Introduction

Prediction of neurological outcome in comatose survivors of cardiopulmonary resuscitation (CPR) has been the subject of several studies in the last three decades [1–7]. In 2006, a practice parameter for the prediction of the outcome of postanoxic coma was published by the American Academy of Neurology [8]. Bilateral absence of cortical N20 responses of median nerve somatosensory evoked potentials (SEP) 24 h after CPR, as well as absent pupillary light responses, absent corneal reflexes, and absent or extensor motor response to pain after 72 h, were all considered reliable predictors of a poor neurological outcome. In daily practice, a patient with postanoxic coma is examined by a neurologist before additional investigations are requested. Some clinical signs may be associated with the presence or absence of cortical N20 responses but to what extent is unknown. Knowledge of specific clinical signs that predict SEP results may optimize SEP requesting policy.

Methods

Data from the previously published multicenter prospective cohort study PROPAC (prognosis in postanoxic coma, 2000–2003) were used [5]. This study was performed before hypothermia was implemented in daily clinical practice. Adult patients who remained in a coma 24 h after CPR were included. Exclusion criteria were confirmed brain death after 24 h, concomitant traumatic brain injury, life expectancy of no more than several months caused by pre-existent disease, and absence of informed consent from a legal representative.

Neurological examination

Neurological examination, consisting of Glasgow Coma Score (GCS; only motor and eye score) and brain stem reflexes (pupillary light responses and corneal reflexes) was performed in every patient 24, 48, and 72 h after CPR. For current analyses, eye and motor scores were dichotomized; E1 (no eye opening) versus E2–4 (eye opening to pain-spontaneous eye opening) and M1–3 (no motor response-abnormal flexion to pain) versus M4–6 (withdrawal to pain-obeys commands). Pupillary light responses and corneal reflexes were defined as present if at least a unilateral response was present. Our hypothesis was that an eye score of E1, a motor score of M1–3, bilaterally absent corneal reflexes, or bilaterally absent pupillary light responses would all predict an absent SEP; and that an eye score of E2–4, a motor score of M4–6, present corneal reflexes, or present pupillary light responses would all predict a present SEP. Complete neurological examination consisted of eye score, motor score, pupillary light responses and corneal reflexes. Our hypothesis was that the combination of an eye score of E1, a motor score of M1–3, bilaterally absent corneal reflexes, and bilaterally absent pupillary light responses reflexes would accurately predict an absent SEP; and that an eye score of E2–4, in combination with a motor score of M4–6, present corneal reflexes, and present pupillary light responses would accurately predict a present SEP.

Somatosensory evoked potentials

SEPs were recorded with standard procedures and were performed 24, 48, and 72 h after CPR [9]. Local clinical neurophysiologists in each contributing hospital assessed the recordings. The results for the cortical N20 response were documented as absent, present, or technically undeterminable. The median nerve SEP was defined as absent if the cortical N20 response was absent on both sides after left- and right-side median nerve stimulation, in the presence of a cervical potential. For logistical reasons, SEP was not always possible on weekends; if the 72-h SEP was due on a weekend day, the recording was postponed to Monday. SEPs with an undeterminable result were excluded from analysis. The results of 24- and 48-h SEPs were not available for treating physicians in order to avoid any influence of the test results on treatment decisions. The results of the 72-h SEP were disclosed to the treating physicians and if the SEP was absent, treatment was usually withdrawn.

Statistical analysis

Patient characteristics were summarized using descriptive statistics. Variables were expressed as mean and standard deviation, or when not normally distributed, as medians and inter-quartile ranges. The positive predictive values (PPV) for absent and present SEPs were calculated to indicate the proportion of the patients who have an SEP result as expected based on the neurological examination. We performed multivariable logistic regression analysis to relate the probability of an absent or present SEP to neurological examination (eye score, motor score, pupillary light responses, and corneal reflexes). Furthermore, the diagnostic accuracy (calculated as an area under the curve of a receiver operating characteristic curve) of neurological examination to predict an absent or present SEP was calculated.

Statistical uncertainty was expressed by the 95% confidence limits when appropriate, with statistical significance defined as p ≤ 0.05. Analyses were performed by SPSS version 18.0 (SPSS Inc. IBM). Differences between areas under the curve (AUC) were analyzed with STATA version 10.

Results

Data of 407 patients were included, 67% were male and mean age was 63 years. The overall mortality was 89.7% (Table 1). A total of 781 SEPs were performed in this group, 401 were present and 299 absent. The remaining 81 were technically undeterminable and were excluded from further analysis. At 24, 48, and 72 h after CPR, respectively, 231, 216 and 253 SEPs were included.

Table 1 Baseline characteristics
Full size table

In patients with a withdrawal response to painful stimuli (M4), we found an absent SEP in 19–31% at the different time points (Table 2). One patient localized painful stimuli at 48 h after CPR, but had an absent SEP. In patients with no motor response or extension to pain at 72 h after CPR, still 46% had a present SEP. Of the 60 patients with spontaneous eye opening, 22 had at least at once at the same moment an absent SEP.

Table 2 Absent and present SEPs related to motor scores at three time intervals after CPR
Full size table

Absent pupillary light responses after 72 h were the best predictor of an absent SEP (PPV 0.91 (0.79–0.96)) (Table 3). Motor score (M4–M6) after 48 and 72 h had the highest PPV for a present SEP (0.83 (0.68–0.91) and 0.83 (0.69–0.91)). All patients with M6 had present SEPs.

