Secondary Logo

Journal Logo

Diagnostic Accuracy of Neuromonitoring for Identification of New Neurologic Deficits in Pediatric Spinal Fusion Surgery

Neira, Victor M. MD, MAEd; Ghaffari, Kamyar MD; Bulusu, Srinivas BSc, RT (EMG), RET, CNIM; Moroz, Paul J. MD, FRCSC; Jarvis, James G. MD, FRCSC; Barrowman, Nicholas PhD; Splinter, William MD, FRCPC

doi: 10.1213/ANE.0000000000001503
Pediatric Anesthesiology: Original Clinical Research Report
Free
SDC
CE
Continuing Medical Education

BACKGROUND: Intraoperative neuromonitoring (IONM) modalities, transcranial motor-evoked potentials (TcMEPs), and somatosensory-evoked potentials (SSEPs) are accepted methods to identify impending spinal cord injury during spinal fusion surgery. Debate exists over sensitivity and specificity of these modalities. Our purpose was to measure the incidence of new neurologic deficits (NNDs) and estimate sensitivity and specificity of IONM modalities.

METHODS: Institutional Ethics Board approval was obtained to review charts of patients younger than 22 years undergoing scoliosis surgery from 2007 to 2014 retrospectively. The definition of true-positive patients included two subgroups: (1) patients with an IONM alert, which did not resolve despite the interventions and had a NND postoperatively; or (2) patients with an IONM alert triggering interventions and the alert resolved with no NND postoperatively. Subgroup 2 of the definition is debatable; thus, we performed a multiple sensitivity analysis with three assumptions. Assumption 1: without interventions, all such patients would have experienced NNDs (assumption used in previous studies); Assumption 2: without intervention, half of these patients would have experienced NNDs; Assumption 3: without intervention, none of these of patients would have experienced NNDs.

RESULTS: We included 296 patients. Patients with incomplete charts (n = 3), no IONM monitoring (n = 11), and inadequate baseline IONM (n = 7) were excluded. The incidence of NND was 3.7% (95% confidence interval, 2.1%–6.5%). Successful IONM in at least one modality was obtained in 275 patients (92.9%), of whom 268 (97.5%) and 259 (94.2%) had successful baseline TcMEP or SSEP signals, respectively. Fifty-one (17%) patients had IONM alerts, 41 were only TcMEP, 5 were only SSEP, and 5 were in both modalities. After interventions, 42 (82%) patients recovered, 41 had no NND (true-positive under Assumption (1), but one developed a NND (false-negative). Of the 9 patients with no alert recovery, 6 had a NND (true-positive) and 3 did not (false-positives). Of the remaining 224 patients with no alerts, 221 had no NND (true-negatives) and 3 did (false-negatives). Sensitivity was estimated to be 93.5%, 92.2%, and 46.7% for TcMEPs, combination (either TcMEPs or SSEPs), and SSEPs, respectively. Multiple sensitivity analysis demonstrated that sensitivity and specificity vary markedly with different assumptions.

CONCLUSION: TcMEPs are more sensitive than SSEP at detecting an impending NND. IONM modalities are highly specific. Both sensitivity and specificity are impacted substantially by assumptions of the impact of interventions on alerts and NND. Properly designed, controlled, multicenter studies are required to establish diagnostic accuracy of IONM in scoliosis surgery.

From the Department of Anesthesiology, Children’s Hospital of Eastern Ontario, University of Ottawa, Ontario, Canada.

Victor M. Neira, MD, MAEd, is currently affiliated with the Department of Anesthesiology, Dalhousie University, Halifax, Nova Scotia, Canada.

Paul J. Moroz, MD, FRCSC, is currently affiliated with Shriners Hospitals for Children, Honolulu, Hawaii.

Accepted for publication June 19, 2016.

Funding: Department of Anesthesiology, Children's Hospital of Eastern Ontario.

The authors declare no conflicts of interest.

Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s website.

Reprints will not be available from the authors.

Address correspondence to Victor M. Neira, MD, MAEd, Department of Anesthesiology, Children’s Hospital of Eastern Ontario, University of Ottawa, 401 Smyth Rd, Ottawa, ON, Canada K1H 8L1. Address e-mail to vneira09@gmail.com.

