In vitro and ex vivo evaluation of the biological performance of sclerosing foams

Since the first reports on foam sclerotherapy, multiple studies have been conducted to determine the physical properties and behavior of foams, but relatively little is known about their biological effects on the endothelial cells lining the vessel wall. Moreover, a systematic comparison of the biological performance of foams produced with different methods has not been carried out yet. Herein, a 2D in vitro method was developed to compare efficacy of commercially available polidocanol injectable foam (PEM, Varithena) and physician-compounded foams (PCFs). Endothelial cell attachment upon treatment with foam was quantified as an indicator of therapeutic efficacy, and was correlated with foam physical characteristics and administration conditions. An ex vivo method was also developed to establish the disruption and permeabilisation of the endothelium caused by sclerosing agents. It relied on the quantitation of extravasated bovine serum albumin conjugated to Evans Blue, as an indicator of endothelial permeability. In our series of comparisons, PEM presented a greater overall efficacy compared to PCFs, across the different biological models, which was attributed to its drainage dynamics and gas formulation. This is consistent with earlier studies that indicated superior physical cohesiveness of PEM compared to PCFs.

www.nature.com/scientificreports www.nature.com/scientificreports/ subintima was observed, accompanied by progressive separation from the tunica media and initial formation of a thrombus.
Erkin et al. treated varicose vein segments with a selected concentration of POL foam produced with the TSS method, at 1:4 liquid:RA ratio. Vein segments were immersed in foam for 5 minutes, and subsequently examined 29 . Treatment with POL foam caused endothelial swelling, necrosis, and intimal thickening. However, these effects were not statistically correlated to the concentration of sclerosant, except for the presence and extent of necrosis.
Whiteley et al. treated ex vivo human varicose veins with 1% or 3% STS and POL, for 1 or 10 minutes 25 . Cell death and medial damage were directly correlated to surfactant concentration and treatment time. POL caused less damage to the endothelium and smooth muscle cells compared to STS.
Overall, these histological studies demonstrated the qualitative effects of the interaction between sclerosing agents and the vessel wall. Quantitative analyses mostly relied on microscopic measurements, which were however limited to regions of interest within the treated vessel. As for the in vitro studies, therapeutic effects were largely dependent on treatment time and sclerosant concentration, although treatment timescales differed between investigations. This could be due to differences in the physical properties of the sclerosing agent used and the experimental conditions. To the best of the authors' knowledge, there is no comparative quantitative analysis between different foam production or administration methods, or attempt to correlate physical with biological performance of sclerosing foams. This is also reflected in the lack of clinical studies comparing efficacy and safety of different foam production methods.
Herein, we propose two methods for quantifying sclerosant-induced disruption of the endothelial layer in vitro and ex vivo. Using the in vitro model, the therapeutic efficacy of different polidocanol-based sclerosing agents was investigated, and correlated with their physical characteristics and administration protocols. Therapeutic efficacy was subsequently evaluated within a more complex ex vivo model. For the first time, a comparison between different foam production techniques has been performed, by employing biological models with different levels of complexity. Results from this study can provide clinicians with some fundamental understanding of how different foam formulations may perform in the body.

Results
In vitro evaluation of the biological performance of sclerosing agents. In the first step of the study, a method replicating the clinical treatment procedure was designed in order to investigate the biological effects of sclerosants on a two-dimensional (2D) endothelial model (see Methods section for additional details). The mechanism of action of sclerosing agents relies on endothelial damage; therefore, endothelial cell attachment was employed as a metrics for therapeutic efficacy. Since detached endothelial cells are known to undergo apoptosis, cell attachment was considered as an indicator of cell viability 19 . Therefore, a lower percentage of attached cells upon treatment indicated a more effective sclerosing agent.
Firstly, the repeatability of the method was assessed by fixing the injection and treatment parameters (PEM foam, 15 seconds of treatment time, and 1 mL of foam injected without needle) and repeating the experiment six times. Results showed consistency of foam performance across multiple independent repeats (see Fig. S1(A)).
Subsequently, the sclerosing efficacy of liquid POL was investigated. A 1% POL solution was serially diluted in PBS in order to identify the minimum effective and 50% inhibitory concentrations (15 seconds treatment duration, and 1 mL of sclerosant injected without needle). Figure 1 shows that POL 1% is still effective even after five serial dilutions (0.03% final volumetric concentration), removing >50% of cells in a well. Concentrations of foam below 0.02% rendered the treatment ineffective (85 ± 10% of attached cells). A 50% inhibitory concentration of 0.024% was determined from these experiments.