Table 3 Positive predictive values for predicting an absent or present SEP in patients after CPR
Full size table

Due to low PPVs, i.e., close to 0.50, the eye score, motor score and pupillary light responses failed to discriminate SEP results in patients 24 h after CPR; only corneal reflexes were discriminative at 24 h for SEP results. Complete neurological examination at 48 (0.78 (0.72–0.85)) and 72 h (0.78 (0.72–0.84)) had the best diagnostic accuracy for an absent, as well as a present SEP, but this values are only considered as “fair” (Table 4).

Table 4 Diagnostic accuracy (as AUC) of neurological examination in predicting SEP results at 24, 48, and 72 h after CPR
Full size table

Discussion

This study has shown that in patients with postanoxic coma, absent or present cortical N20 responses of median nerve SEP cannot be predicted reliably by neurological examination. Absent pupillary light responses after 72 h were the best predictors of an absent SEP and a motor score of withdrawal to pain or better after 48 and 72 h was the best predictor of a present SEP. Complete neurological examination at 48 and 72 h achieved the best diagnostic accuracy for an absent, as well as a present SEP, but this accuracy could only be considered as “fair”. Therefore, when the clinical examination leaves doubt about the prognosis, the SEP has additional value. However, we should also realize that about half of the patients with a present SEP will still have a poor neurological outcome.

There is only one previous study on this subject [10]. In this study, results of the neurological examination and the EEG in 66 patients after cardiac arrest were retrospectively analyzed for their power to predict an absent SEP at day 3. Univariate analysis showed that absent pupillary light responses, absent corneal reflexes, myoclonus, or extensor or absent motor response to pain at day 1 (odds ratio (OR) 5.4–22.5) and day 3 (OR 7.9–22.6); and a malignant EEG at day 3 (OR 6.6) were all significantly associated with an absent SEP after 72 h. After multivariate analysis, absent corneal reflexes, extensor or absent motor response to pain or myoclonus at day 1 (OR 2.7–20.2) and day 3 (OR 4.1–17.3) and absent pupillary light responses or malignant EEG at day 3 (OR 3.1–7.8) remained predictors for an absent SEP. At day 1, the combination of myoclonus, extensor, or absent motor response to pain and absent corneal reflexes had a diagnostic accuracy of 0.89. At day 3, the same three predictors, together with absent pupillary light responses and a malignant EEG, had a diagnostic accuracy of 0.92.

The presence of spontaneous eye opening, but no tracking or blinking to commands, might reflect only subcortical activity (arousal) and does not necessarily imply the impending development of awareness. Therefore, it is assumed to be an unreliable prognosticator [11–13]. This finding was confirmed in our study, as 37% of the patients with spontaneous eye opening had at least once at the same moment an absent SEP.

In our study, one patient who localized painful stimuli (M5) had an absent SEP 48 h after CPR. This was not expected, as localizing painful stimuli implies intact motor and sensory pathways [13]. The cortical N20 response is considered to be an activation of the primary somatosensory cortex following input from the thalamus. Delay or loss of the N20 peak implies an interruption of the connecting pathways between the cervicomedullary junction and the sensory cortex. Therefore, caution in interpreting SEPs should be taken in patients with focal brain lesions [12]. A possible explanation of our observation might be an overrating of the motor score. Previous studies have shown an excellent agreement for the inter-observer reliability of GCS (kappa 0.82–0.85), but only a “good” inter-observer agreement (kappa 0.63–0.77) for motor score alone [14, 15]. The highest degree of inter-observer agreement (0.86) for the motor score was obtained by neurology residents [14]. In our study, the neurological examination was also performed by other physicians than neurologists, which may have had a negative influence on the reliability of the results.

Other important considerations regarding interpretation of the SEP results are the reproducibility of SEP results in anoxic/non-traumatic coma and the reliability of SEP during and after hypothermia treatment. Previous studies mentioned good reproducibility of the SEP results, tested by repeated measurements [5, 16, 17]. Inter-observer disagreement was related to noise level and failure to strictly adhere to the guidelines. Reduction of noise level below 0.25 μV during recordings improved the mean kappa from 0.34 (fair) to 0.74 (substantial) [9, 18]. SEP recorded during hypothermia seems to be a reliable predictor of poor outcome [19, 20]. Also, after treatment with hypothermia, SEP remains a reliable predictor, but some discussion arose after the publication of Leithner et al., who described one patient with an initial absent SEP after hypothermia treatment and good neurological outcome [7, 21–24].

Patient characteristics of the population of our study might have influenced the results. The PROPAC study included patients who were in coma 24 h after CPR without the administration of any sedative drugs. This selection explains the high mortality, and it might also explain the high proportion of patients with absent SEPs, which makes our results more confident.

A limitation of this study might be that the PROPAC study was performed before treatment with therapeutic hypothermia was routinely used. This may limit the usefulness of the results found in current clinical practice, as neurological examination is hampered by sedative medication administered during hypothermia [7, 23, 25]. However, in daily clinical practice, neurological examination will be performed after wearing off of sedative drugs and before requesting a SEP. Furthermore, sedative drugs such as propofol or midazolam only seem to cause marginal effects on latency and amplitude of the cortical N20 responses [26–28].