The incidence of neurological deficits in cervical spine surgery has been relatively low, ranging between 0.2% and 3.2% [1]. Nevertheless, there has been a continuous effort to improve the safety of cervical spine surgery. Intraoperative neuromonitoring (IONM) is one of the adjuncts hypothesized to improve safety [2–5]. The benefits of IONM in deformity surgery have been long established to reduce incidence of major neurological deficits [2,6,7]. However, current evidence on IONM in cervical spine surgery lacks the strength to give conclusive suggestions, despite showing a reduced rate of new neurological deficits [8,9].

IONM during spinal surgery utilizes three different modalities, namely, somatosensory evoked potential (SSEP), motor evoked potential (MEP), and electromyography (EMG) [10–12]. In view of the devastating consequences of a neurological injury, many surgeons advocate the routine use of IONM to detect possible intraoperative insult; hence immediate remedial measures can be taken [13]. This can be seen in the increasing use of IONM in noncomplex spine surgeries [14,15].

However, there are limitations with using IONM in cervical spine surgery. Fehlings et al. [16] in 2010 showed that multimodal IONM had a wide range of sensitivity between 70% and 100% and specificity ranging from 52.7% to 100%. Previous studies on IONM in cervical spine surgeries showed a higher incidence of false-positive alerts, especially in cases with cervical myelopathy [17]. In addition to a wide specificity range and a low positive predictive value (PPV), the high incidence of false-positive signal alerts makes it challenging for the surgeon to make an intraoperative decision.

This study aimed to determine when an IONM signal alert was clinically significant by comparing the incidence of true neurological deficits in different patterns of signal alerts. In addition, we aimed to identify the risk factors for false-positive signal alerts. The secondary objective of this study was to evaluate the correlation of cervical five (C5) palsies with neuromonitoring signal alerts, as it is generally considered to be a separate entity from a true neurological deficit.

This study was approved by the National Healthcare Group-Domain Specific Review Board (NHG-DSRB), Singapore with reference no. 2017/00373. The requirement for informed consent from individual patients was omitted because of the retrospective design of the this study. Clinical and IONM records of a cohort of 257 patients who underwent cervical spine surgeries were retrospectively reviewed. Fellowship-trained spine surgeons performed all procedures. Consecutive adult patients who underwent spinal surgeries involving the cervical spine down to thoracic one (T1) level were selected for analysis. Both MEP and SSEP were monitored intraoperatively. The exclusion criteria were the absence of a recordable baseline MEP or SSEP preoperatively, surgeries involving tumor, and surgeries performed without IONM.

Electronic and paper medical records of each patient were reviewed for demographic data such as gender and age and clinical information such as the presence of myelopathy and American Spinal Injury Association (ASIA) impairment scale preoperatively. Documentation of postoperative neurological assessment done by the surgical team members immediately after the patient recovers from anesthesia and that which was done by the resident physician twice daily throughout inpatient stay were reviewed for deficits. Neurological deficit was defined as any new motor weakness as per Medical Research Council grading or sensory loss which presented immediately after surgery, excluding isolated C5 deficits. The incidence of C5 palsy was recorded and evaluated separately. Surgical data that were collected included the surgical approach and the number of surgical levels. Imaging records of the cervical spine were reviewed for the presence of ossified posterior longitudinal ligament (OPLL). Intraoperative anesthesia charts were also reviewed to identify significant changes in body temperature or intraoperative blood pressure.

A trained neurophysiologist provided by a commercial neurophysiological monitoring service reviewed IONM records of MEPs and SSEPs. Significant signal drops were defined as a 50% loss of amplitude in MEPs or SSEPs after considering all factors stated in the IONM protocol. A sustained alert was defined as a signal change that persisted till the end of surgery, while a nonsustained drop would return to baseline before the end of surgery.

Data was tabulated and stored on a Microsoft Excel spreadsheet without patient identifiers. Two clinicians who were not involved in the patients’ management evaluated all medical records. These two clinicians were medical officers (junior resident equivalent) with at least 3 years’ experience in orthopedic surgery, one of whom was the primary author.

