Animal models for critical-sized long-bone defects and long-bone non-union: a scoping review

Abstract

Background

Critical-sized long-bone defects and non-unions are among the most challenging complications in fracture care, as they lack the capacity for spontaneous healing and often lead to prolonged disability and repeated surgical intervention. Their complex aetiology involves mechanical, biological, and systemic factors that make consistent clinical management difficult. Preclinical animal models remain crucial for understanding the biological and mechanical factors underlying bone repair. This scoping review summaries existing animal models of long-bone defects and non-unions and identifies methodological gaps affecting reproducibility and clinical translation.

Methods

A comprehensive literature search across five databases (PubMed, Web of Science, Embase, Scopus, and BIOSIS) was performed. Preclinical in vivo animal studies with experimentally induced long-bone defects and non-unions that were published prior to October 2024 were included in the search. The following parameters were collected: animal species, number of animals, definitions of “non-union” and “long-bone defect”, specific bone studied, experimental setup (e.g., unilateral or bilateral model), method of defect creation, fixation technique, defect size, defect filling, additional therapeutic interventions, infection, experiment duration, non-union rate, and the employed surgical approach (single-stage or multi-stage).

Results

A total of 117 studies were included for analysis, whereby 54 studies addressed non-union and 63 focused on long-bone defects. Amongst the included studies, the most frequently investigated species were rats (40.2%), mice (21.4%), sheep (15.4%), and dogs (11.1%). Definitions for non-union and critical-sized defects varied widely, with the most commonly applied method determining non-union being radiographic analysis (74.4%) and time-based criteria (35%). The fixation methods applied were intramedullary nails (35%), external fixators (22.2%), and plates (28.2%), with the reported average experimental duration across studies of 116 ± 101 days.

Conclusion

Preclinical models for critical-sized defects and non-union remain highly heterogeneous, limiting reproducibility and translational relevance. Clear, species-appropriate definitions, improved reporting of defect and fixation parameters, and alignment of study duration with species-specific healing dynamics are essential to strengthen comparability. While model selection should remain driven by the research question, greater standardisation and the development of biologically faithful non-union models are needed to improve the clinical predictive value of preclinical research.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13018-026-06697-4.

Introduction

Musculoskeletal disorders are among the leading causes of disability and mortality globally, as highlighted by the Global Burden of Disease study [1]. Among these, significant clinical and socioeconomic challenges tend to result from critical-sized long-bone defects and fracture non-union [2–4]. These conditions may cause loss of function, substantial pain, and reduced quality of life, through lost Disability-Adjusted Life Years (DALYs), as well as imposing a significant economic burden [5–7]. The approximated costs resulting from treatment of persistent non-union is ranging up to £34,000 GBP per patient in the UK [8], whilst in the United States, total annual treatment costs for bone defects are estimated at $5 billion [9]. Another cost-identification query performed by Kanakaris et al. revealed costs of a “best-case scenario” humeral, femoral, and tibial non-union of, respectively, £15,566, £17,200 and £16,300 [10]. The Occupational Safety & Health Administration of the Department of Labour of the United States of America reports substantial losses in work hours and confirms that many Americans are affected by work-related conditions, for example back pain, tendinitis, carpal tunnel syndrome [11]. In addition, the European Agency for Safety and Health at Work (EU-OSHA) identifies musculoskeletal disorders as a leading cause of disability concerning healthcare workers [12].

Over the past decades, animal models have not only clarified fundamental principles of fracture healing but have also directly shaped clinical practice [13, 14]. For example experimental work demonstrated how mechanical stability influences healing, guiding the modern use of load-sharing intramedullary nails versus load-bearing compression plates [15, 16]. During the fracture-healing process, several outcomes can occur. Ideally, the fracture heals as expected, but some injuries progress more slowly, leading to what are known as delayed unions (e.g. Food and Drug Administration (FDA) defines a delayed union as the persistence of a radiologically visible fracture line between three to six months post-injury) [17]. If, after a reasonable period (e.g. FDA defines a non-union as the persistence of a radiologically visible fracture line at greater than nine months post-injury), there is still no clinical, radiological or biological sign of healing, the condition is classified as a non-union [17]. They are traditionally classified as atrophic, oligotrophic or hypertrophic, based on radiographic appearance and callus formation. Atrophic non-union show little or no callus formation, oligotrophic types display limited callus with reduced osteogenesis, and hypertrophic forms exhibit abundant callus formation [18]. However, there is no uniform definition for non-union. Across studies, it has been observed that these benchmarks lack uniformity [19–21].

