The overall goal of this systematic review and meta-analysis was to map and analyze all animal experiments that have been reported to date for in situ TEVG based on resorbable synthetic scaffolds. Moreover, we aimed to identify model-specific parameters that could potentially influence outcome and define points-of-attention in the design of animal experiments for in situ TEVGs and the reporting thereof. Meta-analysis was performed on patency as a clinically relevant and main functional readout parameter for TEVGs. The main findings are that for the preclinical evaluation of in situ TEVGs rats are the most often used, choice of animal sex and implant site are often dependent on animal species, and both health status and animal age are poorly reported. Meta-analysis with subsequent subgroup analysis showed that patency rates for TEVGs are influenced by follow-up time, as well as the choice of animal species and animal age.
Number of publications and species
The use of resorbable synthetic grafts to induce tissue regeneration in situ has gained tremendous momentum over the last decade for a broad variety of applications10,30. This is driven by the recognition that the inflammatory response to an implanted material can be modulated to induce and steer the formation and remodeling of functional new tissue10,16. This trend has placed more emphasis on animal experimentation in order to test the in vivo response to newly developed materials already at an early stage. This is reflected by the mapping of in vivo studies on resorbable synthetic vascular grafts as performed in this study, showing a strong increase in the number of publications on in situ TEVGs in the last decade.
Moreover, our analysis shows a clear shift towards rat and mouse models as the most frequently used animal models to test TEVGs for in situ TE. Particularly, a steep rise in the use of mice over the recent years can be observed. This is probably due to the development of knockout and transgenic techniques, which allow assessment of TEVG functionality in simulated pathologies already early in preclinical stage, or assessment of TEVGs seeded with non-autologous cells31,32. The largest part of the immune-compromised animals included in this systematic review were SCID/beige mice (73% of all immune-compromised animals), which are often used to circumvent the adaptive immune response when testing TEVGs seeded with non-autologous cells such as human cells32. In this systematic review, we only included animals of experimental groups that used TEVGs without cells (this was usually the ‘control group’ in the studies testing TEVGs seeded with human cells), or with on-the-fly seeded cells (i.e., no in vitro cultures), which suggests that the number of immunocompromised mice used for TEVG research in general is even larger.
Interspecies differences
Given that in situ vascular TE is heavily dependent on the intrinsic regenerative capacity of the recipient of the TEVG, the choice of animal model may be of crucial importance for the translational relevance of the outcome. While much emphasis has been placed on the influence of local graft properties (e.g., material choice, mechanical properties, microstructure)7,33, the potential systemic influences that arise from the choice of in vivo model have been largely underrepresented in the literature. Both inter-species differences in the regenerative capacity and processes, as well as intra-species characteristics such as age, sex, and multifactorial disease profiles, are important considerations for the in vivo testing of in situ TEVGs16. The meta-analysis performed in this study revealed an overall median patency of 79%, which is in the same range as the median patency reported in a recent systematic review of Skovrind et al., who observed an overall patency of 83% for TEVGs (seeded/unseeded, biological/hybrid/synthetic and degradable/non-degradable) in large-animal models only34. Albeit that we cannot conclude which animal model is most representative for the clinical situation, we detected inter-species differences in overall patency rates. Interestingly, significantly lower patency rates were found in dogs and pigs when compared to rats, and significantly lower patency rates in pigs were observed when compared to rabbits. The main compromising factors for TEVG patency are thrombus formation and adverse remodeling (i.e., intimal hyperplasia). In this systematic review, we did not quantify to what extent the reasons for graft occlusion were due to either cause and whether this was subject to inter-species differences. As the reporting on the reason and timepoint of graft failure was incomplete in part of the papers, it would be too speculative to make the direct link. Nevertheless, important inter-species differences have previously been reported with respect to thrombogenicity, as well as regenerative potential35,36,37.
With respect to inter-species differences in thrombogenic potential, Grabowski et al. (1977), assessed platelet adhesion of heparinized blood of eight different species on different biomaterials under flow. It was observed that the inter-species differences were biomaterial dependent, meaning there was a much higher platelet adhesion of dog blood compared to human (and cow, baboon, macaque, pig, and sheep) to “Cupropjan” and “Avcothane” biomaterials, but this dog-human difference was absent for GoreTex biomaterials35. Similarly, in one of the earlier papers included in our analysis, Van Der Lei et al. (1989) describe species-dependent thrombogenicity of PU grafts36. Specifically, implantation of PU vascular grafts in the rabbit carotid artery was compared with their previous studies with the same prostheses in the rat aorta38. Their results show that prostheses that remained patent in the rat model, occluded in the rabbit model. Albeit that there were two different sites used, the latter more closely resembled the clinical situation in terms of thrombogenicity according to the authors36, as well as a review by Byrom et al.39. Almost 30 years later, differences in hemostasis and thrombosis between preclinical animal models and humans are still extensively studied, because extrapolating the preclinical findings to humans still results in inaccurate predictions of the clinical scenario37.
