Abstract
-
Objective
To evaluate the predictive value of the final-to-baseline amplitude reduction ratio (FBR) of motor evoked potentials (MEP) and somatosensory evoked potentials (SEP) for early postoperative neurological deficits (PNDs) following intradural extramedullary (IDEM) tumor resection, and to explore a candidate prognostic cutoff for intraoperative neurophysiological monitoring (IONM).
-
Methods
A total of 123 consecutive patients who underwent IDEM tumor resection with multimodal IONM were retrospectively analyzed. Early PND was defined as ≥1-grade increase in the modified McCormick scale at 1 month postoperatively. MEP and SEP FBRs were calculated and analyzed using multivariable logistic regression and receiver operating characteristic (ROC) curve analysis.
-
Results
Early PND occurred in 26.02% (32 of 123) of patients. Multivariable regression identified MEP FBR as the sole independent predictor of early PND (odds ratio, 1.064; 95% confidence interval, 1.040–1.088; p<0.001). ROC analysis demonstrated excellent predictive performance with an area under the curve of 0.922 (p<0.001) and an optimal cutoff value of 61.84%. MEP FBR showed a significant graded association with PND severity (p<0.001).
-
Conclusion
Our findings suggest MEP FBR may serve as a promising predictor of early postoperative neurological deficit among patients undergoing IDEM tumor resection. A multimodal IONM strategy centered on MEP monitoring, with SEP as a supplement appears reasonable for IDEM tumor resection. A 60% reduction in MEP amplitude may represent an exploratory prognostic cutoff warranting prospective validation.
-
Keywords: Spinal cord neoplasms, Intraoperative neurophysiological monitoring, Motor evoked potentials, Somatosensory evoked potentials
INTRODUCTION
Intradural extramedullary (IDEM) spinal tumors represent the most prevalent anatomical subtype of intraspinal tumors, accounting for over 60% of all cases [
1,
2]. IDEM tumors are generally benign, with schwannomas and meningiomas being the most common histopathological subtypes [
3,
4]. Clinically, common presenting symptoms in patients with IDEM tumors include radicular pain, sensory disturbances, and motor weakness, depending on tumor size and the specific site of nerve root or spinal cord compression. Surgical resection is the current optimal therapeutic choice for these patients, which can effectively alleviate compression by removing the space-occupying lesion. However, given the critical role of the spinal cord in maintaining neurophysiological function, iatrogenic spinal cord injury and the consequent functional deficits have emerged as a significant concern in surgical practice [
5,
6].
Intraoperative neurophysiological monitoring (IONM) has been widely applied in modern surgical practice, and its efficacy in preventing iatrogenic neurological injury during IDEM tumor resection has been well established [
7,
8]. Currently, the commonly used IONM techniques to preserve the functional integrity of ascending and descending pathways in spinal surgeries include somatosensory evoked potential (SEP), motor evoked potential (MEP), and D-wave monitoring. Numerous studies have explored the predictive efficacy of the aforementioned IONM techniques for postoperative neurological deficits (PNDs) [
9-
12]. However, most of these studies evaluated the techniques solely from the perspective of monitoring accuracy, without recognizing that IONM differs fundamentally from diagnostic tests. In IONM, false-positive results may actually arise from intraoperative alerts that prompt surgical teams to modify operative strategies, thereby preventing PND [
13]. This protective intervention complicates the objective assessment of the predictive value of IONM parameters, rendering traditional accuracy metrics potentially misleading. In this study, we quantified the percentage reduction in evoked potential amplitude following tumor resection relative to the baseline recorded immediately before dural opening and examined its association with PND. This approach facilitates a more rigorous and clinically meaningful assessment of the predictive utility of IONM parameter changes for PND. Furthermore, based on our findings, we explore a data-informed candidate prognostic cutoff for IONM during IDEM tumor surgery, to inform future development of warning criteria.
MATERIALS AND METHODS
1. Patient Population
A retrospective cohort study was performed by reviewing the medical records of patients who underwent IDEM tumor resection at Beijing Tiantan Hospital, Capital Medical University from July 2021 to October 2022, with all procedures performed by a single dedicated neurosurgical team. IONM was applied in all surgeries. Inclusion criteria were as follows: (1) age ≥16 years; (2) diagnosed with IDEM tumors according to preoperative neuroimaging and postoperative pathological examination; (3) no concomitant spinal or spinal cord disorders that could independently affect neurological function assessment. Exclusion criteria were as follows: (1) insufficient demographic, clinical, or IONM data; (2) lost to follow-up. Initially, 124 patients were screened for eligibility according to the inclusion criteria, and 1 was excluded due to loss to follow-up, yielding a final sample of 123. The data collected for analysis included demographic characteristics, clinical features, radiological findings, intraoperative IONM parameters, and follow-up information. This study was approved by the Ethics Committee of Beijing Tiantan Hospital, Capital Medical University (approval number: KY2022-080-01). Given the retrospective design and the use of de-identified clinical data, the requirement for informed consent for study participation was waived by the Ethics Committee. Written informed consent for the surgical procedure was obtained from all patients or their legal representatives before surgery.
