CSF and MRI findings in patients with acute herpes zoster

Herpes zoster (HZ) is a common disease caused by reactivation of latent varicella zoster virus (VZV) that has remained dormant in sensory ganglia since primary VZV infection (varicella). Rash and pain are due to inflammation of the nerve and skin as the virus travels from the ganglion along the sensory nerve into the skin. In rare cases, HZ is associated with CNS disease, such as myelitis and meningoencephalitis, presumably due to the virus traveling centripetally along the posterior root to infect the spinal cord, meninges, and brain.

In the immunocompetent host, CNS complications are exceedingly rare (0.2% in a population-based study).¹ However, indirect evidence suggests that viral invasion of the CNS during HZ may be more frequent. First, pathologic studies in HZ patients demonstrate inflammatory changes in the spinal cord and brainstem.^2-5 Second, postherpetic neuralgia (PHN) is associated with segmental atrophy of the dorsal horn of the spinal cord.^6,7 Third, motor involvement detected by electromyography occurs in one-half of the patients and sometimes involves segments outside the area of cutaneous involvement, even contralaterally, suggesting a spread to the anterior horns of the spinal cord.⁸

The CSF of patients with HZ myelitis or meningoencephalitis is usually abnormal, with pleocytosis, increased protein concentration, abnormal CSF/serum albumin ratio indicative of blood-brain barrier (BBB) damage and intrathecal anti-VZV antibodies, and VZV DNA.^9-15 MRI in these cases shows high signal abnormalities on T2-weighted images in the cord, pons, and cerebral white matter.^12,16-19 In uncomplicated HZ, pleocytosis is reported in one-third^20,21 and elevated protein concentration in one-fifth²⁰ of the patients. In no study have intrathecal anti-VZV antibodies, VZV DNA, or MRI been systematically determined in HZ patients without CNS complications.

The current study was undertaken to determine CSF and MRI findings in typical cases of HZ, and to correlate the findings with the clinical manifestations of the disease, including motor and ophthalmic complications, and PHN.

Patients and methods. Patients and clinical follow-up. Before commencement of the study, primary care doctors practicing within the catchment area of Tampere University Hospital were fully informed of the aim and nature of the study through a series of personal contacts and lectures. They were asked to refer all future patients with acute HZ, regardless of symptom severity. This study was approved by the local ethics committee. All participating patients gave informed consent.

Patients were examined neurologically at each visit to assess complications or pain, which was graded on a 4-point verbal scale: no pain, mild pain, moderate pain, and severe pain. The initial visit was arranged as soon as possible after onset of rash, and the follow-up visits took place at 2 weeks, 6 weeks, and 3 months. When needed, longer follow-up was provided until clinical recovery. During the recruitment period, 56 patients were seen (figure 1), 50 of whom underwent CSF sampling (46 patients), MRI (16 patients), or both (12 patients).

Cerebrospinal fluid tests. CSF and paired serum samples were obtained from 46 patients: single CSF sample from 24, two samples from 11, and three samples from 11 patients. The first sampling was taken at 1 to 18 days, the second at 7 to 47 days, and the third at 31 to 171 days after onset of rash. No CSF was taken from 10 patients because of warfarin treatment in 2, patient refusal in 5, and delay of the first visit later than 2 weeks after rash in 3.

CSF cell count (normal, <5 leukocytes/µL) and total protein level (normal, <500 to 790 mg/L, depending on age) were determined. The specimens were frozen at -70 °C until further analyzed. Albumin and immunoglobulin (Ig)G in unconcentrated CSF and sera were measured by immunoprecipitation nephelometry. The CSF IgG index was calculated (normal, <0.70). The CSF/serum albumin ratio was used to evaluate BBB damage (normal, 2.7 to 7.3 × 10^-3, depending on age).²² Oligoclonal bands (OBs) were determined by isoelectric focusing.²³ IgG and IgM anti-VZV antibodies in CSF and serum were determined by commercial ELISA "Enzygnost Anti-VZV" (Behringwerke, Marburg, Germany). The VZV antibody to albumin index (CSF/serum VZV IgG)/(CSF/serum albumin) was calculated to determine if anti-VZV antibody in CSF originated from intrathecal synthesis.²²

