Edited by: Silvestro Micera, Sant'Anna School of Advanced Studies, Italy
Reviewed by: Xavier Navarro, Autonomous University of Barcelona, Spain; Kazutaka Takahashi, University of Chicago, United States
This article was submitted to Neuroprosthetics, a section of the journal Frontiers in Neuroscience
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
Complex prosthetic reconstruction has become a standard for providing patients with useful extremity replacement. However, intuitive control of multiple degree of freedom myoelectric prostheses has proven to be a difficult goal (Aszmann et al.,
Since its introduction in the 1940s, the standard approach to interfacing is the use of few muscle-control signals recorded by surface EMG electrodes (Bergmeister et al.,
Implantation of medical devices requires extensive preclinical testing to analyze long-term biocompatibility as well as functionality of the implant (del Valle and Navarro,
When choosing an animal model, behavioral characteristics should be considered, as frequent handling of the animals during testing is often required. Costs and availability as well as ethical considerations and legality play also a major role when choosing the right model. To quote Bernard E. Rollin (Rollin,
The aim of this study was to evaluate animal models for muscle/nerve implants considering behavioral, economical, and ethical aspects in addition to the scientific validity of the model. In addition, we present an overview of models used in the current literature. Based on our analyses, we describe an algorithm for experimental testing of fully implantable devices under natural moving conditions and their advantages and disadvantages.
The authors designed a systematic search strategy for PubMed and Google Scholar according to the PRISMA guidelines (Liberati et al.,
Flow diagram of the selection process of systematic literature analyses, as recommended by PRISMA Guidelines.
This table shows the PICOS (Patient, Intervention, Comparison, Outcome, Study design) inclusion criteria, used for systematic literature search.
Population | Animal Models |
Intervention | Implantation of interfacing device or electrode interacting with peripheral neuromuscular tissue |
Comparison | Animals used |
Outcome | Type of animal |
Study design | Experimental animal study |
All models were analyzed for costs on the basis of the in-house prices of the Center of Biomedical Research at the Medical University of Vienna. For international comparison, relative relations are given to compare among different models. Ethical standards according the FELASA principles were obtained (Guillen,
The most common nerve model, the rat sciatic nerve was evaluated for feasibility at our facility, comparing operation time to accessing the sciatic nerve between surgeons and unexperienced academic staff. This was performed to proof the model as performable for non-surgeons.
All animal models from the authors' institution were conducted with permission of the ethics committee of the Medical University of Vienna and the Austrian Ministry for Research and Science (BMWF: reference number: BMWF-66.009/0309-WF/II/3b/2010, BMWF-66.009/0340-WF/II/3b/2016, BMWF-66.009/0024-WF/II/3b/2018).
A total of 8859 studies were identified with the search terms listed in
Search terms for the systematic literature search.
Animal AND implantable AND electrode |
Animal AND peripheral AND electrode |
Animal AND EMG AND prostheses |
Implantable EM |
Animal model AND prostheses AND control |
Animal model AND extremity AND reconstruction |
Animal model AND EMG test |
Animal model AND electrode testing |
Literature analyses showed an upward trend for publications of new interfacing devices. Shown are all 198 studies included in the final analyses according to the year of publication. Not all studies represent new interfacing devices, but rather studies using animal models to test electrodes or stimulation parameters for interfacing bionic prostheses.
Analyses showed a significant trend toward the development and testing of interfacing devices over the last years.
Most published studies originated from Northern America (55%) and Europe (34%). Studies from Asia and Australia accounted for 10 or 1%, respectively. South America and Africa were not represented in the identified studies.
Analyses of the animals used in the literature showed predominant use of the rat model (40%). Rat trials had a mean duration of 87 ± 105 days (range: 1–390 d) with a mean sample size of 12.5 ± 10.3 animals (range 1–51) per study. Cats were used in 27% of the studies with a duration of 92 ± 158 days (range: 1–900 d) and a mean of 6.9 ± 4.8 animals (range 1–26) per study. Rabbit trials accounted for up to 13% with 11.9 ± 10.2 animals (range 1–40) and 77 ± 132 days (range 1–480 d). Dog models were used in 5% of the publications with 3.3 ± 2.8 animals (range 1–10), analyzed over 97 ± 157 days (range 1–450 d). Pigs [5.3 ± 3.3 animals (range 1–11), 50 ± 88 d (range 1–270 d)] and monkeys [1.8 ± 0.9 animals (range 1–4), 394 ± 314 days (range 1–930 d)] were both used in 4% of the studies. Sheep [2.5 ± 1.5 animals (range 1–4), 102 ± 18 days (range 84–120 d)] and mouse models [12.3 ± 6.8 animals (range 7–22), 84 ± 83 days (range 16–200 d)] only accounted for 2% each. The remaining 3% were made up from unique experiments including raccoons, guinea pigs, frogs, crayfish, and zebrafish (
Mean duration of experiments of different animal models. The red line indicates 90 days of duration, which is considered as long-term (chronic).