An additional experiment was designed to investigate the extent of polidocanol 'depletion' , potentially due to the interaction with cell medium constituents or intercalation within cell membrane fragments. In these experiments, 1 mL of liquid POL was injected into one well and left for 15 seconds. The solution was then transferred into a neighbouring well, and the process was repeated in order to treat five wells in series. As shown in Fig. 2, the 1% polidocanol solution maintained the same efficacy after five serial injections (only 17.5 ± 4.0% of cells remained attached after the 5 th injection). The experiment was repeated using a lower POL concentration of 0.03%. Results demonstrated that depletion of active POL occurred, as the percentage of attached cells after  www.nature.com/scientificreports www.nature.com/scientificreports/ treatment increased from 29.3 ± 2.0% (3 rd injection) to 49.82 ± 11.8% (4 th injection) and 57.7 ± 19.6% (5 th injection). Reducing the POL concentration further (to 0.02%) resulted in a similar trend, although the change in percentage of attached cells was less significant because of the reduced effectiveness of the sclerosing solution (coherently with the results shown in Fig. 1).
The usage of injection needles with different bore size was also investigated, because of their potential effect on foam size and stability. Cells were exposed to 1 mL of PEM foam for 15 seconds, either with or without a needle. Firstly, a needle with the greatest bore size in the range investigated was employed (16G). Fig. S2 shows that the presence of a 16G needle had a negative impact on foam treatment efficacy (i.e., the percentage of attached cells upon treatment increased from 10.12 ± 2.2% to 16.26 ± 3.0%; p < 0.001). Therefore, in order to investigate this effect further, additional needle bore sizes were tested, corresponding to 25G and 30G. These are the types of needle most frequently employed in clinical practice 19 , allowing us to reproduce more faithfully a clinical injection procedure. Overall, decreasing the needle diameter from 16G to 30G resulted in lower cell death (Fig. 3A). In the case of PEM, there was statistically significant difference in foam efficacy between 16G and 30G needles (p = 0.03) (Fig. 3A). Comparing the different foam production methods, statistical difference was found only when using the largest needle (16G), with PEM associated with statistically greater treatment efficacy (% attached cells: 11.8 ± 4.6%) compared to both DSS (% attached cells: 19.5 ± 8.9%) and TSS (% attached cells: 20.0 ± 11.3%) foams.
In order to determine the effect of needles on foam physical properties, bubble size measurements were carried out using the glass-plate method. Figure S3 shows the bubble size distribution of PEM and PCF foams, injected through different needle sizes. Results show that injection through a needle did not significantly impact on the bubble size distribution of all types of foam. Comparing the different foam types, room air PCFs had a narrower bubble size distribution than PEM (in the bubble size range 0-400 μm) for all needle inner diameters investigated. However, PCFs had a greater number of bubbles in the size range 400-510 μm compared to PEM. Despite there was no significant change in bubble size distribution, foam injection through a needle caused visible phase separation between the liquid and gaseous phases. Therefore, an experiment was developed to quantify  www.nature.com/scientificreports www.nature.com/scientificreports/ foam drainage dynamics within a vial, upon foam injection through needles of different bore size. The vial inner diameter was comparable to the one of well plates used for in vitro biological testing. Figure 4 shows the time evolution of the height of liquid POL solution at the bottom of the vial, which was employed as a metrics for drainage.
When injected using the narrowest needle diameter (30G), all foams presented a higher liquid fraction at the beginning of the experiment [liquid height was 0.45 mm (PEM), 0.66 mm (TSS), and 0.67 mm (DSS)], followed by a relatively slow drainage dynamics. After 200 s, the liquid height was 0.77 mm for PEM, 1.67 mm for TSS, and 0.84 mm for DSS. Differences between foams were more evident at the larger needle diameters, with PEM foam undergoing a significantly slower drainage compared to DSS and TSS foams. The largest difference between foam types was observed when using the 16G needle; after 200 s, the liquid height was equal to 0.23 mm (PEM), 1.52 mm (TSS), and 1.26 mm (DSS).