2. Anesthesia
Total intravenous anesthesia was administered for all surgical procedures. Anesthesia induction was achieved with sufentanil (0.3 μg/kg) and propofol (1.5–2.0 mg/kg). For intraoperative anesthetic maintenance, continuous infusions of remifentanil (0.05–0.20 μg/kg/min) and propofol (4–6 mg/kg/h) were delivered. Rocuronium was administered solely for tracheal intubation, and additional muscle relaxants were avoided unless clinically necessary. Primary physiological parameters, including heart rate, body temperature, blood pressure, peripheral oxygen saturation, and end-tidal carbon dioxide partial pressure, were continuously monitored throughout the procedure.
3. Intraoperative Neurophysiological Monitoring
MEP and SEP were monitored for all patients using the Cadwell Cascade Elite System (Cadwell Industries, USA) in accordance with standardized institutional protocols [
14]. For SEP monitoring, bilateral electrical stimulation of the median nerve (at the wrist) and posterior tibial nerve (at the medial malleolus) was performed using needle electrodes. SEP stimulation parameters were set as follows: 300–400 signal averages; stimulation intensity, 15–30 mA; pulse duration, 0.2 ms; and stimulation frequency, 2.1–4.7 Hz. Signals were recorded using a band-pass filter from 100 to 2,000 Hz and a 50-Hz notch filter. Spiral recording electrodes were placed at the international 10–20 electroencephalographic system derivations C3’ and C4’, with Fpz serving as the reference electrode. For MEP monitoring, transcranial electrical stimulation was delivered using corkscrew electrodes placed at C1 and C2 according to the international 10–20 system. Stimulation parameters were as follows: 5–8 pulses per train; individual pulse duration, 50–500 μs; interstimulus interval within the train, 2–4 ms (corresponding to an intratrain pulse rate of 250–500 Hz); and stimulus intensity, 100–400 V. Bilateral needle electrodes were positioned in the deltoid, extensor digitorum communis, and thenar eminence muscles for upper extremity recording, and in the quadriceps femoris, tibialis anterior, and abductor hallucis muscles for lower extremity recording. Recorded signals were filtered with a band-pass filter from 30 to 3,000 Hz and a 50-Hz notch filter.
Baseline MEP and SEP waveforms were acquired prior to dural opening. Intraoperative real-time monitoring was performed at 3-minute intervals, shortened to 1 minute or more frequently during critical surgical maneuvers. Intraoperative warning criteria were predefined as an 80% reduction in MEP amplitude and a 50% reduction in SEP amplitude relative to baseline values. The final-to-baseline amplitude reduction ratio (FBR) was used as a quantitative metric to assess changes in MEP and SEP waveforms during surgery and to predict PND. The final waveform was defined as the one recorded immediately after tumor resection, when surgical interference had ceased. The FBR was calculated using the following formula: FBR=(baseline amplitude − final amplitude)/(baseline amplitude)×100%. Where the final amplitude exceeded the baseline amplitude, the formula yields a negative value; such values indicate amplitude increase rather than deterioration and were truncated to 0%. The FBR therefore ranges from 0% to 100%.
For each patient, MEP FBR was calculated from the muscle showing the greatest amplitude reduction among all monitored muscles, and SEP FBR was calculated from the recording channel showing the greatest amplitude reduction among all monitored SEP channels. Any muscle or channel without reproducible baseline waveforms after technical adjustment was excluded from the FBR calculation for that individual. Moreover, patients with complete loss of MEP or SEP waveforms were assigned an FBR value of 100% and remained included in all analyses.
4. Clinical Assessment and Follow-up
Tumor characteristics, including the involved spinal anatomical segment (cervical, thoracic), tumor location relative to the spinal cord (dorsal or ventrolateral), and the number of involved vertebral levels, were extracted from preoperative magnetic resonance imaging scans. Symptom duration was extracted from patients’ electronic medical records. The extent of surgical resection and intraoperative estimated blood loss were retrieved from operative notes, with resection defined as gross total resection (GTR) or subtotal resection (STR). The histopathological diagnosis was made according to the 2016 World Health Organization classification of central nervous system tumors.
Neurological function was assessed using the modified Mc-Cormick scale (mMCS) via outpatient visits or telephone interviews at 1 week, 1, 3, 6, and 12 months postoperatively. PND was defined as an increase of ≥1 grade in the mMCS score relative to the preoperative baseline. Early PND was defined as such deterioration observed at the 1-month postoperative follow-up visit. For severity-stratified analyses, patients were categorized into 3 mutually exclusive subgroups based on the change in mMCS grade at 1 month postoperatively: (1) the no-PND group (no deterioration in mMCS grade compared with preoperative status, comprising patients whose grade was unchanged or improved), (2) the mild-PND group (a 1-grade increase in mMCS), and (3) the moderate-to-severe-PND group (an increase of ≥2 grades in mMCS).