A total of 257 cases were reviewed and 207 patients were analyzed after excluding those without detectable baseline MEP or SSEP signals preoperatively. The mean and median age of patients was 58 years old; 79% (163/207) of patients were males, 50% (104) had cervical myelopathy, and 18% (38/207) had OPLL; 50% (103/207) of surgeries involved three levels or more, 52% (108/207) were done via the anterior approach, 45% (94/207) were done via the posterior approach, and 2% (5/207) were done via both anterior and posterior approach at the same setting. Only 7% (14/207) of patients had severe preoperative neurological deficit with an ASIA impairment scale of either A, B, or C. Emergent surgeries contributed to 12% (25/207) of cases. The incidence of new neurological deficit after surgery was 1% (2/207).

The increasing use of IONM in cervical spine surgery in recent years has been contested, mainly because of reports of false negatives and the disadvantages of each of the neuromonitoring modality when used separately [17–19]. It should be noted that some of the studies included isolated C5 palsies as neurological deficits that contributed to false negatives [17]. Despite the controversy, Fehlings et al. [16] revealed a consistently high sensitivity of 70%–100%. In view of a variety of evidence, our study aimed to determine the incidence of positive IONM signal alerts and when it was clinically significant.

In our study, we found that IONM signal alerts were relatively common, occurring in 25% of all cases. However, the overall incidence of postoperative true neurological deficit was only 1% in our study, which fell well within previously reported range of 0.2%–3.2% [17,20]. This meant that most (96%) of the signal alerts were false positives. To help the surgeon interpret these high false-positive rates, we identified that group 2 (sustained, bimodal) signal drops in SSEPs and MEPs had the highest rate of neurological deficits at 29%. There was no occurrence of neurological deficits in unimodal signal alerts, regardless of whether the alert was transient or sustained. Hence, the highest level of caution should be taken if there are concurrent signal drops in MMEP and SSEPs intraoperatively as it is prognostic for a postoperative neurological deficit [12,21].

There were no statistically significant differences in variables that could account for false-positive signal alerts (Table 3). However, a trend was found in patients with cervical myelopathy to have false-positive signal alerts. The incidence of false-positive signal alerts in those with myelopathy was higher, at 29%, compared with those who were not myelopathic, at 19%. This result replicates the higher rate of false-positive IONM alerts in patients with cervical myelopathy seen in previous studies [17,20].

We found an incidence of 6% in C5 palsy, which was also within the previously reported range of 0%–30% [22–25]. No correlation was found between occurrence of C5 palsy and positive IONM signal alerts (p=0.73). The sensitivity of IONM in detecting C5 palsy was 17% with a specificity of 26%. This may be artificially low due to the fact that we chose not to look at EMG records as evidence to show that real-time EMG monitoring was sensitive to C5 palsies, while SSEPs and MEPs were not [16].

There are several limitations in our study. The retrospective nature of this study makes it prone to recall bias. A large proportion of cases were excluded based on not having detectable baseline MEP or SSEP signals, some of which utilized unimodal IONM during surgery. Exclusion of these cases may skewer the efficacy of the individual modalities. Despite having 207 patients, the sample size maybe too small to produce significant results and this may be the reason that there were no statistically significant risk factors for false-positive IONM alerts. In addition, certain pertinent factors such as length of surgery or estimated blood loss could not be obtained from available records.

The incidence of positive signal drops in IONM for cervical spine surgeries was 25%. Bimodal IONM using MEP and SSEP provided sensitivity of 100% and specificity of 97.6% in detecting new neurological deficits, while the sensitivity in detecting C5 palsies was only 17%. Our study showed that by grouping the signal alerts into transient or sustained and unimodal or bimodal, the surgeon will be able to better predict which signal drops are clinically significant, thereby allowing for improved intraoperative decision making for remedial measures.