The inclusion criteria include the following: (1) controlled studies that assessed the neurological effects of edaravone on rat models of SCI without any restrictions on language, publication date, or publication status; (2) laboratory rats of any age, sex, or strain that underwent SCI induced by compression or contusion injury; (3) the experimental group was treated by edaravone with no restriction on dosage, formulation, and administration methods; (4) the control group received either vehicle, saline, saline with diluted dimethyl sulfoxide, or no treatment; and (5) reporting one of the following outcomes: Basso, Beattie, and Bresnahan (BBB) locomotor rating scale, spared white matter area, or malondialdehyde (MDA) (Sng et al., 2022).

The following studies were excluded: (1) case reports, in vitro studies, clinical studies, reviews, and commentaries; (2) studies that used a rat model of SCI induced by non-traumatic ischemia, transection, laceration, traumatic root avulsion, or genetic modification; (3) studies combining edaravone with another treatment, without including a group treated with edaravone alone; (4) absence of control group; and (5) outcome assessments did not include BBB score, spared white matter area, or MDA.

Data extraction was independently conducted by two reviewers. Data collection includes the name of the first author, publication year, details of the study population, injury type and level, specifics of the intervention, and measurements. The proposed mechanisms and changes in related molecules were extracted to investigate the neuroprotective mechanism of edaravone. In studies encompassing multiple intervention groups, data from only the edaravone and negative control groups were extracted for analyses. The mean, standard deviations (SD), and sample sizes were collected from each group of animals. If the study data were represented visually without numerical values, we utilized GetData Graph Digitizer 2.26 (http://getdata-graph-digitizer.com) to extract the relevant statistical data from the graphical representations. In the event of missing data, the authors were contacted on a weekly basis. If no response was received from the original authors within 2 weeks, the related study was included but limited to qualitative analyses only (Yao et al., 2015).

The methodological quality of the included studies was assessed using SYRCLE’s RoB tool, a validated instrument for evaluating the risk of bias in laboratory animal experimentation within systematic reviews (Hooijmans et al., 2014). The checklist evaluates 10 items relating to selectivity bias, implementation bias, measurement bias, missed visit bias, reporting bias, and other biases. In each study, these domains were judged as “yes,” “no,” or “unclear,” indicating a low, high, or unclear risk of bias, respectively. The methodological quality assessment was conducted independently by two investigators, and any disagreements were resolved through consensus or the involvement of a third investigator (Zhou et al., 2022).

The data from the included studies were analyzed using RevMan 5.3 software. In studies with multiple intervention groups, we combined these groups to conduct a single pairwise comparison (Higgins and Green, 2011). The mean, SD, and sample size of animals in each group were used for comparisons. The effects of the interventions in each study were summarized using weighted mean differences (WMDs) or standardized mean differences, along with 95% confidence intervals (CIs). Heterogeneity among the included studies was assessed using the χ2 test and the Cochrane I² (Higgins and Green, 2011). The random-effect model was employed due to the expected heterogeneity resulting from the exploratory character of the animal studies. Statistical significance was defined as a P value less than 0.05. Linear graphs were constructed using GraphPad Prism software to illustrate the dynamic WMDs of BBB scores and highlight the variation trend in BBB scores of both groups over time.

Subgroup analyses were conducted to investigate the factors that modify BBB scores, including species, gender, injury type, route of administration, and times and administered dosage. Meanwhile, we conducted sensitivity analyses to evaluate the robustness of our findings and to examine the sources of heterogeneity by excluding studies that lacked outcome assessor blinding, small-sample studies, and individual studies.

In order to investigate the optimal dose, we employed a network meta-analysis approach based on the Bayesian method using Stata 12.0 (StataCorp LP, College Station, Texas, USA) to simultaneously compare the therapeutic effects of different treatment regimens by integrating both direct and indirect evidence. We calculated the probability of each treatment regimen occupying a specific rank and subsequently ranked those treatments based on their surface under the cumulative ranking curve (SUCRA) (Owen et al., 2020).

All studies, except one (özgiray et al., 2012), used BBB scores to evaluate locomotor recovery in rats following treatment with edaravone. The meta-analysis of BBB scores and the variation trend of BBB scores in both groups demonstrated that the edaravone-treated rats showed improved BBB scores compared to the control group. The WMDs between the two groups gradually increased from day 7 (seven studies, n = 246, WMD = 1.96, 95% CI = 1.23 to 2.68, P < 0.00001) to day 28 (seven studies, n = 222, WMD = 4.41, 95% CI = 3.19 to 5.63, P < 0.00001) after injury, and then maintained stably in the remaining time points (Figures 2A–C; Table 3).

Three studies reported the spared white matter areas in the injury epicenter using the Luxol fast blue method (Ohta et al., 2005; Ren et al., 2006; Ohta et al., 2011). All of them presented this outcome as the ratio of spared white matter area to total spinal cord area. Pooled results from those studies indicated an improvement in the spared white matter area in rats after edaravone treatment (three studies, n = 128, WMD = 10.69, 95% CI = 6.24 to 15.13, P < 0.00001; Figure 3A; Table 3).

