Introduction

Method

Search strategy and Selection criteria

The design of the present study followed the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines proposed in 2020(10). For the initial search of relevant keywords in each international database PubMed/Scopus, MEDLINE(Ovid), Cochrane Library was performed between 2015 and 2025 (Table 1), the reason for choosing a ten-year time period was to review new results and compare them. The selection criterion for studies was the PICO strategy (Table 1); a control group was not required in the present study. Any type of Reviews studies, in vivo studies, case reports, and references that were unavailable or incomplete, studies that did not include human subjects, and studies without quantitative data were excluded from the research.

Table 1. Search Strategy and PIO  strategy.

Data Extraction and bias assessment 

In the present study, the process of data extraction and risk of bias assessment was carefully conducted according to the standards of systematic review and meta-analysis by two blinded and independent authors. In the first stage, data were collected from the selected studies using a predefined researcher-made form. The extracted information included the names of the authors, year of publication, type of study, characteristics of the study population (mean age, gender, and sample size), and Type of cervical spine surgery.

Using the Newcastle Ottawa scale (NOS)(11), quality evaluation was carried out in three areas: outcome, comparability, and selection. "High quality" was defined as scores greater than 7 according to the NOS tool.

Two researchers independently and blindly conducted the quality assessment and data extraction process. A third researcher provided final confirmation, and any disagreements were resolved through re-examination and consultation. Also, to ensure the accuracy of the data, a re-examination of key sources and results was conducted. This approach ensured the accuracy and validity of the analyses in the systematic review and meta-analysis.

Statistical analysis

Meta-analysis was performed using STATA 17 software. Effect sizes for sensitivity, specificity, and predictive value of IONM were extracted from the articles, and considering the heterogeneity between studies, which was assessed by the I² index (No heterogeneity : 0.0% < I² < 24.9%; low heterogeneity: 25.0%  < I² < 49.9%;  Moderate heterogeneity: 50.0%  < I² < 74.9%; High heterogeneity: 75.0%  < I² < 100%), the Random Effects model with DerSimonian-Laird method was selected. Egger's regression test and Begg's rank correlation test were used to examine publication bias. A funnel plot was also drawn to visually examine the symmetry of the data. The statistical significance level was considered to be p<0.05. Subgroup analyses were performed based on the type of IONM modality (SSEP, MEP, combined), type of surgery (anterior vs posterior), and number of surgical levels (single-level vs multi-level).

Result

Literature Search

849 articles were found by systematic literature review that matched the search strategy. 146 duplicates, 3 case reports, and 538 studies with other type of study (exclusion criteria) and  based on titles relevant to the exclusion criteria. A total of 162 articles were screened and articles that did not meet the inclusion criteria were excluded (n=114). The full texts of 48 articles were reviewed by two independent, blinded authors and subjected to screening for inclusion and exclusion criteria; only twelve articles met the inclusion criteria, which were selected for review in the present study (Figure 1).

Characteristics of included studies

A total of 3754 patients underwent various cervical spine surgeries. Nine studies were retrospective cohort studies and three were prospective cohort studies. Sample sizes ranged from 46 to 1,622 patients, and the mean age of participants ranged from 49.6 to 66.2 years. This dataset provides broad coverage of patient populations and types of surgeries, allowing for a comprehensive assessment of the effectiveness of intraoperative neurological monitoring in preventing neurological complications in various settings of cervical spine surgery. Table 3 shows the quality of the studies.

Table 2. Summary of characteristics of studies selected for meta-analysis

RCs: Retrospective Cohort study; PPV: positive predictive value; NPV: negative predictive value;  PCs: Prospective Cohort study; EELF: Extended endoscopic lumbar foraminotomy; TLIF: transforaminal lumbar interbody fusion; UBE: Unilateral biportal endoscopic; RCT: randomized controlled study; ACSS: Anterior cervical spine surgery; ACDF: Anterior cervical spine discectomy with fusion; PLIF: Primary posterior lumbar interbody fusion.

Table 3. Newcastle-Ottawa Scale assessment of cohort included studies.

Sensitivity

Using the fixed-effects model and the inverse-variance method, the overall sensitivity of IONM in identifying neurological complications was calculated to be 0.912 (95% CI: 0.888–0.936) (Figure 2), indicating the high accuracy of IONM in detecting intraoperative neurological problems. I² = 61.1% indicates moderate heterogeneity between studies. H² = 2.57 and Q test with χ²(11) = 28.28, p = 0.0029 also confirmed the existence of a significant difference between the results of the studies. A random-effects model with the REML method was used to better account for differences between studies and heterogeneity observed in fixed-effect analysis. The pooled sensitivity was estimated to be 0.916 (95% CI: 0.872–0.960) (Figure 3).

Specificity

Using the fixed-effects model and the inverse-variance method, the overall specificity of IONM was calculated to be 0.851 (95% CI: 0.827–0.876) ) (Figure 4), indicating the high ability of IONM to correctly detect the absence of neurological complications during surgery. I² = 36.6% indicated low heterogeneity between studies. IONM is a reliable tool for identifying patients without neurological complications during surgery.

Figure 4. forest plot showed specificity for IONM in predicting neurological deficits.

Positive predictive value (PPV)

Using the fixed-effects model and the inverse-variance method, the overall PPV of IONM was calculated to be 0.926 (95% CI: 0.902–0.951) (Figure 5). These results indicate the very high ability of IONM to correctly predict patients with neurological complications. I² = 45.6% indicates moderate heterogeneity between studies.

Negative predictive value (NPV)

Using the fixed-effects model and the inverse-variance method, the overall NPV of IONM was calculated to be 0.874 (95% CI: 0.850–0.898) (Figure 6). These results indicate that IONM negative alarms correctly identify patients without true neurological complications in most cases. I² = 87.3% indicates very high heterogeneity between studies. Hence, using the random effects model and the REML method, the overall NPV was calculated to be 0.833 (95% CI: 0.751–0.916) (Figure 7).