Bone defects represent a specific subgroup of fractures where the segmental bone loss does not achieve spontaneous healing. Once the defect exceed a certain threshold (e.g. 1.5–2.5 times the diameter of the affected bone), healing will not occur without additional intervention, this situation is considered as a critical size defect [22–24].

Non-unions and to some extent critical-sized defects have a multifactorial aetiology, which includes mechanical instability, infection, impaired vascularisation, and systemic conditions such as metabolic or inflammatory disorders.

Despite advancements in surgical techniques and innovations in biomaterial-based interventions, there are still inconsistencies in clinical outcomes [25, 26]. Achieving satisfactory bone healing remains a challenge, which is characterised by radiographic closure of the defect, the ability to bear full weight, the absence of discomfort, and no signs of infection. This highlights the urgent need for a deeper understanding of the underlying pathophysiology, particularly in relation to developing more effective treatment strategies [19, 27].

The fixation system used in animal models plays a decisive role in determining experimental outcomes. By defining the mechanical environment at the defect site, fixation influences callus formation, vascularisation, and the overall course of bone regeneration. The evolution from basic intramedullary pins and external fixators to anatomically adapted plates and modular locking systems has expanded the range and precision of preclinical approaches [28–30]. However, this diversity also contributes to variability across studies and complicates comparisons between models. A clearer understanding of how fixation choice affects healing outcomes is therefore essential for improving reproducibility and translational relevance.

Thorough preclinical studies are essential for the framework of fracture healing research and the development of new interventional strategies prior to translation into human applications [31]. Despite the wide variety of animal models used in preclinical research, there is a lack of consistent definitions for non-union and CSDs, as well as varying methodologies across different studies. These inconsistencies pose significant challenges for comparing results and ensuring reproducibility. Consequently, researchers aim to generate more consistent data and establish standardised definitions; however, they often depend on a limited number of comparable studies to achieve this objective [32].

Models targeting small animals, such as mice, rats, and rabbits, are frequently employed due to their ease of handling, low cost, and short study duration. For molecular and genetic investigations, mice are particularly useful, whereas rats are often preferred for tissue engineering applications [33]. Even though these models are extremely advantageous in early-stage research, they lack the biomechanical and physiological characteristics of human bone [34]. In terms of similarity to humans, greater anatomical and biomechanical similarities can be extracted from large animal models, such as pigs, sheep, and dogs, which makes them highly valuable for translational and implant-related studies. However, due to higher costs, extended study durations and logistical constraints, their usage is quite limited [35]. Other complications in comparative research can be observed through variations in defect creation methods, defect size, fixation techniques, and follow-up periods [36]. The reliability and clinical significance of preclinical findings are constrained by the variations in models and methodologies [37].

Animal models for fracture non-union and critical-sized bone defects have progressively evolved from mechanically focused systems to biologically and translationally refined frameworks. The earliest models were primarily designed to study fixation stability in large animals, emphasising mechanical healing with limited insight into biological processes [38]. Subsequent advances in fixation technology and microsurgical techniques allowed for more controlled induction of non-union through periosteal stripping, vascular disruption, or infection, enhancing model reproducibility [27, 39–41]. Later, the introduction of standardized rodent models with defined radiological and time-based criteria [23] and the emergence of tissue-engineering approaches expanded experimental capabilities. More recent developments, including 3D-printed scaffolds, bioactive materials, and high-resolution imaging, have produced anatomically precise, species-specific models that more closely replicate human clinical conditions [16, 42–44]. This continuous refinement reflects the field’s transition from purely mechanical to integrated mechanobiological and translational research paradigms [45].

This issue is addressed by Garcia et al. (2013) through a proposed paradigm to define “union”, “non-union”, and “delayed union”, particularly targeting rodent models. They present an overview of existing models of delayed healing and non-union in rats and mice, with a summary of the benefits of various experimental approaches and evaluation techniques [23]. Garcia et al. defined non-union as a failure of bone bridging after 15 weeks in rats and after 12 weeks in mice [23]. A critical size defect was defined as a segmental bone defect leading to non-union [23]. Delayed union was defined as a delay in bone bridging compared to an adequate control group [23]. Bone union was defined as the initial formation of bone tissue across the fracture gap, as observed through histological or microCT analysis [23].