VZV DNA in CSF was detected by PCR as previously described.^11,24 The DNA was extracted from frozen CSF aliquots with InstaGene purification matrix (BioRad Laboratories, Inc, Hercules, CA). The PCR was carried out in 50-µL mixtures containing 20 µL of extracted DNA, 1 unit of Pful (Stratagene, LaJolla, CA), 100 mM of each deoxynucleoside triphosphate, and 1 µM of each primer Pful buffer (Stratagene). Duplicate samples were amplified for 45 cycles. The amplified DNA was separated using agarose gel electrophoresis and transferred to nylon membranes. The identity of the DNA bands was confirmed by hybridization with a digoxigenin-labeled probe (Boehringer Mannheim, Indianapolis, IN), followed by an antidigoxigenin antibody conjugated to alkaline phosphatase (Boehringer Mannheim) and CSPD chemiluminescent substrate (Tropix, Bedford, MA).

Each assay included two negative water controls and two positive sensitivity controls. Tests were considered valid if the controls yielded the expected results and the patient replicates agreed. This VZV PCR method has a sensitivity of 0.5 viral copies per reaction. The assay was negative for 38 control CSF samples from age- and gender-matched patients who had established neurologic diseases not caused by VZV.

Magnetic resonance imaging. MRI was performed on 16 consecutive patients with cranial (12 patients) or cervical (4 patients) HZ 1 to 5 weeks after the eruption, and MRI was repeated in four cases 2 to 3 months later. Because flow artifacts in the thoracic and lumbar cord were expected to interfere excessively with any HZ-related changes, it was decided by consensus that only patients with cervical and cranial HZ would be scanned, thus taking advantage of the larger size of the cervical medulla and trigeminal brainstem complex.

MRI was performed on a 0.5-Tesla scanner (Philips Gyroscan ACS NT5, Best, The Netherlands). In addition to T1-, T2-, and proton density (PD)-weighted axial images, three-dimensional volume T1 fast-field echo (FFE)-weighted axial and coronal images, thin-slice 3-mm spin-echo (SE) axial images, and gadolinum DTPA-enhanced three-dimensional volume T1 FFE axial and coronal images were taken. Patients with cervical HZ underwent volume and SE imaging of the cervical spine in addition to cranial imaging. The parameters of T1 FFE sequence were repeat time (TR) = 42 milliseconds, echo time (TE) = 14 milliseconds, flip angle = 30, number of signal averages (NSA) = 2, matrix = 212/256, field of view (FOV) = 220, slice thickness = 1.5/0.8 mm. In T2-weighted SE images, the imaging parameters were TR = 2198 milliseconds, TE = 110 milliseconds, NSA = 10, matrix = 512, FOV = 190, slice thickness = 3/0.3 mm.

To quantitate the signal intensity of lesions in the brainstem, intensity profile maps were drawn. MRI with similar protocols was performed on 17 age- and gender-matched control patients. Clinical indications for brain imaging of the control patients were dizziness (9 patients), facial pain (2 patients), unilateral hypacusis (2 patients), transient ischemic attack (2 patients), headache (1 patient), and seizures (1 patient).

Assessment of scans was performed by a neuroradiologist (P.D.) blinded to the clinical course of the disease. Lesions were ascribed to HZ only if the signal in the spinal cord was detected at the level of rash, or in the case of ophthalmic zoster, in the trigeminal nerve or brainstem near or in the ipsilateral trigeminal brainstem complex.

Statistical analysis. Cross-tabulations were analyzed using Fisher's exact test. Log-linear modeling and logistic regression were employed for multivariate analysis. Results were summarized using odds ratios (OR) with 95% confidence intervals. Computation was performed using BMDP (BMDP, Los Angeles, CA) statistical software (release 7.0) on a VAX/VMS mainframe.

