INTRODUCTION
Intramedullary spinal cord tumors (IMSCT) resection is associated with increased long-term survival.[ 1 ] However, it remains a difficult procedure due to the intricate position of these tumors, which is secondary to the difficulty in differentiating between the tumor and the surrounding healthy tissue.[ 2 ] Thus, IMSCT resection surgeries result in a relatively high rate of complication compared to surgeries with incidence rates of 17.5%[ 3 ] –29.3%.[ 4 ] Common neurological complications associated with spinal cord tumor resection surgeries include worsening sensory symptoms, worsening motor deficits, cerebrospinal fluid leaks, spinal deformity, and others.[ 5 ] In addition, a retrospective review found 43.6% of patients exhibited signs and symptoms of dorsal column dysfunction postoperatively.[ 6 ]
The use of somatosensory-evoked potentials (SSEPs) in spinal cord surgery allows for neurological monitoring of the dorsal column pathway during surgical procedures. The value of SSEP in resection of spinal cord tumors has been well described.[ 7 , 8 ] SSEP monitoring has been described to have a 94.4% sensitivity rate and a 96.8% specificity rate in predicting postoperative deficits during spinal cord tumor resection surgery in a study with 203 patients.[ 9 ] However, some also have reported high false-positive and false-negative changes in prediction of deficits with the use of SSEPs.[ 10 , 11 ] Another intraoperative neuromonitoring (IONM) technique for intramedullary tumor resection is motor-evoked potentials (MEPs). Monitorable MEPs are a strong predictor of postoperative motor function; however, they have been reported to be affected more frequently by external factors rendering their predictive value questionable.[ 12 , 13 ] Although SSEP was the standard method of IONM in spinal surgeries before MEP, it is important to determine if SSEP changes along with MEP changes (SSEP/MEP) have better sensitivity and specificity for predicting postoperative neurological deficits when compared to using SSEP alone.
Although individual retrospective studies have examined the effectiveness of IONM in IMSCT resection, the sample sizes of those studies are small. Thus, concerns still exist regarding the effectiveness and necessity of SSEP in spinal cord tumor resection surgeries when compared to SSEP/MEP. Therefore, the purpose of this study was to perform a systematic review and meta-analysis of the relevant literature to identify the reliability of SSEP changes versus SSEP/MEP changes and risk of postoperative neurological complications in patients after IMSCT resection surgery.
METHODS
Protocol and registration
This systematic review followed the guidelines of the preferred reporting items for systematic reviews and meta-analyses procedure [Figure 1 ]. Electronic database searches are included in the supplementary materials [Supplemental Figure 1].
Figure 1: PRISMA Chart – Study elimination process. Preferred reporting items for systematic reviews and meta-analyses, SSEP = Somatosensory evoked potential
Eligibility criteria
Meta-analysis of studies including postoperative outcomes following intramedullary tumor resection surgery with intraoperative SSEP monitoring was included. Reports using other IONM modalities in addition to SSEP were included (ex: electroencephalogram and D-wave); however, the data analyses in this study were limited to SSEP and MEP data alone. Postoperative neurological deficits were the primary outcome studied, consisting of both motor and sensory deficits. Deficits were either described or graded using a Modified McCormick Scale.[ 14 ] Deficits were evaluated immediately after surgery and then during follow-up (1–18 months). Tumors at all levels of the spinal cord were included. Titles and abstracts were independently screened to identify relevant studies. The following keywords are deficit and intramedullary . The study search dated from inception to October 2020.
Search and study selection
The literature search was conducted using PubMed, Web of Science, and Embase database for relevant literature. The following inclusion criteria were established before literature search: (1) randomized control trial, prospective, or retrospective cohort review, (2) conducted in patients with any type of IMSCT in any age group, (3) intraoperative SSEP monitoring during spinal cord tumor resection, (4) postoperative assessment conducted, (5) studies conducted with a sample size ≥10 patients, (6) published in English, and (7) included details regarding postoperative neurological deficits.
Data extraction
The authors (AM and MA) independently screened the title, abstract, and text against the inclusion criteria. Studies that did not meet all the criteria were rejected. All studies were recorded in an Excel spreadsheet with acceptance or rejection and reason for rejection by inclusion criteria not met (1–7). The final list was created after reconciliation by a third author (RR).