This scoping review differs from Garcia et al. (2013) by offering a comprehensive overview of the definitions of non-union and CSD across various animal species and models. The primary goal is to highlight under-explored areas that could benefit from further research. To aid in the planning of future studies on non-union and CSD, a website has been created that presents an overview of the data extracted from the included papers.

Methods

Eligibility criteria

This scoping review adhered to the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines to ensure methodological transparency, reproducibility, and accuracy [46] (Fig. 1). Using a PECO (Population, Exposure, Comparator, and Outcome) framework, the review focused on preclinical long-bone research employing in vivo animal models (P) for the investigation of critical-sized long-bone defects and non-union (E), included studies with any comparators or none (C), and mapped reported model characteristics, methodological parameters, and healing outcomes (O). The detailed inclusion and exclusion criteria are provided in Supplementary material 1.

Fig. 1.

Prisma Flow chart of inclusion of studies

Database search protocol

A systematic literature search was conducted in October 2024, across five electronic databases: PubMed, Scopus, Web of Science, Embase, and BIOSIS. Boolean operators and Medical Subject Headings (MeSH) keywords were utilised to capture relevant studies comprehensively. The detailed search strings employed were documented in Supplementary Material 2. Prior to initiating the full-text screening, the research team convened to define the parameters for data extraction, ensuring alignment with the study objectives. These parameters underwent pilot testing and subsequent refinement to guarantee the comprehensive capture of all relevant data.

Extracted parameters included study metadata, such as author, publication year, DOI, country of origin, animal species, number of experimental animals, control group details, and methodological nuances. Definitions of “non-union” and "’long-bone defect” as reported in each study were systematically documented. Other variables included the specific bone studied, experimental set-up (e.g., unilateral or bilateral), osteotomy method, fixation technique, defect size, and defect filling (if applicable). Additional therapeutic interventions and infection scenarios were recorded, along with experiment duration, non-union rate, and the employed approach (single-stage or multi-stage).

Data processing and organization

Extracted data were organized in an Excel file (Microsoft Corp., USA) To enable reliable comparisons and facilitate synthesis, extracted data were systematically organized into clusters based on their relevance: definitions of non-union, definitions of bone defect, defect creation methods, fixation methods, therapeutic interventions, defect filling methods, and infection models.

Extracted data

Definitions of non-union are categorised into time-based, radiographic, histological, biomechanical, and clinical markers. Further specifications (e.g., the defined periods, the radiological findings, etc.) are described in Supplementary Material 3. Critical-sized bone defects are primarily defined by size, estimated healing potential, semi-quantitative characteristics. When no further information was given the category “undefined entries” is used. The methods employed to create bone defects and induce non-union are categorised into distinct techniques. Fixation techniques are classified into five major categories: intramedullary fixation, external fixation, plate fixation, pin and clip fixation, and minimal or no fixation. Therapeutic interventions are investigated and, if additional therapies are used, classified into stem cell therapies, growth factors, pharmacological treatments, experimental therapies and scaffolds.

Defect filling methods are divided into five distinct groups: untreated defects, bone grafts, synthetic scaffolds, Bone morphogenetic protein (BMP)-enhanced materials, and specialized fillers. Infection parameters are classified based on the presence or absence of infection and specific pathogens used. Some studies explicitly induced infections, while others reported infection outcomes as unplanned events. These clusters are described in Supplementary Material 3.

Animal age and postoperative follow-up schedules were summarised descriptively across the included studies. Animal age was reported in months. Only numeric values were used to calculate means and standard deviations, while qualitative descriptors (e.g. adult, mature, young adult) and studies that did not report age were recorded separately. Radiographic and histological follow-up times were reported in days.

Statistical methods

To synthesise extracted data and identify recurring methodological characteristics across studies, descriptive quantitative analyses are conducted. A summary of categorical variables, such as fixation techniques and therapeutic interventions, are provided using absolute frequencies and relative proportions. Continuous variables such as defect sizes and healing durations are described using means and standard deviations, as measures of central tendency and dispersion. Since the objective is to characterize methodological trends rather than offer an assessment of statistical differences between groups, the study conducted inferential statistical analyses.

Results

This review selected 115 papers for inclusion based on the outcome of screening and confirmation of eligibility. As a result of a separate review of methods, two papers investigating different methodologies were counted twice, leading to a total of 117 reports (Fig. 1). Of the studies included, non-union was addressed by 54 reports, whereas 63 reports focused on long-bone defects.

