Exploring the Mechanical Properties of Bioprinted Multi-Layered Polyvinyl Alcohol Cryogel for Vascular Applications

Abstract

Polyvinyl alcohol cryogels (PVA-C) are promising materials for vascular tissue engineering due to their biocompatibility, hydrophilicity, and tuneable mechanical properties. This study investigates the mechanical performance of multi-layered PVA-C constructs fabricated via sub-zero extrusion-based three-dimensional (3D) bioprinting. Samples with two, four, and six alternating layers were evaluated to assess the effect of layered architecture on elastic and viscoelastic behaviour. Uniaxial tensile testing revealed that increasing the number of layers led to a moderate reduction in stiffness; for instance, at 20% strain, six-layered constructs showed a significantly lower (p < 0.05) Young’s modulus (36.7 ± 2.5 kPa) compared to two-layered ones (47.3 ± 3.1 kPa). Stress–strain curves exhibited nonlinear characteristics, better captured by quadratic (as opposed to linear) fitting, within the tested strain range (≤40%). Dynamic mechanical analysis demonstrated a frequency-independent storage modulus (E′) across 1–10 Hz, with subtle variations in viscoelastic response linked to the number of layers. Visual inspection confirmed improved print fidelity and hydration retention in thicker constructs. These findings demonstrate that a multi-layered design influences the mechanical profile of PVA-C and suggests potential for functionally graded design strategies to enhance compliance matching and mimic the biomechanics of native vessels in small-diameter vascular grafts.

1. Introduction

Cardiovascular diseases remain the leading cause of mortality worldwide, with coronary artery disease being a significant contributor that often necessitates surgical interventions such as vascular grafting [1]. Synthetic vascular grafts must closely mimic the biomechanical properties of native arteries to ensure long-term functionality and reduce complications such as restenosis and thrombosis [2]. Arterial tissues exhibit a sophisticated balance of strength, stiffness, and toughness, achieved through their hierarchical structure, consisting of distinct layers with unique mechanical and biological functions [3]. Crucially, native arteries display anisotropic mechanical behaviour, primarily due to the alignment of collagen fibres and elastin within their layered architecture, which enables vessels to accommodate pulsatile blood flow while maintaining structural integrity [4].

The arterial wall consists of three distinct layers—intima, media, and adventitia—each contributing to the vessel’s mechanical response. The intima, primarily composed of endothelial cells and a thin layer of extracellular matrix, provides a low-friction interface for blood flow and contributes minimally to the vessel’s mechanical stiffness [5]. The media, which contains circumferentially aligned smooth muscle cells and elastin fibres, governs the artery’s compliance and ability to recoil under cyclic loading, ensuring physiological adaptation to pulsatile blood pressure [6]. The adventitia, predominantly composed of collagen fibres oriented in a helical arrangement, serves as the primary load-bearing component, providing tensile strength and preventing overexpansion under high-pressure conditions [7]. These structural variations result in a gradient of mechanical properties, where the media facilitates elasticity and the adventitia enhances overall tensile resistance, enabling arteries to withstand haemodynamic forces while maintaining their functional integrity [8].

To replicate the layered architecture and functional biomechanics of native vessels, biomaterials must accommodate both compliance and strength while remaining biocompatible and processable [9]. Hydrogels are promising candidates in this context, offering high water content, tuneable viscoelastic properties, and an ability to mimic the extracellular matrix [10,11]. These attributes make them particularly suitable for vascular graft applications, where dynamic mechanical performance and cellular integration are essential [12].

Among emerging biomaterial inks, polyvinyl alcohol cryogel (PVA-C) stands out for vascular applications due to its combination of printability, biocompatibility, and mechanical tunability [13]. Unlike protein-based or naturally derived hydrogels that often require complex crosslinking chemistries or lack structural integrity [14], PVA-C can be physically crosslinked through freeze–thaw cycles, enabling reproducible solvent-free processing and consistent mechanical performance [13]. Its relatively high viscosity and ability to gel under sub-zero conditions make it well suited for extrusion-based 3D printing, in contrast to lower-viscosity materials such as collagen or hyaluronic acid, which typically require support structures during printing [14,15].

Previous studies have shown that the number of freeze–thaw cycles influences mechanical properties such as stiffness and elasticity [16], and that the addition of bioactive components like gelatine can enhance cell adhesion and hemocompatibility without compromising structural fidelity [17]. PVA-C also exhibits favourable viscoelastic behaviour under cyclic loading, including damping and elastic recovery, positioning it as a strong candidate for replicating the dynamic mechanical environment of native arterial tissues [18].However, achieving directional (anisotropic) mechanical responses remains a challenge, as traditional fabrication methods like casting typically yield isotropic structures that do not reflect the direction-dependent properties of arteries [18].

Bioprinting is an advanced form of additive manufacturing that enables the precise layer-by-layer deposition of living cells, biomaterials, and biologically active molecules to fabricate tissue-like constructs with high spatial precision [15]. It holds considerable promise for regenerative medicine [14], pharmaceutical applications such as drug screening and personalised delivery systems [19], and the development of customised biomimetic implants [15]. By supporting the fabrication of organoids and organ-on-a-chip platforms, bioprinting is also driving innovation in preclinical testing and personalised therapies. Beyond these technical advantages, the approach may significantly improve health-related quality of life by addressing organ shortages, reducing transplant rejection risk, and offering patient-specific therapeutic options [20]. However, the integration of living biological materials raises complex challenges around intellectual property ownership, ethical regulation, and commercialisation, underscoring the need for adapted legal frameworks [15]. These broader developments provide important context for advancing sub-zero bioprinting strategies aimed at engineering vascular grafts with biomimetic mechanical behaviour.

Sub-zero 3D bioprinting, of PVA-C, has emerged as a promising strategy to fabricate layered constructs with controlled internal architectures [21]. Unlike conventional fabrication techniques, layer-by-layer deposition allows for the accurate spatial organisation of material, enabling controlled anisotropic mechanical behaviour that more closely resembles the natural structure of arterial tissues [18]. However, prior research has largely focused on layer thickness and FTCs, with limited studies addressing the effect of multi-layered architectures.

Beyond anisotropic properties, arterial tissues also exhibit viscoelastic and hyperelastic behaviour, meaning they can store, dissipate, and recover mechanical energy under cyclic and large-strain deformation. Viscoelasticity is essential for accommodating dynamic blood flow and reducing mechanical fatigue over time. The storage modulus (E′) quantifies the material’s ability to store elastic energy and resist deformation under load, directly correlating with structural integrity. The loss modulus (E″) quantifies the material’s viscous behaviour, reflecting the material’s damping capability and ability to absorb mechanical stress. The ratio of these parameters, expressed as tan δ (E″/E′), indicates the damping efficiency of the material. A higher tan δ suggests greater damping, meaning more energy is lost as heat, whereas a lower value reflects a predominantly elastic response with efficient energy recovery. Materials with frequency-dependent moduli exhibit stronger viscoelastic effects, whereas frequency-independent behaviour suggests a dominant elastic response [14,15,19,22].

Hyperelasticity describes materials that undergo significant deformation but return to their original shape without permanent damage, a characteristic essential for soft biological tissues such as arteries [10]. The stress–strain response of hyperelastic materials is nonlinear, requiring constitutive models beyond simple linear elasticity. In previous studies, PVA-C displayed hyperelastic properties due to its crosslinked polymeric network, making it a promising candidate for biomimetic vascular grafts [23].

The aim of this study is to investigate the influence of increasing the number of layers on the elastic response, and the hyperelastic and viscoelastic behaviour of PVA-C samples under a uniaxial load. By assessing stiffness, frequency-dependent mechanical responses, and nonlinear stress–strain behaviour, this work provides insight into the role of layered architectures in 3D-printed cryogels towards the development of functionally graded PVA-C structures for vascular applications.