Results. Clinical course of herpes zoster. The study population was composed of 50 immunocompetent patients with acute HZ: 28 women and 22 men (mean age, 59 years; range, 17 to 84 years; 15 patients younger than 50 years). HZ was cranial in 14, cervical in 7, thoracic in 22, and lumbar in 7 patients, as defined by the most affected dermatome. Cranial HZ consisted of 11 patients with ophthalmic, 1 patient with maxillary, and 2 patients with mandibular involvement. Seven patients with cervical HZ had rash in the distribution of C2 or C3, whereas dermatome was involved in C7 in one. In the thoracic (T1 to T12) and lumbar (L1 to L5) regions, the distribution of HZ was relatively even. The rash covered at least two contiguous dermatomes in 17 patients: two dermatomes in 10 patients, three dermatomes in 6 patients, and four dermatomes in 1 patient.

In the first 2 weeks of the disease, the worst pain experienced was reported as no pain by 2 patients (4%), mild by 2 patients (4%), moderate by 21 patients (42%), and severe by 25 patients (50%). Analgesics were used by 28 patients (56%), antidepressants by 23 patients (46%), and steroids by none. Antiviral treatment was commenced within the recommended 72 hours from onset of rash in 29 patients (58%) and after 72 hours in 5 patients (10%). No patient had signs of meningeal irritation, encephalitis, or myelitis. Complications (other than prolonged pain) developed in 10 patients (20%). Six patients (12%) developed peripheral motor weakness. Three patients had ophthalmic complications (keratitis or iritis), and one patient reported double vision of 2 days' duration. All complications were transient. Ten patients (20%) developed PHN, defined as pain lasting more than 3 months after the onset of rash. The intensity of PHN pain was reported as moderate by 1 patient and mild by 9 patients. Five of the patients with PHN had another complication.

Cerebrospinal fluid findings. CSF findings are summarized in table 1 and figure 2. Leukocytosis (range, 5 to 1,440 µL; in five of the cases exceeding 100/µL, lymphocytic in all but one) was found in the first sample in 17 patients, and tended to decrease in subsequent samples. Leukocytosis did not correlate with delay in obtaining the first sample or with age. Four patients with an initially normal leukocyte count showed slight leukocytosis (range, 5 to 12/µL) in the second sample. Leukocytosis in the first CSF sample, present more frequently in complicated than uncomplicated cases (8/10 versus 9/36, p = 0.003), was seen in all cases (7/7) with HZ-related MRI findings, but only in 1/4 with normal MRI (p = 0.01). However, leukocytosis was not associated with acute pain or PHN.

OBs were found in four patients, one of whom had mild and uncomplicated HZ but OBs in three consecutive samples; an abnormal IgG index was present in his third sample (at 3 months). Other CSF parameters were normal. The three other patients with OBs had clinical complications as well as leukocytosis. Two of these had intrathecal IgG production against VZV, and one had VZV DNA in the CSF. Two of these underwent MRI examination, and both (Patients 2 and 6 in table 2) had HZ-related MRI findings.

VZV DNA was present in the CSF of 10 patients. A positive PCR was not a function of delay in getting the lumbar puncture. In only one patient did the second sample (on day 15) also prove to be PCR positive. She did not receive antiviral treatment, and her disease was uncomplicated and almost painless. The presence of either VZV DNA or anti-VZV IgG in CSF did not correlate with the clinical outcome (i.e., intensity of pain at acute phase, complications, or development of PHN). The only PCR-positive patient who underwent MRI examination had a normal study.

Bivariate analysis indicated an association among leukocytosis, anti-VZV IgG, and VZV DNA in CSF. Multivariate analysis showed that a positive PCR was 10.2 (95% CI 1.18, 57.71) times more likely when leukocytosis was present, and that leukocytosis was 10.5 (95% CI 2.03, 54.3) times more likely when intrathecal VZV antibody was present.

Magnetic resonance image findings. The MRI analyses of brainstem and cranial nerves were normal in six patients (five trigeminal, one cervical HZ). Focal high-signal-intensity areas on T2-weighted scans were found in the brainstem in 10 patients: in 9 patients, one or more were interpreted as HZ-related (table 2). In one other patient with cervical HZ, a solitary small lesion in the brainstem was observed and deemed to be unrelated to HZ (Patient 8). Two patients with cervical HZ had high signal intensity on T2-weighted scans of the spinal cord (Patients 9 and 10). Both had HZ-related brainstem lesions. Enhancement of the trigeminal nerve was seen in three cases, all manifesting concomitant brainstem changes. Examples of positive findings are shown in figures 3 to 5.