5. Statistical Analysis
All statistical analyses and graphical data visualization were performed using R ver. 4.4.2 (R Foundation for Statistical Computing, Austria) and GraphPad Prism ver. 10.1.2 (GraphPad Software Inc., USA). Continuous variables were presented as mean±standard deviation or median with interquartile range (IQR). Categorical variables were reported as counts and percentages. Group comparisons were conducted using the independent- samples t-test, chi-square test, Fisher exact test, or nonparametric test, as appropriate. Multivariable logistic regression analysis was performed to identify independent predictors of early PND, with results reported as odds ratios (OR) and corresponding 95% confidence intervals (CIs). In addition to basic demographic factors (age and gender), variables with clinical relevance or a univariate p-value<0.200 were included in the multivariable logistic regression model. Receiver operating characteristic (ROC) curve analysis was used to evaluate the predictive performance of IONM metrics, with optimal cutoff values determined by the Youden Index. Pairwise comparisons of areas under the ROC curves (AUCs) between MEP FBR, SEP FBR, and the combined logistic regression model were performed using the DeLong test. To address potential overfitting, we performed bootstrap internal validation with 1,000 resamples, in which the full model-building pipeline, including univariate pre-selection (p<0.200), multivariable logistic regression, and final variable selection retaining the single variable with the smallest p-value among those reaching p<0.05, was independently repeated within each resample, to calculate the optimism-corrected AUC and optimism bias. To assess the relationship between IONM metric changes and neurological deficit severity, patients were stratified by mMCS grade changes (PND severity), and differences in IONM metrics among the 3 severity cohorts were analyzed using the Kruskal-Wallis test. The Jonckheere-Terpstra test for ordered alternatives was additionally used to test for a monotonic trend across increasing severity, with correction for ties. A 2-sided p-value<0.05 was considered statistically significant.
RESULTS
1. Patient Characteristics
The final cohort comprised 52 male and 71 female patients, with a mean overall age of 50.00±15.88 years. All surgical procedures were performed with the patient positioned in the lateral decubitus position and via a posterior midline approach. GTR was achieved in all patients except one, in whom STR was necessary due to severe tumor adherence to the spinal cord. The baseline characteristics of the entire cohort were summarized in
Table 1.
2. Temporal Evolution of PND
Neurological function was systematically assessed at predefined postoperative time points during follow-up, and the proportion of patients with postoperative neurological deficit at each time point is presented in
Fig. 1. The incidence of PND exhibited a clear time-dependent pattern, declining to 0.81% (1 of 123) at 1 year postoperatively. Accordingly, early PND (at 1 month postoperatively) was designated as the primary study endpoint for subsequent analysis. The 7-day PND was not adopted as a primary endpoint because of its temporal proximity to surgery, which could increase susceptibility to confounding factors, such as residual anesthetic effects, postoperative edema, and acute surgical stress. In contrast, the 1-month postoperative time point can balance clinical relevance and outcome stability, effectively minimize the impact of transient acute confounders, and thus accurately reflect the early neurological outcomes specific to IDEM tumor resection.
As shown in
Table 1, no significant differences were observed in baseline demographic and clinical characteristics between patients with and without early PND, including age, gender, symptom duration, involved spinal anatomical segment, tumor pathology, postoperative hospital stay, and preoperative neurological status. However, patients with early PND were significantly more likely to harbor ventrolateral tumors (p= 0.001) and experienced greater intraoperative blood loss (p=0.012).
3. Predictive Value of MEP and SEP Changes for Early PND
To investigate the predictive value of IONM metrics for early PND, we compared the FBRs of MEP and SEP between patients with and without early PND (
Fig. 2A and
B). The results demonstrated that the FBRs of both MEP and SEP were significantly higher in patients who developed early PND (p<0.001 and p=0.031, respectively). Complete intraoperative MEP loss (FBR=100%) occurred in 9 patients (7.3%), of whom 8 (88.9%) developed early PND. To further identify independent predictive factors for early PND and to adjust for confounding effects, we included age, gender, tumor location relative to the spinal cord, intraoperative blood loss, and IONM metrics (MEP FBR, SEP FBR) in the multivariable logistic regression model (
Fig. 3). After adjusting for all covariates, MEP FBR was identified as the sole independent predictor of early PND. Specifically, each 1% increase in MEP FBR was associated with a 6.4% increase in the odds of developing early PND (OR, 1.064; 95% CI, 1.040–1.088; p<0.001). Correspondingly, a 10% incremental increase in MEP FBR translated to 85.7% higher odds of early PND (OR, 1.857; 95% CI, 1.483–2.324). In contrast, ventrolateral tumor location, intraoperative blood loss and SEP FBR were associated with early PND in univariate analysis but were not independently associated after adjustment (p>0.05 for all).