Another important contributor to the inter-species differences in patency may be differences in regenerative capacity between animals. For example, Zilla et al. previously reported on the importance of differences in endothelialization potential between animal models. Particularly, transanastomotic endothelialization is dependent on species, senescence, anatomical dimensions, and graft surface, with species being the most important determinant17. Another important determinant of success for in situ TEVG is the balance between tissue regeneration and resorption of the initial synthetic implant material10. Fukunishi et al. recently reported on inter-species differences in the resorption rates of nanofibrous vascular grafts, with a faster graft resorption of the grafts in sheep compared to rat over a 6-month timeframe. In addition, higher levels of ECM, elastin, and mature collagen were found in sheep40, which suggests that inter-species differences in the rates of graft resorption and tissue formation may have contributed to the differences in patency rates as observed in the present study.
Intraspecies characteristics
In addition to inter-species differences, we analyzed model-specific, intra-species characteristics (i.e., follow-up time, animal age, and sex) via subgroup analyses. These model-specific characteristics dominantly influence the inflammatory state and intrinsic regenerative capacity, and are thereby anticipated to affect in situ tissue engineering outcome (see review De Kort and Koch et al.)41. The overall patency rate was dependent on follow-up time, with a significantly higher mean patency for grafts with a long follow-up time when compared to short and medium follow-up times. This suggests that occlusion is most likely to occur within short- to medium follow-up time, and is less likely to occur after a certain period of implantation. Possibly, occlusion due to thrombus mainly affects animals during the earlier stages after implantation. The influence of attrition bias is expected to be low, as the follow-up time for each animal is defined at the start of the experiment42. Another contributing factor may be the temporal course of the in situ remodeling processes, which is more prone to adverse remodeling at earlier stages when graft resorption and tissue deposition are in a highly active state, but much more stable once a state of tissue homeostasis has been achieved. An illustrative example is the study by Drews et al., who reported on the spontaneous reversal of early stenosis after 6 months of implantation through an inflammation-driven mechano-mediated mechanism for TEVGs that were implanted as Fontan conduit to connect the inferior vena cava and the pulmonary artery in sheep43.
To assess the impact of recipient age, we divided the animals in a young and an adult category, dependent on the species. Based on the studies that could be included in this meta-analysis we found that recipient age may bias graft patency rates, although the number of experimental groups in the ‘young’ category (21 experimental groups) was substantially lower than the ‘adult’ category (219 experimental groups). Very little studies have focused on understanding vascular TE with respect to recipient age7. It is generally assumed that young adults, here included in the adult age category, represent an optimal age due to a high regeneration capacity and relative somatic growth stability12. In the field of in situ TEVG, hardly any study has been performed on older/aged animals. Of all included animals, only one study reported on ‘old’ animals (n = 20)26. Aging is a risk factor for cardiovascular diseases, due to changes in vascular biology associated with age, such as arterial stiffening and cellular senescence. Moreover, aging affects the host inflammatory response to an implanted biomaterial, as has been reported for subcutaneously implanted biomaterials44,45. A recent study by Johnson et al. (2021) showed that PCL-gelatin fiber grafts in younger rats had less flow disturbance and healed with more organized ECM structures when compared to aged rats46. Considering that elderly patients are an important clinical target cohort for application TEVGs, in particular for small-diameter artery replacements, the observed influence of animal age on TEVG patency emphasizes the importance of considering animal age when evaluating in situ TEVGs. Of note, additional effects of animal size and weight were not taken into account within this analysis due to poor reporting. Younger animals, which are intrinsically smaller than adult animals, might receive grafts with smaller diameter, which is suggested to influence patency rates as well, especially for small animal models. However, species-dependent sensitivity analysis on the impact of animal age (young or adult) indicated no differences with the exclusion of any of the six species.
