Two investigators independently searched the following databases (inception to Oct 2022): PubMed, Embase, Cochrane Library, Web of Science, and CNKI (China National Knowledge Infrastructure) databases. The electronic search strategy used the following keywords: “Printing, Three-Dimensional”, “Titanium Mesh Cage”, and “anterior cervical corpectomy and fusion”. The search terms were adjusted according to the characteristics of each database, and we also examined the reference lists of the screened full-text studies to identify additional trials that might be eligible. And a third reviewer was consulted when the two reviewers could not reach a consensus.

The inclusion and exclusion criteria for this study followed the PICOS principle. (1) Participants: Patients with Cervical Spondylotic Myelopathy (CSM) or Ossification of the Posterior Longitudinal Ligament (OPLL) or other cervical spondylosis, who need to be treated with single-level anterior cervical corpectomy and fusion (SL-ACCF). (2) Intervention and Comparison: In the operation procedure, the implant is the Titanium Mesh Cage (TMC) or the 3D-printed artificial vertebral body, and no implants other than Titanium Mesh Cage (TMC) or 3D-printed artificial vertebral body and titanium plates were used during the procedure. (3) Outcomes: The study should include at least one of the following data: Operation time, intraoperative blood loss, subsidence of vertebral body; preoperative, and final follow-up C2–C7 Cobb angle, Japanese Orthopedic Association (JOA) scores, or Visual Analog Scale (VAS) scores. (4) Study design: Observational studies and randomized control trials were eligible. Surgery with more than one level, case reports, case series, commentaries, practice guidelines, systematic reviews and meta-analysis were excluded. In addition, duplicate studies with the same cohort or studies considered by consensus to be of low quality were excluded.

Data were extracted from the included studies as follows: (1) study design: first author, publication region, publication time, and study type; (2) sample demographics: number of patients, age, sex, and disease diagnosis; (3) surgery details: Type of implants and their details, operation time, intraoperative blood loss; (4) analysis variables: Severe subsidence of vertebral body; operation time, intraoperative blood loss, preoperative, and final follow-up C2–C7 Cobb Angle, Japanese Orthopedic Association (JOA) scores, and Visual Analog Scale(VAS) scores.

Two reviewers evaluated bias risk in the cohort studies using the Newcastle-Ottawa scale (18). Sensitivity analysis was performed by excluding a single study of each study in turn and reanalyzing the data. Publication bias was analyzed qualitatively by funnel plot.

The continuous variables were estimated by weighted mean difference (WMD), and dichotomous variables were estimated by using odds ratios (ORs) with 95% confidence intervals (CIs). The statistical heterogeneity of the pooled results was determined using the I² statistic. For this meta-analysis, we used the fixed-effect model when I² was greater than 50%, and if I²was less than 50%, a random-effect model was applied. The meta-analysis results were considered statistically significant when the p value < 0.05. The meta-analysis was performed using Review Manager 5.4 (Revman, The Cochrane Collaboration, Oxford, The UK).

A total of 129 articles from PubMed, Web of science, the Cochrane Library, and CNKI were initially identified. The exact number of articles identified in each database is as follows: PubMed (n = 59), Web of Science (n = 39), the Cochrane library (n = 7), CNKI (n = 14). 48 articles were excluded because of duplication, and 73 studies were excluded by screening the titles and abstracts for: irrelevant studies, case reports, non-comparative studies and review articles. Leaving 7 articles that underwent a comprehensive full-text analysis. Finally, 6 studies (19–24) were included in the final meta-analysis. The flow chart used for the new systematic review according to PRISMA 2020 is shown in Figure 1.

A total of 341 patients were enrolled in the 6 studies. The 3D-printed artificial vertebral body included 150 patients, and the TMC group included 164 patients. The characteristics of the included studies are presented in Table 1. The quality of included studies was evaluated according to the Newcastle-Ottawa scale, with scores above 7 (including 7) of high quality, and the evaluation results are shown in Table 2.

Preoperative evaluation indicators indicate the baseline level of patients before surgery, three studies (n = 141 patients; 65 in the 3D group and 76 in the TMC group) provided the preoperative cervical C2–C7 Cobb angle, three studies (n = 161 patients; 75 in the 3D group and 86 in the TMC group) offered Visual Analog Scale(VAS) scores and six studies(n = 314 patients; 150 in the 3D group and 164 in the TMC group) reported the preoperative Japanese Orthopedic Association (JOA) scores. No statistically significant difference was observed in the above indicators between the 3D group and TMC group, Combined effect value and heterogeneity test results: Preoperative cervical C2–C7 Cobb angle: P = 0.80, WMD −0.21 [−1.78, 1.36], Heterogeneity: Chi² = 0.75, df = 2 (P = 0.69); I² = 0%; Preoperative Visual Analog Scale(VAS) scores: P = 0.33, WMD −0.18 [−0.55, 0.18], Heterogeneity: Chi² = 0.86, df = 2 (P = 0.62); I² = 0%; and preoperative Japanese Orthopedic Association (JOA) scores: P = 0.95, WMD 0.01 [−0.32, 0.34], Heterogeneity: Chi² = 5.22, df = 5 (P = 0.39); I² = 4% [Figure 2]. which indicated that the included studies did not differ significantly between the 3D group and TMC group at baseline and that the observations were comparable.