4. Predictive Performance of MEP FBR for Early PND
ROC curve analysis was conducted to evaluate the predictive performance of MEP FBR for early PND (
Fig. 4). The analysis revealed excellent discriminative ability, with an AUC of 0.922 (p<0.001). An optimal cutoff value of 61.84% for MEP FBR was identified by the Youden Index, which was associated with a sensitivity of 71.9% and a specificity of 97.8% for predicting early PND. Bootstrap internal validation with 1,000 resamples confirmed robust model performance, yielding an optimism-corrected AUC of 0.921 with negligible optimism bias (0.0006). MEP FBR was retained by the selection rule in 974 of 977 valid bootstrap iterations (99.7%), indicating high stability of the variable selection process; of the 23 excluded resamples, 20 were excluded because quasi-complete separation prevented any variable from reaching p<0.05, and 3 because the AUC could not be computed for the selected variable. The bootstrap-derived optimal thresholds showed a median of 52.1% (IQR, 40.8%–61.8%), reflecting substantial uncertainty in any specific cutoff estimate.
For clinical practicability, the data-derived optimal cutoff of 61.84% was rounded down to 60%. This rounding reclassified one patient without early PND as test-positive, so that the pragmatic 60% cutoff yielded the same sensitivity (71.9%) but a slightly lower specificity (96.7%, 88 of 91) than the data-derived cutoff (97.8%, 89 of 91). We next applied 2 cutoffs—60% and 80% —
post hoc to the same final MEP FBR values, in order to examine how the choice of cutoff affects classification of 1-month PND (
Table 2). The 80% value was selected for comparison because it corresponds numerically to our institutional intraoperative warning criterion; however, because intraoperative warning events were not retrievable from the retrospective records, this analysis compares 2 cutoffs of the same postresection metric and does not evaluate the performance of real-time intraoperative warning. The 60% threshold demonstrated sensitivity of 71.9% (23 of 32), specificity of 96.7% (88 of 91), positive predictive value of 88.5% (23 of 26), and negative predictive value of 90.7% (88 of 97). In comparison, the 80% threshold showed sensitivity of 43.8% (14 of 32), specificity of 98.9% (90 of 91), positive predictive value of 93.3% (14 of 15), and negative predictive value of 83.3% (90 of 108). The 60% threshold identified 9 additional PND cases (true positives: 23 vs. 14) at the cost of 2 additional false-positive classifications (false positives: 3 vs. 1).
5. Comparative Performance of Individual and Combined IONM Metrics
To evaluate whether combining MEP and SEP metrics could enhance predictive accuracy, we compared the discriminative performance of MEP FBR alone, SEP FBR alone, and a combined logistic regression model. MEP FBR demonstrated excellent predictive performance with an AUC of 0.922 (95% CI, 0.871–0.973), significantly outperforming SEP FBR (AUC=0.629, 95% CI, 0.512–0.746, DeLong test p<0.001). When both metrics were incorporated into a combined logistic regression model, the resulting AUC was 0.906 (95% CI, 0.834–0.979), which did not differ significantly from MEP FBR alone (p=0.372) (
Table 3;
Fig. 5). These findings indicate that MEP FBR captures the majority of predictive information for early PND, with SEP FBR providing minimal incremental value in a combined model.
6. Graded Association Between MEP FBR and PND Severity
To assess the relationship between MEP FBR and the severity of early PND, patients were stratified by the degree of postoperative neurological deterioration (no PND, mild PND, moderate-to-severe PND). A distinct stepwise increase in intraoperative median MEP FBR was observed with escalating PND severity. The median MEP FBR was 1.69% (IQR, 0.00%–18.18%) in patients without PND, 70.39% (IQR, 42.95%–88.99%) in those with mild PND, and 100.0% (IQR, 93.40%–100.0%) in patients with moderate-to-severe PND (
Fig. 6). MEP FBR differed significantly across the 3 severity groups (Kruskal-Wallis H=53.444, p<0.001), with progressively higher median values and no overlap of interquartile ranges. A Jonckheere-Terpstra test for ordered alternatives confirmed a monotonic increasing trend in MEP FBR across increasing PND severity (J=2,821.0; standardized J–T statistic=7.494; p<0.001).
DISCUSSION
The current study investigated the predictive utility of quantitative IONM metrics for early PND in patients undergoing surgery for IDEM tumors. The MEP FBR emerged as the sole independent predictor of early PND and showed a graded positive association with its severity. Furthermore, an MEP FBR threshold of 61.84% was identified as the optimal cutoff for predicting early PND, providing a reference for the future development of refined, physiology-informed warning criteria for IDEM tumor resection.
1. PND Following IDEM Tumor Resection Is Predominantly Transient and Mild
A key clinical observation from the present study is that the PND subsequent to IDEM tumor resection is overwhelmingly transient and mild, which is distinct from the persistent and severe PND observed after intramedullary spinal cord tumor surgery [
14-
16]. Neurological function assessments at scheduled postoperative time points showed a time-dependent decrease in PND incidence, with the overall rate dropping to less than 1% at 12 months postoperatively. From a surgical safety perspective, such a long-term PND rate represents a clinically acceptable outcome. Moreover, among patients with early PND, the vast majority (25 of 32) had only a 1-grade deterioration on the mMCS, further confirming the mild nature of PND in patients with IDEM tumors. The distinct PND profiles following resection of IDEM versus intramedullary spinal cord tumors reflect the fundamental pathophysiological differences between the 2 tumor types. Resection of intramedullary tumors requires direct spinal parenchymal dissection, which can easily cause irreversible structural damage to the corticospinal tract and thus induce persistent, severe PND. By contrast, IDEM tumors are extramedullary and typically benign. The neurological impairment in these patients is not due to spinal cord structural destruction but to transient mechanical stress or mild ischemic injury to the spinal cord during surgical decompression and tumor dissection.