Large animal models (rabbit, cat, dog, monkey, pig, and sheep) had a mean duration of 135 ± 87.2 days compared to small animal models (mouse, rat) with a mean duration of 85 ± 11.2 days.
Considering the length of the trials, experiments with a duration over 3 months (90 days) are defined as chronic (long-term) experiments (Anderson et al.,
Muscular interfaces were tested in only 23% of the studies, thus the vast majority of the selected publications related to neural interfacing (77%). Here, 96% of cats were used for neural interfaces as well as 59.6% of dogs, 89% of pigs, 88% of rabbits, 79.2% of rats, 0% of sheep, and 33.3% of primates. For the rat model, which was the most common animal model used for neural interfacing, 88.5% of the studies focused on neural interfaces using the sciatic nerve. Of the remaining studies on the rat, 11% used the vagal and glossopharyngeal for sensory recordings or stimulation.
To evaluate feasibility of the most commonly used peripheral nerve model in the rat—the sciatic nerve model—we compared the procedure between surgeons and academic staff with low surgical experience at the Center of Biomedical Research of the Medical University of Vienna. The average time from skin incision to accessing and preparing the sciatic nerve, was 2 min ± 35 s for surgeons compared to 3 min ± 40 s for academic staff.
Costs analyses are based on in house costs at the Center for Biomedical Research at the Medical University of Vienna, which significantly varied among different models. Acquisition costs and overall costs of operation (including anesthesia, analgesia, consumption items, instrument sterilization, and required staff costs) for a single animal where ~110€ (124$) for a rat, 800€ (900$) for a rabbit, and up to 5,000€ (5,640$) for a large model like pig or sheep. These costs do not include housing after the operation which varies from 90 Cent (1$) per day for a rat, over 3.75 € (4.23$) per day for a rabbit, and up to 6.5 € (7.33$) for a sheep. Large animal models are therefore more expensive than the rat model by a factor of 7.2 (rabbit) and 45.4 (sheep). Costs were analyzed on the basis of costs of representative trials over the last years. Other institutional charges, such as acquisition or housing costs may vary in different countries, but relative relations should remain comparable (
Relative proportions of the costs of different models are given. The rabbit is by a factor 7.2 more expensive than a rat model. Sheep models are up to a factor 45.5 more expensive than rat models.
Cost analyses of a primate model at the DPZ (German Primate Center) show acquisition costs only of up to 8,000€ (9,080$). As operation requires a team of several people involved, costs for a single animal as calculated above increase tremendously. Housing costs account up to about 25€ (28.37$) per day.
Although modern bionic prostheses are technically very advanced and capable of complex movements, the interface between man and machine is still a limiting factor for sophisticated control. Therefore, research efforts are directed into the development of implantable interfaces, which is evident in the upward trend of recent publications (
The most commonly used model for
Rat handling is comparably easy. It requires basic training and can be accomplished even in small facilities. The rat offers many possibilities for both muscular and neural interface testing due to easy accessible nerves, such as the sciatic nerve, which is considered the gold standard (Vasudevan et al.,
Besides the common use of the rat model for neural implants, muscle electrodes can also be tested with this model. Our literature analyses showed that 20.8% of the studies used the rat model to evaluate muscular interfaces.
Despite many advantages of the rat model, for human-sized implants this model is promptly limited. Implantable interfaces, connected to muscle or nerve tissue, are subjected to high mechanical stress due to movements and muscle contractions. To simulate this mechanical stress, large animal models are advisable, which furthermore eases fitting of the implant due to comparable anatomy to humans.
