Following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, we performed a meta-analysis of clinical studies that compared the timing of surgery after preoperative tumor embolization. A systematic search of PubMed, EMBASE, Web of Science, and the Cochrane Database was conducted on December 23, 2021, independently by two separate reviewers (W.T.Y. and C.H.L.). The search terms used were “spine” (or “spinal”) AND “metastasis” (or “metastases”) AND “embolization” AND “schedule” (or “timing”). We included only papers written in English, and only the largest study was included if there were overlapping study populations. After excluding duplicate articles, the search results were screened by title and abstracts for the following exclusion criteria : reviews, case reports, letters, experiments; no adjuvant TAE; and no tumor resection. After eliminating the excluded papers, full-text articles were obtained, and studies were thoroughly screened again using the same exclusion criteria. We excluded articles for the following reasons: no inclusion of spinal metastasis; a comparison of blood loss with and without TAE; and the use of another evaluation unit for blood loss (e.g., hemoglobin). We also examined the references of all included papers to find other relevant articles. Any inconsistencies between the two reviewers were resolved by discussion and agreement.

The goal of the search was to identify articles that met the following inclusion criteria : direct comparative studies of EBL stratified by the timing of surgery after TAE for spinal metastasis. Studies were included in the data extraction if they reported EBL in an early and late interval between TAE and surgery. When available, we recorded characteristics such as the number of patients, age, sex, country, study period, material of TAE, degree of TAE, and the definition of time interval. Outcomes were extracted from the texts and tables of the studies. If data from individual cases were available, we collected all data, including the time to surgery and EBL.

Quality assessment was done independently by a pair of authors (W.T.Y. and C.H.L.); any inconsistencies between them were resolved through discussion. Study quality was determined for controlled observational cohort studies with the Risk of Bias Assessment Tool for Nonrandomized Studies (RoBANS). Summary estimates were evaluated using the calculated effect size based on the mean difference (MD) and 95% confidence intervals (CIs), using Review Manager (RevMan) version 5.4. The Cochrane Collaboration, 2020. The data from each included study were weighted by the inverse of the variance of the results, thereby accounting for both within- and between-study error. Heterogeneity was assessed in all meta-analyses using the Higgins I2 statistic and the Cochran Q test. According to the heterogeneity, a random-effect or fixed-effect model was chosen. We performed a subgroup analysis to examine whether the effects varied according to the degree of embolization. We assessed publication bias by visual inspection of funnel plots and by calculating the p-value (one-sided) for Egger’s intercept. All tests were two-sided, with p-values less than 0.05 deemed significant.

An initial literature search resulted in six papers in PubMed, eight in EMBASE, 53 in the Web of Science, and 13 in the Cochrane Library. Of these 80 studies, 19 were excluded as duplicates. After screening titles and abstracts, 33 papers were excluded, because 15 were reviews, case reports, letters, or experiments, 14 were not about adjuvant TAE, and four were studies without tumor resection. The 28 remaining studies were subjected to full-text review, resulting in the exclusion of 21 studies. The reasons for the exclusion of these articles were no inclusion of spinal metastasis (n=9), studies comparing EBL with and without TAE (n=11), and the use of another evaluation unit (hemoglobin) for blood loss (n=1). Finally, seven papers were included in this meta-analysis. The detailed results of the selection process are shown in Fig. 1.

Table 1 summarizes the characteristics of the studies included in this analysis, which contained 390 participants [9,10,14,20,21,23,24]. All seven studies were retrospective studies, and the study period was from 1999 to 2014. The degree of tumor embolization was divided into complete, partial, and/or near complete, and the definition of complete embolization was usually 90% (ranging from 75% to 90%). The reported rate of complete embolization was 49.5% to 92.8%. The most frequently used embolization material was PVA particles, followed in descending order by gelatin sponge, coil, and NBCA glue.

Two of the seven studies included both metastatic tumors (n=120) and primary bone tumors (n=90) [23,24]. This may have increased the heterogeneity of the study target. However, these two studies were finally enrolled in this meta-analysis because the patients with primary bone tumors were distributed evenly across each group, and both groups of tumors were diagnosed as hypervascular by the clinicians. The other five studies included only metastatic spine tumors.

The definition and frequency of complete embolization varied among the included studies. Four papers defined complete embolization as a 90% decrease of tumor blush [10,20,21,23], and two papers used thresholds of 80% and 75% [9,14]. One paper did not clearly describe the criterion that was used [24]. The frequency of complete embolization ranged from 30% to 93%.