A permanent new neurologic deficit (NND) after elective spinal surgery is one of the most feared complications for patients, their families, and their physicians. Some uncertainty in the actual incidence of NND and the variables associated with its development continues to exist. Much technology has been developed to reduce the risk and severity of neurologic injury. Intraoperative neuromonitoring (IONM) has been used during spinal fusion surgery (SFS) for early detection of impending spinal cord damage for more than 30 years and has largely replaced the wake-up test.1,2 The incidence of NND varies depending on several factors including the etiology of scoliosis with an incidence of 0.02%–0.7%, 1.0%–1.1%, and 2.0%–2.9% in idiopathic (IS), neuromuscular (NMS), and congenital scoliosis (CS), respectively.3

IONM has changed the anesthesia management paradigm. Initially, somatosensory-evoked potentials (SSEPs) were the only modality used and the anesthesia was managed using a balanced anesthesia technique with sub-MAC concentrations of volatile anesthetics (VA), opioids, muscle relaxants, and controlled hypotension.4 SSEPs monitor integrity of the sensory neural pathways in the dorsal column of the spinal cord (SC). Advances in IONM with concerns about injury to motor tracts in the spinal cord that could not be identified by SSEPs led to the use of transcranial motor-evoked potentials (TcMEPs) in addition to SSEPs. TcMEPs monitor the integrity of the motor tracts and the anterior segments of the spinal cord, which may be at higher risk of ischemic insult during SFS. Currently, most anesthesiologists recommend anesthesia management with total IV anesthesia (TIVA), normotension, and no muscle relaxation to accommodate both SSEPs and TcMEPs.2,5

There is controversy regarding IONM modalities’ diagnostic accuracy. Most authors have identified TcMEPs as the most sensitive modality, but others have reported SSEPs as the most sensitive.6–9 Accepted definitions of true-positive, false-positive, true-negative, and false-negative for IONM used by most authors are shown in Table 1.6–8,10 There is a concern that the common definition of true-positives results in an overestimation of sensitivity of the index test because it assumes that all patients with an alert would not recover unless an intervention is performed.

Table 1.

Table 1.

The objective of this study was to measure the incidence of NND and calculate the sensitivity and specificity of TcMEP and SSEP and their combination in detecting a NND in pediatric SFS. We also studied the association of preoperative comorbidities and intraoperative factors with IONM alerts.

Back to Top | Article Outline

METHODS

Institutional Research Ethics Board approval was obtained for this retrospective chart review (REB approval date: April 24, 2014, Protocol No. 14/71X). Patients undergoing SFS from April 2007 to May 2014 at a single pediatric academic institution, the Children’s Hospital of Eastern Ontario, were included in the study. The institutional review board waived informed consent. The STARD guidelines (Supplemental Digital Content, http://links.lww.com/AA/B470) for reporting diagnostic accuracy studies were followed in the writing process of this study.11 Inclusion criteria were subjects less than 22 years old undergoing SFS. Exclusion criteria were those without IONM or without good baseline in either TcMEP or SSEP modalities or those procedures done for vertical expandable prosthetic titanium rib (VEPTR) device insertion and lengthening, correction of spondylolisthesis, and surgery for spine trauma or tumor resection. If a patient had a two-staged procedure, the patient was considered as a single subject and each surgical event was not considered separately, ie, two surgeries were performed as parts of the same SFS.

The institutional comprehensive perioperative information database was managed using REDCap® software.12 One of the investigators collected the information (KG). Another investigator (VN) independently reviewed 20% of the charts to confirm data quality.

Back to Top | Article Outline

NEUROMONITORING

The neuromonitoring was performed using a 32-channel IONM machine (Cadwell Elite; Cadwell Laboratories, Inc, Kennewick, WA). We used subdermal needle electrodes for recording and stimulating purposes. Baseline SSEPs were obtained by interleaved electrical stimulation (rate of 3.7 Hz; 200-ms pulse width, 20-mA intensity, analysis time was 100 ms, band-pass 50–1500 Hz) from both median and posterior tibial nerves at peripheral, subcortical, and cortical levels.5

TcMEPs were evoked using a constant voltage stimulator (TCS-4; Cadwell Elite, WA) with a train of 5 stimuli having an interstimulus interval of 2.1 ms. The myogenic responses were recorded with a band pass of 50–5 kHz and sweep of 15 ms/div and sensitivity of 200 μV from upper and lower limb muscles for these roots: C3–4 trapezius, C5–6 biceps, C6–7 triceps, C7–8 extensor digitorum communis, T1 abductor pollicis brevis, T7–12 external oblique and rectus abdominis, L1–2 iliacus, L2–4 vastus medialis, L4–5 tibialis anterior, S1–2 medial gastrocnemius. Electroencephalographic signals were obtained by remontaging existing electrodes used for SSEP to derive a 4-channel display with real-time compressed spectral analysis. The 4 channels were between Fpz and Cz, C3 and Cz, Cz and C4, C3′, and C4′ and displayed in a separate window of the same multimodality IONM machine. Electroencephalography was used to assess anesthesia depth by the neuromonitoring technologist and to coordinate the management with the anesthesiologist. BIS was available but was not used routinely. Free-running electromyography and triggered electromyography were used to evaluate pedicle screws placement. Baselines were obtained and verified after induction of anesthesia and immediately after incision of the skin. Good IONM baseline was considered when TcMEP and SSEP signals were consistent and reproducible.

