The study protocol is registered on the PROSPERO website under the registration number CRD42023438831.

We performed a comprehensive search of PubMed, Embase, Web of Science, and the Cochrane Library from their inception to October 2023 following PRISMA guidelines (5). The search strategy included a combination of terms related to anterior and posterior approaches and CSCI: (1) anterior OR anterior cervical corpectomy and fusion OR ACCF; anterior cervical discectomy with fusion or ACDF; (2) posterior OR laminectomy OR LA OR laminoplasty OR LP; and (3) cervical spinal cord injury OR CSCI; and (1) and (2) and (3). The reference lists of all relevant retrieved articles and reviews were manually searched to identify additional studies that might have been missed. Two independent reviewers screened the titles and abstracts, and full-text articles were obtained for further evaluation.

Studies were included based on the following criteria: (1) Study design: Randomized or non-randomized controlled studies; (2) Study population: Persons with a history of traumatic CSCI; (3) Interventions: Comparison of clinical outcomes between anterior and posterior approaches; (3) Primary outcomes: Preoperative and postoperative Spinal Injury Association (ASIA) and Japanese Orthopedic Association (JOA) scores, calculated recovery rates (RRs) (%); (4) Secondary outcomes: Complication rate, operation time, intraoperative blood loss, and length of stay. Exclusion criteria included duplicate reports, non-English studies, studies lacking a control group, studies only with abstract or with unavailable statistical data. Reviews, case reports, letters, comments, animal trials or cadaver studies were also excluded.

Data were extracted independently by two authors, and discrepancies were resolved through discussion with a third author. The following information was extracted: (1) Basic information including title, first author, year of publication, and study design; (2) Sample characteristics such as sample size, sex, and age; and (3) Interventions and clinical data including surgical approach for each group, preoperative and postoperative ASIA and JOA scores, RRs calculated based on JOA score, complication rate, operation time, intraoperative blood loss and length of stay.

The quality of the included studies was independently evaluated. Newcastle Ottawa Scale was introduced for cohort studies, and the Cochrane Risk of Bias tool was applied for randomized controlled trials, respectively. The Newcastle‒Ottawa Scale includes 3 domains: quality of selection, comparability, exposure, and outcome of study participants. A maximum of 9 points was assigned to each study, including 4 for selection, 2 for comparability, and 3 for outcomes. A study with a final score >6 was regarded as high quality (6, 7). The Cochrane risk of bias tool covers six domains of bias: selection bias, performance bias, detection bias, attrition bias, reporting bias, and other bias11. Each of these factors was recorded as yes (“low” risk of bias), no (“high” risk) or unclear.

Data were analyzed using Review Manager (RevMan, Version 5.0). We employed either the inverse-variance method or the Mantel-Haenszel test to calculate effect sizes and their corresponding 95% confidence intervals. Forest plots were generated for visual representation of the data. We adopted a fixed-effects model to assess the overall impact of each approach on the primary and secondary outcomes. Statistical heterogeneity among the included studies was evaluated using the Cochrane Q test and quantified by the I² statistic. In cases where the p value was less than 0.05 or the I² statistic exceeded 50%, indicating significant heterogeneity, a random-effects model was employed for meta-analysis. Otherwise, a fixed-effects model was utilized. A p value of less than 0.05 was considered to indicate statistical significance (8).

Our search yielded 1,433 studies, of which five met the inclusion criteria after screening (Figure 1) (9–13). All five studies were conducted in Asia, with publication years ranging from 2003 to 2023. The sample sizes varied from 47 to 180, totaling 613 persons (320 with the anterior group and 293 with the posterior group, respectively). The follow-up length ranges from 6 months to 17 years (Table 1). The quality of the randomized control study (9) was evaluated by the Cochrane risk of bias tool (Figure 2), and quality of the included cohort studies (10–13) was evaluated using the Newcastle‒Ottawa scale (Table 2).

