This systematic review was conducted by strictly adhering to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, as described in Supplementary Appendix 1 (20). Given the nature of this systematic review of published research articles, this study was waived for the requirement of ethical approval.

After establishing the inclusion and exclusion criteria for this study, two researchers conducted a comprehensive systematic search across several electronic medical databases, including PubMed, EMBASE, Web of Science (WoS), Cochrane Library, and MEDLINE Ovid. Literature searches were continued through November 2023, with articles systematically screened based on titles, abstracts, and full texts. Additionally, manual searches of the reference lists of identified studies ensured that no relevant research studies were overlooked.

The search strategy for this study was developed based on PubMed guidelines. To minimize the risk of omitting relevant literature, we employed “MeSH” terms to define specific search topics. We used Boolean operators AND and OR to combine MeSH terms with other search terms, refining the scope of the search and enhancing the precision of the results. This strategy was adapted to the specific requirements of the other four databases, as detailed in Supplementary Appendix 2.

The study selection process was conducted independently by two researchers, strictly adhering to the inclusion and exclusion criteria. Full-length research articles were progressively screened based on the title, abstract, and full text, followed by cross-checking the results after the initial screening phase. Subsequent refinements and verifications were performed by the third and fourth reviewers. In cases of disagreement, the researchers first discussed the discrepancy to reach a consensus. If the matter was not resolved, an tenth researcher was introduced to the team to make the final decision.

After completing the full-text screening, data extraction was independently carried out by two researchers from the selected studies. The extracted data were then compared and verified for integrity and accuracy by two additional researchers. Any discrepancies were resolved through discussions or consultation with a senior researcher. Collected data included essential details such as the first author, affiliations, publication year, study title, and research objectives, in addition to specifics about the study animals, including species, gender, age, and body weight, as well as general study characteristics, sample size, modeling methods, control groups, observation period, vertebral level, and assessment methods. The primary outcomes were muscle mass, CSA, FI, stiffness, fiber type composition, and fiber CSA. A comprehensive summary of these details is available in Supplementary Table 1.

Two researchers rigorously evaluated the quality and risk of bias of selected studies using the Systematic Review Centre for Laboratory Animal Experimentation's Risk of Bias (SYRCLE's RoB) tool (21), which assessed six domains, namely selection bias, performance bias, detection bias, attrition bias, reporting bias, and other biases. These domains were then divided into ten items and supplemented with a list of additional signaling questions to facilitate the assessment. The ratings of “yes,” “no,” or “unclear” answers corresponded to low, high, and unclear biases, respectively, where a higher number of “yes” responses indicated a higher study quality. Any discrepancies during the evaluation were resolved by a senior researcher. The assessments were performed using Review Manager (RevMan) version 5.4 software. Detailed results can be found in Supplementary Figure 1.

After conducting an extensive systematic search across multiple electronic medical databases, we initially identified 89 relevant studies from PubMed (n = 14), EMBASE (n = 16), WoS (n = 46), Cochrane Library (n = 0), and MEDLINE Ovid (n = 13). Another study was identified through citation searching and included in the screening process. After removing 34 duplicate articles, 56 studies were screened for title and abstract relevance. Then, 20 studies were screened out for the full-text analysis. Next, 11 articles were excluded based on inclusion and exclusion criteria as decided by two researchers. Finally, nine studies met the PRISMA standards and were included in the systematic review (22–30). The complete screening process and specific reasons for exclusions/inclusions at each stage are illustrated in Supplementary Figure 2.

The studies selected for the final analysis encompassed a publication span of 35 years, from 1988 (22) to 2022 (23, 29). Notably, two-thirds of those articles (6 out of 9) shared a common author, Paul W. Hodges (24, 25, 27–30) Detailed characteristics of the included studies are summarized in Supplementary Table 1.

