METHODS

During a period of nine years (January 1992 to December 2001), 21 990 neurosurgical operations were done in our hospital. IOM was applied only in cases that were considered to carry a significant risk of neurological damage. Thus, based on the surgeon’s decision, the type of lesion, and the availability of the monitoring team, some kind of neurophysiological monitoring was undertaken in 884 cases (4% of all operations). From this database 658 consecutive cases in which SEP monitoring was done were selected for further analysis. Cases confined to facial nerve monitoring, monitoring of motor evoked potentials, and monitoring involving spinal cord stimulation were excluded from the analysis in order to avoid confusion, as in modes other than SEP monitoring the criteria for monitoring alarms are different and somewhat ill defined. Some data on a subgroup of this population have been analysed and published elsewhere, with a focus on surgical interventions together with a detailed description of our techniques.⁷

The age of the patients ranged from 0.6 to 83.2 years (mean 45.8 years), and the male to female ratio was 1.3:1. The duration of monitoring ranged from 0.8 to 12.9 hours, with a mean of 3.3 hours. The patients’ diagnoses were as follows: 282 intracranial tumours (156 infratentorial, 126 supratentorial), 158 extramedullary and 51 intramedullary spinal lesion, 139 intracranial vascular lesion, and 28 other non-tumour lesions. The target of monitoring—that is, the structure that was mostly exposed to damage during surgery—was a cerebral hemisphere in 251 cases, the brain stem in 198 cases, and the spinal cord in 209 cases. Median nerve somatosensory evoked potentials (M-SEP) were used for monitoring in 463 cases, and tibial nerve potentials (T-SEP) in 166. In 29 cases both modes were used.

Recording technique

In the earlier cases, balanced anaesthesia with nitrous oxide and isoflurane was employed, while in the later cases total intravenous anaesthesia was used exclusively. Our recent anaesthetic protocol includes induction with propofol, remifentanil, and rocuronium and maintenance of anaesthesia by continuous infusion of remifentanil and propofol. M-SEP and T-SEP were elicited by stimulation of the nerve at the wrist or at the lower ankle, respectively, using square wave electrical pulses (100 ms duration, 20 mA intensity, 5.1 Hz stimulation rate). A stimulation rate of 3.1 Hz was used in some cases, if the amplitudes of the baseline recordings were low. Stimulation and recordings were done using subcutaneous needle electrodes. For recording, the active electrodes were placed at C3′ and C4′ (M-SEP) or at Cz′ (T-SEP) with a reference electrode at Fz according to the international 10-20 system. Proximal recordings were made using either electrode at the second cervical vertebra, at Erb’s point, or the first lumbar vertebra. An earth (ground) electrode was placed at the forehead. Resistance of all electrodes was kept below 5 kΩ. The time base was set to 50 or 100 ms and a filter bandpass of 30 to 3000 Hz was used. Depending on the quality of the recording, 250 to 500 trials were averaged.

The recordings were visually analysed for the presence of the main peaks N20–P25 and P40–N50, respectively, and peak to peak amplitudes as well as peak latencies were measured. After induction and with anaesthesia in steady state, intraoperative baseline values were established and the recording parameters were maintained throughout the surgical procedure. In patients with intraoperative events during IOM or unexpected neurological deficits, additional recordings were made postoperatively within the first days after surgery.

Collection of data

Criteria of an event—that is, a pathological finding in neurophysiological monitoring—were a 50% reduction in amplitude of the cortical complex and/or a 10% increase in peak latency compared with baseline if sustained for two consecutive trials. In all cases of monitoring events, several conditions were checked such as alterations to the anaesthesia regimen, physiological abnormalities, and surgical problems noted by the surgeon.

All patients had a neurological examination postoperatively, done by a member of the staff who was not involved in the monitoring procedure. Patients were classified as suffering from a neurological deficit if a new significant neurological dysfunction occurred corresponding to the monitoring target. For example, if surgery of an intracranial aneurysm was monitored using M-SEP or T-SEP, a new hemiparesis was considered to be a postoperative deficit. However, a patient with a new dysfunction of the oculomotor nerve was not admitted to this group, because monitoring was not targeted at this nerve. For spinal cord monitoring any motor or sensory function of the spinal cord was considered to be the target of monitoring. Thus all patients with new symptoms of spinal cord dysfunction postoperatively were assigned to the group with postoperative deficits.

Classification of IOM

For the purpose of this study patients were assigned to one of four groups corresponding to the information obtained from IOM, the intraoperative findings, and the postoperative neurological examination. These groups were:

• True positives—Patients were assigned to this group if a monitoring event occurred and a new neurological deficit corresponding to the target of monitoring was present. However, patients without a neurological deficit were also included if the surgeon identified and reacted to an intraoperative problem corresponding to the monitoring event in order to prevent neurological damage.

• False positives—Patients were assigned to this group if a monitoring event occurred, but corresponding intraoperative findings or problems were missing and the patient showed no postoperative deficit corresponding to the target of monitoring.