Back to Top | Article Outline

DEFINITION OF NEUROMONITORING ALERTS AND RESPONSE TO INTERVENTIONS

A persistent (for the last 3 test trials) unilateral or bilateral loss of ≥65% of the amplitude of the TcMEP or ≥50% of the amplitude of SSEP or 10% increase in the latency of the SSEP relative to a stable baseline was regarded as an alert.5,7 Recovery was defined as return of the signal in response to interventions to within 25% of the stable baseline value.7

Back to Top | Article Outline

INTERVENTIONS IN RESPONSE TO IONM ALERTS

Alerts were managed according to institutional guidelines, which were developed internally according to available guidelines in 2008 and updated in 2014.13,14 Whenever alerts occurred, the neurophysiology technologist notified the team and the following parameters were recorded: time of incident, anesthetic agents in use and their doses, mean arterial pressure (MAP), hematocrit, base deficit, temperature, and stage of surgery. The neurophysiology technologists registered all actions taken by surgeons, anesthesiologists, and technologists in the IONM machine. Abnormal physiologic parameters were corrected. Once the alert was declared, anesthesiologists used fluids or vasopressors to keep the mean arterial blood pressure above 90 mm Hg.

If the alert was persistent, surgeons would consider reversal of surgical maneuvers that had taken place such as reducing or removing the traction weights and then reversal of any deformity correction instrumentation, which may have been inserted before the alert (eg, distraction or derotational rods and/or pedicle screws, hooks, or sublaminar wires or cables). Such maneuvers allowed the spinal deformity to return to its original conformation and theoretically could reverse whatever potential insult that might have caused the IONM alert. If these measures were not successful, a wake-up test, urgent neurology consult, and methylprednisolone injection were considered.15,16

Back to Top | Article Outline

Independent Variables

Preoperative variables were age, gender, weight, height, etiology of scoliosis, Cobb’s or kyphosis angle, number of levels involved, and comorbidities by organ system. Organ system comorbidities were described as follows: cardiovascular, respiratory, neurologic, muscular, renal, hepatic, hematologic, gastrointestinal, endocrine, connective tissue, malnutrition, and other.

Intraoperative variables were type of surgery and included anterior, posterior, or combined approach in either one or two stages; surgeons, anesthesia technique including inhalation, balanced, or TIVA, hypotension (defined as a 20% decrease of systolic blood pressure from the preoperative baseline for more than 15 min)16,17; use of vasopressors as infusion for more than 15 minutes or 3 or more vasopressor boluses; minimal temperature; maximal base deficit; and minimal hematocrit.

Back to Top | Article Outline

Dependent Variables

Intraoperative variables were the following: (1) success in obtaining a consistent baseline signal in (i) combined IONM modalities (either TcMEP or SSEP), (ii) TcMEPs, and (iii) SSEPs; (2) IONM, TcMEPs, and SSEPs alerts; (3) IONM alert-related factors: (i) anesthesia related were defined as use of VA, muscle relaxants, boluses of anesthetics before the alert and hypotension, and (ii) surgery-related: application of traction weights, ligation of spinal arteries, osteotomies, screw insertion, rod insertion, and distraction for surgery. Alerts were categorized according to the stage of surgical correction: before, during, or after the correction18; and (4) interventions in response to alerts performed by the anesthesiologists and surgeons according to guidelines were other variables of interest.

Postoperative variables were as follows: incidence of NND defined as a sensory or motor deficit that was not present before the surgery. The level of NND was defined according to Reames as total or partial spinal cord, cauda equina, nerve root, brachial plexus, and peripheral nerve defect.3 Functional recovery was classified as complete, partial, or none compared with the preoperative status.3 The follow-up of patients with a NND was up to 18 months.

The primary outcome was the incidence of NND. The secondary outcomes were success in the IONM by modality, incidence of alerts, sensitivity, and specificity of each IONM modality and their combination. Tertiary outcomes were association of perioperative factors with IONM alerts.

Back to Top | Article Outline

Statistical Analysis

Given a true frequency of NND 0.8%–2% and assuming a missing data of 10%, which corresponds to the percentage of cases without successful IONM, a sample size of 300 subjects was estimated to allow estimation of the frequency of NND within ±1.7% with 95% confidence. Treating combined staged procedures as if they were single procedures is at best an approximation. To evaluate the sensitivity of our results to this approximation, we removed all of these combined staged procedures and repeated our analysis.