In each study, the following data were collected when available: study design, IONM modalities used, SSEP alarm criteria, intraoperative SSEP changes, intraoperative SSEP/MEP changes, type of postoperative neurological deficit, number of true positives (intraoperative SSEP and SSEP/MEP changes with new neurological motor or sensory deficit), false positives (intraoperative SSEP and SSEP/MEP change with no new neurological motor or sensory deficit), false negatives (no intraoperative SSEP and SSEP/MEP change with new neurological motor or sensory deficit), and true negatives (no intraoperative SSEP and SSEP/MEP change with no new neurological motor or sensory deficit). A 50% decrease in SSEP amplitude was used as the alarm criteria for SSEP change. A 50% decrease in SSEP or MEP amplitude was used as the alarm criteria for SSEP/MEP change.
True-positive, false-positive, false-negative, and true-negative values were collected for SSEP changes and SSEP/MEP changes. SSEP and SSEP/MEP sensitivity, specificity, and 95% confidence interval (CI) were calculated using a 2 × 2 Table. If a zero cell occurred in the study, sensitivity and specificity were calculated by adding 0.5 to all the cells to allow for statistical inference.
Statistical analysis
The statistical analyses were conducted using the MADA package in R (accessed on May 3, 2021). Analyses for the logit-transformed pairs were conducted using a bivariate normal model of sensitivities and false-positive rates before fitting to a linear mixed model. The linear mixed model preserved the bivariate nature of the data by considering the correlation between sensitivity and specificity. From this model, the mean logit sensitivity, specificity, and covariance were estimated. Forest plots and a summary receiver operating characteristic (ROC) curve with a 95% confidence ellipsoid were constructed.
Secondary analysis included fixed and random effects models, which were fitted for log-diagnostic odds ratios (DORs), of SSEP change and SSEP/MEP change. The bivariate nature of the data was converted to a univariate problem. Q and I 2 statistics were calculated to describe the percentage of variation across studies that is because of heterogeneity compared to chance. Q values indicate heterogeneity but I 2 values were used as the true indicator because they represent the extent of heterogeneity within a sample. A I 2 >0.2 was interpreted as significant heterogeneity. Publication bias and heterogeneity across all studies were assessed using a funnel plot [Supplemental Figure 2]. A McNemar test was performed on both SSEP and SSEP/MEP groups of studies to identify any significant differences in the sensitivities between unimodal and bimodal monitoring.
Overall SSEP changes were used to calculate the positive and negative likelihood ratios to predict postoperative neurologic deficits. The likelihood ratios were used to create a Fagan nomogram with the pretest probability equaling the total incidence of postoperative neurological deficits in the total cohort.
RESULTS
Literature search
Six hundred and forty-five papers were found through database searches. After assessing titles and abstracts, 184 papers remained for full-text screening. Following full-text screening, nine papers remained that reported outcomes following SSEP/MEP changes and SSEP changes alone. Therefore, relevant data from the nine final selected studies were used to create SSEP/MEP change and SSEP change cohorts. Statistical analysis was conducted on these nine studies reporting on intraoperative SSEP changes and SSEP/MEP changes and postoperative neurological outcomes in intramedullary tumor resection.
Study characteristics
The total cohort consisted of 303 patients for the SSEP group and 350 patients for the SSEP/MEP combined group. The rate of postoperative neurological deficits was 34.3% (104/303) for the SSEP group and 35.42% (124/350) for the SSEP/MEP group. Of the total SSEP cohort, 31.3% (95/303) incurred significant intraoperative SSEP changes. Of the total SSEP/MEP cohort, 32.6% (114/350) incurred significant intraoperative SSEP or MEP changes.
In patients with intraoperative SSEP changes, the rate of new postoperative neurological deficits was 53.6% (51/95). The deficit rate was 25.4% (53/208) in patients without intraoperative SSEP changes. In patients with SSEP or MEP changes (SSEP/MEP), the rate of postoperative neurological deficit was 75.4% (86/114). The deficit rate was 13.3% (38/236) in patients without intraoperative SSEP or MEP changes.
Data analysis
Postoperative neurological deficits and all significant somatosensory-evoked potential changes
Study sensitivities ranged from 27% to 92%, whereas study specificity ranged from 59% to 98% [Figure 2 ]. Using a bivariate mixed-effects model, the combined sensitivity across all studies was 55% (95% CI, 38%–71%) and the specificity was 76% (95% CI, 65%–85%). The area under the ROC curve was found to be 72.7% [Figure 3 ]. Using a pooled random effects model, the DOR for an SSEP change predicting postoperative neurological deficits was 5.6 [95% CI, 1.55–20.26, I 2 = 0%, Figure 4 ]. This finding suggests that patients who developed new postoperative neurological deficits following intramedullary tumor resection were ~ six times more likely to have had a significant intraoperative SSEP change compared to those who did not have a postoperative neurological deficit. The pooled log-diagnostic ratios (log-DORs) were calculated using a random effects model to decrease variations in individual CIs of studies.