• False negatives—Patients were assigned to this group if monitoring was completely uneventful, but the patient suffered from a new postoperative deficit corresponding to the target of monitoring.

• True negatives—Patients were assigned to this group if monitoring was uneventful and no new postoperative deficits were found.

Database

The complete dataset of IOM including all recordings and amplitude and latency values were stored in a database. The dataset additionally comprises remarks on the stages of the surgical procedure and, if present, on special intraoperative events. The patient’s preoperative recordings of evoked potentials and, if done, the postoperative recordings were also included. Furthermore the datasets were supplemented with remarks on the patient’s postoperative course and the protocol of the surgical procedure. The database is designed to allow a meaningful retrospective analysis of each monitoring case.

RESULTS

In 527 of the total 658 cases (80%) monitoring was uneventful, and in 131 cases (20%) a monitoring event was noted. In 106 of the cases with a monitoring event, the alterations to the evoked potentials persisted throughout the surgical procedure, whereas they returned to normal during surgery in 25 cases. All the latter, except for four false positive cases, were assigned to the group of true positives. A new postoperative deficit was observed in 112 (17%) of the 658 cases. According to our criteria the patients were assigned to groups as follows: 105 (16%) true positives, 26 (3.9%) false positives, 27 (4.1%) false negatives, and 500 (76%) true negatives. A detailed overview is given in table 1.

Analysis of false negative findings

A summary of the findings in the 27 false negative IOM cases is given in table 2. A careful case by case analysis of these 27 patients identified three different groups with IOM failure. In the first group of five patients a thorough review of clinical, radiological, and neurophysiological findings presented convincing evidence that the neurological deficit has not occurred intraoperatively, but in the early postoperative period after termination of monitoring. There were four patients who underwent clipping of an intracranial aneurysm and one with a large meningioma of the skull base infiltrating the cavernous sinus and the wall of the intracavernous internal carotid artery. Early postoperative neurological examination on the intensive care unit was unequivocally normal in three patients. In two patients a mild deficit was suspected at the time of the first examination postoperatively, but both showed a clear step by step deterioration in the early postoperative period. In all five cases SEP monitoring was completely uneventful until the end of surgery, but all patients showed marked alterations of SEP when recordings were done postoperatively—on the same day in two cases or on the day after surgery in three. In all five cases computed tomography showed an ischaemic lesion that was related to the vascular territory of a vessel exposed during surgery. One may speculate that vasospasm or an embolic event as a consequence of surgical manipulation of the vessel wall was the underlying cause.

The second group comprised 14 patients with uneventful SEP monitoring who suffered from only mild postoperative neurological deficits, which were completely reversible in nine cases. In five patients an incomplete hemiparesis occurred after surgery for a supratentorial tumour in the vicinity of the motor cortex. Two patients suffered from mild deficits following vascular surgery. In two patients with surgery for an intradural spinal meningioma a mild paraparesis occurred postoperatively, and in another three patients with surgery for an intramedullary lesion (one astrocytoma, two angioblastomas) a pre-existing myelopathy was worsened. Furthermore, in one patient a minor lesion of a cervical nerve root occurred during resection of an intradural schwannoma, and in another patient mild signs of brain stem dysfunction were observed after removal of a meningioma of the cerebellopontine angle. An example for this type of IOM failure is given in fig 1.

In the third group of eight patients, severe neurological deficits were observed postoperatively although SEP monitoring was completely uneventful. This group comprised three aneurysm cases, two cases of infratentorial tumours with compression of the brain stem, and one patient each with a brain stem cavernoma, a supratentorial glioblastoma involving the motor cortex, and a supratentorial meningioma. An example from this group is given in fig 2.

For calculation of sensitivity and negative predictive value those five patients with convincing evidence of a delayed neurological deficit were excluded from the group of false negatives and the eight patients with severe neurological deficits were considered separately. A summary of the data is given in table 3.

DISCUSSION

The primary goal of IOM is to identify patients with impending neurological damage early during surgery, to enable the surgical team to react in time. Such an intervention by the surgical team following a monitoring alarm is potentially effective in preventing neurological deterioration in the patient. In our experience a cautious estimation is that successful interventions occur in about 5% of the cases monitored. A more detailed discussion of this topic is given elsewhere.⁷ The rate of false positive findings appears to be relatively high in this series, particularly for cases of brain stem monitoring (table 2). This is a matter of concern because false positive alarms may disturb the progress of surgery. A detailed analysis suggests that false positive alarms are partly attributable to the patient’s semi-sitting position used for different types of surgery at our institution and that there are ways to overcome this problem.⁸

Concerning false negative findings, published reports generally have to be read carefully because criteria for false negative findings are not always clearly defined. For example it is sometimes unclear whether SEP monitoring of the spinal cord is or is not required to detect a pure motor deficit. One could argue that a pure motor deficit with normal SEP monitoring is not a failure of monitoring if dorsal column function remains normal. Another example is cases where SEPs deteriorate intraoperatively and the patient wakes without a new neurological deficit, because the surgeon was able to react to the event successfully. These cases are sometimes considered false positives, but they are in fact correct predictions of an impending neurological deficit. In our opinion it seems reasonable to classify these types of events as true positives. An example is given in fig 3. If these cases were counted as correct predictions, the sensitivity of IOM would appear more favourable.^7,9 In most reported series the sensitivity of neurophysiological IOM is around 80%. However, if distinct subgroups are considered the sensitivity shows much more variation, ranging from 20% to 100%.^10–14 An overview is given in table 4.