Quantitative variables were summarized as mean, standard deviations, and 95% confidence intervals (CIs). Incidence of NND was described as a percentage with 95% Wilson score CI. For the analysis of the diagnostic test properties of alerts, patients with incomplete data, no monitoring, or uninterpretable baseline signals were excluded. IONM alerts (SSEP and TcMEP) were described as percentages with 95% Wilson score CIs.

Definitions of true-positive, false-positive, true-negative, and false-negative subjects were taken based on previous studies as shown in Table 1 to calculate sensitivity and specificity of TcMEP and SSEP monitoring.6–8 Definition of true-positive cases in these studies includes patients in whom: (1) an alert triggers an intervention, signal is recovered, and no NND occurs; or (2) an alert triggers an intervention, signal is not recovered, and NND occurs. Sensitivity and specificity of the combined IONM modalities (either TcMEP or SSEP) and each individual IONM modality (TcMEP and SSEP) in predicting a NND was calculated and reported with 95% CI using Wilson score using the known definition of them. “Sensitivity is the proportion of people with the target disorder in whom the test result is positive.” “Specificity is the proportion of people without the target disorder in whom a test result is negative.”19,20

A multiple sensitivity analysis was performed to address concerns regarding potential overestimation of sensitivity related to inclusion of subgroup 1 of the previously mentioned definition of true-positive patients in calculations. A multiple sensitivity analysis was performed with three assumptions: Assumption 1: without interventions, all such patients would have experienced NNDs (assumption used in previous studies)6–8; Assumption 2: without intervention, half of these patients would have experienced NNDs; Assumption 3: without intervention, none of these of patients would have experienced NNDs.

Anesthesia and surgery-related factors were analyzed using descriptive statistics. A univariate analysis was performed to identify association between independent variables and dependent variables. The number of alerts per case was modeled in terms of presence or absence of specific types of comorbidity using Poisson regression. Data were analyzed using SPSS software version 23 (IBM SPSS Statistics for Windows, Version 23.0; Armonk, NY). Two-sided P values <.05 were deemed significant.

Back to Top | Article Outline

RESULTS

Demographic and clinical characteristics of participants are shown in Table 2. A total of 296 eligible participants who underwent 318 SFS were included. Patients were excluded from the analysis because of no use of IONM (n = 11), no good baseline in any of the modalities (n = 7), and incomplete data (n = 3) (Figure 1).

Table 2.

Table 2.

Figure 1.

Figure 1.

Eleven patients had a NND in the postoperative period; 1 of them was among the patients excluded from the analysis, because this patient did not receive IONM (Case 2 in Table 3). The incidence of NND in our study was 3.7% (95% CI, 2.1%–6.5%). The incidence of NND was 2.1 (95% CI, 0.8%–5.2%) in IS, 6.7% (95% CI, 2.6%–15.9%) in NMS, 14.3% (95% CI, 2.6–51.3) in CS, and 8.3% (95% CI, 2.3%–25.8%) in patients with other etiologies. All the patients except patient 2 had returned to their baseline level of function during the 18-month follow-up. Patients with a NND at the level of peripheral nerve or brachial plexus had complete recovery within the first postoperative week (Table 3).

Table 3.

Table 3.

Good baseline signals in either TcMEP or SSEP were present in 275 (92.9%) patients; of those, 268 (97.5%) patients had successful TcMEP and 259 (94.2%) had successful SSEP monitoring. TcMEP monitoring was successful in 99%, 72%, and 100% in IS, NMS, and CS, respectively (P <.001). SSEP monitoring was successful in 98%, 63%, and 86% in IS, NMS, and CS, respectively (P <.001).

A flowchart of patients with combined IONM (either TcMEP or SSEP) and NND is in Figure 1. The total number of detected alerts was 70, ie, in some patients there was more than one alert. IONM alerts were detected in 51 patients (17%), 42 of them recovered after interventions. Of these responders, 1 woke up with a NND (false-negative) and the other 41 did not (true-positives). Among the nonresponders, 6 patients had NND (true-positives) and 3 did not have NND (false-positives). Among the patients with no alerts, 3 of them had a NND (false-negatives) and 221 did not have a NND (true-negatives). Flowcharts showing results of TcMEPs and SSEPs for NND are in Figures 2 and 3. None of the patients with combined staged surgeries had alerts in both procedures.

Figure 2.

Figure 2.