2. Materials and Methods

2.1. Overview

The bioprinting of hydrogels is well acknowledged to require a large amount of process parameter optimisation in order to obtain a balance between the printability and structural integrity of the construct [24]. The parameters used in this study, for example, PVA concentration and flow rate, were fixed after a large amount of empirical testing by isolating each material and process parameter and exploring the impact of parametric variation on the process. Further to this, the environmental parameters, which, in this study, were not controlled directly, were monitored to assess their impact.

2.2. Biomaterial Ink Preparation

A solution of 11% w/w PVA (146–186 kDa, 99+% hydrolysis, Sigma-Aldrich, St. Louis, MO, USA) was dissolved in deionised water by autoclaving for 1 h at 121 °C, followed by mechanical stirring at 50 °C for 1 h, with additional stirring at room temperature (RT) (22.5 ± 1 °C) for 1 h to ensure complete dissolution and homogenisation. The 11% concentration was selected for its balance between mechanical performance and printability. Autoclaving was specifically incorporated to facilitate dissolution while ensuring that the process can be translated into a sterile manufacturing protocol for future biomedical applications, where sterility is a critical requirement [25].

2.3. Sample Design and Fabrication

Rectangular samples with nominal design dimensions of 12.3 mm × 10 mm and thicknesses of 1, 2, and 3 mm were created using Fusion 360 (Autodesk, San Francisco, CA, USA) and converted to G-code using REGEMAT 3D Designer slicing software (REGEMAT 3D S.L., Granada, Spain). These dimensions were determined based on computer-aided design (CAD) specifications. The 3D input model used for the fabrication process is provided in Appendix A.2 (Figure A1) to illustrate the base CAD geometry. However, to account for potential deviations due to fabrication and processing, the actual dimensions of the samples were measured after the completion of the FTCs and prior to mechanical testing.

A 0–90° infill pattern was implemented to enhance structural stability and ensure uniform mechanical properties across the sample cross-section. Samples were fabricated using a Regemat BIO V1 bioprinter (REGEMAT 3D S.L., Granada, Spain) equipped with a 0.58 mm nozzle. Printing parameters—including flow rate, infill speed, and print bed temperature—were optimised empirically to accommodate the viscosity of the PVA solution. The flow rate was set to 0.6 mm/s, with an infill speed of 2 mm/s. During fabrication, the print bed was maintained at approximately −2.5 ± 1.5 °C, with temperature actively monitored throughout the process. Full details of the monitoring setup and environmental controls are described in Section 2.3, while the recorded temperature profiles during printing are reported in Section 3.6.

Following fabrication, the samples underwent three freeze–thaw cycles (FTCs) alternating between −20 °C and RT (~22 °C), based on protocols established to promote mechanical stability and flexibility [26,27,28]. Each cycle consisted of freezing at −20 °C for 16 h followed by thawing at RT for 8 h to induce physical crosslinking. Post printing, the samples were immersed in deionised water for four days to achieve full swelling prior to mechanical testing. After swelling, final sample dimensions were recorded; width and thickness were measured directly, whereas the length was not included, as tensile testing was performed based on a fixed gauge length of 10 mm (distance between clamps). Full dimensional data are provided in Appendix A.3.

Samples were produced with 2, 4, and 6 alternating layers, with toolpaths alternating between parallel and perpendicular orientations. Each sample type was fabricated in quadruplicate (n = 4).

Table 1. Summary of manufactured sample types.

Sample Type Name	Sample Numbers	Layers	Dimensions (mm3)
2L-3FTC	2L-1, 2L-2, 2L-3, 2L-4	2	12.3 × 10 × 1
4L-3FTC	4L-1, 4L-2, 4L-3, 4L-4	4	12.3 × 10 × 2
6L-3FTC	6L-1, 6L-2, 6L-3, 6L-4	6	12.3 × 10 × 3

provides a summary of the sample types and corresponding identifiers used in this study.

2.4. Dimensional Fidelity Analysis

Prior to the mechanical characterisation, the reproducibility of the printed constructs was assessed; the actual width and thickness of all samples were measured and compared against the CAD specifications for each sample type. The target digital dimensions were 12.3 mm (width) × 1 mm (thickness) for 2L samples, 12.3 mm × 2 mm for 4L samples, and 12.3 mm × 3 mm for 6L samples (Appendix A.2, Figure A1). The average dimensional deviations, calculated from four printed samples per group (Appendix A.3,

Table A2. Width and height (in mm) of PVA-C samples with two, four, and six layers.

Sample	Width (mm)	Height
2L-1	12	0.73
2L-2	11.79	0.83
2L-3	12.49	0.66
2L-4	11.71	0.91
4L-1	12.28	1.8
4L-2	11.6	1.79
4L-3	11.26	1.79
4L-4	12.08	1.76
6L-1	12.41	2.36
6L-2	12.01	2.38
6L-3	11.65	2.44
6L-4	11.53	2.62

), were as follows:

2L samples: width deviation = 1.6 ± 2.8 %, thickness deviation = −23.4 ± 11.5 %;

4L samples: width deviation = −1.4 ± 3.3 %, thickness deviation = −10.5 ± 1.6 %;

6L samples: width deviation = −5.9 ± 3.2 %, thickness deviation = −18.5 ± 3.4 %.

While deviations in thickness were more pronounced—likely due to material spreading and layer compaction—width measurements showed acceptable reproducibility, remaining within ±6% of the target. These findings indicate that, despite the thermal and structural challenges of sub-zero printing, dimensional fidelity in the printing plane was retained across constructs.

2.5. Sample and Environmental Monitoring

The ambient environmental conditions during fabrication were carefully monitored to ensure consistency. The printing bed temperature was tracked using a LabJack T7 thermocouple (Labjack, Lakewood, CO, USA), while ambient temperature and humidity were recorded to account for potential variations in the fabrication environment.

To monitor any visible changes in the structure during fabrication and testing, samples were imaged at three stages: post bioprinting (within 15 min), post FTCs, and during mechanical testing at 40% strain. A Dino-Lite Digital Microscope (Dino-Lite, Torrance, CA, USA) was used to capture high-resolution images, enabling the visual analysis of the structural integrity and deformation patterns of the samples.

Samples were weighed before and after each dynamic mechanical analysis (DMA) test repetition to quantify dehydration effects, as these tests were significantly longer in duration than the ramp tests and, thus, represent a more significant risk of dehydration. The percentage weight change was calculated relative to the initial sample weight to assess hydration loss during testing. Measurements were conducted for the first two DMA test repetitions, and, as weight loss remained consistent between them, further measurements were not performed for the third repetition.

2.6. Mechanical Testing and Data Processing

2.6.1. Uniaxial Tensile Testing

Tensile tests were performed using a Bose ElectroForce 3200 testing system (TA Instruments, New Castle, DE, USA) equipped with a 5 N load cell, with data acquisition managed via WinTest software (TA Instruments, New Castle, DE, USA). Samples were clamped at both ends with a gauge length of 10 mm and subjected to uniaxial ramp testing up to a maximum strain of 40% at a constant displacement rate of 0.25 mm/s. Load–displacement data were recorded and converted to stress–strain curves using standard engineering definitions.

Uniaxial tensile tests were performed on samples with a total length of 30 mm (10 mm gauge length and 10 mm per clamping region) and a width of 12.3 mm. Each sample was subjected to five loading cycles; the first cycle was used for preconditioning and was excluded from subsequent analysis. The average and standard deviation (SD) were calculated from the final four repetitions (T2–T5). Environmental testing conditions were not actively controlled; however, all experiments were conducted under nominal ambient laboratory conditions (approximately 21–23 °C and 40–50% relative humidity).

2.6.2. Stress–Strain Data Processing and Young’s Modulus Calculation

Young’s modulus (YM) was calculated at predefined strain levels (10%, 20%, 30%, and 40%) using a custom MATLAB R2023a (The MathWorks Inc., Natick, MA, USA) script, which incorporated error estimation for stress and strain values. A point-wise approach was applied within ±0.5% strain of each target level to determine YM. The mean, SD, and propagated error were computed for each test repetition to assess variability and reliability.

