EC concentration on distribution area are indicated in the table. (D) The average maximum pressure was calculated for each condition (n=5 for each group). Plus signs (+) indicate the needle gauge that achieved the largest distribution area for each EC concentration. The dashed black lines indicate the ideal pressure range of 70 to 250 kPa. (E) and (F) repeat the statistical comparison for the effect of needle gauge and EC concentration, respectively, on injection pressure. Error bars are standard error (ns = not significant, *p <0.05, **p < 0.005, ***p<0.0005).

We also assessed the impact of EC concentration on the thickness of the gel front (i.e. the rim of the depot that is directly in contact with the phantom/tissue). More gel forms as the injection continues with additional layers being added from the inside of the depot, which push out the outer layers and increase the thickness of the gel front over the course of the injection. Fig. 6A shows how gel front thickness varies with EC concentration (when injected through 22 and 23G needles). The visual results seen in representative images in Fig. 6A reflect the quantitative results seen in Fig. 6B. As seen, gel front thickness significantly increased with EC concentration.

Fig. 6:

(A) Subset of representative images from Fig. 5 showing how gel front thickness was calculated. The white and gray traces from ImageJ highlight the outer and inner borders of the gel front, respectively. Scale bars = 5 mm. (B) The average gel front thickness achieved with 22 and 23G needles. Statistical comparisons for the effect of needle gauge on gel front thickness are indicated by asterisks on the graph. Error bars are standard error (ns = not significant, *p <0.05, **p < 0.005, ***p<0.0005). No statistically significant differences in gel front thickness were observed between 22 vs 23G.

Tissue experiments:

A protocol was developed for visualizing the 3D EC-ethanol distribution volume in intact tissue with CT. Iohexol is more radio dense than cervical tissue and appears white in the reconstructed images (Fig. 7A). Our results indicated that for injected volumes of ≤1000 μL virtually the entire injected volume was retained in the tissue while volumes of 3000 μL led to significant leakage into cracks outside of the injection site (Fig. 7B, C).

Fig. 7:

CT imaging to assess impact of injection volume in tissue. (A) Representative CT images of the gel ethanol distribution achieved in ex vivo swine cervices with different initial injection volumes – 300, 500, 750, 1000, 1500, 2000, and 3000 μL (27G 1.3-cm-long beveled needles, 13 mm needle insertion depth, 10 mL/hr injection rate, 6% EC-ethanol). Images are shown as a side view with the injection site (circled in red) oriented towards the top of the image. Cracks are circled in green. Scale bars = 5 mm. (B) Average volume of deposited EC-ethanol and (C) average ratio of deposited EC-ethanol to initial injection volumes into ex vivo swine cervices (n=4–5 for each group). Error bars represent standard error. Statistical significance between conditions is indicated by a bar and asterisk (*p<0.05, **p < 0.005, ***p<0.0005). For (B) there were statistical differences between all groups except 300/500, 750/1000, 750/1500, 1000/1500, 1500/2000, 1500/3000, and 2000/3000.

Results from Experiments 2–5 guided the development of a speculum-compatible hand-held injector to deliver ethanol into the cervix, requisite for clinical translation. To assess the repeatability of the injection depot achieved with the hand-held injector, we injected two volumes (500 and 1000 μL) of EC-ethanol-iohexol into ex vivo swine cervices as described in methods section : 1) automated control: the injector was connected to a syringe that was placed in a syringe pump to control injection rate (10 mL/hr) and volume (500 or 1000 μL), or 2) manual control: the injector was connected to a syringe that was used to manually deliver volumes (500 or 1000 μL). For both volumes, we found that manual and automated control provided comparable distribution volumes and variability (Fig. 8), suggesting that a syringe pump may not be required, which would greatly simply the clinical protocol.

Fig. 8:

CT imaging of visualize distribution volume achieved with handheld injector. (A) Representative CT images of EC-ethanol distribution achieved in ex vivo swine cervices with different initial injection volumes – 500 and 1000 μL – and with automatic (syringe pump set at 10 mL/hr) vs manual (standard syringe) injection (22G 8.9-cm-long beveled needles, 13 mm needle insertion depth, 6% EC-ethanol). Images are shown as a side view with the injection site (circled in red) oriented towards the top of the image. Scale bars = 5 mm. (B) Average volume of deposited EC-ethanol and (C) average ratio of deposited EC-ethanol to initial injection volumes into ex vivo swine cervices (n=6 for each group). Error bars represent standard error. Statistical significance between conditions is indicated by a bar and asterisk (*p<0.05, **p < 0.005, ***p<0.0005).