Type of animals

Table 1 provides a summary of the distribution of animal species in the methods and categorised data of the numbers of animals used. Additionally, it summarizes the distribution, frequency and proportion of specific bones studied within each animal group, reflecting the focus and preferences in research methodologies.

Table 1.

Distribution of animal species in studies and categorized data (study details) and bone types and counts for each animal species

Animal species	n (%)	Total reported animal count	N° of studies with reported animal count	N° of studies with missing counts	Mean ± SD (animals per study)	Min. –Max	Bone	Relative use (%)
Rat	47 (40.2%)	1634	40	7	40.9 ± 31.7	5; 137	Femur Tibia Ulna Radius Fibula	32/47 (68.1%) 12/47 (25.5%) 1/47 (2.1%) 1/47 (2.1%) 1/47 (2.1%)
Mouse	25 (21.4%)	777	17	8	45.7 ± 28.9	14; 106	Femur Tibia	19/25 (76%) 6/25 (24%)
Sheep	18 (15.4%)	331	15	3	22.07 ± 11.41	8; 41	Femur Tibia Metatarsus	2/18 (11.1%) 15/18 (83.3%) 1/18 (5.6%)
Dog	13 (11.1%)	306	13	0	23.54 ± 20.2	3; 69	Femur Tibia Radius Ulna	4/13 (30.8%) 4/13 (30.8%) 3/13 (23.1%) 2/13 (15.4%)
Rabbit	9 (7.7%)	258	9	0	28.67 ± 20.7	5; 73	Femur Tibia Radius	2/9 (22.2%) 4/9 (44.4%) 3/9 (33.3%)
Axolotl	2 (1.7%)	36	1	1	36.0	36; 36	Femur Fibula	1/2 (50%) 1/2 (50%)
Pig	2 (1.7%)	22	2	0	11 ± 7.1	6; 16	Tibia	2/2 (100%)
Frog	1 (0.9%)	0	0	1	0.0	n.a	Hind limb	1/1 (100%)
Total	117 (100%)	3364	97	20	34,7 ± 27.47	3; 137

As indicated in Table 1, out of the 115 papers analysed, the most frequently used species are rats (47 studies, 40.2%). Sprague–Dawley rat is the most prevalent strain, followed by Wistar, Long Evans, and Fischer F344 rats. The mouse is the second most frequently used species, with investigations spanning over 25 papers (21.4%). The included mice are: C57BL/6, Balb/c, double transgenic (Col1/Col2), CD-1, and nude (Nu/J) strains. Sheep studies are covered in 18 papers (15.4%), targeting various breeds, such as Merino, North Holland, Bergamasca-Massese crossbred, and Charolais-Swifter. Canines are investigated across 13 papers (11.1%), through breeds such as beagles, mongrel dogs, coonhounds, and red tick hounds. Nine studies (7.7%), investigate rabbits predominantly New Zealand White rabbits. Across two studies (1.7%), the amphibian axolotl and the pig are each referenced. A single paper (0.9%) focuses on the description of a frog model.

Definition of non-union and bone defect

Out of the 54 studies that investigated non-union, a definition for non-union is not provided in one study (1.9%). Categorisation of non-union for the remaining 53 studies is based on a predefined classification system outlined in the methodology. Twenty-five studies employ a time-based definition, whereas 48 studies use a radiographic or imaging-based definition, 33 studies apply histological criteria, and 16 studies use a biomechanical-based definition. Additionally, specific supplementary conditions (e.g. pseudarthrosis with synovial fluid and critical-sized defects without regeneration) are incorporated in two studies (Fig. 2a).

Fig. 2.

A Distribution of non-union definitions; B Distribution of long-bone defect definitions

Among the 53 studies that provided reported definitions, eight (15.1%) use a single-category definition. Twenty-two studies (41.5%) employ a dual-category approach, most with combinations of radiological/imaging and time-based criteria. Additionally, in the definition of non-union, 20 studies (37.7%) use a combination of three categories. Three studies (5.7%) adopt a comprehensive approach that integrated various aspects: time-based, radiological, histological, and biomechanical assessments (Fig. 2a).

In 18 out of the 63 studies included the definition of “bone defect” is not provided (28.6%). Of the remaining 45 studies (71.4%), 38 provide definitions on the basis of the defect size, which is defined based on the diameter of the bone or the total length of the bone. In 11 studies, a bone defect is classified as critical based on healing criteria outlined in Supplementary Material 3. Quantitative characteristics are employed in seven studies, including criteria such as segmental or bone loss exceeding 50%, as specified in Supplementary Material 3.

