Beyond dichotomy: patterns and amplitudes of SSEPs and neurological outcomes after cardiac arrest

Reliable early prognostication of neurological recovery in comatose cardiac arrest survivors receiving therapeutic hypothermia (TH) remains a major clinical challenge [1]. Sedation with or without paralysis during TH may delay neurological recovery [2]. Therefore, these confounders might limit the value of a neurological examination and necessitate other multimodal prognostic approaches.

Of these modalities, somatosensory evoked potential (SSEP) measurement is a noninvasive bedside technique that can even be used for unstable critically ill patients and is less confounded by sedation or hypothermia than electroencephalography (EEG). The bilateral absence of N20 on days 1–3 or later after the return of spontaneous circulation (ROSC) is currently the most reliable predictor of a poor outcome, with a false-positive rate (FPR) close to 0% [3‐7]. However, the current interpretation of SSEPs has several limitations. As the N20 interpretation is dichotomous (absent/present), discrimination between a low-amplitude and absent N20 is crucial. High noise levels and artifacts may affect the ability to discern a low-amplitude N20 and impede reliable interpretation [8]. These problems could also produce interobserver disagreement and false-positive cases. Furthermore, the absence of N20 has a low sensitivity for predicting a poor outcome [9‐12]. A few studies have suggested 0.1–0.3 μV as a reliable noise level for cortical SSEPs [8, 13‐15]. Several features have recently been studied for increasing the sensitivity of SSEPs [16]. Endisch et al. performed a large prospective study on the relationship between cortical SSEP amplitudes and short-term neurological outcomes and found that absent or very low-amplitude SSEPs appear to be highly predictive of a poor outcome, with high sensitivity [17]. Kim et al. reported that the absence of P25/30 has better sensitivity than the absence of N20 for predicting a poor neurological outcome at hospital discharge [18].

However, neither of these recent analyses has been validated, especially in patients who were followed for a long period. We hypothesized that the absence of P25 and the amplitude of the N20–P25 complex have higher sensitivity than the absence of N20 for predicting a poor neurological outcome after 6 months. The aims of this study were to evaluate and compare the predictive values of SSEPs by using pattern and amplitude analyses of cortical short-latency SSEPs.

In this retrospective, single-center registry-based study, when we focused on the P25 peak or N20–P25 component, the sensitivity for predicting a poor outcome and the specificity for predicting a good outcome simultaneously increased more than twofold compared to those of the traditionally used pattern (absence of N20). On the other hand, the amplitude analysis was suitable for detailed prediction. An N20–P25 amplitude < 0.64 μV was a good predictor of a poor outcome, with prognostic value similar to that of the pattern analysis, while an amplitude > 2.31 μV was more capable of predicting a good outcome, with high specificity.

SSEP interpretations are traditionally dichotomous, with the absence or presence of N20 predicting a poor outcome or not. In our results, the absence of N20 or P25 guaranteed a poor long-term neurological outcome. However, the sensitivity for predicting a poor outcome was different between the two waves. This result is in line with the finding of a study by Kim et al. [18]. Although there was an insufficient number of type III SSEPs, all patients who experienced a good outcome showed the presence of both N20 and P25. The difference in prognostic value between the absence of N20 and P25 originated from the higher incidence of N20 than P25 in the poor outcome group (69.5% and 29.8%, respectively). Bauer et al. have reported that the extent of hypoxic-ischemic brain damage in cardiac arrest survivors increases along the afferent sensory pathway, with pronounced vulnerability of the thalamic and cortical brain regions [3]. Cortical N20 and P25 peaks were preserved in all patients with a CPC score of 1–2, whereas a stepwise decrease in the N20 and P25 cortical peaks was detectable in 63% and 59% of patients with a poor outcome, respectively. In a study by Kim et al., a significant decrease in detectable N20 and P25 peaks was observed in 30.9% and 11.1% of patients with a poor outcome, respectively [18].

However, the reason why the absence of P25 is more common than the absence of N20 in patients with a poor outcome is unknown. Both the N20 and P25 peaks are now widely accepted to be generated in the posterior bank of the central sulcus, corresponding to Brodmann area 3b [23‐30]. However, it is possible that a minor contribution to the P25 component arises from the anterior bank of the central sulcus in Brodmann area 4 [29‐31]. The issue of an additional contribution to changes in the P25 component by an anterior source in area 4 or by an additional radial source residing in area 1 may be clarified further only by applying multidipole localization algorithms to multichannel recordings [29‐31].

The findings of the amplitude analysis are consistent with those of previous reports in the literature [17, 32, 33]. Various pathophysiological processes may decrease amplitudes in the acute phase of brain ischemia [34, 35]. However, amplitude assessment has largely been ignored in most clinical studies because the absolute value of the N20 amplitude varies by recording location and due to individual variability [36]. Some investigators have indicated that in the acute phase of stroke, the absolute N20 amplitude has significant predictive value regarding long-term functional recovery [37]. For comatose cardiac arrest patients, the N20 and N35 amplitudes were associated with functional outcomes at the time of discharge [32]. Among 49 anoxic coma patients, none with a baseline-to-peak N20 amplitude < 0.6 μV or a peak-to-peak short-latency wave amplitude < 1.2 μV recovered consciousness [33].