Discussion

While many groups are interested in developing in situ gelling hydrogels, the influence of injection parameters on the resulting distribution of injectable gels have yet to be thoroughly studied and characterized. Here we sought to identify the dominant delivery parameters that govern the resulting distribution volume of phase-transitioning therapeutic injections. Specifically, we selected ethanol as our active pharmacological ingredient (API), the additive EC as a component of the injectate that modifies its rheology, and delivery into the cervix as a candidate clinical focus to treat cervical dysplasia and cancer. While our experiments involved a particular pathological and pharmacological context, the methodologies presented and trends that were established here are applicable to a variety of other drugs embedded in hydrogel delivery systems and other therapeutic targets. These experiments, the results of which are summarized in Table 1, focused upon which injection parameters should be tuned, and which are less important, in the rational design of injection protocols. Phantoms were used as a high throughput platform for characterizing injection parameters because they offer high optical contrast between the EC-ethanol gel and the surrounding agarose without the need for additive dyes, are easy to create, store, and dispose of, are easy to section and image, are low-cost, and emulate the poroelasticity of native tissue[28, 35]. Poroelasticity influences aspects of infusion such as pressure, backflow, and ratio of distribution volume to initial injection volume[27]. While studies in phantoms enabled us to identify multiple parameter combinations that maximized distributions around the injection site, they do not capture spatial heterogeneity in tissue, which contributes to leakage away from the injection site. Previously, we had found in phantoms that injection pressure scaled with injection volume, and that there was a critical pressure beyond which leakage occurred[23]. Here, we studied the impacts of injection volume on the ratio of distribution volume in tissue to the initial injection volume (i.e. amount retained in tissue) and improved upon our earlier method of injection images[23]. Specifically, we used iohexol to provide contrast during micro-CT imaging, which did not require tissue sectioning and enabled reconstruction into 3D volumetric objects for visualization and quantitative analysis[36].

Experiments in tissue-mimicking phantoms indicated that varying needle type and injection rate did not have significant impacts on overall distributions. However, needle type impacted the orientation of the distribution, particularly the orientation of the largest cross-sectional area. While injections with beveled and Whitacre needles yielded the largest cross-sectional areas parallel to the bevel or side-hole, respectively, injections with blunt needles yielded unpredictable orientations of the largest cross-section area that were slanted to differing degrees relative to the needle (Fig. 3). Thus, for our application in the cervix, beveled needles were a preferred choice due to their more repeatable and predictable injection distributions vs, those of blunt needles, and their much lower cost, compared to Whitacre needles. Ultimately, needle selection depends on the application of interest and the orientation of the injection cloud that is needed relative to the plane of needle insertion. We also observed that increasing injection rate from 10 to 30 mL/hr did not significantly impact distribution area for 300 μL injections of 6% EC-ethanol. This was consistent with our previous studies, which indicated that increasing EC concentration from 3% to 6% minimized leakage regardless of the increase in injection rate from 1 to 10 mL/hr[23]. We investigated higher injection rates up to 30 mL/hr since faster injection rates are clinically preferable. We found no significant differences between 10 and 30 mL/hr for 6% EC-ethanol (Fig. 4). This suggests that if injectate viscosity is sufficiently high (on the order of a 1000 cP for 6% EC-ethanol compared to 100 cP for 3% EC-ethanol as previously determined in Morhard et al [22]) and gelation occurs in target tissue, then injections can be delivered more quickly with little to no impact on API distribution.

While the experiments in phantoms revealed that needle type and injection rate are less important to control, needle depth and gauge significantly impacted gel formation and drug distribution. In particular, increasing the needle depth from 10 mm to its full length of 13 mm led to a significant increase in distribution area (Fig. 4). Previous studies have shown that maximizing compressive stress between the sample and needle interface minimizes backflow to the surface[37, 38]. Increasing needle depth increases the amount of surface area between the sample and needle thereby increasing compressive stress, decreasing backflow, and increasing distribution area. Thus, depending on the application, deeper needle depths should be selected to maximize injected drug distribution. We also observed that needle gauge significantly impacted API distribution; this should be carefully selected in relation to injectate viscosity (which here was modified by EC concentration). Specifically, we observed a critical relationship between needle gauge and EC concentration, which was governed by pressure buildup in the injector system. Lower viscosity injectates (6% EC-ethanol) needed to be delivered through higher gauge needles (i.e., 27G) while higher viscosity injectates (12% EC-ethanol) needed to be delivered through lower gauge needles (i.e., 22G) to achieve maximal distribution volumes (Fig. 5). Both needle gauge and EC concentration directly impacted the pressure build up within the injection system – higher gauges and more viscous solutions yielded higher pressures. Interestingly, the combination of needle gauge and viscosity was key to increasing distribution area and was associated with some pressure building up within the injection system. Thus, needle gauge should be carefully selected in relation to injectate viscosity and subsequent gelation in tissue to ideally produce pressures in the range of ~70–250 kPa in order to maximize API distribution.