Error propagation was employed to quantify measurement uncertainties in the stress–strain data at each recorded point. Standard error propagation techniques were applied, incorporating uncertainties from multiple sources. The force measurement uncertainty of ±0.025 N was derived from the specifications of the load cell, while the displacement uncertainty of ±0.002 mm was based on the mechanical testing machine specifications. Additionally, the cross-sectional area uncertainty was determined from the calliper precision (±0.02 mm).

Linear and quadratic regressions were applied to the stress–strain data to determine the most accurate representation of the material’s mechanical response. The linear fit used the equation σ = aε + b, while the quadratic fit used σ = cε² + dε + e, where σ is the engineering stress, ε is the engineering strain, and a–e are fitting parameters. The best-fitting model was selected based on the highest coefficient of determination (R²) and minimal residuals. Details of the R² values and regression analysis are provided in Appendix B.2.

2.6.3. Dynamic Mechanical Analysis

DMA was performed to assess the viscoelastic behaviour of PVA-C samples by measuring the storage modulus (E′) and loss modulus (E″) across a physiologically relevant frequency range. These parameters characterise the material’s ability to store and dissipate mechanical energy. Additionally, the complex modulus (E*) was calculated as follows:

E* = sqrt (E′² + E″²),

(1)

where E* represents the total mechanical stiffness, combining both the elastic and viscous contributions. The phase angle (δ), which is derived from the ratio of E″ to E′, indicates the degree of viscoelasticity, with larger δ values representing increased damping.

The loss tangent (tan δ), also referred to as the damping factor, was calculated as follows:

tan δ = E″/E′,

(2)

where tan δ provides insight into the balance between elastic and viscous behaviour. A lower tan δ suggests a more elastic response, whereas a higher tan δ indicates increased energy dissipation.

DMA tests were conducted at RT under controlled hydration conditions. Samples were subjected to sinusoidal displacement at frequencies ranging from 1 to 10 Hz, simulating cyclic deformations experienced by arterial tissues [29]. The storage modulus (E′) represents the elastic component of the response, while the loss modulus (E″) reflects the viscous contribution. The testing parameters, including frequency, mean displacement levels, and dwell times, are summarised in

Table A3. Testing condition file parameters for DMA of samples.

Condition Number	Frequency (Hz)	Mean Level (mm)	Mean Level Dwell (sec)	Dynamic Amplitude (mm)	Hold Value (mm)	Hold Value Dwell (sec)
1	1	3	2	0.4	0	0
2	2	3	2	0.4	0	0
3	3	3	2	0.4	0	0
4	4	3	2	0.4	0	0
5	6	3	2	0.4	0	0
6	8	3	2	0.4	0	0
7	10	3	2	0.4	0	0
8	1	3	2	0.4	0	0

in Appendix B.1.

Each sample underwent three repeated tests on different days. Between tests, samples were rehydrated in deionised water for 24 h to maintain consistent swelling behaviour and mechanical properties. Lower frequencies, such as 0.5 Hz, were excluded to minimise dehydration effects. After completing the frequency sweep, samples were returned to the initial frequency of 1 Hz to assess any changes in mechanical response over the course of testing.2.7. Data Analysis

All experimental data were processed using MATLAB and Microsoft Excel (Microsoft, Redmond, WA, USA). Engineering stress and strain, along with dynamic mechanical properties, were computed using the equations summarised in

Table 2. Summary of mechanical equations used for data analysis.

Equation No.	Formula	Description	Variables
(3)	σ = F/A	Engineering stress	σ: stress (Pa); F: force (N); A: cross-sectional area (mm2)
(4)	A = w × T	Cross-sectional area	w: width (mm); T: thickness (mm)
(5)	ε = d/L	Engineering strain	ε: strain; d: displacement (mm); L: gauge length (mm)
(6)	E = σ/ε	Young’s modulus	E: modulus (Pa); σ: stress (Pa); ε: strain
(7)	K* = F/d	Dynamic stiffness	K*: dynamic stiffness (N/mm); F: force (N); d: displacement (mm)
(8)	K′ = K* × cos(δ)	Storage stiffness	K′: storage stiffness (N/mm); δ: phase angle (radians)
(9)	K″ = K* × sin(δ)	Loss stiffness	K″: loss stiffness (N/mm); δ: phase angle (radians)
(10)	S = (w × T)/h	Shape factor	S: shape factor; h: sample height (mm)
(11)	E′ = K′/S	Storage modulus	E′: storage modulus (Pa)
(12)	E″ = K″/S	Loss modulus	E″: loss modulus (Pa)

. These include calculations for stress, strain, dynamic stiffness, storage and loss stiffness, shape factor, and the final derivation of storage (E′) and loss (E″) moduli.

Force and displacement data from uniaxial tensile testing were used to calculate stress and strain using Equations (3)–(5). Stiffness values derived from DMA were processed via Fourier analysis to determine phase lag (δ), from which storage and loss stiffness values were obtained using Equations (8)–(10). These were normalised using the shape factor (Equation (10)) to compute E′ and E″ (Equations (11) and (12)).

Statistical analysis was conducted to evaluate intra- and inter-sample variability, with data presented as mean ± SD. A one-way analysis of variance (ANOVA) was performed to determine whether significant differences existed among sample groups. ANOVA assesses whether the means of multiple groups are statistically distinct by analysing the variance within and between groups [30]. The significance of observed differences was evaluated using p-values, which quantify the evidence against the null hypothesis; smaller p-values indicate stronger evidence that the groups are different [31]. If significant differences were detected (p < 0.05), Tukey’s post-hoc test was applied to conduct pairwise comparisons. Tukey’s test accounts for multiple comparisons, allowing for a more reliable identification of significant differences among sample groups [32].

To assess the effect of layer number on viscoelastic properties, ANOVA and Tukey’s post-hoc tests were conducted at three distinct frequencies: 2 Hz, 6 Hz, and 10 Hz. This approach allowed for us to evaluate whether differences in storage modulus (E′) and loss modulus (E″) were frequency-dependent and whether the mechanical response varied significantly between the 2L, 4L, and 6L samples at these key frequencies.

3. Results

3.1. Sample Imaging

A representative example of Sample 4L-3 is shown in Figure 1. Figure 1a shows the sample 15 min post bioprinting, with a uniform grid-like structure and well-defined printed filaments. Figure 1b presents the sample after three FTCs, where an increase in opacity and a slight reduction in pore size can be observed. Figure 1c captures the sample at maximum strain (40%) during tensile testing. These images illustrate the structural changes in the samples throughout the fabrication and mechanical testing processes. Images of all samples at each stage are provided in Appendix A.1.

Despite reaching 40% strain, no visible filament breakage or detachment was observed, suggesting that the material maintained structural continuity under tensile loading. Differences in filament separation and structural uniformity between sample types were evident across all three stages, highlighting the influence of layer number on mechanical integrity.

3.2. Stress–Strain Curves and Curve Fitting

Force–displacement data from the ramp tests were converted into stress–strain curves, with YM at a specific strain used for comparisons between different samples. Figure 2 shows a representative uniaxial tension response up to 40% strain for Sample 2L-1. The experimental data include the propagated error and a linear (σ = aε + b) and quadratic (σ = cε² + dε + e) line of best fit. The quadratic model demonstrates a superior fit based on the coefficient of determination (R²) value.

The R² was used to evaluate the goodness of fit, with values consistently higher for the quadratic model as opposed to the linear model.

Table A4. Coefficient of determination (R²) values for linear and quadratic fits across all samples and test repetitions.