Two researchers independently assessed the quality of nine animal model-related studies using the SYRCLE's RoB tool, which revealed a lower risk of bias in 36.67%, an unclear risk in 58.89%, and a higher risk in 4.44% of cases. Concerning the selection bias, all studies failed to adequately describe random sequence generation and allocation concealment, thereby receiving an “unclear risk of bias” rating. Of these, three studies (33.33%) (24, 27, 29) reported balanced baseline characteristics between the control and experimental groups, meriting a “low risk of bias.” In the area of performance bias, four studies (44.44%) (23–25, 27) ensured that experimental animals were housed under identical conditions, aligning with randomization principles and thus were also rated as “low risk of bias.” However, three studies (33.33%) (26, 29, 30) did not disclose the living conditions of the experimental animals, resulting in an “unclear risk of bias”, and two studies (22.22%) (22, 28) reporting their study animal housing in different environments were designated with a “high risk of bias.” Furthermore, the majority (88.89%) (23–30) of studies did not specify whether blinding was implemented for animal caregivers and researchers, thus resulting in an “unclear risk of bias” rating. Only one study (11.11%) (22) explicitly mentioned the inability to blind personnel due to the special design of cages and was marked as “high risk of bias”. Regarding the detection bias, none of these studies detailed the implementation of proper randomization methods for selecting animals for outcome assessment, leading to an “unclear risk of bias.” Additionally, four studies (44.44%) (22, 26, 29, 30) which blinded their outcome assessors, were classified as “low risk of bias,” while the remaining five (55.56%) studies (23–25, 27, 28) did not clarify this point, thus receiving an “unclear risk of bias” recognition. Regarding the attrition bias evaluation, one study (11.11%) (28) reported having an incomplete dataset due to unexpected death or exclusion of animals without addressing the impact on study integrity, earning a “high risk of bias.” Four studies (22, 25, 27, 30) reported complete outcome data and were thus assessed as “low risk of bias,” whereas the other four (23, 24, 26, 29) lacked explicit reporting on data completeness, resulting in an “unclear risk of bias.” All studies were evaluated as “low risk” for reporting and other biases. Detailed results of the bias risk assessments are available in Supplementary Figure 1.

The selection of appropriate animal models is pivotal for effective IVD degeneration research. Commonly, rats, mice, guinea pigs, hamsters, rabbits, dogs, cattle, sheep, goats, domestic pigs, chickens, horses, camels, kangaroos, and primates, such as baboons and macaques are exploited to develop the disease model (18, 35, 36). However, the majority of these models are quadrupeds, and their habitual postures and locomotive behaviors, influenced by gravity, significantly differ from those of humans. Even bipeds, such as primates and kangaroos, exhibit variations in posture and movement from humans; for example, kangaroos leverage their tails for spinal support, potentially mitigating IVD and PSM degeneration. Therefore, these animal models cannot fully replicate the natural degenerative processes observed in human IVD degeneration pathology.

In 1929, Colton et al. recognized the potential of bipedal models and developed a method to induce bipedalism in rats by forelimb amputation (37). Since then, this approach has gained traction, with numerous researchers refining the model to significantly enhance IVD degeneration outcomes (38–40). For instance, Liang et al. established a similar model using brachial plexus root avulsion (41, 42), while Ao et al. induced a bipedal posture in mice by leveraging their aversion to water, placing them in containers with a water depth of 5mm (43). While, Cassidy et al. opted for a bipedal rat model with ligation and excision of upper limbs above the humerus, and coupled with housing in specially designed cages that required standing for access to elevated food and water sources (22). By six weeks of age, these rats were capable of walking entirely upright and spent considerable time each day in this posture, thus closely mimicking the natural progression of IVD degeneration observed in humans.