Figure 2: Forest plot of the sensitivities and specificities for significant intraoperative SSEP changes in predicting postoperative neurological deficit. SSEP = Somatosensory evoked potential, CI = Confidence interval
Figure 3: HSROC curve for sensitivity and specificity of intraoperative SSEP in predicting postoperative neurological deficit. SSEP = Somatosensory evoked potential. HSROC = Hierarchical summary receiver operating characteristic
Figure 4: Forest plot of log-DOR for significant intraoperative SSEP changes in predicting postoperative neurological deficit. SSEP = Somatosensory evoked potential, log-DOR = log-diagnostic odds ratios
The positive and negative likelihood ratios depicted in the Fagan nomogram were 2.29 and 0.59, respectively [Figure 5 ]. The pretest probability was found to be 34.3%. This value was calculated based on the total incidence of postoperative neurological deficits in the SSEP cohort. The estimated posttest probability of developing a new postoperative neurological deficit increased to 54% for patients with a significant intraoperative SSEP change. The estimated posttest probability of developing a new postoperative neurological deficit decreased to 24% for patients without a significant intraoperative SSEP change.
Figure 5: Fagan nomogram depicting pre- and post-SSEP change probability for developing a new neurological deficit. SSEP = Somatosensory evoked potential
Postoperative neurological deficits and all significant somatosensory evoked potential/motor evoked potential changes
Study sensitivities ranged from 40% to 94%, whereas study specificity ranged from 25% to 99% [Figure 6 ]. Using a bivariate mixed-effects model, the combined sensitivity across all studies was 80% (95% CI, 62%–91%) and the specificity was 86% (95% CI, 68%–94%). The pooled area under the ROC curve was estimated to be 89.0% [Figure 7 ]. Under a pooled random effects model, the DOR for an intraoperative SSEP/MEP change predicting postoperative neurological deficits was 30.32 [95% CI, 6.92–132.85, I 2 = 0%, Figure 8 ]. The DOR suggests patients who developed new postoperative neurological deficits following intramedullary tumor resection were ~30 times more likely to have a significant intraoperative SSEP change compared to those who did not have a postoperative neurological deficit. The pooled log-DORs were calculated using the random effects model to reduce variability in CIs of individual studies.
Figure 6: Forest plot of the sensitivities and specificities for combined intraoperative SSEP and MEP changes in predicting postoperative neurological deficit. SSEP = Somatosensory evoked potential, CI = Confidence interval, MEP = Motor evoked potential
Figure 7: HSROC curve for sensitivity and specificity of combined intraoperative SSEP and MEPs in predicting postoperative neurological deficit. SSEP = Somatosensory evoked potential, HSROC = Hierarchical summary receiver operating characteristic, MEP = Motor evoked potential
Figure 8: Forest plot of log-DOR for SSEP and MEP changes in predicting postoperative neurological deficit. SSEP = Somatosensory evoked potential, CI = Confidence interval, MEP = Motor evoked potential, log-DOR = Log-diagnostic odds ratios
The McNemar test showed statistically significant difference in the sensitivities of SSEP compared to SSEP/MEP in two of the studies included in this analysis (Cannizzaro et al ., 2019,[ 15 ] P = 0.017; Park et al ., 2018,[ 16 ] P = 0.044). The remaining seven studies showed no significant difference in the sensitivities of SSEP compared to SSEP/MEP.
The positive and negative likelihood ratios were 5.71 and 0.23, respectively, and depicted in the Fagan nomogram [Figure 9 ]. The pretest probability was found to be 35.4% and was calculated based on the total incidence of postoperative neurological deficits in the SSEP/MEP cohort.
Figure 9: Fagan nomogram depicting pre- and postcombined SSEP or MEP change probability for developing a new neurological deficit. SSEP = Somatosensory evoked potential, MEP = Motor evoked potentia
The estimated posttest probability of developing new postoperative neurological deficits increased to 76% for patients with a significant intraoperative SSEP or MEP change. The estimated posttest probability of developing new postoperative neurological deficits decreased to 11% for patients without a significant intraoperative SSEP or MEP change.
DISCUSSION
The results of this systematic review show that intraoperative SSEP/MEP has better sensitivity and specificity compared to SSEP alone during intramedullary tumor resection surgery to accurately predict the risk of postoperative neurological deficits. Our analysis found that patients with new postoperative neurological deficits were nearly 30 times more likely to have an intraoperative SSEP/MEP change and six times more likely to have had only an intraoperative SSEP change compared to patients without new postoperative neurological deficits.