Sample	Repetition	R2 (Linear)	R2 (Quadratic)
2L-1	T2	0.9947	0.9997
2L-1	T3	0.9948	0.9997
2L-1	T4	0.9957	0.9997
2L-1	T5	0.9964	0.9996
2L-2	T2	0.9962	0.9994
2L-2	T3	0.9961	0.9996
2L-2	T4	0.9962	0.9997
2L-2	T5	0.9962	0.9997
2L-3	T2	0.9974	0.9995
2L-3	T3	0.9965	0.9997
2L-3	T4	0.9963	0.9997
2L-3	T5	0.9965	0.9997
2L-4	T2	0.9918	0.9995
2L-4	T3	0.9908	0.9996
2L-4	T4	0.9899	0.9996
2L-4	T5	0.9897	0.9996
4L-1	T2	0.9995	0.9996
4L-1	T3	0.9976	0.9997
4L-1	T4	0.9945	0.9998
4L-1	T5	0.9937	0.9998
4L-2	T2	0.9973	0.9995
4L-2	T3	0.9967	0.9997
4L-2	T4	0.9964	0.9997
4L-2	T5	0.9963	0.9997
4L-3	T2	0.9919	0.9997
4L-3	T3	0.9906	0.9997
4L-3	T4	0.9903	0.9998
4L-3	T5	0.9903	0.9998
4L-4	T2	0.9981	0.9995
4L-4	T3	0.9971	0.9998
4L-4	T4	0.9968	0.9998
4L-4	T5	0.9972	0.9998
6L-1	T2	0.9985	0.9997
6L-1	T3	0.9981	0.9997
6L-1	T4	0.9984	0.9997
6L-1	T5	0.9985	0.9997
6L-2	T2	0.9972	0.9995
6L-2	T3	0.9951	0.9998
6L-2	T4	0.9957	0.9998
6L-2	T5	0.9960	0.9998
6L-3	T2	0.9987	0.9994
6L-3	T3	0.9984	0.9996
6L-3	T4	0.9989	0.9996
6L-3	T5	0.9989	0.9996
6L-4	T2	0.9942	0.9997
6L-4	T3	0.9940	0.9998
6L-4	T4	0.9946	0.9998
6L-4	T5	0.9943	0.9998

in Appendix B.2 presents the R² values for both fitting methods across all samples and test repetitions. All quadratic R² values exceeded 0.999, while the values for the linear fits were slightly lower.

3.3. Young’s Moduli

YM were extracted at 10%, 20%, 30%, and 40% strain. The average YM values, calculated across four test repetitions for each sample type (2L, 4L, and 6L), are shown in Figure 3, Figure 4 and Figure 5. The values remained consistent across repetitions, as demonstrated by the low SD values, indicating low intra-sample variability. However, significant inter-sample variability was observed across all sample designs.

For each strain level, average YM values and their corresponding SDs for all samples and repetitions are shown in Figure 6, with numerical values provided in

Table A5. Young’s modulus and standard deviation across strain levels for 2L, 4L, and 6L samples.

Strain Levels (%)	Mean Young’s Modulus, 2L (kPa)	Standard Deviation, 2L (kPa)	Mean Young’s Modulus, 4L (kPa)	Standard Deviation, 4L (kPa)	Mean Young’s Modulus, 6L (kPa)	Standard Deviation, 6L (kPa)
10	47.52	8.07	39.61	5.46	34.75	6.93
20	48.62	7.54	40.75	6.23	35.26	7.53
30	51.27	7.50	42.51	7.55	36.72	8.21
40	54.99	7.47	44.99	8.94	38.75	8.86

(Appendix B.3). The analysis reveals an increase in stiffness with strain, with higher modulus values observed at 40% strain compared to 10%, supporting the evidence for a nonlinear (as opposed to linear) mechanical response.

To assess whether layer number significantly influenced YM at different strain levels, a one-way ANOVA was conducted for each strain level (10%, 20%, 30%, and 40%). The analysis revealed statistically significant differences in YM across the sample types (2L, 4L, and 6L) at all strain levels. The full ANOVA results are provided in

Table A7. Results from one-way ANOVA tests (95% confidence level) investigating inter-sample variability for YM at different strain levels.

Strain Level (%)	p-Value	Statistical Difference Between Sample Means
Inter-Sample Variability of Two-Layered (2L) Samples
10	0.907	p > 0.05, no statistical difference
20	0.884	p > 0.05, no statistical difference
30	0.075	p > 0.05, no statistical difference
40	0.143	p > 0.05, no statistical difference
Inter-Sample Variability of Four-Layered (4L) Samples
10	0.018	p < 0.05, Sample 4L-2 different from 4L-1 and 4L-3
20	0.015	p < 0.05, Sample 4L-2 different from 4L-1 and 4L-3
30	0.137	p > 0.05, no statistical difference
40	0.029	p < 0.05, Sample 4L-2 different from 4L-1
Inter-Sample Variability of Six-Layered (6L) Samples
10	0.008	p < 0.05, Sample 6L-1 different from 6L-4
20	0.008	p < 0.05, Sample 6L-1 different from 6L-4
30	0.069	p > 0.05, no statistical difference
40	0.147	p > 0.05, no statistical difference

and

Table A8. Results from one-way ANOVA tests (95% confidence level) investigating data linearity by checking for statistical differences between YM values at different strain levels (10%, 20%, 30%, 40%).

Sample Type	p-Value	Statistical Difference
2L	p < 0.001	p < 0.05, significant differences between strain levels
4L	p = 0.0039	p < 0.05, significant differences between strain levels
6L	p = 0.0805	p > 0.05, no significant differences between strain levels

(Appendix C.1). Following this, Tukey’s post-hoc test was applied to determine which sample types differed significantly. At 40% strain, 2L samples were significantly stiffer than both 4L and 6L samples, while no significant difference was found between 4L and 6L. The complete Tukey comparison results across all strain levels are provided in

Table A9. Post-hoc one-way ANOVA results comparing YM at individual strain levels (10%, 20%, 30%, 40%).

Sample Type	p-Value	Statistical Difference
2L	<0.001	p < 0.05, 40% strain different from 10%, 20%, 30%
4L	<0.001	p < 0.05, 40% strain different from 10%, 20%
6L	0.798	p > 0.05, no statistical difference

and

Table A10. Tukey’s multiple comparison test results for YM at 40% strain.

Comparison	Lower Bound	Mean Difference (Estimate)	Upper Bound	p-Value	Significance
2L vs. 4L	2.50	8.63	14.77	0.0039	Significant
2L vs. 6L	8.09	14.23	20.36	3.38 × 10⁻6	Highly Significant
4L vs. 6L	−0.54	5.59	11.73	0.0805	Not Significant

(Appendix C).

Within each sample group, some inter-sample consistency was observed. In the 2L group, samples 2L-1 and 2L-3 exhibited similar behaviour, as did samples 2L-2 and 2L-4. In the 4L series, 4L-2 and 4L-3 responded similarly, while, in the 6L series, samples 6L-3 and 6L-4 demonstrated a comparable mechanical response. Although slight variations were observed between individual samples, none were classified as anomalous.

3.4. Dynamic Mechanical Analysis Results

The average storage moduli (E′) and loss moduli (E″) within the range of 1–10 Hz for all of the 2L, 4L, and 6L samples are shown in Figure 7 and Figure 8. The storage and loss moduli for each sample, for all repetitions, are shown in Appendix B.4,

Table A6. Individual storage (E′) and loss modulus (E″) values for each sample across 1–10 Hz.

. Figure 7 and Figure 8 show the trends of a decrease in storage moduli and increase in loss moduli with increasing frequency. However, the SDs of the data are large enough to suggest that these trends are not significant. Looking at each sample in isolation, in Appendix B.4, it can be shown that there are large differences in the average moduli between samples, creating very large SD in the combined data sets overall.

Comparing Figure 7 and Figure 8, the storage modulus is consistently higher than the loss modulus across all sample designs and frequencies. At 2 Hz, the E′ increased with the number of layers: the two-layered (2L) samples exhibited the lowest (~8422 Pa), followed by the four-layered (4L) samples (~11,343 Pa) and the six-layered (6L) samples (~10,813 Pa). Similar trends were observed at 6 Hz and 10 Hz, though the differences were less pronounced at higher frequencies.

Statistical analysis confirmed these observations. One-way ANOVA at 2 Hz, 6 Hz, and 10 Hz indicated significant differences between sample types for E′ (Appendix C.2,

Table A11. Results from one-way ANOVA tests (95% confidence level) for storage modulus at 2 Hz, 6 Hz, and 10 Hz.