All five studies included in this meta-analysis reported both initial ASIA scores at admission (preoperative) and final ASIA scores at the end of the follow-up period (postoperative). Statistical analysis using the Chi-square test revealed no significant heterogeneity for either preoperative or postoperative ASIA scores (I² = 0%, p = 0.44 and I² = 0%, p = 0.43, respectively). Furthermore, no significant differences were observed between the anterior and posterior surgical approaches in terms of preoperative (WMD = −0.04, 95% CI −0.12 to 0.03, p = 0.26; Figure 3) or postoperative (WMD = −0.03, 95% CI −0.12 to 0.07, p = 0.59; Figure 4) ASIA scores.

Similarly, the initial JOA scores for persons undergoing either anterior or posterior surgical approaches were comparable (WMD = −0.01, 95% CI −0.29 to 0.26, p = 0.93; Figure 5). Postoperative JOA scores also showed no significant difference between the two groups (WMD = 0.44, 95% CI −0.27 to 1.15, p = 0.23; Figure 6). It is noteworthy that while the preoperative JOA data exhibited low heterogeneity (I² = 9%, p = 0.35), the postoperative JOA data demonstrated high heterogeneity (I² = 87%, p < 0.0001).

RR was quantified using the formula:RR(%)=JOAscoreatlastfollow-up-PreoperativeJOAscore17-PreoperativeJOAscore×100This metric was employed in three of the studies to gauge the extent of neurological functional improvement (10, 11, 13). The meta-analysis revealed no significant difference in neurological RR between the anterior and posterior surgical approaches (WMD = −0.75, 95% CI −2.22 to 0.72, p = 0.32; Figure 7). Additionally, the Cochrane Q test indicated an absence of heterogeneity among these studies (I² = 0%, p = 0.69).

In the present meta-analysis, a cohort of 445 persons was evaluated across four studies (9, 11–13). Of these, 61 persons (25.8%) who were participated in the anterior surgical approach and 50 persons (23.9%) who underwent the posterior approach reported postoperative complications. The predominant complications encompassed odynophagia, hoarseness, neck pain, and wound infection, as delineated in Table 3.

Upon statistical scrutiny, no discernible difference was observed in the frequency of complications between the anterior and posterior surgical approaches (OR = 1.07, 95% CI 0.41–2.80, p = 0.88; Figure 8). It is noteworthy that a significant level of heterogeneity was detected among the included studies (I₂ = 74%, p = 0.009).

To further validate these findings, a sensitivity analysis was executed, focusing on universally reported complications such as infected wounds and hoarseness. This analysis corroborated the stability of our primary results, revealing no heterogeneity (I² = 0%, p = 0.90) and an OR of 0.90 (95% CI 0.43–1.87, p = 0.77; Figure 9).

In this meta-analysis, operation time and intraoperative blood loss were evaluated across three studies, encompassing a total of 398 persons—216 treated with the anterior approach and 182 with the posterior approach (11–13). Statistical analysis revealed no significant disparities in either operation time (p = 0.72; Figure 10) or intraoperative blood loss (p = 0.09; Figure 11) between the two surgical approaches. However, it is imperative to note the presence of substantial heterogeneity in these datasets (I² = 98% for operation time, and 100% for intraoperative blood loss). This heterogeneity is postulated to emanate from variations in the specific surgical techniques employed, precluding the possibility of subgroup analyses due to the limited scope of the included literature.

Two studies endeavored to elucidate the differences in hospital stay duration between the anterior and posterior surgical approaches (11, 12). Statistical analysis yielded a high degree of similarity (WMD = −2.72, 95% CI −6.80 to 1.35, p = 0.19; Figure 12), albeit with pronounced heterogeneity (I² = 99%, p < 0.0001). We posit that this heterogeneity is likely attributable to variations in hospital environments, as the studies sourced their samples from different settings, thereby introducing significant disparities.