Out of 45 studies, 36 (80.0%) apply a single criterion whereas seven studies (15.5%) employ two criteria. The application of quantitative characteristics and size-based definitions are the most prevalent combination. Two studies (4.4%) observe the application of all three criteria for the definition of long-bone defects (Fig. 2b).

Fracture and defect creation

Out of 117 studies, six (5.1%) do not provide sufficient methodological details about fracture or defect creation. Out of the 111 reported methods, 67 papers (60.4%) employed sawing or cutting techniques, with oscillating saws as the most frequently used tool. Fourteen studies (12.6%) employ drill-hole osteotomy. Twelve studies (10.8%) report mechanically induced fractures, either using externally fabricated devices or three-point bending mechanisms. Nine papers (8.1%) employ specialized micro-instrumentation (Fig. 3a). A combined approach is utilized in one study (0.9%), whereby the initial performance of a drill-hole osteotomy is followed by manual fracture completion.

Fig. 3.

A Methodology fracture/defect creation; B Techniques of non-union creation

Non-union creation

In 32 studies, the method for non-union creation is not specified. The induced defect is not treated in 48 studies, resulting in non-union. Twenty-eight studies document the surgical procedure of periosteal stripping. In seven studies, it is observed that non-union was caused by an infectious process. Nine studies induced non-union mechanically. Additionally, to impede bone healing, 17 studies employ a mechanical barrier such as glove or muscle. More details are reported in supplementary material 3.

Out of the 117 studies, 63 (53.8%) provide a description of a single method for establishing non-union. Twenty studies (17.1%) use a two-method combination, most often involving an untreated defect paired with periosteal stripping. A combination of mechanical instability, untreated defect, and periosteal stripping was used in two study to induce non-union (1.7%) (Fig. 3b).

Defect size

In 27 out of the 117 studies (23.1%), no report on defect size is provided. In contrast, explicit defect size information is included in 90 (76.9%) papers. Considering all animal models included, the average defect size is 10.14 mm, with a maximum size of 50 mm and a minimum size of 0.45 mm. The average defect sizes observed within the datasets are detailed in Table 2.

Table 2.

Average defect size by animal model

Animal model	Average defect size (mm)	Minimum (mm)	Maximum (mm)
Axolotls	1.82	0.7	3.5
Dogs	14.79	2.0	40.0
Mice	2.02	0.45	4.28
Rabbits	11.08	2	30
Pigs	26.67	5.0	50.0
Rats	4.45	1.0	10.0
Sheep	25.18	3.0	50.0
Frog	No data

Among the 117 studies included, 100 studies (85.5%) utilised a monolateral approach (i.e. experimentation on a single limb), whereas a bilateral design is reported in 15 studies (12.8%). Specifications on the usage of monolateral or bilateral approach are not provided in two studies (1.7%).

Species-specific non-union rates in animal models

The evaluation and stratification of the non-union rates of the models is divided into the corresponding studies: non-union and long-bone defect studies. Table 3 provides an illustration of the distribution of non-union rates (reported as percentages) by animal species across the 54 non-union studies, whereas Table 4 provides a summary of the non-union rates (reported as percentages) from the 63 long-bone defect studies.

Table 3.

Number of non-union studies by non-union rate (%), organized by animal model

	Non-union rate
	[0–20%]	[20–40%]	[40–60%]	[60–80%]	[80–100%]	Undefined	Total
Animal
Dog	0	0	1	0	4	1	6
Mouse	0	0	0	1	11	2	14
Rabbit	0	0	0	0	5	0	5
Rat	0	0	0	0	23	0	23
Sheep	0	1	0	2	3	0	6

Table 4.

Number of long-bone defect studies by non-union rate (%), organized by animal model

	Non-union rate
	[0–20%]	[20–40%]	[40–60%]	[60–80%]	[80–100%]	Undefined	Total
Animal
Axolotl	0	0	0	0	1	1	2
Dog	0	0	0	0	4	3	7
Frog	0	0	0	0	0	1	1
Mouse	0	1	0	1	3	5	10
Pig	0	0	0	0	1	1	2
Rabbit	1	0	0	0	1	2	4
Rat	1	0	0	0	9	15	25
Sheep	0	0	0	1	6	5	12

Axolotls, frogs, and pigs are not represented in the non-union studies. All detailed results of the species-specific animal models presented in Tables 3 and 4 are also displayed in the open-access online tool (Supplementary Material 8).