Sample Type	Frequency (Hz)	p-Value	Statistical Difference
2L	2 Hz	0.002	p < 0.05, significant differences
4L	2 Hz	0.003	p < 0.05, significant differences
6L	2 Hz	0.005	p < 0.05, significant differences
2L	6 Hz	0.002	p < 0.05, significant differences
4L	6 Hz	0.003	p < 0.05, significant differences
6L	6 Hz	0.005	p < 0.05, significant differences
2L	10 Hz	0.002	p < 0.05, significant differences
4L	10 Hz	0.003	p < 0.05, significant differences
6L	10 Hz	0.005	p < 0.05, significant differences

). Post-hoc Tukey’s Honestly Significant Difference (HSD) tests showed that the 2L samples were significantly different from both 4L and 6L samples at all frequencies, while the difference between 4L and 6L was not significant (Appendix C.2, Table A10). This suggests that increasing the number of layers from two to four significantly increases the modulus, but adding more layers beyond four does not lead to a further significant increase in E′.

For the loss modulus, ANOVA results showed significant differences at 2 Hz and 10 Hz (Appendix B,

Table A12. Results from one-way ANOVA tests (95% confidence level) for loss modulus at 2 Hz, 6 Hz, and 10 Hz.

Sample Type	Frequency (Hz)	p-Value	Statistical Difference
2L	2 Hz	0.002	p < 0.05, significant differences
4L	2 Hz	0.042	p < 0.05, significant differences
6L	2 Hz	0.300	p > 0.05, no statistical difference
2L	6 Hz	0.002	p < 0.05, significant differences
4L	6 Hz	0.042	p < 0.05, significant differences
6L	6 Hz	0.300	p > 0.05, no statistical difference
2L	10 Hz	0.002	p < 0.05, significant differences
4L	10 Hz	0.042	p < 0.05, significant differences
6L	10 Hz	0.300	p > 0.05, no statistical difference

), but not at 6 Hz. Tukey’s HSD test revealed that, at 2 Hz and 10 Hz, the 2L samples were significantly different from both the 4L and 6L samples (p < 0.05), whereas the difference between 4L and 6L remained statistically insignificant. This suggests that layering affects energy dissipation at lower and higher frequencies but has a diminished effect at intermediate frequencies. At the end of each test, when the frequency was returned to 1 Hz, the storage and loss modulus values remained within the initial measurement range, with differences falling within the SD.

3.5. Dehydration Analysis

The percentage weight change was calculated to quantify dehydration, with observed losses ranging from −8.21% to −21.63% (outlined in full in

Table A19. Sample weight changes before and after DMA testing.

Sample Number	Weight Before DMA (g) (Day 1, 1st DMA Test)	Weight After DMA (g) (Day 1, 1st DMA Test)	% Change (Day 1)	Weight Before DMA (g) (Day 2, 2nd DMA Test)	Weight After DMA (g) (Day 2, 2nd DMA Test)	% Change (Day 2)
2L-1	0.103	0.081	−21.63%	0.103	0.080	−22.08%
2L-2	0.131	0.108	−17.94%	0.126	0.103	−18.00%
2L-3	0.125	0.101	−18.88%	0.122	0.103	−15.53%
2L-4	0.139	0.114	−17.73%	0.134	0.114	−15.35%
4L-1	0.262	0.230	−12.28%	0.259	0.230	−11.23%
4L-2	0.267	0.245	−8.38%	0.264	0.236	−10.45%
4L-3	0.264	0.234	−11.18%	0.255	0.227	−11.10%
4L-4	0.274	0.246	−10.45%	0.279	0.237	−14.79%
6L-1	0.408	0.371	−8.99%	0.414	0.378	−8.51%
6L-2	0.408	0.374	−8.21%	0.413	0.377	−8.69%
6L-3	0.422	0.371	−12.08%	0.432	0.389	−10.00%
6L-4	0.419	0.378	−9.67%	0.416	0.377	−9.37%

, Appendix D.1). The average weight change varied across sample types and between test days. Pre-test weight measurements on the second testing day were consistent with those recorded before the first DMA test, indicating that samples had sufficient time to reabsorb water and return to their swollen state between test repetitions.

Table 3. Mean and standard deviation of dehydration percentage for each sample type across test days.

Sample Type	Mean % Dehydration (Day 1)	SD % Dehydration (Day 1)	Mean % Dehydration (Day 2)	SD % Dehydration (Day 2)
2L	−19.05	1.79	−17.74	3.14
4L	−10.57	1.64	−11.89	1.96
6L	−9.74	1.67	−9.14	0.68

summarises the mean percentage dehydration and SD for each sample configuration on Day 1 and Day 2 of testing.

The results highlight that 2L samples exhibited the highest dehydration levels, while 6L samples retained the most water, suggesting that increased layering may enable water retention.

3.6. Environmental Monitoring

The printing bed temperature, laboratory ambient temperature, and humidity were recorded before and during each print to assess environmental stability throughout the fabrication process. The laboratory ambient temperature remained stable at 23 ± 1 °C, while humidity was maintained at 40 ± 1%, indicating minimal fluctuations in environmental conditions.

The printing bed temperature was monitored at a time step of 0.5 s, capturing temperature variations throughout the printing process. The recorded environmental conditions for each sample configuration were as follows:

• Two-layered (2L) samples: Printing bed temperature: −1.84 ± 1.49 °C;
• Four-layered (4L) samples: Printing bed temperature: −2.95 ± 1.24 °C;
• Six-layered (6L) samples: Printing bed temperature: −2.81 ± 0.92 °C.

4. Discussion

4.1. Elastic Response

The YM values calculated at 10%, 20%, 30%, and 40% strain showed a decreasing trend with increasing layer number (Figure 6). The 2L samples exhibited the highest YM values across all strain levels, averaging 45.7 ± 3.2 kPa at 40% strain, while 4L and 6L samples measured 38.2 ± 2.9 kPa and 34.1 ± 2.4 kPa, respectively.

Statistical analysis using one-way ANOVA confirmed that YM values differed significantly between sample types at all strain levels (p < 0.05) for the 2L and 4L samples. Tukey’s post-hoc test further revealed that 2L samples had significantly higher YM values than both 4L (Table A8, Appendix C.1) and 6L samples at 40% strain, while no significant difference was found between 4L and 6L. These results indicate that layering impacts stiffness, but beyond four layers, additional layering does not further reduce YM in a statistically meaningful way.

The low intra-sample variability (i.e., across repeated tests on the same sample) demonstrates the reproducibility of the characterisation methodology, with the material response being predominantly elastic after the preconditioning cycle. However, high inter-sample variability was observed, particularly in the 2L and 4L samples, despite the extensive empirical testing used to optimise the fabrication process. This suggests that fabrication-related factors, such as variations in printing bed temperature, polymer deposition accuracy, and localised hydration differences, may contribute to mechanical performance variability between samples.

Visual inspection revealed that 2L samples exhibited the least uniform structure, with non-homogeneous polymer distribution and localised drying effects (

Table A1. Samples imaging post printing, after FTCs, at 40% (maximum) strain.

2L-1			2L-2
2L-3			2L-4
4L-1			4L-2
4L-3			4L-4
6L-1			6L-2
6L-3			6L-4

, Appendix A.1). During tensile testing, one sample exhibited visible filament breakage (Sample 2L-4), suggesting weak points which could be associated with dehydration and polymer redistribution. In contrast, 6L samples exhibited minimal filament separation, better hydration retention, and greater structural integrity under an applied load (Table A1, Appendix A). These imaging results confirm that layer number influences both uniformity and mechanical performance, with thinner constructs (2L) exhibiting greater structural inconsistencies, while thicker samples (6L) demonstrate improved stability under mechanical loading.

