CSF neurofilament levels: A potential prognostic marker in Guillain–Barré syndrome

Guillain–Barré syndrome (GBS) was originally described as a demyelinating polyradiculoneuropathy with a good prognosis. Axonal degeneration may adversely affect recovery.

Axonal degeneration is usually diagnosed electrophysiologically. However, results may be mild or inconclusive early in the disease course,^1,2 indicating the need for additional early markers of axonal degeneration.

Disintegration of the axonal membrane results in the release of axoskeletal proteins such as neurofilaments (NfHs) into the CSF (reviewed by Petzold³). Our interest in CSF NfH levels in patients with GBS originated in the ad hoc observation of a bimodal distribution in a small cohort of patients with retrospective evidence of axonal damage in those with the higher levels.⁴ We investigated the prognostic value of CSF NfH^SMI35 in GBS.

Methods.

This prospective study was approved by the Joint Research and Ethics Committee of the Institute of Neurology and the National Hospital for Neurology and Neurosurgery (NHNN). Informed consent was obtained.

From 118 patients with the referral diagnosis of GBS admitted to the NHNN between January 2000 and December 2003, 23 fulfilled the inclusion criteria (diagnosis of GBS consistent with published criteria,² availability of a CSF sample, and no comorbidity that could influence outcome).

Motor function and disability grade were quantified using the Medical Research Council grading system (MRCS) and F score.⁵

Nerve conduction studies (NCSs) and EMG were performed using standard protocols. Patients were classified according to criteria based either on NCSs alone⁶ or supplemented with EMG.⁷ The latter criteria were found to be more accurate and used for further analyses. Dependent on the disease course, between one and three subsequent assessments were performed.

CSF NfH^SMI35 levels were measured in duplicates using a standard ELISA with the analyst being blinded to all other data.⁴ The upper limit of normal is 0.73 ng/mL and was determined in a reference population of 416 patients with a median age of 41.9 (interquartile range [IQR] 31.2 to 55.8) years.⁴

All statistical analyses were based on predefined variables for predicting outcome.

Results.

The baseline characteristics of the 23 patients (seven women, 16 men; median age 47 years) are summarized in table E-1A and B (on the Neurology Web site at www.neurology.org). There was no significant difference to the previously published reference population.⁴ The first neurologic symptom was leg weakness in 14 patients (61%), more diffusely distributed weakness in two (9%), and hand/foot numbness in six (26%). One patient presented with diplopia and developed Miller–Fisher GBS overlap syndrome.

Treatment with IVIg (0.4 g/kg over 5 days) was initiated in 15 (71%) patients. One patient with severe disease also received subsequent plasmapheresis. The mortality was 4% (1/23). This patient died of a myocardial infarction 44 days after onset of GBS while being ventilated in the intensive care unit. The median follow-up time was 175 (IQR 90 to 253) days.

The results based on the Albers' classification⁷ are summarized in table 1. At the last assessment, performed 48 days (median) after onset, there was evidence of axonal degeneration in 17 of 23 (74%) of the patients.

The median time from onset to lumbar puncture was 8 days. Patients with axonal degeneration⁷ had 12.5-fold higher median CSF NfH^SMI35 levels (1.00 ng/mL) compared to those with demyelinating GBS (0.08 ng/mL, p = 0.0135, Wilcoxon two-sample test, table 2). None of the other CSF or serum results differed between the two groups (table E-2). There was no correlation of CSF NfH^SMI35 with time from onset or age. The one patient who died had a CSF NfH^SMI35 concentration at admission, which was 13 times the upper limit of normal (18.6 ng/mL).

Patients with evidence of axonal involvement were slightly older than those with demyelination (not significant, supplementary data on the Neurology Web site at www.neurology.org). However, there was an age difference for patients with high NfH^SMI35 levels (75 years) compared to those with normal NfH^SMI35 levels (41 years, p = 0.0202, Wilcoxon two-sample test, see supplementary data). Although this significance was lost after Bonferroni correction, the finding was taken into account for additional age-adjusted analyses. The age-adjusted analyses did not alter the findings of the study (see supplementary data).

Outcome.