The biological effect of changing the foam volume was also investigated, by injecting either 0.5 mL, 1 mL, or 2 mL (which are comparable to clinically injected volumes, if normalised to the treated surface area) 19 . In these experiments, the treatment time was fixed to 15 seconds. Results showed a significant reduction in the percentage of attached cells with increasing the volume of foam from 0.5 mL to 2 mL (Fig. 5). Moreover, PEM had significantly greater efficacy compared to PCFs when using 0.5 and 1 mL of foam. Increasing the foam volume further (2 mL) resulted in comparable percentage of attached cells between PEM and PCFs (<10% in all cases).
The effect of varying the exposure time of HUVECs monolayers to sclerosing agents was investigated. Earlier in vitro and ex vivo studies have reported on treatment times in the range 5 s-1 hr, whilst it is usually recognised to be in the order of a few seconds in vivo 32 . In this study, the treatment time was varied in the range 15-120 s, which is consistent with our previous determinations of foam plug persistence within an artificial vein model 17 . As shown in Fig. 6, the efficacy of a 120 s long treatment (PEM = 6.5 ± 0.9%, DSS = 10.5 ± 2.6%, TSS = 9.7 ± 2.3%) was significantly higher compared to shorter treatments. Overall, PEM was statistically more effective than both DSS and TSS, at all treatment times investigated.
In addition to the above quantitative assays, histopathologic observations of treated HUVECs were performed. Images of cell monolayers exposed to various sclerosing agents were captured, using an optical microscope with phase contrast. The untreated (control) cells displayed a normal EC morphology for confluent monolayers, and were adherent to the substrate ( Fig. 8(D)). Following treatment, cell morphology changed to a more rounded appearance; the monolayer became disrupted, where a large number of cells detached from the substrate and, in some cases, only fragments of cells were present. Figure 8(A-C) show images of cells after exposure to foam generated using different production methods (15 seconds treatment duration, and 1 mL of foam injected without needle). It is evident that PEM ( Fig. 8(A)) and DSS RA ( Fig. 8(B)) foams caused greater cell detachment compared to TSS foam ( Fig. 8(C)), which is coherent with the quantitative determinations (Fig. S1B).
Ex vivo evaluation of the biological performance of sclerosing agents. In order to investigate the sclerosing performance of foams in a more realistic biological model, ex vivo experiments were established. The www.nature.com/scientificreports www.nature.com/scientificreports/ reliability of the method was initially evaluated by quantifying endothelial damage induced by Type I collagenase, an enzyme that removes EC from the vessel wall by proteolysis of underlying collagen. The vein was exposed to the enzyme for 10 minutes. The same procedure was repeated using liquid POL (1% v/v), and a physiological saline as a control (Fig. S4). Following exposure to Evans Blue-conjugated BSA, control cords showed no leakage into the tissue surrounding the vein (the quantity of EB extravasated was 0.5 ± 0.2 mg EB/g tissue). The collagenase solution (positive control) showed a level of disruption equivalent to 42 ± 4.5 mg EB/g tissue, whereas liquid POL caused 21 ± 1.2 mg EB/g tissue of extravasation. Upon verification of the method, the effect of treatment time was investigated. The vein was treated with liquid POL 1% for 1, 5, and 10 minutes. Figure 9(A) shows that endothelial disruption is directly proportional to exposure time (extravasation ranged from 1.55 ± 2 to 21 ± 1.2 to mg EB/g tissue).
The same experiment was subsequently performed using PCFs and PEM, using a constant exposure time of 1 minute. Figure 9(B) shows that PEM was more effective in disrupting the endothelium compared to DSS PCF and liquid POL; whilst no significant difference was observed between PCFs and liquid POL. The DSS method in this test produced less disruption than PEM (corresponding to 1.3 ± 0.8 and 3.9 ± 2.1 mg EB/g tissue, respectively) but the variation in the measurement was such that this was not statistically significant. www.nature.com/scientificreports www.nature.com/scientificreports/

Discussion
Effect of foam production methods and administration-related parameters on foam efficacy in vitro. Since the introduction of foam sclerotherapy as a treatment method against varicose veins, numerous studies have been conducted in order to further the understanding of the physical properties and behavior of foams 20,23,33,34 . However, a relatively limited body of work has focused on the biological effects of sclerosants on endothelial cells and the vessel wall 7,11,25,26 . Earlier studies have revealed that sclerosing efficacy is directly correlated to treatment time and sclerosant concentration 29,35 . However, the lack of quantitative analyses and the difference between the physical properties of sclerosing agents investigated, have both hindered the ability to draw generalized conclusions about the efficacy of different foam production and administration methods.
