Intraoperative Neurophysiological Monitoring for Intradural Extramedullary Spinal Tumours

The use of intraoperative neurophysiological monitoring (IONM) has been an integral part of spinal surgery for several years.^1,2 Introduced in the 1970s,³ somatosensory-evoked potentials (SSEPs) were the first modality used in spinal surgery.⁴ Later, methods of monitoring motor-evoked potentials (MEPs) and D-waves were discovered, providing a more reliable way to predict postoperative paralysis.^5,6 Newer advances, such as dorsal column mapping^6,7 and the double train stimulation technique,⁸ enable greater preservation of neurological function.⁴

Although the clearest application of these techniques is for intramedullary spinal cord tumours, the accuracy and clinical value of IONM in intradural extramedullary (IDEM) spinal tumours remains controversial. This is primarily due to many articles grouping all spinal tumour types in their studies, leaving few that focus specifically on IDEM tumours. Additionally, study sizes are often small, resulting in high potential variability.

Development of postoperative motor deficits is a primary risk in IDEM tumour surgery. Whilst in most cases, postoperative deficits are temporary and patients generally recover motor function,^9,10 studies have found that permanent neurological deficits arise in approximately 6-7%.^11,12 The use of IONM seeks to alert the surgeon of impending neurological injury before permanent deficits are inflicted.

In recent years, the frequency of published articles analysing the use of IONM in IDEM tumour surgery has doubled.^1,2,10,13-22 However, the use of IONM for this spinal tumour type remains controversial amongst the neurosurgical community, with a lack of focused studies. At our centre, use of IONM in patients with IDEM tumours has been variable over the past 10 years. The aim of this study is to compare post-operative outcomes in this cohort, distinguishing between those in whom IONM was used and not used.

Adult (≥18 years) patients who underwent surgery for histologically- and radiologically-defined IDEM tumours at our tertiary neurosurgery unit between 2010 and 2020 were retrospectively identified. Exclusion criteria were age <18 years and incomplete or missing surgical or follow-up data. All data being retrospectively collected and anonymised, neither informed consent nor institutional board review approval were required for this study.

Patients were divided into two groups: IONM and non-IONM. Two factors affected availability of IONM: (1) early on in the 10-year study period, IONM was not ‘standard of care’ at our centre, meaning its use was more sporadic, and (2), unavailability of IONM ‘out-of-hours’ precluding its use in many emergency cases. The primary outcome measured was postoperative American Spinal Injury Association (ASIA) score. Secondary outcomes investigated were use of intraoperative ultrasound (iUS), presence of cord oedema, maximal tumour spinal canal occupancy in the axial plane, placement of surgical drains, incidence of postoperative complications, 3- and 6-month Eastern Cooperative Oncology Group (ECOG) scores, extent of resection, length of stay, discharge location and tumour recurrence. For the IONM group, information on monitoring modality and changes in the amplitude and latency of MEPs, SSEPs and D-waves (when performed) were assessed.

Location of tumours in this study were defined in relation to the spinal cord in the axial plane as anterior, posterior, anterolateral, posterolateral or circumferential. The presence of spinal cord oedema was assessed using preoperative sagittal T2-weighted MRI scans. The canal occupancy of the tumour was measured from preoperative axial gadolinium-enhanced T1-weighted images.

The neurophysiologists at our hospital used either a comprehensive Inomed Medizintechnik GmbH ISIS system for multimodal IONM or a Medtronic NIM-ECLIPSE E4 NS 32-channel amplifier (Medtronic Xomed, Jacksonville, FS) sampled at 10Khz. Both systems comprise state-of-the-art recording and stimulating hardware, with amplification and analysis of the raw data possible throughout the procedure via a display screen. When IONM was used, a decrease in either MEPs (abrupt disappearance at threshold, ≥80% drop at supramaximal stimulation or 100V increase in stimulation intensity)²³ or SSEPs (≥50% amplitude drop and/or 10% latency increase)²⁴ was considered significant by the surgical team: the resection was paused, the blood pressure was increased by a minimum of 20mmHg of systolic blood pressure and the spinal cord was bathed in warm saline. A period of 5-10 minutes was given for IONM signals to recover before the resection was resumed. If no improvement was detected at that point, surgery was either resumed or abandoned depending upon the informed consent process that took place with the patient.

IONM was measured using several methods. For cervical lesions, MEPs were measured using the deltoid, biceps, extensor and abductor pollicis brevis or abductor digiti minimi, the quadriceps, tibialis anterior, gastrocnemius and abductor hallucis. For thoracic and lumbar cases, the same muscles were used apart from deltoid and biceps. The bulbocavernosus reflex was measured by stimulating the dorsum of the penis or clitoris and recording from anal sphincter. SSEPs were measured using the median or ulnar nerve and bilateral posterior tibial nerves. For free-run electromyography (EMG), performed in the muscle groups mentioned above, ‘A-train’ patterns triggered a warning. Modalities used to map the cauda equina were the same as the thoracic and lumbar regions, with additional monitoring of the anal sphincter and urinary sphincter (the latter with an electrified urinary catheter). Mapping and free-run EMG was performed with a concentric bipolar probe in the aforementioned muscle groups.