SSEPs do offer some advantages compared to other modalities. SSEPs can be used throughout the surgery without risk of patient movement and SSEP changes have been correlated postoperative sensory deficits.[ 17 ] In cases of intramedullary tumor resection, the function of the motor and sensory pathways can be damaged separately, leading to potentially worse outcomes using only SSEP monitoring.
SSEP monitoring only assess the function of the spinal cord through the dorsal column. The anterior two-thirds of the spinal cord , supplied by the anterior spinal artery, can be damaged and not recognized through SSEP monitoring. MEP monitoring can alleviate this risk by monitoring descending motor pathways which run in the lateral columns of the spinal cord , capturing a more holistic picture of spinal cord function during surgery, which was supported by our findings.
In our study, SSEP changes reported a lower sensitivity compared to specificity (55% vs. 76%, respectively). Prior studies have found SSEP measurements remain intact when serious postoperative motor deficits have occurred, which may be a result of low sensitivity of SSEP for postoperative motor deficits.[ 11 , 18 , 19 ] In cases where motor parameters may not remain stable, Deletis and Sala advised against solely using SSEP monitoring in IMSCT surgeries because surgical manipulations can damage either the motor or the sensory pathways.[ 20 ] Thus, for intramedullary tumor resection, SSEP may not be best used alone, but rather in conjunction with another form of IONM for better specificity and sensitivity of postoperative neurological deficits. Studies have found that MEPs may be present when SSEPs are lost or poorly defined.[ 21 ] MEPs are highly variable and can easily be altered by the effects of anesthesia and paralytics. SSEP is more specific for postoperative motor deficits.[ 22 ] In cases where both MEP and SSEP are not used simultaneously, SSEP use alone is helpful in predicting postoperative neurological deficits.[ 23 ]
In some cases, new postoperative neurological deficits can occur without IONM changes. Wilkinson et al . found patients with cervical radiculopathy or spondylotic myelopathy undergoing one- or two-level anterior cervical discectomy and fusion did not have decreased rates of neurological deficits with IONM use.[ 24 ] They attribute the discrepancy to a variety of factors including surgeon experience, stringent definitions of neurological deficits, and a false sense of security provided by IONM.[ 24 ] It is important that IONM not be used as a crutch for surgeons as the efficacy of IONM is under investigation. Outside of spinal tumor resection, IONM reports a low sensitivity to predict postoperative neurological deficits.[ 25 ] The absence of IONM changes may contribute to a false sense of security altering the surgeon’s clinical judgment.
This study does have several limitations. First, there are not many published studies documenting postoperative neurological deficits and SSEP changes in detail. In addition, there is little data outlining the types of SSEP and MEP changes that occur during surgery. Therefore, more investigation is needed to determine if certain types of SSEP and MEP changes correlate to different postoperative outcomes. For the studies that included another IONM modality in addition to SSEP and MEP, the predictive power of SSEP was likely altered.
Financial support and sponsorship
Nil.
Conflicts of interest
There are no conflicts of interest.
REFERENCES
1. Garcés-Ambrossi GL, McGirt MJ, Mehta VA, Sciubba DM, Witham TF, Bydon A, et al. Factors associated with progression-free survival and long-term neurological outcome after resection of
intramedullary spinal cord tumors:Analysis of 101 consecutive cases. J Neurosurg
Spine 2009;11:591–9.
2. Nakamura M, Ishii K, Watanabe K, Tsuji T, Takaishi H, Matsumoto M, et al. Surgical treatment of
intramedullary spinal cord tumors:Prognosis and complications.
Spinal Cord 2008;46:282–6.
3. Patil CG, Patil TS, Lad SP, Boakye M. Complications and outcomes after
spinal cord tumor resection in the United States from 1993 to 2002.
Spinal Cord 2008;46:375–9.
4. Imajo Y, Taguchi T, Yone K, Okawa A, Otani K, Ogata T, et al. Japanese 2011 nationwide survey on complications from
spine surgery. J Orthop Sci 2015;20:38–54.
5. Safaee MM, Lyon R, Barbaro NM, Chou D, Mummaneni PV, Weinstein PR, et al. Neurological outcomes and surgical complications in 221 spinal nerve sheath tumors. J Neurosurg
Spine 2017;26:103–11.
6. Manzano G, Green BA, Vanni S, Levi AD. Contemporary management of adult
intramedullary spinal tumors-pathology and neurological outcomes related to surgical resection.
Spinal Cord 2008;46:540–6.
7. Gonzalez AA, Jeyanandarajan D, Hansen C, Zada G, Hsieh PC. Intraoperative neurophysiological monitoring during
spine surgery:A review. Neurosurg Focus 2009;27:E6.