Given the statistically significant reduction in YM with increasing layer number, it is likely that mechanical decoupling at layer interfaces and polymer crosslinking variations contribute to the observed stiffness reduction. This could be due to the increase in the distance from the print bed and, therefore, the increase in the printing temperature as the number of layers increases (discussed further in Section 4.6). However, the lack of a significant difference between 4L and 6L suggests a saturation effect, where, beyond a certain layering threshold, additional layers do not further weaken the construct.

These findings suggest that increased layering mitigates dehydration effects and improves structural homogeneity but reduces overall stiffness. Further investigations should explore whether modifications in fabrication conditions, such as refining deposition accuracy or implementing controlled printing environments, could enhance consistency in mechanical properties and polymer crosslinking.

4.2. Dynamic Response

The DMA results indicate that layering influences the storage moduli (E′) and loss moduli (E″), but that the samples do not exhibit significant frequency dependence. While subtle trends were observed across frequencies, statistical analysis confirmed that the differences fell within the SD, reinforcing the frequency-independent nature of these samples.

Layering had a clear effect on mechanical properties, with 2L samples exhibiting significantly lower E′ values compared to 4L and 6L samples at 2 Hz and 10 Hz (p < 0.05, Appendix C.2,

Table A13. Results from Tukey’s HSD storage modulus at 2 Hz.

Comparison	Mean Difference	p-Value (Adjusted)	Lower Bound	Upper Bound	Statistical Significance
2L vs. 4L	4950.1	0.016	1032.38	8867.819	Significant
2L vs. 6L	6854.267	0.002	2936.547	10,771.99	Significant
4L vs. 6L	1904.168	0.402	−2013.55	5821.887	Not Significant

,

Table A14. Results from Tukey’s HSD storage modulus at 6 Hz.

Comparison	Mean Difference	p-Value (Adjusted)	Lower Bound	Upper Bound	Statistical Significance
2L vs. 4L	4368.505	0.018	832.893	7904.117	Significant
2L vs. 6L	6011.1	0.0027	2475.488	9546.712	Significant
4L vs. 6L	1642.595	0.4314	−1893.017	5178.207	Not Significant

and

Table A15. Results from Tukey’s HSD storage modulus at 10 Hz.

Comparison	Mean Difference	p-Value (Adjusted)	Lower Bound	Upper Bound	Statistical Significance
2L vs. 4L	4025.866	0.025	560.388	7491.345	Significant
2L vs. 6L	5454.201	0.004	1988.722	8919.68	Significant
4L vs. 6L	1428.335	0.509	−2037.144	4893.814	Not Significant

). However, increasing the number of layers beyond four did not result in a statistically significant increase in stiffness. This may suggest that additional layering reaches a saturation point where further gains in mechanical reinforcement are minimal. A similar pattern was observed for E″, where 2L samples showed significantly different values from 4L and 6L at 2 Hz and 10 Hz, but not at 6 Hz (p < 0.05, Appendix C.2,

Table A16. Results from Tukey’s HSD loss modulus at 2 Hz.

Comparison	Mean Difference	p-Value (Adjusted)	Lower Bound	Upper Bound	Statistical Significance
2L vs. 4L	175.4693	0.0197	31.0644	319.8743	Significant
2L vs. 6L	270.3606	0.0014	125.9556	414.7656	Significant
4L vs. 6L	94.8912	0.2131	−49.5138	239.2962	Not Significant

,

Table A17. Results from Tukey’s HSD loss modulus at 6 Hz.

Comparison	Mean Difference	p-Value (Adjusted)	Lower Bound	Upper Bound	Statistical Significance
2L vs. 4L	133.634	0.262	−87.581	354.848	Not Significant
2L vs. 6L	239.381	0.035	18.167	460.596	Significant
4L vs. 6L	105.748	0.413	−115.467	326.963	Not Significant

and

Table A18. Results from Tukey’s HSD loss modulus at 10 Hz.

Comparison	Mean Difference	p-Value (Adjusted)	Lower Bound	Upper Bound	Statistical Significance
2L vs. 4L	107.092	0.716	−269.422	483.606	Not Significant
2L vs. 6L	223.610	0.272	−152.815	600.214	Not Significant
4L vs. 6L	116.608	0.675	−259.907	493.122	Not Significant

). This suggests that layering affects energy dissipation at lower and higher frequencies, but the effect diminishes at intermediate frequencies.

Despite minor fluctuations, Figure 7 and Figure 8 further illustrate this trend, showing that loss modulus (E″) remains substantially lower than storage modulus (E′) across all configurations, confirming their predominantly elastic response. This predominance of elastic behaviour is consistent with the rheological profile expected from shear-thinning materials such as PVA-based hydrogels, where the flow behaviour index is typically <1, reflecting non-Newtonian behaviour under shear [33].

Whilst a trend between frequency and the storage and loss moduli was observed, it was not statistically significant due to the high level of variability, particularly for inter-sample differences. Similarly, the results presented for all samples in Appendix B.4 did not show a distinct trend beyond the inherent variability between sample repeats. This contrasts with native arterial tissues, which exhibit pronounced frequency-dependent viscoelastic properties due to their complex hierarchical structure and active cellular components [29]. The absence of significant frequency dependence in the bioprinted PVA-C samples suggests that, while they provide a stable mechanical response under oscillatory loading, they do not fully replicate the dynamic behaviour of arterial tissues, which progressively stiffen at higher frequencies to accommodate pulsatile blood flow.

While individual 2L, 4L, and 6L samples exhibited frequency-independent behaviour, their differing absolute responses suggest the potential for functionally graded materials (FGMs). By designing alternative spatial distributions of the PVA-C toolpath, normalised for the continuity of thickness, for example, embedded in another material, or dual 3D printing, it may be possible to engineer constructs that enable a change in storage and loss modulus through space. This could enable the tuning of mechanical gradients to enhance biomimetic performance and optimise vascular graft designs for improved physiological compatibility.

4.3. Nonlinear Response

To assess any nonlinearity in the mechanical response, linear and quadratic data fits were undertaken. As shown in Figure 2, while the linear model provided a reasonable fit across most of the strain range, it deviated at both lower and higher strain levels, falling outside the error bars. In contrast, the quadratic model offered a better representation of the stress–strain response (Appendix B.2, Table A4). The improvement in R² was minimal, however, and so a hyperelastic framework such as Mooney–Rivlin or Ogden is not required within the tested strain rate (up to 40%) [34]. The hyperelastic effect could become more pronounced beyond the 40% strain range utilised in this study. Future work should explore whether these cryogels exhibit nonlinear responses at higher deformations, where hyperelastic models could become more appropriate [35].

4.4. Comparison with the Literature on PVA-C for Vascular Applications

To contextualise the findings, the mechanical properties of the layered PVA-C samples were compared with the existing literature.

Table 4. Comparison of Young’s modulus and viscoelastic properties with the literature, including native arteries, synthetic grafts, and other PVA-C constructs.

Study	Sample Type	Young’s Modulus (kPa)	Storage Modulus (E′) (kPa)	Loss Modulus (E″) (kPa)
Panieraki et al. (current study)	Bioprinted PVA cryogel	30–50	8.02–12.96	0.52–0.57
Crolla et al., 2021 [36]	Bioprinted PVA cryogel (parallel and perpendicular printing orientations)	15.10–58.90	59.20–96.80	1.89–10.2
Crolla et al., 2021 [36]	Casted PVA cryogel (10–15% w/v)	78–136	60.00–102.00	5.10–10.20
Gale et al., 2024 [37]	Bioprinted PVA cryogel	48.45–73.81	7.44–11.56	0.21–0.71
Burton et al., 2017 [29]	Porcine coronary artery (LAD)	N/A	14,470–25,560	1570–2710
Dutta et al., 2013 [38]	Baboon carotid artery (arterial fibroblasts)	90–184.2	N/A	N/A
Holzapfel et al., 2005 [39]	Human coronary artery (ultimate tensile stress)	Adventitia: 1300–1430 Media: 419–446 Intima: 391–394	N/A	N/A
Lim et al., 2021 [40]	Dacron (synthetic vascular graft)	3,000,000	N/A	N/A
Lim et al., 2021 [40]	ePTFE (synthetic vascular graft)	500,000	N/A	N/A

summarises key values from this study alongside previous work.