Four studies reported the MDA content in the spinal cord tissue (Ishii et al., 2018; Ohta et al., 2005; Ren et al., 2006; özgiray et al., 2012). The pooled results demonstrated that edaravone intervention reduced MDA levels in the spinal cord compared to the placebo group (four studies, n = 71, WMD = −2.07, 95% CI = −3.56 to −0.58. P = 0.007; Figure 3B; Table 3).

Subgroup analyses revealed no significant differences in locomotor recovery between rats based on gender, strain, number of injections (single versus multiple), or injection routes (intravenous versus intraperitoneal); however, locomotor recovery may be influenced by the type of injury (Figures 4, 5; Supplementary Table S1). Dose was argued to critically impact the efficacy of the pharmacological intervention. While our analysis suggested that edaravone at doses of 5–10 mg/(kg·d) only exhibited a weak tendency to improve BBB scores compared with doses of <5 mg/(kg·d), there was no statistically significant difference. Additionally, administering edaravone at doses exceeding 10 mg/(kg·d) did not result in any further improvement in BBB scores (Figure 5C; Supplementary Table S1). The relatively insufficient discrimination in doses may potentially impact the results of subgroup analyses; further analyses are required to explore the appropriate dose.

Because BBB scores trended stably from the fourth week after SCI, we conducted network analysis using BBB score data from day 28 to day 42 post-injury to compare the effects of edaravone at different doses (Zhou et al., 2019). At day 28 after injury, the forest plot indicated 5–6 mg/(kg·d) edaravone increased BBB scores compared with controls, and increasing the dose beyond 5–6 mg/(kg·d) did not produce better results (Figure 6A). The SUCRA values gradually increased as the daily dose increased initially, then trended stably from 5–6 mg/(kg·d) to the maximal dose (Figures 7A, B). The analysis of BBB scores on day 35 and day 42 after SCI yielded highly consistent results (Figure 6B, 7C; Supplementary Figure S1).

Sensitivity analyses were conducted by excluding studies that lacked outcome assessor blinding, small-sample studies, and individual studies. The heterogeneity of the inter-study statistics remained to be moderate or high after using all exclusion methods. The results of sensitivity analyses showed that the improvement of BBB scores after edaravone intervention remained largely unchanged across various exclusion strategies (Supplementary Table S2).

Table 4 shows the proposed mechanisms of edaravone in the included studies. Collectively, the neuroprotective mechanisms of edaravone for SCI primarily involve the mitigation of oxidative stress, suppression of neuroinflammation, reduction in neuronal loss, and attenuation of ferroptosis.

Following SCI, a substantial accumulation of free radicals occurs rapidly and persistently at the site of injury, triggering a cascade of pathological events during the secondary injury phase (Zrzavy et al., 2021; Peng et al., 2023; Yu et al., 2023). Superoxide dismutase (SOD), a type of metalloenzyme, plays a pivotal role in scavenging free radicals of superoxide anions and serves as a crucial component in the detoxification of ROS (Wang et al., 2018). While MDA is mainly produced as a product of the lipid peroxidation reaction triggered by ROS, it subsequently amplifies the occurrence of secondary harmful reactions and the generation of ROS (Wang et al., 2019). The antioxidative capacity of edaravone has been proved in numerous central nervous system (CNS) diseases (Jiao et al., 2015; Ahmadinejad et al., 2017; Jaiswal, 2019). Five of the included studies reported a decreased level of MDA or an enhanced SOD content in rats with SCI treated with edaravone (Ishii et al., 2018; Ohta et al., 2005; Ren et al., 2006; Ohta et al., 2011; özgiray et al., 2012). The meta-analysis revealed that edaravone resulted in a reduction in MDA levels in the spinal cord compared to the placebo group, verifying the favorable antioxidative role of edaravone in treating SCI.

The combination of a direct and indirect role in free radicals gives edaravone a powerful antioxidative property (Watanabe et al., 2018; Ismail et al., 2020). At a physiological pH, approximately half of edaravone exists as edaravone anions, allowing this agent to directly trap hydroxyl radicals and quench active oxygen by electron transfer or reaction with oxygen molecules (Watanabe et al., 2018; Higashi et al., 2006). In addition, various signaling pathways act as intermediary links for the antioxidant mechanism of edaravone (Ismail et al., 2020; Zhang et al., 2019). Nuclear factor erythroid 2-related factor 2 (Nrf2) and sirtuin 1 (SIRT1) can modulate mitochondrial homeostasis and antioxidant enzyme activity to defend against oxidative stress, while inactivation of these molecules occurs in SCI and related CNS diseases (Samarghandian et al., 2020; Jiang et al., 2023). In a pheochromocytoma cell 12 (PC12) model of neurotoxicity, edaravone was shown to enhance Nrf2 expression, reduce the generation of ROS, and promote cell survival, which was abolished by Nrf2 gene knockdown (Shou et al., 2019). Meanwhile, edaravone treatment has been shown to activate Nrf2, heme oxygenase 1, and SIRT1 in experimental models of traumatic brain injury, amyotrophic lateral sclerosis, and stroke (Ismail et al., 2020; Cui X. et al., 2022). Collectively, the unique structure of edaravone, along with the involvement of Nrf2 and SIRT1, potentially mediates its role in attenuating oxidative stress to mitigate disorders in neurological function and tissue structure in SCI.