This transient and mild nature of PND after IDEM tumor resection may seem to imply that IONM has limited clinical significance for such patients. However, clinical medicine should pursue higher therapeutic goals, and sparing more patients from short-term PND remains a worthwhile aim, which means the IONM strategy for IDEM tumor resection still demands continuous refinement and improvement.
2. MEP Monitoring as a Central Component of a Multimodal IONM Strategy for IDEM Tumors
The current study provides evidence supporting the role of MEP monitoring as a central component of the multimodal IONM strategy for IDEM tumor resection. MEP FBR was identified as the sole independent predictor of early PND among all clinical and neurophysiological parameters assessed. Furthermore, a statistically significant, stepwise positive association was demonstrated between MEP FBR and the severity of early PND. Collectively, these results suggest that MEP FBR not only reliably predicts early PND but also quantitatively reflects the degree of postoperative neurological impairment.
In contrast, while SEP FBR was significantly higher in patients with early PND in univariate analysis, it failed to retain independent predictive value for early PND in the multivariable analysis after adjustment for confounding factors. This finding is consistent with our previous research on intramedullary spinal cord tumors, revealing that the predictive utility of SEP is consistently inferior to that of MEP across different spinal tumor subtypes [
14]. The superior sensitivity and earlier detectability of MEP compared with SEP in IONM can be explained by anatomical and neurophysiological mechanisms. First, the corticospinal tract is predominantly perfused by the anterior spinal artery with limited collaterals, making it more susceptible to ischemia, traction, and compression than the posterior columns, which have a richer blood supply and superior ischemic tolerance. Second, MEP reflects the integrity of the entire motor pathway, including spinal anterior horn synaptic transmission, which is a process highly susceptible to mild metabolic or ischemic disturbances. In contrast, SEP primarily represents axonal conduction along the posterior columns, which is more resistant to subtle physiological changes [
17,
18]. Our finding that the combined MEP-SEP model did not outperform MEP alone warrants comment. This phenomenon occurs when a strong predictor (MEP) is paired with a substantially weaker one (SEP), where the latter contributes minimal independent information and may introduce statistical noise. Importantly, this does not diminish the clinical value of SEP monitoring for comprehensive assessment of dorsal column integrity and overall spinal cord function; rather, it clarifies that SEP changes provide limited incremental predictive value specifically for motor deficits beyond what MEP already captures. Accordingly, SEP monitoring is recommended as a complementary method alongside MEP monitoring, enabling a more comprehensive evaluation of spinal cord function [
19,
20].
As for D-wave monitoring, previous studies have demonstrated its high predictive value for postoperative motor outcomes in patients undergoing spinal surgery [
21,
22]. As a specialized application of MEP monitoring, D-wave enables safe continuation of tumor resection after MEP loss during surgery and exhibits high accuracy in predicting long-term PND. However, D-wave monitoring was not performed in the present cohort. Whether D-wave would provide additional predictive value beyond MEP and SEP in this specific surgical context requires prospective comparative investigation.
Collectively, a multimodal IONM strategy centered on MEP monitoring, with SEP monitoring as a supplement, appears reasonable for IDEM tumor resection. In our cohort, ventrolateral tumor location and greater intraoperative blood loss were associated with early PND in univariate analysis but were not independently associated after adjustment. From a mechanistic perspective, ventrolateral tumors are anatomically adjacent to the corticospinal tract, and excessive intraoperative blood loss may induce spinal cord hypoperfusion and mild ischemic injury. One plausible explanation for the attenuation of these associations after adjustment is therefore that their effect on outcome is largely captured by intraoperative MEP change: both factors may increase the risk of corticospinal tract compromise, which is itself reflected in MEP FBR. These variables may thus be regarded as markers identifying patients in whom intraoperative MEP change is more likely, rather than as independent predictors of outcome, and this interpretation requires prospective confirmation.