Results of multiple sensitivity analysis for combined and individual IONM modalities are shown in Table 4. We examined the effect of the assumptions regarding definition of true-positive patients on modalities’ test specifications. Looking at our flowcharts, as the assumption moves from Assumption 1 (total acceptance of the definition) to Assumption 3 (total rejection of the definition), the number of true-positives decrease and false-positives increases. This is reflected by a decrease in sensitivity, specificity, and positive predictive value (PPV) of each modality. Negative predictive value remained unchanged for each modality. Sensitivity was high for TcMEPs (93.5%; 95% CI, 82.5%–97.8%) and combined (92.2%; 95% CI, 81.5%–96.9%) (Figure 2; Table 4). SSEPs with their high false-negative rate had poor sensitivity (46.7%; 95% CI, 24.8%–69.9%) (Tables 3 and 4; Figure 3). All modalities were highly specific. Two-by-two tables and calculations of diagnostic accuracy are in the Supplemental Digital Content (Appendix 1, http://links.lww.com/AA/B469). There were 22 patients with combined staged procedures in our database who were treated as a single SFS. We repeated the estimates after excluding these 22 patients from the database. Results were minimally changed.

Table 4.

Table 4.

Figure 3.

Figure 3.

Preoperative comorbidities and IONM alerts were analyzed (Table 5). The average number of TcMEP alerts in patients with cardiovascular comorbidities was approximately 2.6 times higher than in patients without them (95% CI, 1.3–4.9; P = .005). The average number of SSEPs alerts in patients with gastrointestinal comorbidities was approximately 5.6 times higher than in patients without them (95% CI, 1.64–19.11; P = .006). Other preoperative factors were not significantly associated with IONM alerts.

Table 5.

Table 5.

Figure 4.

Figure 4.

Figure 5.

Figure 5.

Description of anesthesia and surgery-related factors at the time of alerts are in Figure 4. Blood pressure was reduced in more than 50% of alerts. Vasopressors were in use at the time of alerts in 52%, 60%, and 100% of the first, second, and third alerts, respectively. There was no association between use of volatile agents for anesthesia maintenance and alerts, although in more than 30% of alerts, a volatile agent was in use. No other studied variables were associated with alerts. Interventions are in Figure 5. Anesthesiologists’ intervention responses to IONM alerts were as follows: administration of vasopressors, IV fluids, blood transfusion, decreasing anesthetic depth, or repositioning the limb in 15% to 44% of the occasions. Surgeons modified the surgery, released the traction, or decreased the distraction in 11% to 16% of the alerts.

Back to Top | Article Outline

DISCUSSION

The overall incidence of NND (3.7%) in our study was greater than that reported by Qiu (1.89%) among their 1074 pediatric scoliosis patients and Reames (1%) in 19,360 patients from the Scoliosis Research Society Morbidity and Mortality Database.3,21 A potential explanation for our higher incidence of NND is that we were able to capture all cases with NND independently of the severity. Another possible explanation is that large databases have the limitations of volunteer reporting with potentially incomplete or inaccurate data and focusing on major complications, which may underestimate the true incidence.

IONM was successful in at least one modality in 97.5% of patients. TcMEP and SSEP monitoring were successful in 95.0% and 91.8% of patients, respectively. Our results on SSEP or combined monitoring are consistent with a previous study with a similarly mixed diagnosis population (n = 162) by Vitale who reported success rates of 92% and 93% in SSEP and combined modalities, respectively, but we were more successful in TcMEP monitoring.10 We found significant differences in success rate for both TcMEP and SSEP between IS and NMS scoliosis subjects, whereas Vitale found significant differences only for TcMEPs. The success rate in TcMEPs and SSEPs in NMS patients was 73% and 62%, respectively, which is similar to Vitale (64% and 82%, respectively).10

Our incidence of alerts was 18.5%, which is higher than others (3.4%–8%),7,8,10 but similar to Pelosi (15%).22 Our patients and those of Vitale and Pelosi were heterogeneous but Vitale used TIVA for maintenance of anesthesia, whereas Pelosi did not use a uniform technique as in our study.10,22 The different anesthetic techniques may explain the observed differences in alert incidences.

There is little agreement regarding the criteria to declare a TcMEP alert. Authors have used decreases in amplitude between 50% and 80%.5–7,22,23 We used 65%, which is the value used in a similar multicenter study.7 These variations in the threshold affect diagnostic accuracy of the test. If the threshold is set low, eg, 50%, the number of alerts (false-positives) increases and the perioperative team will have increased interruptions to address these alerts. If the threshold is set high, eg, 80%, the number of alerts decreases, but the risk of a missed reversible NND increases. Also, others such as Bhagat defined a true alert as the one with IONM criteria, once anesthesia, positioning, and technical factors were all ruled out.23 Most authors consider an alert returning to baseline as a true-positive. They explain the temporal relationship between the corrective interventions and signal recovery indicates an impending spinal cord injury successfully prevented by the interventions. However, other authors defined these cases as probably positives.23

