The items of this meta-analysis were reported according to the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) (Additional file 3).
Article selection
Approximately, 2784 articles were found after the initial search. 2674 irrelevant studies were excluded via browsing the abstract and 45 articles were precluded because of unexpected outcomes and interventions. For example, some of the excluded studies used the cavity of the injured spinal cord, the function of bladder, motor evoked potential (MEP) and (SEP) sensory evoked potential as the endpoints. Unfortunately, these nonfunctional measures are not primary goals. Models used in the rest of the excluded studies are constructed by heat, ultrasound, electricity and magnetism and did not meet the preset criteria (damage caused by ischemia and force).
Finally, 65 experimental trials were taken into analysis, including 46 articles investigating rats and 19 articles investigating mice (Additional file 4).
Characteristics of included studies
The primary objectives utilized in SCI models were rats and mice. In rats, the most injured segment ranged from T8-T11(Liu, Huang, Jia, Li, Liang and Fu 2015;Qi, et al. 2019), however, the injured segment of mice was ranging from C5 to T12(Wan, et al. 2020;Wang, et al. 2019;Zhou et al. 2020). Four SCI models, including ischemia, strike, compression, and transection model, were performed in these studies. The agents were manually synthesized oligonucleotides or other vectors carrying these oligonucleotides. Agents were delivered by either tail vein injection or intrathecal injection. The drug was always immediately injected or continuously injected for 3 days (Table 1).
Assessment of Locomotor function in rats
We analyzed the SMD value of BBB score between SCI and miRNA treatment group at the time when the first and the last assessment of hindlimbs locomotor function were performed. It was demonstrated that significant locomotor function had been discovered at the first assessment (SMD 0.67, 95% CI: 0.43 to 0.90, p<0.01) in the treatment group. At the last assessment, a greater SMD value of BBB score was observed (SMD 3.90, 95% CI: 3.08 to 4.73, p<0.01) (Figure 2A). Then, the subgroup analysis demonstrated that none of these factors, including injection manner, vectors, and gender, significantly impacted the outcome from the final assessment (Figure 2B).
Subsequently, we calculated the SMD value of BBB score from the 1st to 35th day post-injury. Pooled analysis showed that miRNAs can effectively boost the locomotor function of rats and the SMD value of BBB score between two groups gradually increased over time (Figure 1C). Furtherly, we identified that miRNAs were effective in transection model, and strike model but not in the ischemia model (Figure 1D). As for the ischemia model, significant recovery of locomotor function was identified 7-days post-injury but this significance vanished at the 14th ,21st, and 28th day post-injury.
Analysis of up-and down-regulated miRNAs in rats
miRNA therapy aims to up-regulate or down-regulate the specific miRNA level in vivo. Hence, these articles were categorized based on the miRNA alteration. The outcome showed that miRNA therapy can remarkably promote the locomotor function of rats from the 1st to the 35th day post-injury (Figure 2).
Analysis of miRNA families in rats
We classified the miRNAs in rat modes of SCI into 9 families, (named miRNA miR-21-5p/34a-3p/124-3p/126-3p/223-3p/543-3p/30-3p/136-3p/15-5p) miRNA miR-21-5p/34a-3p/124-3p/126-3p/223-3p/543-3p families showed potent capacity in recovering strength of hind limbs of rats over the whole observation period (Figure 3). While, the other three families (miRNA-30-3p/136-3p/15-5) did not show any treatment effect.
Assessment of Locomotor function in mice
The recovery of locomotor function in mice was measured by the evaluation of the strength, BBB score and BMS. First, we analyzed the effect of miRNAs in recovering the strength of limbs in mice (Figure 4A). Interestingly, we found these miRNAs were effective in treating mice in SCI models at the 28th day post-injury; however, non-effectiveness was identified in the pair (SMD 4.88, 95% CI: -0.33 to 10.10, p=0.07), left (SMD 3.56, 95% CI: -0.68 to 7.80, p=0.10), and right (SMD 3.82, 95% CI: -0.33 to 7.98, p=0.07) forelimb(s) at the 14th day post-injury. Then, the analysis of BBB score and BMS showed similar results (Figure 4B). No significant promotion in BBB score and BMS of mice was identified at the 3rd and 7th day post-injury, respectively. Significant promotion of locomotor function of mice was observed until the 21st day post-injury at the earliest.
Analysis of quality score for rats
The 10 items of SYRCLE’s tool symbols 10 scores. If one item is marked as “+”, this study will get 1 score, otherwise, this study gets 0 (Table 2). To evaluate whether article quality impacts the final results, we divided all articles assessing the locomotor function of rats into poor quality (1-4 scores), middle (5-7 scores) and high-quality (8-10 scores) article manually.
Finally, 2 articles are marker as high quality, 16 articles are marker as low quality and 28 articles are marked as middle quality. The pooled results from both low quality and middle-quality articles yet demonstrated that miRNAs were potent in recovering hindlimbs locomotor function of rats whenever the BBB scale was performed at the 3rd, 7th, 14th, 21st, 28th, 35th day post-injury (Figure 5). On the first-day post-injury, only low-quality studies showed that rats receiving miRNA therapy had higher BBB score than SCI rats. As for high-quality studies, no hindlimbs locomotor function was discovered from the first day to the end.
Trial sequential analysis
TSAs were performed for rats at the end of the follow-up day in a random-effects model meta-analysis with an overall significance level (α) of 0.05 and a type II error risk (β) of 0.1 (i.e., power 90%) preset (Additional file 5). The cumulative Z-curve for rats crossed the upper monitoring boundary for benefit and the adjusted required information size was calculated as 423 accrued rats, confirming a beneficial effect of exosomes on locomotor recovery (Additional file 5A). Furtherly, we set the value as the least difference of BBB score and Z-curve crossed the upper monitoring boundary for benefit before reaching the adjusted information size 2117 (Additional file 5B).
Publication bias and quality evaluation of included articles
Publication biases for BBB scores of rats assessed at first and last measurement of hindlimbs locomotor function (Additional file 6 and 7), and strength (at the 35th day post-injury), BBB score (at the 28th day post-injury), BMS score of mice (at the 35th day post-injury) (Additional file 8-10) were exhibited as funnel plots. We evaluated the article quality using SYRCLE’s tool. The results showed that the randomness and blindness were nicely performed by most studies, while the rest merely reported either the randomness or the blindness. Other bias indexes were low risks. Overall, most of the included articles were middle and low quality, and only four articles were considered as high quality. The outcome of the quality assessment was provided in Table 2.