Fixation method

The method of bone fixation is not reported in six studies (5.1%). Forty-one studies (35.0%) employed intramedullary fixation, whereas 26 studies (22.2%) use external fixators. Thirty-three studies (28.2%) employed the use of plate fixation and pin and clip fixation in three methods (2.6%). Fourteen studies (12.0%) did not report fixation or unstable fixation. Furthermore, a combination of two different fixation methods was employed in six studies (5.1%). Figure 4 presents an overview of the fixation methods across different animal models.

Fig. 4.

Overview fixation methods by animal

Duration of trial

The duration of the experiment is not reported in one study (0.9%). The reported average duration, for the remaining 116 (99.1%) studies, is 116.27 days, with the shortest study lasting 21 days and longest 730 days. A detailed overview is presented in supplementary material 4.

Animal age and radiographic or histological postoperative time points

Animal age and postoperative follow-up schedules differed considerably between studies and across species, and in several cases age or follow-up timing was not reported.

The mean age was 78.0 months for axolotls; 26.1 ± 19.3 months for dogs; not reported for frogs; 3.5 ± 3.6 months for mice; 14.2 ± 16.6 months for pigs; 6.7 ± 4.2 months for rabbits; 3.9 ± 2.6 months for rats; and 49.8 ± 32.0 months for sheep (Supplementary material 5).

The mean time to first radiographic assessment was 10.5 ± 14.8 days for axolotls; 14.1 ± 21.5 days for dogs; 7.8 ± 9.1 days for mice; 14.0 ± 19.8 days for pigs; 24.1 ± 24.7 days for rabbits; 16.2 ± 29.2 days for rats; and 31.9 ± 88.1 days for sheep (Supplementary material 6). Further radiographic assessment time points are shown in Supplementary material 6.

The mean time to first histological assessment was 25.5 ± 6.4 days for axolotls; 61.5 ± 40.4 days for dogs; 28.8 ± 24.3 days for mice; 84.0 days for pigs; 58.0 ± 55.9 days for rabbits; 42.0 ± 33.2 days for rats; and 119.1 ± 99.8 days for sheep (Supplementary material 7). Further histological follow-up time points are shown in Supplementary material 7.

Discussion

This scoping review, which includes 115 papers, aims to provide a comprehensive overview of animal models related to CSDs and non-union.

Despite the use of numerous models, definition criteria and surgical techniques, the considerable methodological heterogeneity across studies poses a major challenge. These variations greatly impede reproducibility, hinder cross-comparison, and limit the generalisability of the findings. Substantial variations were observed in the definitions of CSDs and non-unions. The most common criterion for CSDs, was a defect size of 1.5–2.5 times the diameter of the bone, though there is no guarantee that this standard is universally applicable [23, 24]. Additionally, there was reliance on inconsistent combinations of time-based criteria, imaging techniques, and histological or biomechanical assessments when defining non-union. This variability mirrors the numerous challenges encountered in clinical research, where a universally accepted definition of non-union remains elusive [21, 47].

It is important to note that many preclinical studies replicate non-union by creating a critical-sized defect, although this approach does not fully capture the biological complexity of a true non-union [23, 48]. Genuine non-union models require deliberate disruption of the healing process, through impaired vascularisation, periosteal stripping, instability, or infection, and are therefore more difficult to standardize and reproduce [48]. While critical-sized defects offer insight into the limits of spontaneous bone regeneration, they do not inherently represent pathological non-union [49]. Distinguishing between these concepts is essential to improve the translational validity of preclinical findings and their alignment with clinical pathophysiology [23].

A collection of well-characterised rodent models and species-specific definitions proposed by Garcia et al. aims to improve comparability in research [23]. However, the limited adoption of standardized models can be attributed to the increasing trend of researchers using diverse and inconsistent models. This ongoing heterogeneity underscores the need for greater alignment on validated, reproducible models and standardized definitions.

Species-specific anatomical and physiological differences present challenges when it comes to translating research findings across various animals. For example, a defect measuring 4.45 mm on average may be deemed critical-sized in mice and rats. However, this measurement does not necessarily apply to larger animals, such as sheep, where an average defect size of 25.18 mm is considered critical. Furthermore, healing timelines tend to differ significantly between species. Dorafshar et al.’s suggestion of a definition system, formulated on the basis of bone length or volume rather than diameter alone, may provide a more consistent and translatable framework [50]. In addition, as highlighted in the consensus reported by Histing et al., animal age can influence bone healing and should therefore be carefully considered when selecting an animal model [51].