Various methods have been developed by several research groups to model IVD degeneration animals beyond the bipedal rats, as described previously. These include the needle puncture method to injure the IVD; the blade injury model using surgical blades to disrupt the IVD structure; and the spinal instability model, which involves separating the posterior muscle complex, excising spinous processes, and severing ligaments. Other techniques include the tail curling/ring model, inducing force imbalance in IVD; the axial compression model, applying pressure under compressed and angled conditions; and the vibration model, subjecting animals to continuous vibrational forces. Furthermore, the injection of specific chemicals can also induce IVD degeneration and a combined application of aforementioned methods can also be exploited to establish an effective IVD degeneration model in rodents. Genetic modifications, such as in SPARC-null mice, SM/J mice, and LG/J mice, can disrupt IVD metabolism and induce loss of structural integrity, closely simulating natural degeneration processes in humans. Besides, the aging model of IVD degeneration focuses on the age-related structural and pathological changes in hamsters (18, 35, 36, 44–58).

This study aimed to underscore the importance of structural changes in the multifidus muscle in animal models of IVD degeneration. Of note, it is critical to minimize any impact on the multifidus muscle during the model construction. Consequently, the posterior approach is inappropriate as it inevitably damages the PSM, at least to some extent. On the other hand, the method analyzed in this study involved the slow insertion of needles through the rat's back into the L4/5 and L5/6 IVD segments. For the sham group, needles were inserted to pierce only the PSM without penetrating the IVD structure, thereby isolating the effect of PSM puncture itself (23). This technique simulates common clinical manifestations of human posterior lateral IVD lesions. Nonetheless, in most cases, artificial disturbances in PSM could be a major confounding factor leading to PSM degeneration secondary to the IVD degeneration.

Although the bipedal rat model closely resembles the natural degenerative processes of human IVD degeneration, however, is limited to its widespread application due to ethical concerns, modeling challenges, and prolonged observation periods. Recently, genetically modified and aging models have become the most effective disease animal models with spontaneous IVD degeneration without requiring any external physical or chemical interventions (28). However, the multifactorial nature of IVD degeneration pathology and the non-specific triggering of unknown pathophysiological mechanisms are the major bottlenecks in the widespread application of gene knockout IVD degeneration models (44). In contrast, anterior approach needle puncture and blade injury methods are globally recognized for IVD degeneration modeling due to their straightforward techniques, significant IVD degeneration post-modeling, and minimal interference with PMS damage.

When selecting the segmental level of IVD for model construction, it is important to consider the physio-anatomical structure of the modeling animal due to significant interspecies variations. In humans, degeneration of lumbar IVD predominantly occurs at the L4-L5 and L5-S1 levels, where 90% of disc herniations manifest, largely because of the sacro-lumbar intermediate position. Unlike lumbar vertebrae, sacral vertebrae in adults typically fuse into a single sacrum, lacking physiological movement and buffering capabilities. Therefore, subjecting the L4-L5 and L5-S1 IVDs to increased mechanical pressures and motion ranges, particularly during flexion and extension, can promote the IVD degeneration onset (59–62). Thus, it is advisable to select the two most distal segments of the lumbar spine to model IVD degeneration in animals. Experimental studies cited in this article, informed by relevant zoological literature, note that rodents usually possess 6 lumbar vertebrae, whereas rabbits have 7, and sheep and domestic pigs 6–7 (63–71). In all the nine studies reviewed, segments near the distal lumbar spine, especially L4 and L5, were consistently selected.