Next the impact of injection volume on distribution volume and fluid leakage was evaluated in excised swine tissue. We observed that the ratio of distribution volume to initial injection volume (i.e. ratio of the amount retained) was close to 1.0 for injection volumes ≤1000 μL (Fig. 6). At volumes greater than 1000 μL, the ratio dropped first to ~0.8 for 1500 and 2000 μL injections and then to ~0.6 for 3000 μL injections at which point fluid leakage outside of the depot into cracks was clearly visible (Fig. 6A). Thus, for shallow injections (i.e. ~13 mm depth) in cervical tissue, injection volumes ≤1000 μL led to maximal retention within target tissue and minimal fluid leakage.

Results from phantom and tissue experiments informed the development of a speculum-compatible hand-held injector to deliver EC-ethanol into the cervix (Fig. 2). Needle selection was informed by clinical injection procedures used in the cervix. Specifically, a 22G 8.9-cm-long beveled needle attached to a needle extender is typically used to inject a local anesthetic such as lidocaine into the cervix prior to gynecological excisional treatments. The elongated slim form factor of the needle and needle extender enables the gynecologist to visualize how to guide the needle through the speculum with the naked eye. Because the 8.9 cm needle is longer than the needles used in the phantom experiments, the pressure built up more quickly with 6% EC-ethanol. Thus, injections of 6% EC-ethanol (as opposed to 12% EC-ethanol) met remained within our ideal pressure range. The handheld injector could be manually operated (i.e., a user pushing on a plunger) or automatically operated (i.e., connected to a syringe pump) to control injection rate. Because injection rate is less important to control when injectate viscosity is sufficiently high, we evaluated if manual injections yielded comparable results to automated injections in which injection rate was controlled. Testing in swine cervices indicated that automatic and manual injections of both 500 and 1000 μL achieved statistically similar distribution volumes. Interestingly, the ratios of EC-ethanol distribution volume/ initial injection volume increased from 500 to 1000 μL with the 1000 μL injections forming a more symmetrical depot around the injection site. This was observed for both automatic and manual injections (Fig. 7), indicating that 1000 μL consistently yielded optimal depots in the cervix. Further, these results confirmed that injection rate does not need to be controlled when injecting 6% EC-ethanol into the cervix, eliminating the need for a syringe pump. Thus, manual injections (without a syringe pump to control injection rate) can be used if the injectate viscosity is high and gelation is sufficient.

There were several limitations of this study. First, it was difficult to measure the backflow of EC-ethanol because ethanol evaporates as soon as it is exposed to air (i.e. as soon as it reaches the tissue surface); consequently this study focused on the resulting distribution area/volume around the injection site rather than measuring backflow directly. Second, while phantoms enabled high-throughput experiments to establish trends between injection parameters and gel distributions, the size of the distributions and fluid leakage (particularly backflow) will not directly translate to the tissue environment. Significantly more backflow occurs in phantoms compared to tissue because the phantom tears when the needle is inserted resulting in much lower compressive stress between the needle-phantom interface vs the needle-tissue interface. Third, while a range of injection parameter values were explored in this study, not all may be clinically practical. For example, while 27G Potocky needles have been used previously for cervical injections, they are not as commonly used as other needle gauges[31]. Additionally, while an injection rate of 30 mL/hr may be practical for small injection volumes (e.g. 1 mL would take 2 minutes to deliver), it may not be clinically practical for larger injection volumes. Thus, once our initial trends were established, we focused on more practical parameter values in our final experiment (experiment 6, Figure 7) where a 22G needle and manual injection rate were used. From this experiment, we found that manual and automated control (where the syringe pump was set to 10 mL/hr) provided comparable distribution volumes and variability (Fig. 7). This suggests that injections can be delivered manually, which is more clinically practical. Fourth, normal swine cervices were selected for this study because they are similar in size and morphology to human cervices[39]; however, the tissue was from healthy pigs without cervical disease – to our knowledge a large animal model of cervical dysplasia does not exist[39]. Precancerous and cancerous lesions are stiffer and more heterogeneous, which could make crack formation more likely. Further studies should assess the impact of such tissue heterogeneity on injected drug distribution. Additionally, studies in excised tissue do not account for the impact of vascular or lymphatic clearance. In a pilot in vivo study in swine, we found that an injection of 500 μL led to a necrotic volume of ~200 μL 24 hours after ablation[24]. Vascular clearance in vivo or some leakage to the surface or into the vaginal canal could have contributed to the differences between the initial injected volume and the resulting necrotic volume. Larger follow up studies are underway to further investigate the translation of EC-ethanol in vitro injection results to in vivo conditions. In particular, additional studies are needed to evaluate how increased viscosity impacts tumor retention and response. Finally, we note that this study focused on trends from single injections; multiple sequential injections or devices that can inject at multiple sites around the cervix simultaneously, producing larger distributions, might be necessary clinically, especially in treating more advanced lesions.