To assess the impact of IONM independently for schwannoma, meningioma, and ependymoma patients, we classified each patient’s procedure into (1) ‘no signal changes’, where no warnings were generated during the procedure, (2) ‘reversible signal changes’, were a neurophysiological warning appeared as described above, reversing after surgical action was taken (eg, pause in resection, blood pressure increase and/or warm irrigation), or (3) ‘irreversible signal changes’, where the above actions failed to restore baseline signals.

This was then developed into a contingency table. Given that intraoperative neuromonitoring is a continuous monitoring technique in theatres, rather than a diagnostic test, contingency tables were 3 × 2 rather than 2 × 2. Due to some fields containing empty values, Chi-square analysis could not be applied. Instead, likelihood ratio was calculated by dividing the number of patients with signal change (reversible or irreversible) that developed a worsening of ASIA score postoperatively by the number of patients with no signal change that had a worse postoperative ASIA score.

Statistical analysis was performed using STATA 13.1 (StataCorp, College Station, TX). T-test, multilinear regression, logistic regression, ordered logistic regression and multinomial logistical regression tests were used to study the primary and secondary outcomes. A P value of <.05 was considered significant. To perform the mathematical analysis of our data, we also used Bayesian network analysis. Bayesian networks are directed acyclic graphs. A graph has nodes, and edges that connect the nodes. A Bayesian network does not permit cycles between its nodes. There are three different types of nodes in Bayesian networks: chance nodes, decision nodes and utility nodes. Bayesian networks contain a set of conditional probability distributions, of the relationship of its nodes and their potential causal dependence, represented by the network’s edges. The edges define every possible outcome of the preceding causal nodes in the form of conditional probability distributions. Given some observed evidence, the prior probability of that evidence would be defined as its probability of occurrence before new data is collected. Bayesian networks aim to use prior probabilities to estimate the conditional probability distributions of each of its nodes in the presence of new, unobserved data.

We built a Bayesian network based on chance and decision nodes that were correlated to each other with edges according to expert domain knowledge on causal inference (see Figure 1). Two counterfactual scenarios were further assessed to describe postoperative ASIA and length of stay without the causal inference of IONM.

Much of the controversy surrounding IONM usage in surgery for IDEM spinal tumours revolves around its effectiveness for this tumour type, in comparison to its known efficacy for intramedullary tumours.² IONM usage comes with drawbacks including increased duration of surgery, previously suggested to be a risk factor for surgical site infections.²⁵ There remains debate about IONM for IDEM tumours, especially if canal occupancy is low, as they are considered extra-axial lesions. We believe our study is a good comparison of divergent surgical strategies in a single neurosurgical unit regarding IONM use. Our use of Bayesian network analysis further aids in inter-group comparison, since Bayesian networks, unlike frequentist statistics, enable assessment of causal inferences of individual factors in the network through counterfactual questions, where the links between these factors and other nodes is broken.

We have been able to predict that the use of IONM is beneficial for patients with IDEM spinal tumours. Prior studies have shown positive outcomes for IDEM tumours regardless of IONM use, achieving gross total resection (GTR) rates of >90%, postoperative complication rates of <20%, and >80% of patients having either stable or improved neurological function at follow-up.^26-29 Even when IONM alarms have sounded intraoperatively, suggesting immediate deficit and the need to use standard measures to improve microvascular perfusion as described above, studies have found improved patient outcomes with IONM usage.³⁰ However, a few studies report permanent new neurological deficits as a result of surgery, regardless of IONM use.^26-29,31,32 The good overall neurological outcomes in surgery for IDEM tumours might have been partially responsible for a high prevalence of anterolateral lesions in the non-IONM group, since patients were not randomised, the use of IONM instead being determined by surgical team preference. Furthermore, in a significant number of patients (related with preoperative ASIA score), MEPs, SSEPs or both could not be recorded. Nevertheless, IONM generally provides good sensitivity and specificity in IDEM tumours: amongst papers that combine modalities, sensitivity is mostly >80% and specificity >90%.^{15,18,21,22,33} Few patients who exhibit no change in potentials intraoperatively later develop neurological deficits, suggesting IONM can effectively predict positive outcomes. On the other hand, approximately 10% of patients (19% in our study) do exhibit changes intraoperatively.^{10,15,18,19,21,22,33} In our study, there was a higher probability of remaining without new neurological deficit in the IONM group compared to the non-IONM group, whilst the probability of neurological deterioration was lower in the IONM group.

Tumour recurrence rates were low, with only four cases in the IONM group and six cases in the non-IONM group. IONM has facilitated high rates of GTR in prior studies, with low rates of tumour recurrence in cases where GTR was not achieved.^28,31,32 In our cohort, GTR was comparable amongst both groups: 83% in IONM patients and 85% in non-IONM patients.