Microglia, the resident immune cells in the CNS, respond to the initial injury of the spinal cord within a few hours and facilitate the activation of astrocytes as well as infiltration of peripheral immune cells at the injury site. Subsequently, an inflammatory cascade is initiated (Hu et al., 2023). Scavenging of ROS is of significance for ameliorating neuroinflammatory responses (Zrzavy et al., 2021). Correspondingly, Pang et al. (2022) demonstrated that edaravone effectively suppresses the activation of microglia and astrocytes, thereby mitigating inflammatory responses in SCI. Moreover, the expression of anti-inflammatory factors, such as IL-10 and IL-13, was upregulated by edaravone to effectively attenuate neuroinflammation (Pang et al., 2022). Nuclear transcription factor-κB (NF-κB) and the NOD-like receptor family pyrin domain-containing 3 (NLRP3) inflammasome function as downstream signals of ROS, playing pivotal roles in regulating persistent inflammatory responses during SCI (Gao et al., 2019; Liu et al., 2020). In HT22 cells stimulated by amyloid β or hydrogen peroxide, edaravone was shown to ameliorate inflammatory response and cell apoptosis by diminishing ROS generation, inhibiting NF-κB p65 phosphorylation, and suppressing NLRP3 inflammasome activation (Zhao Z. Y. et al., 2013; Guo et al., 2023). Edaravone reduced the expression of NLRP3, caspase-1, and associated speck-like protein containing a caspase-recruitment domain to blunt the microglial activation induced by lipopolysaccharide (Li J. et al., 2021). Evidence from the in vivo studies confirmed the inhibitory effect of edaravone on ROS generation, NLRP3 inflammasome, and nuclear transcription of NF-κB in multiple neurological disorders (Akiyama et al., 2017; Miao et al., 2020; Meng et al., 2021). Therefore, the suppression of neuroinflammation by edaravone may be attributed to its ability to inhibit free radicals and downstream signaling pathways, including NF-κB and NLRP3 inflammasome.

The generation of large numbers of free radicals and a vigorous inflammatory cascade response following injury can notably disturb the organelle function and homeostasis in neuronal cells, leading to numerous neuron apoptosis (Zrzavy et al., 2021; Sun et al., 2024). In the experimental treatment of SCI, it was reported that edaravone increased the survival rate of neuron cells in the lesion area (Ishii et al., 2018; Pang et al., 2022). Our meta-analysis also revealed edaravone can mitigate tissue damage in the lesion area. The B-cell lymphoma-extra-large (Bcl-XL) and caspases are involved in the cell apoptosis event resulting from oxidative stress (Liu et al., 2022). Administration of edaravone could result in a significant decrease in caspase-3 positive cells and a rise in Bcl-XL levels in rats with SCI (Wang et al., 2013). As mentioned above, Nrf2 and SIRT1 are important regulators of oxidative stress (Samarghandian et al., 2020; Jiang et al., 2023). Interestingly, edaravone was suggested to effectively boost the activation of Nrf2 and SIRT1 to improve the neuron apoptosis caused by excessive generation of ROS during brain injury (Shou et al., 2019; Li X. et al., 2021; Cui X. P. et al., 2022). Collectively, a rational molecular chain comprising Nrf2 (or SIRT1), ROS, and Bcl-XL (or caspase-3) may account for the anti-apoptotic mechanism of edaravone in treating SCI and related CNS diseases.

The rapid and persistent oxidative stress can induce iron overload, impairing the activity of glutathione peroxidase 4 and reinforcing the buildups of ROS in return. This leads to structural damage to the cellular membrane and ultimately causes cell death (Feng et al., 2021; Zhang et al., 2022). Notably, edaravone treatment after SCI was suggested to upregulate anti-ferroptosis proteins, including glutathione peroxidase 4 and system Xc-light chain, and exert inhibitory effects on pro-ferroptosis proteins (Pang et al., 2022), implying that an anti-ferroptosis mechanism participates in the neuroprotective effect of edaravone. The detailed molecular mechanism behind the anti-ferroptosis of edaravone remains to be further illustrated in future studies.

Taken together, we hypothesize that the core mechanism through which edaravone exerts its neuroprotective effects is by attenuating oxidative stress. Partly due to its antioxidant effects, edaravone then counteracts neuroinflammatory responses, attenuates neuronal cell apoptosis, and suppresses ferroptosis through diverse molecular pathways for the treatment of SCI (Figure 8). However, more high-quality studies are required to verify this hypothesis.