Regarding the timing of surgery, the criteria used to distinguish between early surgery and late surgery were roughly 1–2 days after TAE. Three papers defined the early surgery as an operation within the same day after embolization; two papers defined it as an operation within 24 hours; one paper defined it as an operation within 2 days; and one paper defined it as an operation within 48 hours. In a subgroup analysis, we collected data from individual patients in which the timing of surgery after embolization and the EBL were recorded.

Among the 390 patients from seven studies, 196 patients underwent surgery early after TAE, and 194 patients had surgery later. The proportion of male patients was 60.3%, and the mean age at surgery was 56.8 years in the included studies. Six studies of 261 patients reported the exact histologic type of tumor, and renal cell carcinoma was the most common tumor pathology (n=104, 39.9%), followed by thyroid cancer (n=50, 19.2%). In all seven studies, there were a total of five reported complication cases : two cases of acute stroke and one case each of cord ischemia, severe hypertension, and transient visual field defect. All complications occurred within several hours after TAE.

In the overall comparison, the early surgery group underwent surgery within 1–2 days after TAE, and the late surgery group received surgery at least 1 day after TAE. All studies demonstrated a consistent tendency for less EBL in the early surgery group, but without statistical significance (MD, 86.3 mL; 95% CI, -95.5 to 268.1 mL; p=0.35) as shown in Fig. 2. We divided the results according to the timing of surgery (within 1 day/24 hours and within 2 days/48 hours), and neither group showed statistical significance. Five studies used within 1 day as the standard of timing, and the MD of EBL was 64.6 mL (95% CI, -122.9 to 252.1 mL; p=0.50) [9,10,14,20,23]. The others used 2 days, and the MD of EBL was 425.0 mL (95% CI, -315.4 to 1165.3 mL; p=0.26) [21,24].

We selected individual cases of complete embolization with a time interval of within 1 day in three studies [9,10,21]. In the data from individual participants who underwent complete embolization and surgery within 1 day after TAE, the early surgery group showed 248.7 mL (95% CI, 84.4 to 413.0 mL) less bleeding than the late surgery group. This difference was statistically significant (Fig. 3). One study that reported a 93% rate of complete embolization was also included in the subgroup analysis [23]. This analysis also showed a significant difference in EBL (MD, 233.3 mL; 95% CI, 76.0 to 390.5 mL; p=0.004), as presented in Fig. 3.

To clarify the effect of degree of TAE, we divided the data into complete and partial embolization. Seventy-three patients from three studies were described as undergoing complete embolization [10,20,21]. Overall, the complete embolization group showed significantly less EBL than the partial embolization group (Fig. 4), and the MD of EBL was 219.8 mL (95% CI, 39.3 to 400.2 mL; p=0.02). Among these patients, those who underwent early surgery showed a significant difference in EBL (MD, 326.7 mL; 95% CI, 73.9 to 579.5 mL; p=0.01), but a significant difference was not found in the late surgery group (MD, 134.7 mL; 95% CI, -133.6 to 403.0 mL; p=0.32).

Single elimination of each study did not affect the overall results of the meta-analysis. The Egger’s test result for the summary estimates of all papers was 0.29 (p=0.60) for EBL. The funnel plot was symmetric, suggesting an absence of significant publication bias within the studies. These results imply no substantial evidence of publication bias in the dataset.

Some limitations need to be acknowledged and addressed regarding the present study. First, there was substantial heterogeneity in primary cancer, vascularity, definition of complete embolization, and kinds of surgery. Research on spinal metastasis usually faces this problem. In this field, a metaanalysis may be a good method that can integrate data from multiple studies and reduce confounding variables. This study included only direct comparative studies in which factors other than the timing of surgery were controlled. Second, definitions of completeness of embolization vary among studies. In four out of seven studies [10,20,21,23], the definition of complete (or successful) embolization was 90%, while in two other studies, it varied from 75% to 80% [9,15]. The degree of embolization was evaluated by the amount of tumor blush reduction, which is determined subjectively by the operator. There is no objective and consistent tool for evaluating the degree of TAE. This may weaken the clinical effectiveness of this paper. Finally, we did not conduct a bottom-up analysis to determine the appropriate timing for surgery, but instead conducted an analysis based on the standards that previous papers used. The findings, therefore, must be interpreted with caution. A large-scale prospective study is needed to determine the optimal surgical timing after embolization.