Also what is unclear is the definition of a clinically significant recovery from an alert. Although there is a range for amplitude recovery from 25% to 100% of the baseline, we followed Schwartz and accepted a 25% return to baseline as recovery.7,23,24 Under Assumption 1, we found TcMEPs to have high sensitivity (93.5%). Combination of IONM modalities had almost the same sensitivity (92.2%). As noted in Figure 1, 275 patients had good baseline in either TcMEPs or SSEPs, whereas only 265 patients had good baseline in TcMEPs (Figure 2). These populations of 275 and 268 patients were not completely overlapped and explain the slight differences in sensitivity. SSEPs had poor sensitivity (46.7%). Other authors reported 100% sensitivity for TcMEP or combined modalities with similarly low sensitivity for SSEPs.7,8,23 Our TcMEP sensitivity is different from others because we had 3 false-negative patients; patients 1 and 3 had purely sensory deficit, and patient 5 had a peripheral nerve sensory deficit (Table 3). Patient 3 had TcMEP and SSEP alerts in the right arm that responded to repositioning but the patient had C5 to T1 motor and C5, C6 sensory neurapraxia that resolved 6 days later.

SSEP modality did not identify correctly any of patients with NNDs in our study. Reports of patients with a NND and a falsely negative SSEP exist.25,26 We found that patients with a new purely sensory or peripheral nerve deficit may have a negative TcMEP result, and patients with a new purely motor or peripheral nerve deficit may have a negative SSEP result. In a recent study of 477 patients with IS or CS undergoing SFS using inhalational anesthetics and muscle relaxants, SSEP was reported as the sole monitoring modality. Authors used described criteria for alerts and found a sensitivity of 95.0% and a specificity of 99.8%.9

It is unknown how many patients with an alert will recover without a NND if no intervention is utilized. We performed a multiple sensitivity analysis to examine the effect of 3 theoretical/plausible recovery patterns. The sensitivity, specificity, and PPV decrease because cases move from true-positives to false-positives as we transition from Assumption 1 to Assumption 3. Previous studies considered those cases as true-positives as explained earlier. We consider that Assumption 1 may overestimate the real sensitivity of the test and that the sensitivity is most likely between Assumptions 1 and 2 (combined 71.1%–96.9%; TcMEPs 72.8%–97.8%, and SSEPs 13.8%–69.9%). Some authors consider alerts secondary to hypotension, anesthesia, and technically-related factors as false-positives.23,25

Anesthetic and physiologic factors impair SSEPs and TcMEPs differently.2,5,27–29 TcMEPs are affected by VA, including N2O and muscle relaxants. SSEPs are less affected by VA and the use of muscle relaxants decreases interference and improves signal-to-noise ratio. Most authors recommend TIVA to optimize TcMEP signaling. However, recently authors noticed that below MAC concentrations of desflurane did not compromise TcMEP monitoring as long as they increased stimulating voltage.30 Moreover, hypotensive anesthesia has been extensively used to decrease bleeding and blood transfusion in SFS, but hypotension may be associated with spinal cord hypoperfusion.31,32 There is no consensus regarding the best anesthesia technique and hemodynamic targets in SFS.

Patients’ diagnoses influence IONM. Among NMS patients, a high-risk population for NND, IONM was unfortunately more difficult to obtain. In contrast, Hermanns obtained equally effective IONM in IS and NM using propofol–remifentanil anesthesia.33 Patient comorbidities may impact IONM alerts. Vitale reported cardiopulmonary comorbidities to be associated with IONM alerts.10 We were able to only identify cardiovascular and not pulmonary comorbidities as the associated factor. More specifically TcMEPs were affected and not SSEPs. TcMEPs are reflective of anterior spinal cord blood flow. Thus, it is not surprising that patients with a vascular and/or cardiac problem would be at increased risk of spinal cord ischemia.

The current study has several strengths. The study population was diverse and a typical SFS caseload. As well, the study size was adequate for detailed univariate analysis of subgroups. Only 1% (3 patients) were excluded because of incomplete information when compared with other studies where as much as 45% of charts were excluded because of inadequate records.10

One of the limitations of the study is the retrospective design with the possibility of incomplete or inaccurate documentation. The source of information to identify NND was the surgeons’ pre- and postoperative notes. Surgeons’ and anesthesiologists’ interventions in response to alerts were extracted from physicians’ notes and annotations in the IONM machine. Heterogeneity in patient population, anesthetic management, and surgical and IONM techniques may introduce confounder factors that affect results.