Operation time and intraoperative blood loss were used for evaluating the surgical trauma. Five studies (n = 274 patients; 130 in the 3D group and 144 in the TMC group) provided operation time and four studies (n = 223 patients; 110 in the 3D group and 113 in the TMC group) provided intraoperative blood loss. No statistically significant difference exists in intraoperative blood loss between the two groups (P = 0.65, WMD −0.59 [−3.10, 1.92], Heterogeneity: Tau² = 1.63; Chi² = 3.96, df = 3 (P = 0.27); I² = 24%) [Figure 3]. In contrary, the 3D group showed significantly shorter operation time compared with TMC group (p = 0.04, WMD −6.77 [−13.15, −0.40], Heterogeneity: Tau² = 44.12; Chi² = 39.94, df = 4 (P < 0.00001); I² = 90%) [Figure 4], It may indicate that 3D-printed artificial vertebral body can shorten the operation time compared with conventional Titanium Mesh Cage(TMC), but there is no significant difference in the amount of blood loss during surgery.

Vertebral subsidence was used to assess the ability of the implant to maintain vertebral height. All included articles defined severe subsidence as a decrease of more than 3 mm in the height of the fused segment during the follow-up period after surgery compared with that immediately after surgery (25, 26). A total of 5 studies reported the severe subsidence of vertebral body (n = 254 patients; 120 in the 3D group and 134 in the TMC group). Statistically significant difference was observed in 3D group and TMC group (p < 0.0001, OR = 0.12 [0.05, 0.32], Heterogeneity: Chi² = 1.59, df = 4 (P = 0.81); I² = 0%) [Figure 5]. The incidence of severe subsidence of vertebral body in the 3D group was significantly lower than that in the TMC group, indicating that 3D-printed artificial vertebral body performed significantly better than traditional TMC in maintaining vertebral body height after operation.

The last follow-up time was defined as 1 year after surgery for all included studies. There studies (n = 161 patients; 75 in the 3D group and 86 in the TMC group) provided final follow-up VAS scores and four studies (n = 274 patients; 130 in the 3D group and 144 in the TMC group) reported the final follow-up JOA scores. There was no statistically significant difference between the two groups regarding final follow-up VAS scores (P = 0.98, WMD = −0.42 [−0.91, 0.06], Heterogeneity: Tau² = 0.10; Chi² = 4.71, df = 2 (P = 0.09); I² = 58%) [Figure 6] and final follow-up JOA scores (P = 0.16, WMD = 0.64 [−0.25, 1.54], Heterogeneity: Tau² = 0.84; Chi² = 30.11, df = 4 (P < 0.00001); I² = 87%) [Figure 7]. Although Japanese Orthopedic Association (JOA) scores and Visual Analog Scale (VAS) scores improved in the last follow-up in each study compared with preoperative assessment, there was no significant difference in the improvement between the two groups.

Similarly, the last follow-up time in the included studies was 1 year after surgery. Two studies (n = 101 patients; 45 in the 3D group and 56 in the TMC group) provided the final follow-up cervical C2–C7 Cobb angle. There was statistically significant difference in cervical C2–C7 Cobb angle between the 3D group and TMC group at final follow-up cervical C2–C7 Cobb angle (P < 0.0001,WMD 5.88 [3.04, 8.73], Heterogeneity: Tau² = 2.93; Chi² = 3.10, df = 1 (P = 0.08); I² = 68%) [Figure 8], The results showed that the cervical C2–C7 Cobb angle in the 3D group was larger than that in the TMC group at the final follow-up.

The results of this analysis showed that the heterogeneity of the operation time, final follow-up VAS scores, JOA scores and cervical C2–C7 Cobb angle was high. A random-effect model was used to partially eliminate the effect of heterogeneity, but the results showed that the heterogeneity was still high. After removing one of the included studies for each index (final follow-up VAS scores, JOA scores and cervical C2–C7 Cobb angle), we found that heterogeneity did not decrease significantly, but it did not affect the results, this shows that the results of this meta-analysis are relatively reliable. When it comes to operation time, when we removed each included article separately, the combined effect value shows a significant change or even no significance. Therefore, the conclusion that the 3D group had a shorter operation time than the TMC group is not reliable, and the large heterogeneity may come from the technical level of the surgeons.

Funnel plots were constructed to assess publication bias, and the results were largely symmetrical, indicating acceptable publication bias in our analysis. However, the distribution of JOA scores was significantly asymmetric at the last follow-up, suggesting that publication bias is likely (Figures 9–15).