8. Sutter M, Eggspuehler A, Grob D, Jeszenszky D, Benini A, Porchet F, et al. The validity of multimodal intraoperative monitoring (MIOM) in surgery of 109
spine and
spinal cord tumors. Eur
Spine J 2007;16 Suppl 2 S197–208.
9. Forster MT, Marquardt G, Seifert V, Szelényi A.
Spinal cord tumor surgery –Importance of continuous intraoperative neurophysiological monitoring after tumor resection.
Spine (Phila Pa 1976) 2012;37:E1001–8.
10. Hilibrand AS, Schwartz DM, Sethuraman V, Vaccaro AR, Albert TJ. Comparison of transcranial electric motor and somatosensory evoked potential monitoring during cervical
spine surgery. J Bone Joint Surg Am 2004;86:1248–53.
11. Minahan RE, Sepkuty JP, Lesser RP, Sponseller PD, Kostuik JP. Anterior
spinal cord injury with preserved neurogenic 'motor'evoked potentials. Clin Neurophysiol 2001;112:1442–50.
12. Chen X, Sterio D, Ming X, Para DD, Butusova M, Tong T, et al. Success rate of motor evoked potentials for intraoperative neurophysiologic monitoring:Effects of age, lesion location, and preoperative neurologic deficits. J Clin Neurophysiol 2007;24:281–5.
13. Pelosi L, Stevenson M, Hobbs GJ, Jardine A, Webb JK. Intraoperative motor evoked potentials to transcranial electrical stimulation during two anaesthetic regimens. Clin Neurophysiol 2001;112:1076–87.
14. McCormick PC, Torres R, Post KD, Stein BM.
Intramedullary ependymoma of the
spinal cord . J Neurosurg 1990;72:523–32.
15. Cannizzaro D, Mancarella C, Nasi D, Tropeano MP, Anania CD, Cataletti G, et al.
Intramedullary spinal cord tumors:The value of intraoperative neurophysiological monitoring in a series of 57 cases from two Italian centres?. J Neurosurg Sci 2019 doi:10.23736/S0390-5616.19.04758-1.
16. Park JH, Lee SH, Kim ES, Eoh W. Analysis of multimodal intraoperative monitoring during
intramedullary spinal ependymoma surgery. World Neurosurg 2018;120:e169–80.
17. Cheng JS, Ivan ME, Stapleton CJ, Quinones-Hinojosa A, Gupta N, Auguste KI. Intraoperative changes in transcranial motor evoked potentials and somatosensory evoked potentials predicting outcome in children with
intramedullary spinal cord tumors. J Neurosurg Pediatr 2014;13:591–9.
18. Lesser RP, Raudzens P, Lüders H, Nuwer MR, Goldie WD, Morris HH 3rd, et al. Postoperative neurological deficits may occur despite unchanged intraoperative somatosensory evoked potentials. Ann Neurol 1986;19:22–5.
19. Paradiso G, Lee GY, Sarjeant R, Hoang L, Massicotte EM, Fehlings MG. Multimodality intraoperative neurophysiologic monitoring findings during surgery for adult tethered cord syndrome:Analysis of a series of 44 patients with long-term follow-up.
Spine (Phila Pa 1976) 2006;31:2095–102.
20. 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.
21. Costa P, Bruno A, Bonzanino M, Massaro F, Caruso L, Vincenzo I, et al. Somatosensory and motor-evoked potential monitoring during
spine and
spinal cord surgery.
Spinal Cord 2007;45:86–91.
22. Pelosi L, Lamb J, Grevitt M, Mehdian SM, Webb JK, Blumhardt LD. Combined monitoring of motor and somatosensory evoked potentials in orthopaedic spinal surgery. Clin Neurophysiol 2002;113:1082–91.
23. Kang H, Gwak HS, Shin SH, Woo MK, Jeong IH, Yoo H, et al. Monitoring rate and predictability of intraoperative monitoring in patients with intradural extramedullary and epidural metastatic spinal tumors.
Spinal Cord 2017;55:906–10.
24. Wilkinson BG, Chang JT, Glass NA, Igram CM. Intraoperative
spinal cord monitoring does not decrease new postoperative neurological deficits in patients with cervical radiculopathy or spondylotic myelopathy undergoing one or two level anterior cervical discectomy and fusion. Iowa Orthop J 2021;41:95–102.
25. Ibrahim T, Mrowczynski O, Zalatimo O, Chinchilli V, Sheehan J, Harbaugh R, et al. The impact of neurophysiological intraoperative monitoring during
spinal cord and
spine surgery:A critical analysis of 121 Cases. Cureus 2017;9:e1861.