The wake-up test was considered the standard for identification of NND during SFS. It has multiple limitations and complications including: observation at a single point in time, increased operative time, patient movement, accidental extubation and vascular lines dislodgement, pain, and recall. Its use has decreased with the implementation of IONM.5,34

The Scoliosis Research Society (SRS) statement on neuromonitoring supports IONM use.35 However, the level of evidence is limited to retrospective case series and expert opinion (level of evidence C). IONM may require important changes in the anesthesia management, and currently there are no clear guidelines. IONM also requires sophisticated and expensive equipment and a dedicated technologist with specific training and credentialing to interpret IONM signals. We identified multiple gaps and inconsistencies in techniques, definitions, and interpretation of IONM. Monitoring selected neural structures, inability to obtain a good baseline in all patients, and poor association between transient changes and postoperative NNDs are other limitations. Clinicians should consider all these factors. Prospective randomized controlled trials with a multicenter design and larger number of patients are required to further establish performance characteristics and diagnostic accuracy of IONM in SFS. Moreover, reporting NND by major subspecialty societies such as the SRS should become more rigorous and mandatory like the registry for total joint arthroplasties.

Back to Top | Article Outline

CONCLUSIONS

We observed the incidence of NNDs in elective pediatric SFS to be low but not insignificant. TcMEP alone or in combination as multimodal monitoring is more sensitive than SSEP monitoring alone. All the IONM modalities are highly specific (99%–100%). There is a lack of agreement in the definitions of the threshold for TcMEP alerts and recovery. Including patients with recovered alerts not related to surgical factors as true-positive may overestimate sensitivity. Properly designed controlled multicenter studies are required to establish performance characteristics and diagnostic accuracy of IONM in SFS in the future.

Back to Top | Article Outline

DISCLOSURES

Name: Victor M. Neira, A, MD, MAEd.

Contribution: This author helped design the study, conduct the study, analyze the data, and write the manuscript.

Name: Kamyar Ghaffari, MD.

Contribution: This author helped conduct the study, analyze the data, write the manuscript, and collect the data.

Name: Srinivas Bulusu, BSc, RT (EMG), RET, CNIM.

Contribution: This author helped design the study, conduct the study, analyze the data, and write the manuscript.

Name: Paul J. Moroz, MD, FRCSC.

Contribution: This author helped design the study, conduct the study, analyze the data, and write the manuscript.

Name: James G. Jarvis, MD, FRCSC.

Contribution: This author helped design the study, conduct the study, analyze the data, and write the manuscript.

Name: Nicholas Barrowman, PhD.

Contribution: This author helped design the study, analyze the data, and write the manuscript.

Name: William Splinter, MD, FRCPC.

Contribution: This author helped design the study, analyze the data, and write the manuscript.

This manuscript was handled by: James A. DiNardo, MD, FAAP.