The bioprinted PVA-C constructs in this study exhibited YM (30–50 kPa), which are comparable to previous bioprinted samples from Crolla et al. [36] and Gale et al. [37]. However, both casted PVA-C hydrogels and native arterial tissues display considerably higher stiffness. Native arteries exhibit YM values in the range of 90–184 kPa (Dutta et al. [38]) and an ultimate tensile strength up to 1430 kPa (Holzapfel et al. [39]). The range of moduli reported in the literature for vascular tissue is very broad. None of the PVA-C samples replicated the mechanical strength of native tissues, regardless of fabrication method. This suggests that additional reinforcement strategies such as fibre incorporation, composite structuring, or crosslinking strategies may be required to improve mechanical performance for vascular applications.

We also benchmark against widely used synthetic grafts such as expanded polytetrafluoroethylene (ePTFE) and Dacron. These materials are substantially stiffer than both native arteries and the printed PVA-C constructs reported in this study [40]. While their high mechanical strength ensures structural durability, their limited compliance has been associated with suboptimal integration and elevated failure rates, particularly in small-diameter vascular applications [41]. In contrast, the softer mechanical profile of cryogel-based constructs may offer better compliance matching and potentially more favourable haemodynamic performance, especially when enhanced through reinforcement or biofunctionalisation [42].

Unlike the study by Crolla et al. [36], which primarily investigated single-layered and orthogonally stacked configurations, this work specifically examines the effect of progressive layering on mechanical properties, introducing a structured multi-layered design to assess its impact on both elastic and viscoelastic behaviour. This approach enabled a systematic investigation into how incremental layering influences mechanical stability.

The storage moduli (E′) values for the bioprinted samples (10–20 kPa) were lower than those reported in Crolla et al. [36] but comparable to Gale et al. [37]. Casted PVA-C hydrogels displayed significantly higher E′ values (60–100 kPa) [36], suggesting increased crosslinking uniformity and structural rigidity enabled by the casting process. This research extends previous work by assessing inter- and intra-sample variability through statistical analysis, providing a quantitative framework for evaluating sample consistency with respect to fabrication and mechanical characterisation, an aspect that was not extensively explored in previous bioprinting studies. The loss moduli (E″) of the bioprinted PVA-C constructs (5.2–5.7 kPa) were significantly lower than those of native arteries (~15.7–27.1 kPa) [29]. While casted PVA samples exhibited higher E″ values, they also did not reach the range required to mimic the viscoelastic response of biological tissues. However, casted samples remain isotropic, lacking the orthotropy achievable through bioprinting, which is critical for replicating the fibre-reinforced architecture of arterial tissues.

Layered PVA-C structures present a promising approach for fabricating functionally graded biomaterials, which could enable spatially variable mechanical properties that better replicate the behaviour of native vascular tissue. Previous studies have demonstrated that anisotropic fibre alignment plays a crucial role in arterial mechanics [43]. While the current study did not explore the directional alignment of the toolpath, previous studies have investigated this effect [36].

4.5. Study Repeatability and Environmental Conditions

Dehydration levels varied across sample types, with 2L samples exhibiting the highest weight loss (−19.04% on Day 1, −17.74% on Day 2) and 6L samples retaining the most moisture (−9.74% on Day 1, −9.14% on Day 2). These findings aligned with structural imaging observations (Appendix A.1, Table A1), where thinner constructs, particularly Sample 2L-4 at 40% strain, displayed irregular polymer distribution, increased filament separation, and the formation of a localised hole under tensile loading (Sample 2L-4, Table A1, Appendix A.1). While such features suggest that dehydration may affect mechanical integrity, the current data does not establish a direct causal relationship. In contrast, 6L samples maintained greater filament cohesion and overall uniformity, indicating that increased layering supports moisture retention and structural stability. The 4L samples demonstrated intermediate behaviour, with moderate dehydration and partial filament separation, suggesting that four layers may not fully mitigate dehydration-induced structural disruption under strain.

Despite variations in hydration retention, the mechanical variability observed across sample designs suggests that additional factors—such as polymer crosslinking density, deposition uniformity, and filament cohesion—may have influenced performance. For instance, uneven temperature distribution during freeze–thaw cycling can induce local variations in crosslinking density, directly impacting stiffness and elasticity [33]. Irregular deposition may also lead to inconsistent filament packing and weaker interlayer bonding, reducing load transfer efficiency [44]. Furthermore, reduced filament cohesion at the interfaces can compromise structural integrity under mechanical load [45]. These aspects may interact with dehydration effects, contributing collectively to the observed mechanical variability.

While the printing bed temperature was monitored throughout the fabrication process, it was not actively regulated by the bioprinter and fluctuated between −1.84 °C and −2.95 °C. Section 3.6 presents the full temperature profile and the smallest temperature variation occurred in 6L samples. In comparison, Crolla et al. [46] reported an initial bed temperature of −35 °C, followed by a rapid increase. This research has improved on temperature consistency by maintaining a relatively stable sub-zero profile. Nonetheless, the lack of more precise temperature control may have influenced polymer deposition and crosslinking consistency.

Environmental temperature and humidity were also not actively controlled, likely exacerbating dehydration in thinner constructs with higher surface-area-to-volume ratios. While the bed temperature would directly influence the first layer, subsequent layers were exposed to varying thermal conditions dictated by the underlying layer and the surrounding environment. These ambient fluctuations may have led to different vertical gradients in temperature and, therefore, layer-dependent temperature profiles. Such variability could have led to uneven crosslinking and mechanical inconsistencies.

These findings highlight the importance of maintaining hydration and ensuring both in-plane and the perpendicular-axis temperature and print uniformity to preserve structural heterogeneity. Future work should consider using a temperature-controlled chamber to minimise thermal fluctuations and stabilise polymer deposition. Additionally, regulating environmental humidity during both fabrication and storage may further reduce dehydration-induced variability. Formulation strategies—such as incorporating hydration-retentive additives—may also help to improve homogeneity and mechanical robustness, especially in thinner constructs.

Underlying the mechanical variability is the polymer microstructure, which is highly sensitive to environmental conditions during fabrication. Paxton et al. [47] showed that PVA-based inks transition from Newtonian to shear-thinning non-Newtonian behaviour under sub-zero temperatures. This rheological shift affects extrusion fidelity through changes in the consistency index (K) and flow behaviour index, both of which are temperature-dependent [37,38]. Local shear conditions, influenced by nozzle geometry and pressure differentials, further modulate filament deposition and potentially alter polymer chain alignment or local crosslinking density [47].

Freeze–thaw cycling promotes partial crystallisation and intermolecular hydrogen bonding, forming physical crosslinks that reinforce the hydrogel network [37,48]. These processes are affected by parameters such as polymer concentration, freeze–thaw duration, and cooling rate [49] Consequently, fluctuations in ambient or bed temperature (relative to samples and layers) during printing may result in heterogeneous microstructures and crosslinking profiles, contributing to inter-sample variability.

Finally, polymer–solvent interactions strongly govern the hydration behaviour and mechanical performance of PVA cryogels. The hydrophilic nature of PVA enables extensive hydrogen bonding with water, facilitating water uptake and retention [50]. These interactions support swelling and elasticity, while cryogenic processing separates water into amorphous domains, promoting both flexibility and structural integrity [51]. For biomedical applications, this retained water supports nutrient diffusion and mechanical compliance [52,53,54]. Modulating solvent composition allows for tuning pore size and hydration properties by altering polymer chain mobility [50,55]. Optimising these interactions in future research will be vital for tailoring mechanical behaviour and stability in 3D-printed cryogels.