In this study, we have employed two simple quantitative methods to compare the microscopic and macroscopic effects of different foam production techniques on the endothelium. With respect to the microscopic effects, we proposed an in vitro model that allows the quantification of sclerosant-induced endothelial disruption, by determining the number of cells attached to a substrate after treatment. In this method, monolayers  The repeatability of the method was initially evaluated, showing significant consistency across multiple independent repeats (Fig. S1A). In a first step of the study, the model was utilised to investigate the sclerosing  www.nature.com/scientificreports www.nature.com/scientificreports/ efficacy of liquid POL. Only at volumetric concentrations <0.02% the surfactant was rendered ineffective, confirming the potency of this detergent at disrupting the endothelial cell membrane and inducing cell death 27 (Fig. 1). Serial treatments using the same POL solution were performed to assess whether polidocanol deactivation occurred. Reducing the number of active molecules (i.e., by lowering the POL concentration) caused reduced efficacy after a certain number of treatments, which was dependent upon the POL concentration (see Fig. 2). Depletion of active polidocanol over consecutive treatments was likely due to its intercalation within lysed membrane fragments. However, the 1% POL solution (employed to manufacture both PCF and PEM foams) maintained its potency across multiple treatments, and its in vitro biological performance was not affected by polidocanol depletion.
Upon verification of polidocanol efficacy in vitro, the effect of administering sclerosing foams with needles of different bore diameter was investigated. The needle bore size is typically selected based on the vein to be treated, with smaller veins often requiring the smaller 25-30G needles 19 . The performance of different sclerosing foams was statistically different only when using the larger needle (16G) (Fig. 3). Employing narrower needles (i.e., 25G and 30G), foam efficacy reduced and differences between foam types were not statistically significant. This observation may be due to changes in the physical properties of foams when they were conveyed through a needle. Bubble size measurements however revealed that the bubble size distribution of all types of foam was virtually unaffected by the needle inner diameter (Fig. S3). Previous studies have shown that as foam flows through a pipe, the change in bubble diameter is dependent on the pressure drop across the pipe 36 . It can therefore be inferred that the pressure drop required to administer foams manually through clinical needles -and the resultant shear rate -were not sufficient to cause a significant change in the bubble size of PEM and PCF foams. Thus, the observed changes in foam therapeutic efficacy could not be directly related to the foam bubble size distribution. For this reason, additional experiments were performed to quantify the effect of needle injection on foam drainage dynamics, where drainage describes the flow of liquid through a foam 37 . During free drainage, the liquid volume fraction increases monotonically from the top to the bottom of a foam column. This bottom liquid layer is depleted of surfactant molecules, as the surfactant preferentially stabilises the gas-liquid interface of bubbles located in the upper foam layer. The liquid then continues to drain downward over time, until the liquid height reaches a steady state 38 (as shown in Fig. 4). Given that drainage is strongly affected by the size and shape of the foam container, a vial with inner diameter comparable to the well plate used in biological tests was employed. By injecting foams through the narrowest needle (30G) caused visible separation of the liquid and gaseous phases upon injection; thus, the ejected foam experienced only limited drainage (Fig. 4). Phase separation may occur because of the liquid POL travelling at a different velocity compared to the gas bubbles, as observed for other multi-phase systems delivered through needles, such as pastes and cements 39 . The extent of phase separation reduced with increasing the needle inner diameter (corresponding to lower injection velocity), and was almost absent when foams were administered using the largest 16G needle (Fig. 4). When foam separation occurred (i.e., using the 25G and 30G needles), the biological efficacy of foams was dominated by their 'static' liquid fraction, and differences between foam types were not statistically significant (Fig. 3). Conversely, when phase separation was significantly reduced (as in the 16G needle experiments), the ejected foams displayed distinct drainage dynamics (see Fig. 4) that in turn led to differences in their biological efficacy. Notably, the slower drainage of PEM foam resulted in statistically greater therapeutic efficacy compared to PCF foams (Fig. 3), which instead presented a faster initial drainage dynamics. The more rapid drainage of room air PCF foams could be attributed to: (i) the greater liquid:gas volume ratio compared to PEM foam 22 , with previous studies reporting on a direct correlation between foam www.nature.com/scientificreports www.nature.com/scientificreports/ drainage velocity and liquid fraction 40 . (ii) The lower average bubble diameter combined with the presence of a greater proportion of bubbles with diameter >400 μm (see Fig. S3). Notably, higher pressure within the smaller bubbles drives diffusive gas exchange towards the larger bubbles, and the resulting coarsening of the foam accelerates its initial drainage dynamics 41 .