Back to Top | Article Outline

REFERENCES

1. Thuet ED, Winscher JC, Padberg AM, et al.Validity and reliability of intraoperative monitoring in pediatric spinal deformity surgery: a 23-year experience of 3436 surgical cases.Spine (Phila Pa 1976)20103518801886
2. DiCindio S, Schwartz DMAnesthetic management for pediatric spinal fusion: implications of advances in spinal cord monitoring.Anesthesiol Clin North America200523765787
3. Reames DL, Smith JS, Fu KM, et al.Scoliosis Research Society Morbidity and Mortality CommitteeComplications in the surgical treatment of 19,360 cases of pediatric scoliosis: a review of the Scoliosis Research Society Morbidity and Mortality database.Spine (Phila Pa 1976)20113614841491
4. Hersey SL, O’Dell NE, Lowe S, et al.Nicardipine versus nitroprusside for controlled hypotension during spinal surgery in adolescents.Anesth Analg19978412391244
5. Padberg AM, Thuet EDIntraoperative electrophysiologic monitoring: considerations for complex spinal surgery.Neurosurg Clin N Am200617205226
6. Hilibrand AS, Schwartz DM, Sethuraman V, Vaccaro AR, Albert TJComparison of transcranial electric motor and somatosensory evoked potential monitoring during cervical spine surgery.J Bone Joint Surg Am200486-A12481253
7. Schwartz DM, Auerbach JD, Dormans JP, et al.Neurophysiological detection of impending spinal cord injury during scoliosis surgery.J Bone Joint Surg Am20078924402449
8. Kundnani VK, Zhu L, Tak H, Wong HMultimodal intraoperative neuromonitoring in corrective surgery for adolescent idiopathic scoliosis: evaluation of 354 consecutive cases.Indian J Orthop2010446472
9. Thirumala PD, Bodily L, Tint D, et al.Somatosensory-evoked potential monitoring during instrumented scoliosis corrective procedures: validity revisited.Spine J20141415721580
10. Vitale MG, Moore DW, Matsumoto H, et al.Risk factors for spinal cord injury during surgery for spinal deformity.J Bone Joint Surg Am2010926471
11. Bossuyt PM, Reitsma JB, Bruns DE, et al.STARD GroupSTARD 2015: an updated list of essential items for reporting diagnostic accuracy studies.BMJ2015351h5527
12. Harris PA, Taylor R, Thielke R, Payne J, Gonzalez N, Conde JGResearch electronic data capture (REDCap)—a metadata-driven methodology and workflow process for providing translational research informatics support.J Biomed Inform200942377381
13. Ziewacz JE, Berven SH, Mummaneni VP, et al.The design, development, and implementation of a checklist for intraoperative neuromonitoring changes.Neurosurg Focus201233E11
14. Vitale MG, Skaggs DL, Pace GI, et al.Best practices in intraoperative neuromonitoring in spine deformity surgery: development of an intraoperative checklist to optimize response.Spine Deform20142333339
15. Society for Pediatric AnesthesiaPediatric Critical Event Checklists201317Available at: http://www.pedsanesthesia.org/wp-content/uploads/2015/02/Critical_Event_Checklists.pdf. Accessed April 2, 2016
16. Nafiu OO, Voepel-Lewis T, Morris M, et al.How do pediatric anesthesiologists define intraoperative hypotension?Paediatr Anaesth20091910481053
17. Walsh M, Devereaux PJ, Garg AX, et al.Relationship between intraoperative mean arterial pressure and clinical outcomes after noncardiac surgery: toward an empirical definition of hypotension.Anesthesiology2013119507515
18. Jarvis JG, Strantzas S, Lipkus M, et al.Responding to neuromonitoring changes in 3-column posterior spinal osteotomies for rigid pediatric spinal deformities.Spine (Phila Pa 1976)201338E493E503
19. Jaeschke R, Guyatt H, Sackett DLang TA, Secic MDeterminimg the presence of absence of disease. Reporting performance characteristics of diagnostic tests.How to Report Statistics in Medicine20062nd ed.Philadelphia, PAAmerican College of Physicians;125147
20. Jaeschke R, Guyatt GH, Sackett DUsers’ ‘guides to the medical literature.’ How to use an article about a diagnostic test.JAMA1994271703707
21. Qiu Y, Wang S, Wang B, Yu Y, Zhu F, Zhu ZIncidence and risk factors of neurological deficits of surgical correction for scoliosis: analysis of 1373 cases at one Chinese institution.Spine (Phila Pa 1976)200833519526
22. Pelosi L, Lamb J, Grevitt M, Mehdian SM, Webb JK, Blumhardt LDCombined monitoring of motor and somatosensory evoked potentials in orthopaedic spinal surgery.Clin Neurophysiol200211310821091
23. Bhagat S, Durst A, Grover H, et al.An evaluation of multimodal spinal cord monitoring in scoliosis surgery: a single centre experience of 354 operations.Eur Spine J20152413991407
24. Mohamad F, Parent S, Pawelek J, et al.Perioperative complications after surgical correction in neuromuscular scoliosis.J Pediatr Orthop200727392397
25. Ginsburg HH, Shetter AG, Raudzens PAPostoperative paraplegia with preserved intraoperative somatosensory evoked potentials. Case report.J Neurosurg198563296300
26. Malhotra NR, Shaffrey CIIntraoperative electrophysiological monitoring in spine surgery.Spine (Phila Pa 1976)20103521672179
27. Banoub M, Tetzlaff JE, Schubert APharmacologic and physiologic influences affecting sensory evoked potentials: implications for perioperative monitoring.Anesthesiology200399716737
28. Lotto ML, Banoub M, Schubert AEffects of anesthetic agents and physiologic changes on intraoperative motor evoked potentials.J Neurosurg Anesthesiol2004163242
29. Glover CD, Carling NPNeuromonitoring for scoliosis surgery.Anesthesiol Clin201432101114
30. Martin DP, Bhalla T, Thung A, et al.A preliminary study of volatile agents or total intravenous anesthesia for neurophysiological monitoring during posterior spinal fusion in adolescents with idiopathic scoliosis.Spine (Phila Pa 1976)201439E1318E1324
31. Raw DA, Beattie JK, Hunter JMAnaesthesia for spinal surgery in adults.Br J Anaesth200391886904
32. Malcolm-Smith NA, McMaster MJThe use of induced hypotension to control bleeding during posterior fusion for scoliosis.J Bone Joint Surg Br198365255258
33. Hermanns H, Lipfert P, Meier S, Jetzek-Zader M, Krauspe R, Stevens MFCortical somatosensory-evoked potentials during spine surgery in patients with neuromuscular and idiopathic scoliosis under propofol–remifentanil anaesthesia.Br J Anaesth200798362365
34. Mendiratta A, Emerson RGNeurophysiologic intraoperative monitoring of scoliosis surgery.J Clin Neurophysiol2009266269
35. Spine Research SocietySRS IONM Statement 20092009Available at: http://www.srs.org/professionals/advocacy_and_public_policy/nueromonitoring_information.htm. Accessed August 11, 2015

Supplemental Digital Content

Back to Top | Article Outline
Copyright © 2016 International Anesthesia Research Society