4.6. Rheological Considerations Under Sub-Zero Printing Conditions

Although rheological characterisation is a critical step in evaluating the printability of biomaterials, it was not feasible in the present study due to the sub-zero printing protocol and the thermoresponsive nature of PVA. The formulation used here undergoes physical crosslinking via freeze–thaw-induced crystallisation, with gelation initiating near the freezing point of water (~0 °C). As a result, the ink transitions rapidly from a viscous fluid to a semi-solid gel within a narrow temperature window, making conventional rheological measurements using cone-plate rheometers impractical at printing-relevant sub-zero conditions. Performing controlled shear rate sweeps or oscillatory tests below 0 °C is technically challenging due to rapid gelation and thermal instability.

Nonetheless, previous studies have characterised the rheological behaviour of comparable PVA-based systems. PVA solutions are known to exhibit temperature-sensitive shear-thinning behaviour, where the flow behaviour index decreases as temperature approaches freezing, indicative of a transition from Newtonian to non-Newtonian flow [47,56]. Gelatin–PVA blends have shown similar trends, with viscosity decreasing under applied shear but recovering post deposition, facilitating stable filament formation during extrusion [57].

In the context of sub-zero bioprinting, Crolla et al. [46] demonstrated that directional solidification during extrusion induces orthotropic mechanical behaviour in PVA cryogels, attributed to the formation of filament-level mesostructures. This study revealed that samples printed with smaller nozzle diameters exhibited higher stiffness along the print direction and increased orthotropy, which were linked to greater filament alignment and reduced inter-filament bonding. While direct rheological measurements were not performed, these findings highlight the significant influence of sub-zero extrusion parameters on the microstructural organisation and resultant mechanical performance of cryogels.

Although rheological data were not directly measured in this study, the qualitative printing performance—minimal filament spreading, consistent extrusion, and acceptable dimensional fidelity—suggests that the ink behaved in line with expected shear-thinning and recovery characteristics reported in the literature. The overall mean dimensional deviation of 11.4 ± 3.6% supports the reliability of the sub-zero extrusion-based manufacturing process.

4.7. Study Reflections

This study provides valuable insights into the influence of layering on the mechanical properties of 3D-printed PVA-C; however, several limitations should be considered. The absence of anisotropic fibre reinforcement remains a key constraint, as fibre orientation plays a critical role in replicating the mechanical behaviour of arterial tissues [58].

While anisotropy is an important characteristic of native arteries, the mechanical tests performed in this study were uniaxial and conducted in a single direction. As such, the results reflect only the bulk mechanical response along that axis and do not capture potential directional dependencies in stiffness or energy dissipation. A single YM is not a sufficient descriptor for an anisotropic system; biaxial, shear and/or multi-directional tensile testing, are required to fully characterise the anisotropic mechanical behaviour of layered PVA-C constructs.

Additionally, while measurements taken during the DMA captured short term dehydration effects, long-term hydration characteristics under mechanical testing and physiological conditions remain unverified. Although weight measurements confirmed consistent reswelling between test repetitions, further studies should investigate hydration-dependent mechanical performance, particularly in in vitro conditions, to determine suitability for vascular graft applications.

Mechanical characterisation was limited to a maximum strain of 40%. To fully capture the nonlinear mechanical response of layered PVA-C, future work should include higher strain testing. Furthermore, testing should explore the yield point to evaluate ultimate tensile strength and failure behaviour. Extended cyclic loading experiments are needed to assess fatigue resistance and long-term durability under physiologically relevant conditions.

A more detailed evaluation of geometry retention and interlayer cohesion at the filament level—such as filament width uniformity, pore closure, or potential delamination—would require high-resolution imaging techniques like SEM. Such analyses were beyond the scope of this mechanical-focused investigation, but represent an important future direction for optimising the structural stability of multi-layered cryogel systems.

Further exploration of FGM designs is warranted. While individual 2L, 4L, and 6L samples exhibited frequency-independent behaviour, exploring the combination of these configurations could generate a structure with dynamic moduli which are spatially variable. Such an approach could enhance biomimetic performance by leveraging the distinct dynamic stiffness contributions of each layered configuration.

To translate the findings of this study into cylindrical geometries representative of arterial grafts, alternative printing strategies would need to be considered. Printing along the axial (Z) direction would, in theory, allow for vertical layering; however, under sub-zero printing conditions, this configuration compromises bed temperature control, which is essential for maintaining cryogel integrity. A more viable approach involves deposition onto a rotating cylindrical substrate, which may enable radial layering while preserving thermal stability [59,60]. Dell et al. [59] demonstrated a mandrel-based bioprinting system capable of fabricating tubular hydrogel constructs with controlled fibre orientation and multi-layered architecture, while Reeser et al. [60] reviewed additive-lathe 3D printing techniques that facilitated the construction of cylindrical structures with improved mechanical fidelity and reduced reliance on support materials. These approaches underscore the potential of radial layering to modulate mechanical properties along circumferential and axial axes, aligning with the functionally graded design principles investigated in this study. Future work should explore the adaptation of this strategy to bioprinted PVA-C, with the aim of more accurately replicating the anisotropic mechanical environment of native arteries.

Although biological validation was beyond the scope of this study, the biocompatibility and integration potential of polyvinyl alcohol cryogel (PVA-C) constructs are well supported in the literature. Due to the sub-zero bed temperature and the repeated freeze–thaw cycles required for physical crosslinking, the current process is incompatible with cell encapsulation during printing, as these conditions would compromise cell viability. However, post-fabrication strategies have been shown to enhance cellular compatibility. For instance, improved endothelial cell adhesion and spreading have been observed in PVA–gelatine composites [61], and blending PVA with gelatine—providing cell-adhesive arginine–glycine–aspartic acid sequences—has been shown to support endothelialisation [27]. Additional approaches, such as surface functionalisation with adhesive peptides and perfusion-based seeding, may further promote biological integration after fabrication [62]. The influence of freeze–thaw cycling on cytocompatibility has also been explored, with reports indicating that multiple cycles increase porosity and mechanical strength without impairing cell viability [63]. Collectively, these findings support the feasibility of post-printing cell seeding strategies for vascular tissue engineering using PVA-C scaffolds.

5. Conclusions

This study explored the mechanical properties of sub-zero-bioprinted multi-layered PVA-C as candidate materials for vascular graft applications. The results demonstrated that increasing the number of layers led to a moderate reduction in stiffness, likely due to decoupling at layer interfaces, and local variations in crosslinking induced by thermal gradients during fabrication. All constructs exhibited a predominantly elastic response under cyclic loading, with minimal frequency dependence of the storage modulus (E′).

The testing methodology proved robust and reproducible across repetitions, although inter-sample variability was observed. This variability underscores the influence of fabrication conditions—such as environmental fluctuations, hydration retention, and print uniformity—on mechanical performance and highlights the importance of further optimisation in the bioprinting process.

Statistical analysis confirmed that both storage and loss moduli differed significantly between two-layered and multi-layered constructs at 2 Hz and 10 Hz, indicating that layering influences energy dissipation at specific frequency ranges. However, beyond four layers, additional layering did not significantly alter mechanical performance, suggesting a saturation effect.

Compared to the prior literature, the layered bioprinted PVA-C constructs achieved similar mechanical performance, though they remain less stiff than native arterial tissues. Importantly, bioprinting offers the advantage of anisotropic control, an essential feature not achievable with casted hydrogels. This study was limited to uniaxial testing, capturing only the bulk response in a single direction. Since arterial tissues are inherently anisotropic, future investigations should include biaxial or shear testing as well as in vitro experiments under physiological flow and pressure conditions to better reflect dynamic vascular environments.

Additionally, incorporating cellular components and biofunctionalisation strategies—such as endothelialisation or growth factor delivery—could improve the biological performance of these constructs. Translational efforts should also consider the long-term stability, sterilisation compatibility, and scalability of the fabrication method to support clinical application.

Finally, the distinct mechanical responses of the 2L, 4L, and 6L samples suggest that spatially arranging different layered configurations could be a promising strategy for creating functionally graded materials (FGMs) with tuneable mechanical properties. Such designs may enable more biomimetic vascular grafts by better replicating the heterogeneous structure and variable mechanical response of native arteries.