The observed lack of uniformity encompassed fixation techniques, experimental durations, postoperative imaging schedules, and histological assessment of healing. Although a healing period of 12 weeks is commonly used in mice and rats models due to their relatively rapid bone regeneration, larger animals may require observation periods of approximately 6–12 months. Aligning study durations and follow-up timepoints with the specific healing dynamics of each species could improve comparability and enhance the translational value of preclinical findings [28, 52].

The considerable heterogeneity observed across studies is not solely the result of methodological inconsistency but also reflects the evolution of available resources and technologies that have shaped preclinical orthopaedic research. Advances in fixation systems, such as external fixators, intramedullary nails, and anatomically contoured plates, have evolved alongside the progress of the orthopaedic device industry, enabling increasingly biomechanically relevant and species-adapted models [16, 43]. However, the implementation of these techniques is often influenced by institutional infrastructure, local expertise, and available funding, leading to substantial variability in the applied experimental approaches. Importantly, model selection should therefore be guided by the specific research hypothesis rather than by rigid standardisation. While harmonised reporting improves comparability, controlled diversity allows researchers to tailor models to distinct biological mechanisms and translational goals [53]. Additionally, practical considerations such as local availability of animal species and adherence to region-specific ethical committee guidelines further influence model selection and experimental design [38]. Thanks to advances in technology, researchers can now design and produce implants specifically for small animals. This development enables many research questions to be answered without relying on large animal models, thereby supporting both ethical standards and animal welfare. However, testing human implants still requires large animal studies. Not only are these studies more expensive and resource-intensive, but they also face growing ethical and political scrutiny surrounding animal use in research. Together, these factors often influence the choice of model at an early stage of the research process [16].

Fixation systems represent a cornerstone of experimental design in long-bone defect and non-union models, as they determine the mechanical environment that governs bone regeneration. The stability and configuration of the fixation influence callus formation, vascularisation, and the biological pathway of healing, ultimately defining whether repair occurs via endochondral or intramembranous ossification [28, 29]. The diversity of fixation techniques observed in the analysed studies, including intramedullary nails, external fixators, and plate systems reflects not only differences in anatomical site and animal size but also the rapid evolution of orthopaedic implant technologies. Contemporary approaches increasingly integrate modular and 3D-printed devices that allow fine-tuning of stiffness and load distribution, offering more physiologically relevant mechanical conditions [16, 43]. However, this technological variety introduces methodological heterogeneity and complicates direct comparison between studies. Future preclinical research should therefore report fixation parameters in detail, such as device type, material, mechanical properties, and applied load, to enhance reproducibility and enable meaningful cross-study interpretation. Systematic characterization of fixation-related variables will strengthen the translational bridge between preclinical findings and clinical orthopaedic practice.

Methodological transparency requires thorough documentation of surgical techniques and postoperative care, particularly regarding the various fixation methods that can independently affect outcomes, such as the rate of non-union. Within each fixation method, factors such as the type of implant, its size, diameter, and material also play a significant role in the outcome and should be reported in detail. Variability in these aspects can hinder the replication of findings and complicate the interpretation of results [28, 54, 55]. Therefore, future studies should prioritize adherence to standardized reporting guidelines.

A heterogeneous combination of radiographic, histological, and biomechanical methods was used to evaluate healing outcomes. However, across the studies, the specific parameters measured, and the thresholds applied varied significantly [56, 57]. While advanced imaging techniques such as micro-Computed Tomography (CT) and Magnetic Resonance Imaging (MRI) can provide high-resolution structural data, they are often costly (resource-intensive) and not universally accessible. Additionally, there is a general lack of standardisation in both protocols and the interpretation of molecular markers and biomechanical testing. To improve comparability in outcome assessments, it is important to integrate a combination of imaging, histology, and biomechanical evaluation, considering feasibility and following clearly defined thresholds. The reliability of preclinical findings could be further enhanced through the use or adoption of standardized scoring systems and consistent time points for evaluation [23, 58–60].