3. Candidate Prognostic Cutoff for MEP FBR for IDEM Tumor Resection
The optimal warning criteria for IONM remain a highly controversial issue to date. It is generally accepted that a 50% amplitude reduction serves as the standard warning criterion for evoked potential monitoring in cranial surgery, whereas an 80% reduction or an “all-or-none” pattern is more commonly adopted in spinal surgery [
23,
24]. Our team has long held the view that the warning criteria for IONM should be differentiated by disease type and characteristics, moving toward refinement and individualization. IDEM tumor resection likewise deserves a monitoring threshold tailored to its unique anatomical features and injury mechanisms. In the present study, we found that MEP FBR yielded the highest predictive accuracy for early PND when the cutoff value was set at 61.84%. Considering clinical practicability, a 60% reduction in MEP amplitude may serve as a candidate reference threshold for IDEM tumor resection. Applied
post hoc to the same metric, a 60% cutoff would have identified more patients who subsequently developed early PND than an 80% cutoff. Whether adopting a lower real-time warning criterion would actually prevent deficits cannot be inferred from these data, because the 60% level was never applied intraoperatively in this cohort. The higher sensitivity of the 60% cutoff may be attributed to the fact that IDEM resection mainly causes cumulative mild traction, transient vascular disturbance and subclinical microtrauma that trigger moderate MEP decline instead of severe amplitude collapse, whereas the conventional 80% cutoff originates from surgeries featuring abrupt severe spinal cord damage, making 60% better matched to IDEM-specific injury features. However, an important caveat warrant emphasis. The threshold comparison was conducted retrospectively on a cohort where the 80% criterion was actually applied during surgery, and the prognostic cutoff derived from FBR analysis is conceptually distinct from a real-time intraoperative warning criterion. The proposed 60% threshold should therefore be recognized as an evidence-based reference rather than a definitive clinical criterion, given that real-time modification of warning criteria directly influences intraoperative decision-making. Prospective validation is required before this threshold can be recommended for clinical implementation, and further dedicated investigations are still needed to establish the optimal real-time warning criterion specifically for IDEM tumor resection.
4. Limitations
Several limitations of the present study should be acknowledged. First and foremost, the retrospective nature of data collection restricted the acquisition of granular information, including detailed intraoperative neuromonitoring events and itemized postoperative neurological functional status, which prevented further stratified analyses. In addition, complete control of certain potential confounding variables could not be achieved under retrospective study conditions. Second, the study was conducted at a single center with a relatively limited sample size. Accordingly, the overall findings of this study, particularly the proposed 60% MEP amplitude reduction cutoff, should be interpreted with considerable caution. Furthermore, the proposed 60% threshold carries considerable uncertainty attributable to the limited number of PND events (n=32), and should not be interpreted as a precisely determined value. Future large-sample, multicenter, prospective studies are warranted to further validate and optimize this threshold for clinical application.
CONCLUSION
Our findings suggest that MEP FBR may serve as a promising predictor of early PND in patients undergoing IDEM tumor resection. A multimodal IONM strategy centered on MEP monitoring, with SEP as a supplement, appears reasonable for IDEM tumor resection. A 60% reduction in MEP amplitude may represent a candidate reference threshold, but prospective multicenter validation is warranted before it can be recommended as an intraoperative warning criterion.
NOTES
-
Conflict of Interest
The authors have nothing to disclose.
-
Funding/Support
This research was supported by Capital’s Funds for Health Improvement and Research (2022-2-1073) and the National Key R&D Program of China (2025YFC2511502).
-
Author Contribution
Conceptualization: XF, XT, HQ; Data Curation: KL, JL, XW, JY, JS, YL; Formal Analysis: KL, JS, XT, HQ; Funding Acquisition: HQ; Methodology: KL, XF, JL, XW, YL, HQ; Project Administration: WJ, HQ; Visualization: KL; Writing – Original Draft: KL, XF, JL, XW, JY, HQ, JS, YL; Writing – Review & Editing: KL, XF, XT, WJ, HQ.
Fig. 1.Time-dependent changes in the proportion of patients with PND after IDEM tumor resection. The proportion of patients with PND peaked at 1 week postoperatively, then gradually declined over the follow-up period, with nearly complete resolution observed at 1 year after surgery. PND, postoperative neurological deficit.
Fig. 2.Comparison of MEP FBR (A) and SEP FBR (B) between patients with and without early PND. Both showed significantly higher values in patients with early PND than in those without (p<0.001 and p=0.031, respectively). MEP, motor evoked potential; FBR, final-to-baseline amplitude reduction ratio; SEP, somatosensory evoked potential; PND, postoperative neurological deficit.
Fig. 3.Forest plot of the multivariable logistic regression analysis for early PND. Reference categories were female for sex and dorsal location for tumor location. Blood loss was modeled per 100-mL increment; age, MEP FBR and SEP FBR were modeled per 1-unit increment. MEP FBR was the only variable independently associated with early PND (odds ratio [OR] 1.064 per 1% increase; 95% confidence interval [CI], 1.040–1.088; p<0.001). MEP, motor evoked potential; FBR, final-to-baseline amplitude reduction ratio; SEP, somatosensory evoked potential; PND, postoperative neurological deficit.
Fig. 4.Receiver operating characteristic (ROC) curve of MEP FBR for predicting early PND. The AUC, which measures the overall discriminative ability of MEP FBR to distinguish patients with and without early PND, was 0.922 (p<0.001). The optimal cutoff value of 61.84% for MEP FBR was identified via Youden Index, corresponding to a sensitivity of 71.9% and a specificity of 97.8%. MEP, motor evoked potential; FBR, final-to-baseline amplitude reduction ratio; PND, postoperative neurological deficit; AUC, area under the ROC curve; CI, confidence interval.