One patient demonstrated a vascular loop around the ipsilateral trigeminal nerve. Interestingly, during the acute phase of HZ, she experienced shooting pains reminiscent of tic douloureux. None of the brainstem or cord lesions showed enhancement with gadolinium DTPA. The findings normalized on repeat MRI in two patients (Patients 3 and 5 in table 2), and remained unchanged in two others. Patients who had HZ-related MRI abnormalities in the brainstem or cervical cord developed PHN more frequently than did those without findings (5/9 versus 0/7; p = 0.03). These patients also tended to have other complications more frequently (4/9 versus 0/7, p = 0.09). No correlation was observed between the MRI findings and acute pain.

Although supratentorial abnormalities (i.e., small changes in deep gray and white matter) were found in 7/16 patients (44%), these were interpreted as unrelated to HZ. They occurred in patients with and without HZ-related brainstem lesions and had no correlation with age. MRI of brainstem and cranial nerves was normal in the 17 control patients.

Discussion. This prospective study was performed in a population of patients with typical HZ who were referred by primary care physicians fully aware of the investigation aims. The spectrum of clinical manifestations seen in these patients was similar to that reported in a population-based study.¹ Pain and rash ranged from slight to intense, and no cases of severe complications were seen. Although all patients recovered fully, including those with PHN, a significant number of patients manifested CSF and MRI abnormalities.

The inflammatory changes found in CSF suggest direct spread of reactivated VZV from the dorsal root ganglion to the meninges and spinal cord or brainstem. There was strong evidence of VZV in CSF in 14/46 patients (35%), either in the form of a positive PCR or anti-VZV IgG. The antibody-to-albumin index excluded BBB damage as the cause for elevated antibody levels. Persistence of VZV DNA in CSF was unusual, and none of the 12 initially PCR-negative patients became PCR positive in the second sampling 6 to 34 days later. Disappearance of VZV DNA was related to antiviral therapy. There were no PCR-positive cases among patients who received antivirals for at least 2 days before their CSF sampling (0/11), whereas 10/31 patients (32%) with no or less than 2 days treatment were positive (p = 0.04).

MRI changes were best seen on thin-slice 3-mm T2 SE images, the use of which remarkably improved the visualization of the lesions in five patients. No gadolinium DTPA enhancement was seen. This accords with mild inflammation and minimal necrosis, which do not usually lead to permanent changes or scar formation.

Although nonspecific, we think the depicted changes were due to HZ. First, none of the control patients showed similar changes. Second, signal intensity profile maps were drawn to exclude artifacts. Third, the lesions regarded as HZ related were located in the segments corresponding to the rash. Finally, in two cases, MRI changes disappeared on repeat investigation.

A few pathologic studies support the results of the current study, indicating involvement of CNS parenchyma in the acute stages of HZ. Head and Campbell² described degenerative changes in the ipsilateral posterior columns of the spinal cord in 9/21 HZ patients. In two further studies, inflammatory changes were detected in the meninges, spinal cord, and mesencephalic trigeminal sensory nucleus 11 to 117 days after rash.^3,4 Watson et al.⁷ reported one HZ case with marked inflammatory changes in the spinal cord 3 months after the onset of rash. In a retrospective study of patients with HZ, 12 of whom had survived less than 7 weeks after rash, most had inflammatory changes in the brainstem or spinal cord, although VZV was detected in only one patient with meningoencephalitis.⁵

An interesting association was observed between pain and MRI findings. Acute pain was not associated with MRI abnormalities, whereas development of PHN was. The mechanism of pain in the two conditions may not be identical.²⁵ In the latter, CNS changes appear to be crucial in maintaining pain and allodynia.²⁶ The results of the current study suggest that in cases with involvement of the sensory projection areas of the CNS intense enough to be detected by MRI, neuronal damage with cell death and atrophy may occur.^6,7 However, the current study did not include any patients developing chronic PHN, so direct evidence for this is lacking.

Although CSF pleocytosis was associated with the occurrence of HZ complications, and MRI abnormalities with the development of PHN, the predictive value of either abnormality is not high enough to be clinically useful. In general, CSF studies are rarely indicated in acute uncomplicated HZ.