Patients with high (>0.73 ng/mL⁴) CSF NfH^SMI35 levels were more disabled on both outcome measures at the last follow-up visit (figure, A and B, filled columns). Patients with high CSF NfH^SMI35 levels did not achieve independent mobility (F score ≥2, odds ratio 14.40, 95% CI: 1.38 to 150.81) and had impaired motor function (MRCS ≤45, odds ratio 15.00, 95% CI: 1.33 to 169.87). The CSF NfH^SMI35 levels correlated with the final F score (R = 0.47, p = 0.024) and MRCS (R = −0.57, p < 0.01, data not shown).

• Download figure
• Open in new tab
• Download powerpoint

Figure. Primary outcome measures. The F score and Medical Research Council grading system (MRCS) in patients with Guillain–Barré syndrome (GBS) taken at the last follow-up visit, classified into subgroups according to neurophysiologic criteria (open boxes, left side of each graph) and CSF NfH^SMI35 levels (filled columns). For the neurophysiologic assessment, the Albers' criteria⁷ were used to define axonal and demyelinating GBS. CSF NfH^SMI35 levels were classified as high if >0.73 ng/mL and as normal otherwise.⁴ For both scales (F score, MRCS), high CSF NfH^SMI35 levels on admission predicted a worse outcome (p < 0.01 for each comparison, Wilcoxon two-sample test). The mean ± the SD and numbers are shown. NfH = neurofilament; N.S. = not significant.

In contrast there was no significant difference on either outcome scale if patients were classified by neurophysiologic criteria⁷ (figure, A and B, open columns). This finding was reinforced by the post hoc observation that six (35%) of those patients with neurophysiologic evidence of axonal degeneration achieved a surprisingly good outcome (F score = 0 to 2). None of the other predefined variables was of significant prognostic value.

Discussion.

Patients with neurophysiologic evidence of axonal involvement had higher CSF NfH levels compared to those with demyelinating GBS. Pathologically high CSF NfH^SMI35 levels (>0.73 ng/mL⁴) on admission predicted a worse functional and motor outcome.

The findings may be explained by the probable anatomic source for NfH^SMI35. Proximal axonal degeneration at the level of the nerve roots rapidly releases NfH^SMI35 into the CSF. Proximal axonotmesis requires axonal regrowth over a long distance, with the risk of losing chemical and anatomic guidance cues. Importantly, distal axonal damage would not be detectable in an early CSF sample. However, distal axonal damage may not be of prognostic relevance because axonal sprouting provides the potential for recovery.

The important difference between distal and proximal axonal damage may also explain the neurophysiologic results. Both may result in loss of compound motor action potential (CMAP) and EMG signs of denervation. Those 35% of patients with neurophysiologic evidence of axonal damage who made a good recovery may have had distal axonal damage, thus explaining why the Albers' criteria⁷ failed to predict the long-term outcome. Our results are consistent with those of studies that found that reduction in CMAP and EMG signs for denervation were not good indicators of prognosis.^8,9

The neurophysiologic assessment is also limited by the observation that reliable signs for axonal degeneration typically occur late in the disease process with a peak incidence 6 to 10 weeks after disease onset.⁷ Moreover, isolated nerve conduction studies without supplementary EMG information resulted in reclassification of 90% of cases originally thought to be “axonal,” with 25% of all patients remaining unclassifiable.⁶ This is possibly because conduction block, physiologic impairment of axonal conduction, or very distal axonal degeneration/demyelination can cause an axonal pattern on nerve conduction studies.

Comparing the present to other studies, the F score at nadir was similar,^8,10 but the proportion of cases with axonal degeneration (17 of 23, 74%) was high, reflecting a selection bias caused by triaging referrals to a specialized tertiary referral center. Given the magnitude of the findings, a larger proportion of cases with mild demyelinating GBS would probably have strengthened rather than weakened the results, but this would need to be confirmed independently. CSF NfH^SMI35 levels may be useful in deciding who would benefit most from early transferral to a specialized center and potentially who would benefit from targeted neuroprotection, once such treatment options become available.

Acknowledgment

The authors thank the patients for their participations and the physicians who allowed them to study their patients. They also thank Dr. V. Tan and Dr. A. Gaitatzis for helpful comments.