The mechanism for which the slower foam drainage of PEM leads to greater therapeutic efficacy in vitro, is not fully understood yet. However, it could be attributed to the persistence of gas bubbles in the vicinity of the cell membrane, with higher concentration of active polidocanol located at the gas-liquid interface. Conversely, when a fast-draining foam is employed, cells are exposed to the liquid phase that has been depleted of polidocanol, particularly in the shorter term. Depletion is greater in N 2 -containing foams, given to the lower 'mobility' of surfactant molecules in these foams 42 .
The effect of the injected foam volume was also investigated, as it represents a parameter that is varied in the clinical practice. Generally, the volume injected is dependent on the diameter and length of the vein to be treated 43 . There was a significant difference between foam production methods when injecting 0.5 and 1 mL of foam, whereas all treatments had very similar biological performance and became more effective when injecting a greater volume of foam (2 mL) (see Fig. 5). Earlier studies have reported that the dependence of drainage time on the foam liquid fraction reduces with increasing the height of a foam column 44 , which may explain the comparable efficacy of PEM (liquid fraction: 12.5%) and PCFs (liquid fraction: 20%) at 2 mL. The positive correlation between the injected volume and treatment efficacy may be attributed to increased gravitational effects at the higher foam heights 40 , which favors downward motion of active polidocanol towards the cell monolayer. It should be noted that a foam volume ≤1 mL is more representative of a clinical injection procedure, considering the volume of foam normalised to the area of the treated endothelial layer 19 . At these lower volumes, drainage dynamics is governed by both capillarity and gravitational effects.
The effect of varying the treatment time was also investigated. The exposure time was defined based on the predicted persistence of a foam plug in vivo 17 , and values investigated were 15, 30, 60 and 120 seconds (Fig. 6). Overall, there was significant difference in biological efficacy between 15, 30, and 60 seconds of exposure. However, for all types of foam, efficacy significantly increased at 120 seconds of exposure. Notably, bubble collapse in the longer term causes a release of active polidocanol, and biological effects thus become dependent on the liquid POL solution. Further investigations are required to fully elucidate the interplay between foam drainage and the temporal dynamics of membrane disruption upon exposure to the surfactant agent. Overall, PEM maintained superior performance across the all range of exposure times investigated and was more effective over longer term exposures, likely due to its sustained drainage dynamics compared to PCFs 17 .
Considering the potency of the 1% POL solution over multiple treatments in vitro (as illustrated in Fig. 2), the greater therapeutic efficacy of PEM compared to PCFs may not be solely attributed to differences in foam stability and drainage dynamics. Previous studies have demonstrated that the diffusion velocity (or mobility) of water-soluble surfactants in foams is affected by the gas formulation, and that it is greatest in CO 2 foams, followed by O 2 foams and N 2 foams 42 . Experiments were thus conducted using PEM manufactured using different gas formulations (Fig. 7), to assess whether changes in surfactant mobility may influence its therapeutic efficacy. Coherently with these previous findings, the commercial PEM formulation (35:65 CO 2 :O 2 ) had greater efficacy than both 100% O 2 PEM and N 2 -containing (RA) PEM. These results suggest that polidocanol is more readily available for interaction with cell membranes, when N 2 -free foams are employed.
Comparing the ex vivo performance of different foam production methods. In order to evaluate the performance of different sclerosing agents in a more realistic biological model, experiments were performed ex vivo using umbilical cord veins. Sclerosant-induced disruption of the endothelium was determined from extravasation of a BSA-conjugated dye. Initially, the ability of the method to provide a quantification of endothelial disruption was assessed, using collagenase to actively cause endothelial damage. Collagenase, an endopeptidase that digests native collagen 45 , was left in the vein for 10 minutes (Fig. S4). The same procedure was performed using liquid POL (1% v/v) or physiological saline as a control. As expected, saline did not cause tissue damage, whereas the collagenase solution caused greater endothelial disruption compared to liquid POL. It is well known that the enzyme cleaves collagen bonds causing a removal of the endothelium and potential damage to the underlying tissues, compared to a surfactant agent that interferes with the cell membrane only, causing cell death 27 . After method's validation, more clinically relevant exposure times were applied. Veins were treated with liquid POL 1% for 1, 5 and 10 minutes. A direct correlation between contact time and endothelial disruption was observed (Fig. 9A), consistently with in vitro experiments using sclerosing foams.