Generally, several different reasons for failure of IOM can be identified. For example, an event leading to neurological damage may occur after termination of monitoring. In this series such a delayed deficit was responsible for monitoring failure in five of our 27 cases (19%). A stringent definition of false negative findings should not include this type of failure.²⁴ Furthermore, a failure to monitor the pathway at risk seems to be a not uncommon mistake.²⁵ If, for example, M-SEPs are used to monitor surgery of an anterior communicating artery aneurysm, then ischaemia in the territory of the anterior cerebral artery is beyond the scope of monitoring. T-SEPs have to be used in addition in these cases. Another example is the use of bilateral stimulation of the tibial nerve for spinal cord monitoring, which may easily overlook a unilateral lower extremity weakness.²⁶ Furthermore, brain stem auditory evoked potentials (BAEP) are sometimes used to monitor brain stem function, but may prove insensitive to lesions of the upper brain stem.

A different reason for monitoring failure is difficulty in adequately interpreting SEP changes. Temporary changes in evoked potentials after vascular clamping, for example, are sometimes judged to be false negatives if the potentials return to baseline following reperfusion during surgery but the patient wakes with new neurological damage. For a correct interpretation of this finding one has to take into account the fact that white matter is more resistant to ischaemia than grey matter. Thus following temporary hypoperfusion white matter pathways may recover completely with restitution of evoked potential conduction, but grey matter areas may suffer damage, leading to neurological deficits. Consequently, one must bear in mind that restitution of evoked potentials does not guarantee normal postoperative function.^{15,23,25,27,28}

Finally, there are failures of neurophysiological monitoring that are obviously unavoidable. Basic physiological knowledge suggests that with SEP monitoring only parts of the lemniscal system are covered. Thus it is not unexpected that SEPs are insensitive to isolated damage to structures outside the lemniscal system such as the motor pathways. This is a fundamental concern about SEP monitoring, because in the clinical setting it is usually desirable to monitor function in larger anatomical areas such as the whole spinal cord, the brain stem, or the motor cortex. Nevertheless, for practical clinical use SEP monitoring is acceptable because monitoring of the confined lemniscal system functions as an indicator of damage to the larger adjacent area at risk during surgery. The practical value of this basic principle is supported by our findings of an overall sensitivity of 79%, and a 91% sensitivity for severe neurological deficits.

Analysis of our data suggests that circumscribed lesions outside the monitored pathways are the main reason for monitoring failure. In most of our false negative cases neurological morbidity was mild, probably because the majority of lesions were small, as documented by postoperative radiological findings. In this series this was true for 14 patients with mild neurological deficits, which were completely reversible in nine cases. On the other hand, if the lesion is located in an eloquent area, severe permanent neurological deficits may occur. This latter type of severe monitoring failure occurred in eight patients in this series.

In practice it important to identify the clinical settings that are at increased risk of monitoring failure. Our data suggest that such settings are indeed identifiable. These were: surgery of intracranial aneurysms with a risk of damage to small vessels causing ischaemia in the area of the basal ganglia; some cases of brain stem damage in infratentorial tumours, possibly caused by damage to small perforating vessels; and cases with confined damage to the motor cortex in supratentorial tumours. These findings are supported by observations of other investigators reporting results of surgery for aneurysms^6,13,29 or skull base tumours.¹⁵

Finally, if the monitoring team is aware of an increased risk of monitoring failure the question about the consequences arises. In these cases measures to enhance the sensitivity of IOM should be considered. A first approach may be to monitor both M-SEP and T-SEP, on the basis that an enlarged area monitored within the lemniscal system will provide better sensitivity. However, we are at present unable to provide evidence that this approach is effective, and we are currently investigating it at our institution. In cases where isolated damage to the motor cortex is of concern, monitoring of motor evoked potentials in addition to SEP seems a good suggestion.³⁰ If damage to the brain stem is the main risk, then a simultaneous recording of SEPs and BAEPs should be considered. The latter suggestion is based on the findings of Manninen et al,³¹ who showed that in surgery for posterior fossa aneurysms combined monitoring increases the sensitivity from 47% and 37%, respectively, to 84% for both SEP and BAEP. But the surgeon needs to take into account that there are certain types of damage occurring during surgery that may not be detected by IOM.

Conclusions

The use of SEP monitoring should be accompanied by an awareness of its possible limitations. Certain clinical settings can be identified where there is an increased risk of monitoring failure. In these cases additional measures to enhance the sensitivity of monitoring should be considered.