This review proposes several recommendations. First, it is essential to establish consensus definitions for critical-sized defects and non-union in animal studies that consider the specific anatomical and physiological characteristics of different species. These definitions should be based on measurable parameters, such as defect length or volume of the defect in relation to the dimensions of the bone, rather than relying solely on diameter [61]. Second, it is needed to align the duration of experiments with species-specific healing dynamics. For example, mice and rats models typically require 8 to 12 weeks, while large animal studies need approximately 6 to 12 months to fully capture the process of bone regeneration [62, 63]. When selecting fixation methods, it is important to consider their biomechanical relevance to the species and model being used. Furthermore, detailed descriptions of these methods must be provided to ensure reproducibility in research. Third, to enhance values for the research community, a centralized, open-access database should be established. This database would provide a catalogue of validated animal models, defect characteristics, fixation methods, and outcome metrics. In this manner, it would facilitate more effective study designs, minimize duplication of efforts, and promote the use of best practices.

Ethical considerations remain of utmost importance. It is important to prioritize efforts to reduce the use of animals by implementing rigorous study planning and incorporating alternative methods. These methods may include computational modelling and in vitro systems [64, 65]. Improved scientific rigor may be achieved through more efficient and targeted study designs which also support the 3Rs principle (Replacement, Reduction, and Refinement) [66]. This aligns with national and EU-level policies aimed at reducing the use of large animals in research, which in turn influences the availability and further development of certain animal models [67]. The potential in the refinement of bone defect research can be significantly improved through innovative approaches such as hybrid models. Hybrid models offer a combination of traditional in vivo studies with in silico simulations, organ-on-a-chip platforms, and 3D bioprinting. These technologies allow for a more precise investigations of biological mechanisms and may reduce reliance on animal models during the early stages of experimentation [42, 68]. Ultimately, the translation of preclinical findings into therapeutic innovations can be accelerated through the integration of multi-modal platforms.

The translation from preclinical to clinical research remains critically important [69–71]. In the development of effective treatments, animal models that mimic systemic human conditions, such as diabetes or osteoporosis, tend to be quite crucial to achieve positive progress [72–75]. To effectively align preclinical models with clinical needs, greater collaboration among basic scientists, clinicians, and regulatory agencies is necessary.

A practical outcome of this review, aimed at assisting researchers planning animal studies, could be the development of a digital tool (Supplementary material 4). This tool, which would utilize data retrieved from over 100 studies, could provide species-specific recommendations for fixation method, defect size, outcome measures, and observation periods. By promoting methodological consistency, this approach could lead to reduced resource use and enhance the translational potential of preclinical research.

Despite its comprehensive scope, several limitations are evident in this review. For instance, there may be language bias due to the exclusion of literature in languages other than English and German, which could lead to the omission of relevant findings. Additionally, the clustering of diverse methodologies for comparison purposes might have resulted in a loss of nuanced details. The review also shows the underrepresentation of specific long-bone models (humerus, ulna and radius), indicating a need for broader exploration of biological and biomechanical scenarios.

The variability observed in the reporting of methodological details limits the ability to draw universally applicable conclusions. This review emphasizes the urgent need for standardized protocols and more consistent reporting standards in preclinical research [76].

Conclusion

This scoping review demonstrates that preclinical models for critical-sized long-bone defects and long-bone non-union remain highly heterogeneous, with substantial variability in definitions, defect creation techniques, fixation strategies, and outcome assessments across species. Such diversity reflects the evolution of research goals but also creates barriers to reproducibility and limits the translational value of preclinical findings. Clear and species-appropriate definitions of non-union and critical-sized defects, together with systematic reporting of fixation parameters, defect characteristics, and follow-up methodologies, are essential for strengthening evidence generation in this field. Notably, model selection should remain driven by the research question, allowing flexibility when specific biological mechanisms or available technologies require a tailored approach. Small animal models, particularly mice and rats, continue to play a central role in preclinical fracture research. Advances in fixation systems and implant technologies have expanded experimental possibilities and increased model precision. Nevertheless, there remains a lack of a true non-union model that replicates the biological mechanisms of failed healing without relying solely on the creation of a critical-sized defect.

Enhancing methodological consistency, improving transparency in reporting, and aligning preclinical designs more closely with clinical challenges will be crucial for accelerating the development of reliable, effective, and clinically relevant strategies for managing long-bone defects and non-union.

Supplementary Information

Below is the link to the electronic supplementary material.

Funding

The authors declare that no funds, grants, or other support were received during the preparation of this manuscript.

Data availability

All data are provided in the supplementary material. Additionally, an open-access tool was developed to support future experimental planning (www.animal-models.org).

Declarations

Competing interests

The authors declare no competing interests.