The same procedure was performed using PCFs and PEM, with an exposure time of 1 minute. PEM was more effective at disrupting the endothelium compared to DSS PCF, as expected from the results obtained in vitro. There was also a significant difference between the efficacy of foamed and liquid POL, suggesting that the dynamics of foam drainage and the 'local' surfactant concentration levels may become even more influential over foam therapeutic efficacy within a 3D environment. Interestingly, despite TSS foam being less effective in generating endothelial wall damage compared to PEM, differences between mean values were not statistically significant. This finding is in contrast with the in vitro results, where DSS foam was consistently superior to TSS, although differences between PCFs significantly reduced with increasing the treatment time both ex vivo (Fig. 9B) and in vitro (Fig. 6).

Conclusive remarks.
To the best of the authors' knowledge, the present study represents the first systematic comparison of the biological performance of different sclerosing foam formulations, and a first attempt to correlate biological performance with foam physical properties.
Overall, analyzing the results obtained using both the in vitro and ex vivo models, PEM was the most effective foam for disrupting the endothelial layer in a variety of tests and over different timescales of treatment. This was www.nature.com/scientificreports www.nature.com/scientificreports/ attributed to the slower drainage dynamics of PEM compared to PCFs, and -potentially -to the enhanced polidocanol mobility conferred by its gas formulation. It was also shown that reducing the injection needle diameter, increasing the volume of injected foam, and increasing the treatment time, all contributed towards increasing treatment efficacy (for all types of foam).
It should also be highlighted that PCFs made from room air have associated risks, with persistent nitrogen bubbles in the circulation, whereas PEM, made with a low-nitrogen CO 2 :O 2 gas mixture, is not associated with the risks of high-nitrogen content 46 .
In conclusion, in this study we have developed a simple 2D in vitro method to quantify the efficacy of foam sclerotherapy. The method allows for the investigation of different clinical parameters such as exposure time, injected volume, concentration of sclerosant, and needle bore size amongst others. In addition, we utilized a more realistic biological model, i.e. a three-dimensional ex vivo vein model, as a further method of evaluation. However, we are aware that the both in vitro and ex vivo models do not fully reflect the clinical setting for foam sclerotherapy, because they are employed in static conditions and thus do not replicate foam-induced blood displacement, and also do not provide a faithful replication of the varicose vein architecture. Our group is therefore working to overcome this limitation by developing a 3D in vitro platform applied in dynamic conditions, moving closer to models that mimic the physiological and clinical environments, ultimately, as an alternative to animal testing. Despite the current limitations, the two models generated reliable and reproducible results, and they can be employed in parallel in order to compare the performance of sclerosing treatments. In our series of comparisons using both models, we confirmed findings from our previous physical studies 17,22,47 that PEM presents a better overall performance compared to PCFs across a variety of biological efficacy tests.

Methods
Foam production methods. In this study, the commercially available Varithena 1% varicose vein treatment (referred to as polidocanol injectable foam or PEM) was employed, and its performance compared with physician compounded foams (PCFs) made using different foam generation methods.
With respect to PCFs, POL (Croda, Goole, UK) at a concentration of 1% (v/v in buffered saline) was employed as a surfactant agent. Foams were produced by mixing liquid and room air (at a volume ratio of 1:4, respectively) as this is the most widely used formulation adopted in clinical practice 48 . Two methods of PCF production were investigated: (i) DSS and (ii) Tessari. In the DSS method, foam was produced by passing the POL solution from a 5 mL syringe, ten times into and out of a 10 mL syringe. Silicon-free syringes (BD Biosciences, USA) were connected via a Combidyn adapter (B. Braun Melsungen, Germany). In the Tessari method, the straight connector was replaced with a three-way valve that was set at a 30° off-set. Polidocanol endovenous microfoam (PEM) Varithena is a commercially available microfoam combination produced by Provensis Ltd (a BTG International group company, London, UK) consisting of a proprietary 35:65 CO 2 :O 2 gas mixture with ultralow nitrogen content (<0.8%) and 1% POL solution. The foam is contained within a pressurized canister combined with a transfer unit, which can be connected to a 10 mL silicone-free syringe. Once connected, the syringe is filled with 5 mL of foam. Experiments were conducted at room temperature (23 °C), after foam production, and foams were produced by the same operator.