Conclusion

Key injection-based delivery parameters that have significant impact on local drug distribution were identified through a series of experiments in tissue and tissue mimicking phantoms. Needle insertion depth, needle gauge, and injection volume need to be carefully considered when designing optimal delivery protocols in order to maximize drug distribution and minimize leakage. In particular deeper needle depths and staying below the critical pressure by controlling injection volume maximized distributions. Needle gauge and length should be carefully selected, with respect to injectate viscosity, to manage pressure build up within the injection distribution. Conversely, needle type and injection rate are less important to control (with some caveats) since they did not significantly impact overall drug distribution. Notably needle type can be selected to change the orientation of the largest cross-sectional distribution area. Additionally, injection rate does not have to be controlled if injectate viscosity is sufficiently high. By answering key questions related to injection-based parameters, this study provided insights into the optimization of injectable gels. Although this study focused on local delivery of EC-ethanol with application to treating cervical lesions, it has broader implications in local delivery of other therapeutics for a variety of medical conditions.

Supplementary Material

supp1-3340613

Supplementary Fig. 1. (A) Representative images of phantoms injected with 6% EC-ethanol mixed with various amounts of dye (0, 3, 4, 5, 7, or 10 drops). The yellow trace from ImageJ highlights the borders of the distribution. Scale bars = 5 mm. (B) The average distribution area was calculated for each condition (n=5). Error bars represent standard error. Statistical significance between conditions is indicated by a bar and asterisk (*p<0.05). Due to the large variability in results when dye was added, subsequent experiments were conducted without the addition of dye to the EC-ethanol mixture. (C) Representative image of phantom injected with 6% EC-ethanol mixed with 5 drops of dye. The blue food dye diffused farther than the EC-ethanol depot (i.e., not co-localized well with the chemical agents) increasing the variability when analyzing area distribution with the dye.

Supplementary Fig. 2. (A) Diagram of tubing system used for recording pressure in Experiment 4. (B) The average tubing compliance was measured by plugging the end while injecting fluid and measuring the pressure. (C) The tubing expanded with a compliance of 1.4 to 2 μL/kPa (determined by the slope of the graph), which was used to account for the effect of tubing expansion on distribution area in experiment 4.

Supplementary Fig. 3. (A) Representative images of phantoms injected with 6% EC-ethanol mixed with various amounts of iohexol (8, 30, 40, or 50 mg/mL). Scale bars = 2.5 mm. (B) The average signal-to-background ratio (SBR) was calculated for each condition (n=3). Error bars represent standard error. SBR was determined based on signal and background mean gray values determined in ImageJ. Statistics were not calculated because of the small sample size. Visually, minimal difference was found between an SBR of 40 mg/mL and 50 mg/mL, indicating 40 mg/mL is sufficient to achieve depot visualization.

Supplementary Fig. 4. Representative images of the thickness of phantoms injected with different needle depths (4, 8, 10, and 13 mm) and injection rates (10 and 30 mL/hr) for experiment 3. Scale bars = 5 mm.

Supplementary Fig. 5. Pressure curves showing pressure over time associated with injections of 6%, 9%, and 12% EC-ethanol into phantoms with various needle gauges (22, 23, 25, and 27G).

Supplementary Fig. 6. Average maximum pressure versus distribution area achieved with injections of (A) 6%, (B) 9%, and (C) 12% EC-ethanol into phantoms with various needle gauges (22, 23, 25, and 27G).