Next to age, there are differences in the manifestation of CVD between male and female patients. Sex-dependent differences within the cardiovascular system in general47,48,49 and in outcome after treatment with a cardiovascular prosthesis are known50,51,52. In our subgroup analysis on sex, we found no significant differences in patency rates of in situ TEVGs between male and female animals. It is plausible that the effect of sex on the in situ regenerative response is more subtle, and is not directly reflected by patency rates. For example, Blum et al. observed lower cellularity, less collagen deposition and maturation, but higher graft resorption rates in PGA + PCLA grafts when implanted as IVC interposition grafts in male mice, compared to female mice14. Another consideration is that for females in particular, cardiovascular pathophysiology is closely related to age, as the high estrogen levels in pre-menopausal women have a cardioprotective effect, which is lost after menopause53. Traditional animal models typically do not reflect such estrogen-dependent effects due to differences in hormonal regulation between species, and require more dedicated models, such as ovariectomy models54. Even though cardiovascular research has classically been male-oriented, awareness of sex and gender differences in the timing of diagnosis, the disease process, as well as the disease presentation is increasing over the last years55,56. The importance of at least reporting animal sex, let alone include both male and female animals to address the sex-specific bias, has already been described in earlier reviews on TEVGs, e.g., by Bergqvist and Jensen (1985)57.
Hemodynamics and graft characteristics
Next to intrinsic animal-specific biological variables, patency is determined by the hemodynamic conditions, which are determined by the blood flow and blood pressure, as well as physical graft characteristics including graft diameter, length, and wall thickness. Therefore, we subcategorized the experimental groups into low-pressure venous and high-pressure arterial groups, further subdivided by small (<6 mm) and large (>6 mm) diameter. Interestingly, despite the often described low patency rates for small diameter blood vessels17,27,28,29,58, our results did not show a significant influence of the implant site (arterial, venous) and inner diameter (small < 6 mm < large) on the overall patency rate. This may be explained by the fact that most tested grafts are relatively short in length. In rodent models specifically, rapid transanastomotic endothelialization may favor patency rates in short grafts, while this does not occur in humans, as emphasized by Zilla et al.17. Clinically, small-diameter by-pass grafts have a typical length exceeding 20 cm, pre-clinical assessment of short small-diameter grafts might falsely suggest a higher patency rate, as also described by Skovrind et al.34. Therefore, especially for small-diameter grafts, a graft length/diameter ratio of 10 was proposed to reduce false-positive patency rates27,28,29. This threshold was only met in few studies (39/182) within this dataset.
Several groups have performed preclinical experiments to directly assess the influence of implantation site59. One example is a recent study by Sologashvili et al. (2019), in which PCL grafts were implanted in both the carotid and the aortic position in rats. Even though the carotid grafts showed better endothelialization, cellular infiltration, and compliance, as well as lower calcification rates compared to the aortic grafts, overall patency was only 65% in the carotid, versus 100% in the aortic grafts15, which is in line with our results of the subgroup analysis on implant site. The authors suggested that this difference in patency may come from differences in anatomical position, compliance mismatch, and flow conditions15. Indeed, the local hemodynamic conditions have been reported to play a key role for the functional remodeling of in situ TEVGs43, as well as thrombogenicity60, emphasizing the importance of the implant site when setting up animal experiments and interpreting the results thereof.
Although not specifically included in the present meta-analysis, other graft characteristics, such as wall thickness and compliance, will most certainly influence graft patency rates as well. As shown in our mapping analysis there is a great variety in synthetic materials used (Fig. 6), as well as physical properties of the graft, especially graft wall thickness and length show a large variability (Fig. 5). Given this large variability, we did not systematically analyze the effect of these graft properties on patency in the current study.
Recommendations for animal study design and reporting
A striking observation is that for 76% of all included animals (148 studies), no health or immune status was reported. Despite the publication of the ARRIVE (Animals in Research: Reporting In Vivo Experiments) guidelines, first published in 2010 to improve reporting quality of preclinical experiments61, no clear improvement in reporting of health status can be observed in the studies that we assessed. Of all 148 studies that did not report on health/immune status of the animals, 59 studies were published before 2010, and 89 studies after 2010. A similar alarming conclusion was made in a systematic evaluation on reporting quality of the assessment of pulmonary heart valve prosthesis in large animals62. The authors concluded that the introduction of the ARRIVE guidelines did not improve reporting quality over the last 20 years in the field of heart valve research62. Of note, reporting of animal sex in the studies included in our review was much better after 2010, even though animal sex was not reported for 25% of all included animals (32% of all included studies). Only four studies published after 2010 did not report animal sex. Overall, we strongly recommend improved reporting of animal characteristics, following the ARRIVE guidelines.