In vitro test method to evaluate performance of sclerosants. A method was designed to test the efficacy of sclerosants, in both their liquid and foamed form. A monolayer of human umbilical vein endothelial cells (HUVECs) was cultured until confluence into 24 well-plates (Sigma-Aldrich Co. LLC., USA). The following steps were designed to mimic different phases of sclerosant's injection: (i) the HUVECs media (HM) (Thermo Fisher Scientific Inc., USA) was removed from the wells, in order to achieve direct contact between cells and sclerosants; (ii) the cell monolayer was exposed to various sclerosing agents during a fixed time of approximately 15 seconds, reproducing the injection process; (iii) sclerosants were removed using a pipette, and cells were washed once using a HBSS buffer (Hanks Buffered Saline Solution, Sigma-Aldrich Co. LLC., USA) mimicking the sclerosant's displacement and dilution caused by blood flow; and (iv) fresh medium was added. Sclerosants' injection was performed manually using a 5 mL syringe (BD Biosciences, USA), with and without a needle. The syringe was kept perpendicular to the bottom plane of the well, and the sclerosing agent was injected from the centre of the well. The standard procedure was carried out under these conditions: 1 mL of liquid/foamed sclerosant, 15 seconds of exposure time, and a 16 G needle employed for injection. Following treatment, the medium was removed and cells were washed gently in warm HBSS, which was subsequently removed by aspiration.
Cells were subsequently fixed with the addition of 0.7 mL of a 10% formyl saline solution (Sigma-Aldrich Co. LLC., USA). Fixative was then removed by aspiration, and a methylene blue solution (MB) 1% (w/v methylene blue in 0.01 M-borate buffer pH8.5) (Sigma-Aldrich Co. LLC., USA) was added to each well.
The MB solution was then transferred to a 96 well flat-bottomed plate (Sigma-Aldrich Co. LLC., USA), with 0.1 mL being added in duplicate wells. A control set of untreated cells was used to generate a standard curve of MB equivalent to serial dilutions of 100% cells. MB absorbance was then measured using a plate reading spectrophotometer, at a wavelength of 650 nm. Absorbance values of treated cells were then converted into a percentage of attached cells, using a calibration function. The number of cells attached is a measure of the number of live cells upon treatment. The latter was derived from linear regression of experimental data points, using Prism software (GraphPad Software, Inc., USA).
During the study different parameters were varied, such as (i) volume injected, (ii) exposure time, (iii) needle bore size, and (iv) gas formulation. The volumes of injected sclerosant investigated were 0.5, 1, and 2 mL, whilst the exposure times investigated were 15, 30, 60 and 120 seconds. The needles employed were selected based on the common clinical practice, and had an inner diameter of 30G, 25G and 21G, corresponding to 0.16, 0.26 and 0.51 mm, respectively (BD Biosciences, USA). In order to investigate the effect of the gas formulation, PEM foam (2019) 9:9880 | https://doi.org/10.1038/s41598-019-46262-0 www.nature.com/scientificreports www.nature.com/scientificreports/ was produced using different gas constituents (in addition to the commercial formulation), including 100% O 2 and room air.

Measurement of foam drainage dynamics.
A transparent glass vial (outer diameter: 10.9 mm) was placed within a custom-built photographic chamber with a black background. A charge-coupled device (CCD) camera (Canon EOS06) was positioned in front of the vial. The foam was produced and injected (2 mL) inside the vial, using different types of needle (16G, 25G, and 30G) and different foam production techniques (PCFs and PEM). The experiment was repeated five times, for each condition investigated.