The severity of both LBP and IVD degeneration is closely correlated with the quality of the PSM (72). Consequently, establishing an effective evaluation system for PSM degeneration is crucial for numerous animal models of IVD degeneration to accurately assess the model's effectiveness. The PSM, comprising critical muscle groups surrounding the spine, including the multifidus, psoas, and erector spinae, plays a vital role in maintaining spinal stability and supporting physiological movement. A decline in PSM function can disrupt biomechanical balance and structural stability of the spine, leading to uneven stress on the IVD and accelerating its degenerative processes (73–76) PSM degeneration primarily manifests as reduced muscle mass and CSA, increased FI and stiffness, decreased muscle fiber diameter, changes in muscle fiber type compositions, and alterations in various inflammatory cytokines and gene expression levels. Patients with PSM degeneration exhibit decreased muscle strength, endurance, and plasticity (77–81). Among the nine studies analyzed, five studies evaluated the multifidus muscle from a macroscopic perspective. Overall, the study results support our initial hypothesis that at specific time points, IVD degeneration animals would exhibit significant levels of atrophy of the multifidus muscle, with marked reductions in mass and CSA, compared to sham animals. MRI and IHC revealed significant enlargement of multifidus muscle gaps and increased intramuscular lipid droplets in these animals. Furthermore, mechanical passive dynamic testing showed a significant increase (107%) in the stiffness of the multifidus in damaged IVD segments and their adjacent segments, compared to individual fibers (34% increase compared to control) (23, 25–27, 30). However, contrary to the expected hypothesis, Hodges et al. could not detect any significant atrophy of the multifidus at three and six months post-operative injury to the left IVD of Merino sheep via a retroperitoneal approach. The authors suggest that the influence of pro-inflammatory cytokines might prevent muscle atrophy during the transition from the subacute (3 months) to early chronic stages (6 months) of IVD injury (27). Some studies indicate that FI is more significantly associated with high-intensity pain, functional impairments, and abnormalities in spinal structure than muscle CSA, thus potentially serving as an effective indicator for evaluating muscle degeneration (12, 13, 82–84).

At the microscopic level, muscle quality assessment predominantly focuses on the CSA of muscle fibers and their type of distribution. Skeletal muscle fibers are categorized into two major subtypes: type I (slow-twitch) fibers, which primarily utilize aerobic metabolism for energy production, and exhibit higher endurance levels appropriate for prolonged physical activities such as posture maintenance; and type II (fast-twitch) fibers rely on anaerobic metabolism for rapid energy generation and are essential for activities that require short bursts of power like jumping or sprinting (85–87). In bipedal upright species such as humans and primates, lumbar muscles, bearing significant weight, predominantly consist of type I fibers. In contrast, quadrupeds like rodents, dogs, sheep, and horses mainly possess type II fibers (88–92). Research shows that the proportion of type I fibers in human multifidus muscles ranges from 54% to 63% (24, 80, 93), whereas in sheep, it is only 16% to 28% (24, 27, 29), and in rodents, it drops to 5% to 15% (28, 94). Experiments by Cassidy et al. on bipedal upright rats demonstrated a significant increase in type I fibers in their multifidus muscles, while muscle biopsies of human LBP patients indicate a lower proportion of type I fibers and thus poorer fatigue resistance, negatively correlating with the duration of pain (28, 95, 96). Recent research has expanded to include gene expressions of cytokines such as TNF-α and IL-6 to explore the specific pathophysiological mechanisms of LBP and the potential links between IVD and multifidus degeneration (23, 24, 27, 97, 98). Moreover, several studies incorporate behavioral assessments, including gait analysis, von Frey pain tests, hot plate tests, radiant heat tests, acetone tests, spontaneous pain tests, grip tests, and tail suspension tests, to comprehensively evaluate pain symptoms and muscle dysfunction in animal models (23, 99–104).

In summary, we propose the inclusion of several criteria in future evaluations of PSM degeneration: Firstly, radiologically, the assessment of CSA and FI are essential parameters. Secondly, the transition of muscle fiber types and their proportionate composition should be evaluated using immunohistochemical staining. Additionally, a behavioral assessment of muscle degeneration is warranted, encompassing strength, coordination, and motor function, as measured by the Basso-Beattie-Bresnahan Locomotor Rating Scale, grip strength test, hot plate test, and tail suspension test, to fully understand the impact on animal behavior. Finally, post-mortem, changes in inflammatory cytokines and gene expression levels, including TNF-α, IL-6, IL-1β, miRNA, and TGF-β, should be examined to assess muscle degeneration. We posit that these multidimensional indicators can form a scientifically robust evaluation system, offering a reference and theoretical foundation for research into IVD and PSM degeneration models.