Our findings emphasize the importance of adequate selection of the animal model and study characteristics when evaluating in situ TEVGs in a clinically relevant manner. While large animals are generally considered to better resemble the clinical situation in terms of hemodynamics and hematological profiles, small animals are easily available, require low handling and housing efforts, and are relatively affordable in larger scale cohorts37,63,64. Therefore, small-animal models are most often used to initially screen potential TEVGs designs and to perform early functional tests. One important observation is that the choice of animal model tends to come with a bias in other study characteristics. For example, our analyses show that, especially for small animals, implantation site is heavily biased by the choice of animal, with a clear dominance for the abdominal aorta as the preferred implantation site in rats (70% of all experiments involving rats) and rabbits (65%), and the vena cava in mice studies (70%). Additionally, almost all included male animals were rats (95%), and a vast majority of all included female animals were mice (68%). Additionally, subcategorization of age showed that the ‘young’ group was mainly derived from pigs (52% animals included in this subgroup were pigs) and sheep (20% of the total young animals). These biases are typically due to model-specific practical limitations (e.g., animal housing, surgical limitations, earlier investment in specific surgery, and specific animal model), rather than clinical relevance. In order to derive translationally meaningful data, the results of animal models should be interpreted in their appropriate context and therefore we recommend that rather than based on practical limitations, the choice of implant model should be based on clinical relevance and the proposed research question. Dedicated animal models can be applied to study specific aspects of the biological mechanisms underlying in situ vascular tissue regeneration. For example, models that have previously been described to shield transanastomotic cell ingrowth65,66 or animal models to mimic a multifactorial disease profile, such as vascular replacement in diabetic conditions67. The patient population requiring vascular replacements will mainly be the adult and elderly patient, often suffering from multifactorial diseases affecting the immunological state and regenerative capacities. Hence, we recommend to shift the scaffold-centered focus to a more graft-independent, host-dependent focus16. In our review, only a limited number of studies using aged and diseased animals were found, highlighting that this is a knowledge gap in the field. Even though aged and diseased animals involve more practical difficulties as well as higher costs, testing in situ TEVGs in clinically relevant animal models is highly important for clinical translation7. In accordance with the 3Rs (reduce, refine, replace)-perspective68, we advise proper design and reporting of animal studies which is essential in order to maximize the value and impact of animal studies. Especially, reporting of non-significant results should be encouraged to limit publication bias. Unfortunately, within this study, it was not possible to determine this bias due to the single arm nature of study designs and the event rate as primary measure for meta-analysis. Additionally, complementing in vivo studies with hypothesis-driven human in vitro models69,70 and in silico models71,72 can be a powerful strategy to maximize impact and accelerate translation by improving the fundamental understanding of clinically relevant processes underlying in situ vascular tissue regeneration.
Study limitations
It is important to note that the present analysis is restricted to the collected data and to the chosen statistical evaluation methods. Despite the relatively large yield of papers, still five relevant articles20,73,74,75 were not retrieved via the search string, but found later via reference lists of relevant reviews7,10,34,57,76,77,78. Most likely this was because these five articles were not indexed at the time of retrieving the article, highlighting the effectiveness of the search string. We therefore cannot guarantee that a few eligible studies have been undetected by our search strategy. Furthermore, we only included and analyzed synthetic-based, off-the-shelf available TEVGs. We thereby exclude a wide variety of less and more successful (in situ) TEVG approaches, for example, involving decellularization of biological materials79, ‘bio-tubes’80 and pre-culturing in dedicated bioreactors81, for which different biological mechanisms might be involved (e.g., less dependent on inflammatory response, different thrombogenic potential). It should be noted that this distinction between natural materials and resorbable synthetic materials can be hard to establish, for example in the case of hybrid materials like Hyaff-1182, which is of natural origin (and thus excluded from the present study) but heavily chemically processed.
Finally, we only performed meta-analyses on patency, as the primary outcome parameter. While patency represents the functionally most important readout for a vascular graft, analyzing patency as a standalone parameter does not allow for more in-depth analysis of species-specific biological processes. The current meta-analysis was primarily intended to reveal any potential biases in outcome (in this case patency) independent of more specific graft design features, such as microstructural or biochemical design, which would be highly interesting in its own right. We consider the database that we built for this systematic review as a starting point for a wider variety of research questions for the field of in situ vascular TE. Further (meta-)analysis of other readout parameters, such as cellularity, ECM formation, calcification, or endothelialization, is topic of active investigation, although this is highly challenging due to large variations in analysis methods and reporting methodologies. Heterogeneity in study design, reliability of patency assessment methods, and poor reporting, might reduce the strength of the meta-analysis, as the quality of the meta-analysis relies on the quality of the primary studies. For example, we had to exclude some important subgroup categories from the meta-analysis, such as recipient old-age, due to limited number of experimental groups. This again stresses the need for more clinically relevant models.