The time between foam injection and the beginning of the video recording was approximately 23 seconds. Videos were recorded for 5 minutes (25 frames per second), and subsequently analysed using a Phyton script developed in-house. The script loads the video and extracts its individual frames. It then performs the following steps in a semi-automated fashion: (i) User selection of a region of interest for analysis. (ii) Calibrating the image dimensions, by converting pixels into physical units. This is carried out by user selection of a feature of known length (for instance, the diameter of the vial). (iii) Converting the image into a black and white binary format, where black corresponds to the liquid phase and white corresponds to foam. (iv) The centerline of the selected region of interest is determined, and a rectangular window for analysis is defined. The width of this window extends 5 pixels away from the centerline, at both sides. It was decided to analyse foam drainage within an interrogation window (as opposed to a line), as data would be less sensitive to experimental noise. (v) Automated counting of the number of black pixels along the height of the interrogation window. An average height was determined, which corresponded to the height of liquid POL in the vial (upon dimensional calibration). (vi) Steps (iii)-(v) were performed automatically on each image frame, and a plot of the liquid height (in mm) vs. time was generated. This provided a quantitative measure of foam drainage dynamics.

Measurement of bubble size distribution. The bubble size distribution was measured using an in-house
glass-plate method, as described in our earlier study 22 . Briefly, an aliquot of freshly generated foam (volume: 49 μL) was placed on a glass plate and immediately covered by another. The plates were thick enough not to bend, and were separated by a 32 μm thick gap. A flattened foam monolayer was thus created, which comprised 32 μm high, flat cylindrical bubbles. A light microscope and camera (AxioCam ICc 1, Carl Zeiss Microscopy, Cambridge, UK), with lighting adjusted to create sharp images of circular boundaries, were employed to capture sequential image fields. A built-in software was used to "stitch" fields together. Each individual bubble was identified and the bubble diameter measured using the image analysis (AxioVision, Zeiss) programme, with bespoke BubbleSizerMeasure macro. Approximately 2000-3000 bubbles per sample were measured using this procedure. The experiment was repeated five times, for each condition investigated.

Microscope imaging of treated cells.
Bright field images of HUVECs were acquired with an optical microscope (Olympus, CKX41, Japan). Images were taken of live samples immediately after treatment, with phase contrast microscopy (objective magnification 4x).
Ex vivo test method to evaluate performance of sclerosants. This part of the study was carried out in accordance with the Human Tissue Act (2004) and the recommendations of Southampton & South West Hampshire Research Ethics Committee B with Governance provided by the University of Southampton Research Governance Office. Umbilical cords were collected from the Princess Anne Hospital (Southampton, UK) from non-complicated natural vaginal births following agreed ethical collection protocols (Local Research Ethical Committee (LREC); Ref: 07/H0502/83). The umbilical cord was cut from the placenta and sectioned into 10 cm long segments. A steel feeding cannula (16G) was inserted into the vein. The cannula was clamped in place and attached to a 30 mL syringe filled with a physiological saline "cord buffer". The vein was washed until the fluid exiting the other end of the cord was clear. The treated umbilical sample was then cut into 5 vein segments. The vein was filled with a collagenase solution at 0.1% in phosphate buffered saline (PBS, Worthington Biochemical Corp., USA) or with different types of sclerosing agent. The cord segment was then incubated at 37 °C for 10 min. After incubation, the vein was washed again with cord buffer, and filled with 2 mL of Evans blue (EB) (0.33% EB and bovine serum albumin, BSA). The cord was then incubated at 37 °C for 20 min. After incubation, Evans blue was washed out using the cord buffer.
Each cord segment was cut in smaller pieces (0.5 cm long), which were weighed and inserted in 1.5 mL tubes. A formamide solution was added into each tube, and all tubes were transferred into a 62 °C water bath and incubated overnight in order to extract EB from the tissue. Tubes were centrifuged at 13000 rpm for 20 minutes at 20 °C. The supernatant was then transferred into a 96 well flat-bottomed plate. The EB stock solution was serially diluted to generate a standard curve. EB absorbance was then measured on a plate reading spectrophotometer, at a wavelength of 610 nm. The absorbance from a calibration standard curve was used to calculate unknowns, using the Prism software (GraphPad Software, Inc., USA) and applying a hyperbolic interpolation and regression.
Afterwards, the amount of extravasated Evans blue (in mg) per gram of tissue was calculated.
www.nature.com/scientificreports www.nature.com/scientificreports/ statistical analysis. The comparisons between treatments were performed using unpaired Student's t-test with Welch's correction, with appropriate post-hoc tests. Statistical significance was assumed for p < 0.05. All statistical tests were performed with Prism software. Data were reported either as the mean ± standard deviation, or in the form of a Tukey's box plot (comprising 25 th percentile, median, and 75 th percentile).