Many of these requirements can be achieved in a rabbit model, which is ~7 times more expensive than a rat model, but still significantly cheaper than sheep or pigs. Also, the rabbit has the advantage of easy handling compared to larger animals. The sciatic nerve in the rabbit is, with some experience and knowledge of anatomical landmarks, as accessible as in the rat, but significantly larger. A big disadvantage of the rabbit model, however, is the high susceptibility to infections, which is especially relevant for long-term studies with implants. Sterility during the procedures is vital and strict post-operative care as well as antibiotic prophylaxis obligatory. Self-mutilation is frequent and implies high dropout rates. Appropriate suturing and animal cones as well as environmental enrichment can decrease stress for the animals and thereby the dropout rate (Wheeler et al.,
A significantly larger and therefore more robust animal is the sheep, which has the advantage of allowing implanting devices of real size for human application (Sartoretto et al.,
Similar to sheep, pigs offer comparable anatomy—especially regarding size- to the human body but with thicker skin and consequently possible influences on data transmission for telemetry studies. Weight gain has to be considered, especially in long-term experiments, as this can induce higher mechanical stress on implants compared to humans. Also, daily handling of the animals is hampered by their weight, particularly if frequent measurements are needed in awake animals. But even anesthetized animals require more trained staff for handling than other animals.
Easier to handle are cats and dogs, which have been widely used especially in Northern America. Behavioral training to perform certain tasks and reproducible gait behavior are advantages of these models, which can be achieved with less effort than with other models. Contrary to their popularity in Northern America, some European countries have banned these models, or restricted surgical interventions in these models. Therefore, international reproducibility and comparability is limited.
Walter et al. (
Very close to human nature is the
The use of complex tasks involving dexterous movements, for which the primate model is ideally suited, comes with costs, though. Training for a complex experiment needs to be done extensively and new tasks require re-training. For complex tasks, animals need to be continuously trained to maintain their level of performance. Training is mostly needed for making use of the high number of precisely repeated movements and the experimentally well-controlled timing, though, which in other animal models is not even an option. This training effort can be reduced with increasing naturalism of the behavior to be studied, e.g., walking in a freely moving animals (Berger and Gail,
Lastly, testing of feedback devices or electrodes for sensory feedback remains difficult in animal models, although it has already been demonstrated in human studies (Flesher et al.,
Our findings that only about 37% of the trials exceeded more than 3 months duration seems alarming for long-term safety of patients (
The costs of larger animals are significantly greater compared to smaller animals. This fact should not result in fewer animals per trial and shorter durations of the trials. Ultimately, the costs of preclinical trials are minor, compared to the potential costs of complications, device failure, and patient's burden (Fuller et al.,
Implantation of a full-sized system requires larger animals. As dogs are not available everywhere, international comparability among new interfaces would benefit from standardized models. Depending on the size of the implant, either a rabbit or sheep model should be used. A rabbit model is cheaper and easier during daily handling, in contrast to the sheep, which instead has advantages of fitting large implants. Anatomical size of involved nerves and muscles, as well as biomechanical stress to the implant is more realistic in the sheep model, which therefore is generally recommended. A large animal trial for a device for clinical application must at least have a duration of 3 months to be considered long-term (Anderson et al.,
Final testing could be conducted in an NHP model if the complexity of the interface requires this model (brain interfaces or peripheral interfaces with a large number of degrees of freedom for high-DOF hand prostheses, for example). However, for standard electrodes, a long-term application in a standard large animal (sheep) provides enough data on biocompatibility and mechanical stability (Bergmeister et al.,
Testing of new devices is important to improve the components of future bionic interfaces. To ensure patient safety, and to prevent complications in long-term use of novel devices, extensive preclinical testing is crucial. For this purpose, various animal models are available and can be used over different time points. As it is of utmost importance to find the right model for preclinical evaluation, we provide an overview of existing models, and describe the common approach for testing new implants. An iterative approach should be used from short-term rat models, to assess compatibility, safety and functionality of the implant, up to long-term large animal models, such as sheep, to evaluate data acquisition as well as long-term safety. International comparability is granted by a sheep model but not for cat and dog models, as these models are not available in every country. An NHP model can be chosen for complex trials, however limited availability due to ethical concerns as well as costs must be taken into account.
The ultimate goal of testing new interfacing devices is the implantation in humans for sophisticated prosthetic control and feedback, which should last for a lifetime and therefore be tested extensively in the correct animal model for biological and mechanical safety.
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.
All animal models from the authors' institution were conducted with permission of the ethics committee of the Medical University of Vienna and the Austrian Ministry for Research and Science (BMWF: reference number: BMWF-66.009/0309-WF/II/3b/2010, BMWF-66.009/0340-WF/II/3b/2016, BMWF-66.009/0024-WF/II/3b/2018).
MA, KB, and OA designed the concept. All authors analyzed data, contributed their specific expertise, wrote the manuscript, and revised it critically.
MR was employed by the company Otto Bock. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
The Supplementary Material for this article can be found online at:
Federation of European Laboratory Animal Science Associations
Non-human primate.