Fig. 5.Receiver operating characteristic curves comparing MEP FBR, SEP FBR, and combined model for predicting early postoperative neurological deficits. p-values denote the comparison of each AUC against 0.5. MEP, motor evoked potential; FBR, final-to-baseline amplitude reduction ratio; SEP, somatosensory evoked potential; AUC, area under the receiver operating characteristic curve.
Fig. 6.The association between MEP FBR and early PND severity. Violin plots showing MEP amplitude reduction according to the severity of postoperative neurological deficit. Violin shapes reflect data distribution density, dashed lines indicate medians and dotted lines indicate quartiles, with individual data points overlaid. MEP FBR showed a graded positive association with early PND severity (Jonckheere- Terpstra test for trend, p<0.001). MEP, motor evoked potential; FBR, final-to-baseline amplitude reduction ratio; PND, postoperative neurological deficit.
Table 1.Baseline characteristics and univariate analysis of patients with and without postoperative neurological deficit at 1 month
Table 1.
|
Variable |
Total (n=123) |
Postoperative 1 month
|
p-value |
|
With PNDa) (n=32) |
Without PND (n=91) |
|
Age (yr) |
50.00 ± 15.88 |
47.59 ± 16.96 |
50.85 ± 15.49 |
0.321†
|
|
Sex |
|
|
|
0.142‡
|
|
Male |
52 (42.3) |
10 (31.3) |
42 (46.2) |
|
|
Female |
71 (57.7) |
22 (68.8) |
49 (53.8) |
|
|
Duration of symptoms (mo) |
7.00 (3.00–14.00) |
10.50 (3.25–24.00) |
6.00 (3.00–14.00) |
0.220§
|
|
Involved spinal segment |
|
|
|
0.886‡
|
|
Cervical segment |
59 (48.0) |
15 (46.9) |
44 (48.4) |
|
|
Thoracic segment |
64 (52.0) |
17 (53.1) |
47 (51.6) |
|
|
Tumor location |
|
|
|
0.001‡
|
|
Dorsal |
82 (66.7) |
14 (43.8) |
68 (74.7) |
|
|
Ventrolateral |
41 (33.3) |
18 (56.3) |
23 (25.3) |
|
|
No. of involved vertebral levels |
2.00 (2.00–2.00) |
2.00 (1.00–2.00) |
2.00 (2.00–2.00) |
0.234§
|
|
Blood loss (mL) |
100 (50–150) |
100 (100–200) |
100 (50–100) |
0.012§
|
|
Tumor pathology |
|
|
|
0.435‡
|
|
Schwannomas |
58 (47.2) |
12 (37.5) |
46 (50.5) |
|
|
Meningiomas |
51 (41.5) |
16 (50.0) |
35 (38.5) |
|
|
Othersb)
|
14 (11.4) |
4 (12.5) |
10 (11.0) |
|
|
Postoperative hospital stay (day) |
6.00 (5.00–7.00) |
6.00 (5.00–8.00) |
6.00 (5.00–7.00) |
0.979§
|
|
Preoperative mMCS |
|
|
|
0.769∥
|
|
I |
20 (16.3) |
7 (21.9) |
13 (14.3) |
|
|
II |
69 (56.1) |
16 (50.0) |
53 (58.2) |
|
|
III |
30 (24.4) |
8 (25.0) |
22 (24.2) |
|
|
IV |
4 (3.3) |
1 (3.1) |
3 (3.3) |
|
Table 2.Performance comparison of 60% versus 80% final MEP FBR cutoffs for classifying 1-month postoperative neurological deficit
Table 2.
|
Metric |
60% Threshold |
80% Threshold |
|
True positives |
23 |
14 |
|
False positives |
3 |
1 |
|
True negatives |
88 |
90 |
|
False negatives |
9 |
18 |
|
Sensitivity (%) |
71.90 |
43.80 |
|
Specificity (%) |
96.70 |
98.90 |
|
Positive predictive value (%) |
88.50 |
93.30 |
|
Negative predictive value (%) |
90.70 |
83.30 |
Table 3.Comparative ROC analysis of IONM metrics
Table 3.
|
Parameter |
AUC (95% CI) |
Optimal cutoff |
Sensitivity (%) |
Specificity (%) |
p-value vs. MEP |
|
MEP FBR |
0.922 (0.871–0.973) |
61.84% |
71.90 |
97.80 |
- |
|
SEP FBR |
0.629 (0.512–0.746) |
24.63% |
65.60 |
60.40 |
< 0.001 |
|
Combined model |
0.906 (0.834–0.979) |
0.553 |
75.00 |
97.80 |
0.372 |
REFERENCES
- 1. Puac-Polanco P, Guarnizo A, Cruz JP, et al. Intradural extramedullary tumors and associated syndromes. Neuroimaging Clin N Am 2023;33:407-22.
- 2. Czyżewski W, Szymoniuk M, Litak J, et al. Unmasking benign intradural-extramedullary spinal tumors: a systematic review on treatment outcomes and radiological features. Ann Med Surg (Lond) 2025;87:8732-47.