This study additionally showed the use of IONM can more effectively produce good neurological outcomes compared to no IONM usage in the presence of preoperative spinal cord oedema. This is important to highlight, as an oedematous cord is more likely to result in obscuration of margins, intraoperative injury and poor postoperative neurological outcomes.

We acknowledge that it is possible the number needed to treat for ASIA A, B and C patients is higher than the current number of cases we had in our study, and that we therefore cannot yet demonstrate the potential benefit of IONM in these patients. Ideally, centres should report IONM outcomes for IDEM tumours as a separate entity and provide outcomes of patients with and without IONM usage to allow for large scale meta-analyses so that higher evidence of efficacy can be achieved. We attempted to do this by assessing canal occupancy in our cases and still found a positive difference in the patients with IONM vs those without. The reason for using canal occupancy is because it has previously been demonstrated to predict prognostic outcomes in IDEM tumours, with higher occupancy resulting in more preoperative disability as well as generally worse postoperative outcomes.³⁴ We found similar such findings in this study, where the probability of postoperative ASIA E was lower in patients with higher degree of canal occupancy.

In patients with greater baseline neurological deficit, there may be difficulty in obtaining IONM readings from the outset. Whilst we had a low percentage of preoperative ASIA A, B and C patients, precluding statistical analysis of this subset, a difference in recordable signals was identified between preoperative ASIA E (75% recordable MEPs, 68% recordable SSEPs) compared to ASIA D (78% MEPs, 62% SSEPs). During this study, there were 15 patients who had originally been planned for IONM usage but no successful readings could be obtained. These patients remained in the IONM group as this was considered to be significant in terms of spinal cord impairment by the tumour and therefore could be seen as a sign of severity of baseline spinal cord injury. Whilst data on these patients could not be analysed due to the low numbers, their information has been collated (Supplementary Material 1). This group of patients requires further study to better understand the longer-term neurological outcomes and the implications of having unobtainable IONM signals.

MEPs and SSEPs have individual physiological pathways and therefore should be used together rather than in isolation for maximal benefit. For instance, two studies have reported high false positive/false negative ratios when SSEPs were used as a standalone technique.^35,36 MEPs, meanwhile, are in practice seldom monitored continuously—despite this being the gold-standard—because of patient movement during stimulation, which is highly disruptive and impractical during microsurgical dissection.⁴ Furthermore, both techniques are extremely sensitive to inhaled anaesthetics, which suppress the evoked potentials.^37,38

In addition to patient outcomes in patients who do exhibit changes, it is important to consider financial aspects. The cost of IONM is $1535 on average per operation,³⁹ but the length of stay of patients with IONM is significantly lower than that of patients without IONM (8.7 days average length of stay in the IONM group vs 14.2 days in the non-IONM group) which significantly increases healthcare costs above the initial cost of monitoring itself. The reason for reduced length of stay in the IONM can be seen in our Bayesian network analysis. When comparing preoperative ASIA D patients, there was a higher probability that patients who underwent surgery with IONM would remain stable or improve to ASIA E whereas non-IONM patients had a higher probability of worsening postoperatively. This was even more apparent when comparing preoperative ASIA E patients, where probability to remain ASIA E was 20% higher in the IONM group.

The cost-effectiveness of IONM use in spinal surgery has been previously analysed.⁴⁰ Two papers found that the annual cost of treating a single patient rendered paraplegic following spinal tumour resection was more than that of using IONM in 100 and 210 patients respectively.^2,41 As such, these papers found IONM cost-effective. However, in two other studies, IONM was found not to be cost effective.^42,43 In particular, Traynelis and colleagues found that avoiding IONM saved $1 million without compromising patient safety.⁴³ Notably, none of these papers exclusively focused on IDEM tumours, diminishing the strength of derived conclusions.

Finally, it is important to discuss the psychological and medicolegal aspects of using IONM, impacting both the surgeon and patient. We have found no studies examining the psychological impact of IONM usage in spinal surgery, though we speculate that it may provide reassurance to both the surgeon, especially if less experienced, and the patient undergoing surgery. For instance, a pilot study comparing perioperative stress of patients undergoing thyroidectomies with and without IONM usage found significantly lower rates of anxiety among the IONM group.⁴⁴ The argument that it is medicolegally safer for surgeons to use all available intraoperative resources—including IONM—to reduce the risk of postoperative deficit is counterbalanced by instances where IONM data has been used as an opportunity for patients to seek financial compensation.⁴⁵ Therefore, the informed consent and overall preoperative counselling of the patient are crucial in this process, with clarification of the patient’s views regarding potential compromises between postoperative neurological deficits and oncological resection.

The use of IONM appears to have significant benefits in surgery for IDEM spinal tumours, especially if the preoperative ASIA score is D or E. We have shown that our IONM cohort were less likely to develop new neurological deficits and had shorter postoperative hospital stays.