Our results for the amplitude threshold of short-latency waves were similar to those reported by Endisch et al. [17]. However, several differences in methodology deserve further mention. In our study, WLST, which could lead to an elevated low-amplitude threshold in patients with a good outcome, was not considered. Furthermore, the predictability of the SSEP amplitude might be overestimated in analyses considering short-term outcomes. We used the CPC score at 6 months after ROSC as the outcome and classified CPC 3 as the threshold for a poor outcome. Four patients with a CPC 4 at hospital discharge improved their neurological outcomes to CPC 3 after 6 months, and their lowest amplitude was 0.42 μV; two of them had type I SSEPs, and the rest had type II SSEPs. This result reveals that the lower amplitude threshold for a good outcome could be elevated after long-term follow-up. Although one patient in our cohort showed an improvement in neurological status from a CPC score of 3 to a CPC score of 2 after 6 months, the amplitude in his case was 2.34 μV, which was higher than the lowest amplitude in a survivor with a good outcome (0.62 μV) reported by Endisch et al. [17]. Interestingly, despite these differences, our absolute threshold value for the N20–P25 amplitude for predicting a poor outcome was comparable to the findings reported by Endisch et al. [17].

When we assumed that amplitudes above 2.31 μV confirmed the absence of severe hypoxic injury, the specificity of this threshold for a good outcome was 96.5%. Through a case review of 5 patients with a poor outcome who showed an amplitude above 2.31 μV, we found that one patient temporarily recovered consciousness after rewarming, but this patient died of multiple organ failure. Two patients showed status epilepticus (SE) on continuous electroencephalography immediately before the SSEPs were recorded. Their DWI scans indicated diffuse hypoxic injury at 24 and 8 h after the SSEPs were recorded, and these patients ultimately died. The remaining 2 patients with a CPC score of 3 showed diffuse injury on DWI, and one of them showed SE before the SSEPs were recorded. In conclusion, the majority (3/4) of patients with an N20–P25 amplitude above the upper threshold but with severe hypoxic injury showed SE. Although many previous studies have described giant evoked potentials coincident with epileptiform discharges [38‐41], the SSEP amplitudes in those cases were considerably larger than the amplitudes in our cases. Therefore, we are not sure whether epileptiform discharges serve as a confounding factor for the N20–P25 amplitude or produce additional neuronal injury.

The amplitude and morphology of N20 and P25 depend on the electrode location and recording time [36, 42]; however, we analyzed only the amplitudes of short-latency waves to determine whether N20 or P25 was present. Serial and multichannel cortical recordings would facilitate the determination of the exact amplitude threshold through further study. In addition, further analysis to determine the morphology of the meaningful SSEP wave might be needed. Therefore, we believe that our lower amplitude threshold should not be recommended as a guideline for WLST.

Amorim et al. recently proposed that WLST and a self-fulfilling prophecy considerably impact the validity of using SSEPs for the prediction of a poor outcome [43]. As cases of absent SSEPs in patients with a good neurological outcome have been reported, WLST is a potential confounder in many studies. They reported that the FPR for an absent N20 in predicting a poor neurological outcome, adjusted for a WLST rate of 80%, is substantially higher than generally believed. However, it remains unclear whether our results present some evidence against their hypothesis. First, although WLST is prohibited by law in South Korea, the treating physicians were not blinded to the prognosis, and DNR orders were accepted. Finally, among in-hospital deaths, the DNR permission rate was 37.2% (29 patients), and the mode of 15 deaths was the physician’s decision not to escalate therapy according to the family’s demand. Second, SSEP interpretation may be affected by various factors [13], and the interrater reliability of SSEP testing is limited, especially for patterns predicting a poor outcome [8, 13, 43]. In the present study, 24 of 216 recordings (11.1%) were judged to be unsuitable due to a noise level too high to allow reliable interpretation or absence of the N13 peak and were excluded from the analysis. Thus, using SSEPs in isolation to predict a poor outcome is unwarranted, and SSEPs should only be used as one test of many in a multimodal approach to predict the outcome and decide on WLST, as recommended by current guidelines [1]. However, amplitudes above the higher thresholds can be used to suggest further or advanced treatment.

Our study should be interpreted considering the following limitations. First, this study was performed in a single hospital and was a retrospective, registry-based study, which may decrease the generalizability of the results, including the amplitude thresholds. The test machine parameters and recording protocols, such as the filter bandwidth and the stimulus intensity, might variably alter the amplitude of cortical SSEPs [44]. Second, as mentioned above, the SSEP results were not blinded, which could potentially have influenced decisions regarding withholding advanced treatment. Third, because the SSEPs were recorded for clinical purposes, they were not obtained from patients who were awakened immediately after TH. This profound selection bias may limit any interpretation of these data for predicting a good outcome. The recording times were also inconsistent. Although we included these cases based on claims in the recent literature, mild TH does not have a significant impact on the amplitude or presence of SSEPs [45, 46], and a quarter of the SSEPs were recorded before rewarming. Therefore, a prospectively designed multicenter study is needed to increase the generalizability of our results.

In this retrospective, single-center registry-based study, the absence of P25, especially the absence of the N20–P25 component, improves the sensitivity for predicting a poor outcome without jeopardizing the specificity. On the other hand, amplitude analysis was suitable for detailed prediction. The N20–P25 amplitude threshold for predicting a poor outcome (< 0.64 μV) had a prognostic value similar to that observed for pattern analysis, and an amplitude > 2.31 μV provided a sensitivity and specificity for predicting a good outcome of 52.9% and 96.5%, respectively. However, because of some limitations, results should be viewed with caution regarding the possibility of false positives and interpreted in a multimodal approach. A further larger multicenter prospective study is needed to increase the generalizability of our results.