- 3. Fiorante A, Woolman M, Munoz DG, et al. A molecular array for 10-second diagnosis of common spinal tumor types with picosecond infrared laser mass spectrometry. Neuro Oncol 2025;27:2060-72.
- 4. Scheffler P, Neidert N, Straehle J, et al. Artificial intelligence-based analysis and diagnosis of intradural extramedullary spinal tumors by stimulated Raman histology. Neurooncol Adv 2025;7:vdaf211.
- 5. Raco A, Pesce A, Toccaceli G, et al. Factors leading to a poor functional outcome in spinal meningioma surgery: remarks on 173 cases. Neurosurgery 2017;80:602-9.
- 6. Randhawa AS, Srivastava A, Agrawal S, et al. Functional outcomes in intradural extramedullary spinal tumors. Surg Neurol Int 2024;15:114.
- 7. Baig Mirza A, Vastani A, Syrris C, et al. Intraoperative neurophysiological monitoring for intradural extramedullary spinal tumours. Global Spine J 2024;14:1304-15.
- 8. Antkowiak L, Putz M, Sordyl R, et al. Predictive value of motor evoked potentials in the resection of intradural extramedullary spinal tumors in children. J Clin Med 2022;12:41.
- 9. Ghadirpour R, Nasi D, Iaccarino C, et al. Intraoperative neurophysiological monitoring for intradural extramedullary spinal tumors: predictive value and relevance of D-wave amplitude on surgical outcome during a 10-year experience. J Neurosurg Spine 2018;30:259-67.
- 10. Kobayashi K, Imagama S, Ando K, et al. Characteristics of Tc-MEP waveforms for different locations of intradural extramedullary tumors: a prospective multicenter study of the Monitoring Committee of the Japanese Society for Spine Surgery and Related Research. Spine (Phila Pa 1976) 2022;47:172-9.
- 11. Ushirozako H, Yoshida G, Imagama S, et al. Efficacy of transcranial motor evoked potential monitoring during intra- and extramedullary spinal cord tumor surgery: a prospective multicenter study of the Monitoring Committee of the Japanese Society for Spine Surgery and Related Research. Global Spine J 2023;13:961-9.
- 12. Tropeano MP, Rossini Z, Franzini A, et al. Multimodal intraoperative neurophysiological monitoring in intramedullary spinal cord tumors: a 10-year single center experience. Cancers (Basel) 2023;16:111.
- 13. Liu J, You H, Guo D, et al. Application of intraoperative evoked potential monitoring in patients with anterior cerebral artery aneurysms. J Stroke Cerebrovasc Dis 2023;32:107214.
- 14. Li K, Tao X, Liu J, et al. The application value and improved warning criterion of D-wave monitoring in intramedullary spinal cord tumor surgery. Spine J 2024;24:1704-11.
- 15. Hersh AM, Patel J, Pennington Z, et al. A novel online calculator to predict nonroutine discharge, length of stay, readmission, and reoperation in patients undergoing surgery for intramedullary spinal cord tumors. Spine J 2022;22:1345-55.
- 16. Skrap B, Tramontano V, Faccioli F, et al. Surgery for intramedullary spinal cord ependymomas in the neuromonitoring era: results from a consecutive series of 100 patients. J Neurosurg Spine 2021;36:858-68.
- 17. Kothbauer KF. Intraoperative neurophysiologic monitoring for intramedullary spinal-cord tumor surgery. Neurophysiol Clin 2007;37:407-14.
- 18. Deletis V, Sala F. Intraoperative neurophysiological monitoring of the spinal cord during spinal cord and spine surgery: a review focus on the corticospinal tracts. Clin Neurophysiol 2008;119:248-64.
- 19. Toleikis JR, Pace C, Jahangiri FR, et al. Intraoperative somatosensory evoked potential (SEP) monitoring: an updated position statement by the American Society of Neurophysiological Monitoring. J Clin Monit Comput 2024;38:1003-42.
- 20. Addanki RND, Lee BB, Anetakis KM, et al. Precision in practice: a systematic review and meta-analysis of intraoperative neurophysiological monitoring for optimizing outcomes in extramedullary spinal cord tumor resection. J Pers Med 2025;15:513.
- 21. Shigematsu H, Ando M, Kobayashi K, et al. Efficacy of D-Wave monitoring combined with the transcranial motor-evoked potentials in high-risk spinal surgery: a retrospective multicenter study of the Monitoring Committee of the Japanese Society for Spine Surgery and Related Research. Global Spine J 2023;13:2387-95.
- 22. Park H, Kim W, Sim J, et al. The role of D-Wave monitoring in motor-evoked potential loss during intramedullary spinal cord tumors resection. Neurospine 2025;22:650-62.
- 23. Journée HL, Berends HI, Kruyt MC, et al. The percentage of amplitude decrease warning criteria for transcranial MEP monitoring. J Clin Neurophysiol 2017;34:22-31.
- 24. MacDonald DB. Overview on criteria for MEP monitoring. J Clin Neurophysiol 2017;34:4-11.