**3. Results**

#### **3.1 Clinical CN palsies and MRI-detected CN Involvement**

Of 924 patients with NPC, 82 (8.9%) patients initially presented with CN palsy. A total of 134 CNs were found to be paralyzed, and 31 patients were detected with multiple CN palsies. In these 82 patients with CN palsy, 79 patients had one or more MRI findings demonstrating CN involvement, and CN V was the most common nerve to be involved. Of all patients, 46 (5.0%) had clinical evidence of trigeminal nerve palsy at presentation, 92 (28%) divisions of palsied trigeminal nerves were clinically identified. Of these 46 patients, 43 (93.5%) had unilateral paresthesia and 3 (6.5%) had bilateral paresthesia. In 5 of these patients with palsied nerves, MRI findings did not correspond to clinical findings. For example, a patient was detected with palsied CN V3, but MRI demonstrated only evidence of CN involvement of CN V2.

A total of 333 (36%) patients demonstrated MRI-detected CN involvement and of the 514 patients with local advanced (T3-4) disease, 332 patients demonstrated MRI-detected CN involvement. The incidence of MRI-detected CN involvement in T3 patients was lower than that in T4 patients (43.8% vs 97.5%, *P* = 0.000). Only 1 patient with T2 disease demonstrated MRI-detected CN involvement. Clinical signs and symptoms of ipsilaterally affected CNs were absent in 259 (77.8%) of patients with MRI-detected CN involvement. The correlation between MRI-detected CN involvement and ipsilateral CN palsy is shown in Table 1.

#### **3.2 Imaging basis of CN palsies**

Of the 134 paralyzed CNs, 98 (73.1%) demonstrated ipsilateral intracranial or orbital MRIdetected CN involvement; 21 (15.7%) demonstrated basicranial MRI-detected CN involvement. No MRI evidence of CN involvement was observed in the remaining 15

Patients were followed up at least every 3 months during the first 2 years; thereafter, patients were followed up every 5 months until death. The median follow-up period for the whole group was 40 months (range, 2-56 months). All events were measured from the date of commencement of the treatment. The following endpoints (time to the first defining event) were assessed: overall survival (OS); local relapse-free survival (LRFS); and distant metastasis-free survival (DMFS). Local recurrence was established by fiberoptic endoscopy and biopsy and/or MRI. Distant metastases were diagnosed based on clinical symptoms, physical examination, and imaging methods including chest radiograph, bone scan, CT, and

Statistical analyses were performed using the Statistical Package for the Social Sciences (SPSS Inc., Chicago, IL) 12.0 software. The actuarial rates were calculated by the Kaplan-Meier method [8], and the differences were compared by using the log-rank test. Multivariate analyses with the Cox proportional hazards model were used to test independent significance by backward elimination of insignificant explanatory variables [9]. Host factors (age and sex) were included as the covariates in all tests. The chi-square test was used to analyze the relationship between MRI findings and CN palsies. The criterion for

Of 924 patients with NPC, 82 (8.9%) patients initially presented with CN palsy. A total of 134 CNs were found to be paralyzed, and 31 patients were detected with multiple CN palsies. In these 82 patients with CN palsy, 79 patients had one or more MRI findings demonstrating CN involvement, and CN V was the most common nerve to be involved. Of all patients, 46 (5.0%) had clinical evidence of trigeminal nerve palsy at presentation, 92 (28%) divisions of palsied trigeminal nerves were clinically identified. Of these 46 patients, 43 (93.5%) had unilateral paresthesia and 3 (6.5%) had bilateral paresthesia. In 5 of these patients with palsied nerves, MRI findings did not correspond to clinical findings. For example, a patient was detected with palsied CN V3, but MRI demonstrated only evidence

A total of 333 (36%) patients demonstrated MRI-detected CN involvement and of the 514 patients with local advanced (T3-4) disease, 332 patients demonstrated MRI-detected CN involvement. The incidence of MRI-detected CN involvement in T3 patients was lower than that in T4 patients (43.8% vs 97.5%, *P* = 0.000). Only 1 patient with T2 disease demonstrated MRI-detected CN involvement. Clinical signs and symptoms of ipsilaterally affected CNs were absent in 259 (77.8%) of patients with MRI-detected CN involvement. The correlation

between MRI-detected CN involvement and ipsilateral CN palsy is shown in Table 1.

Of the 134 paralyzed CNs, 98 (73.1%) demonstrated ipsilateral intracranial or orbital MRIdetected CN involvement; 21 (15.7%) demonstrated basicranial MRI-detected CN involvement. No MRI evidence of CN involvement was observed in the remaining 15

statistical significance was set at a = 0.05. The *P* values were based on 2-sided tests.

**3.1 Clinical CN palsies and MRI-detected CN Involvement** 

**2.3 Statistical analysis** 

abdominal sonography.

**3. Results** 

of CN involvement of CN V2.

**3.2 Imaging basis of CN palsies** 

(11.2%) paralyzed CNs; among these 15 CNs, carotid sheath (n=13) or retropharyngeal lymph node (n = 1) invasion could be identified along the course of the CNs paralyzed on the ipsilateral side; no tumor could be detected along the course of the remaining paralyzed CNs on the MRI images. Of the 92 divisions of palsied trigeminal nerves, 91 trigeminal nerve palsies (98.9%) showed MRI evidence of 1 or more trigeminal nerve involvements. However, no evidence could be found for the remaining 1 palsied V1 case on MR images.


\*: Sites number means the number of MRI-detected CN involvement

Table 1. The Correlation between MRI-detected CN Involvement and the Ipsilateral Related CN Palsy.

## **3.3 Prognosis of MRI-detected CN Involvement in T3-4 disease**

In 514 T3-4 patients, significant differences were observed in 3-year OS (75.7% vs 89.2%, *P* = 0.001) and DMFS (77.1% vs 87.8%, *P* = 0.002) rates, with better rates observed in the 182 patients who did not demonstrate MRI-detected CN involvement. No significant difference was observed in 3-year LRFS (86.8% vs 91.8%, *P* = 0.067) rates between the patients with and without MRI-detected CN involvement in T3-4 patients.

Because most (97.5%) T4 patients had MRI-detected CN involvement, the prognoses between the T3 patients with and without MRI-detected CN involvement were also compared. All patients with local advanced disease (T3 and T4) in this series were classified into 3 groups: Group 1, T3 disease without MRI-detected CN involvement; Group 2, T3 disease with MRI-detected CN involvement; and Group 3, T4 disease. Significant differences were observed in OS and DMFS rates between Groups 1 and 2 (*P* = 0.009 and *P* = 0.011,

MRI-Detected Cranial Nerve Involvement in Nasopharyngeal Carcinoma 129

Clinical findings were not consistent with the MRI findings of CN involvement in some cases. Nerves are resistant to tumor, and perineural tumor spread is an insidious and often asymptomatic process. The incidence of MRI-detected CN involvement is high in NPC; patients are often asymptomatic [10]. According to our observation, the incidence of MRIdetected CN involvement was much higher than that of clinical presentation of CN palsy in

With improvement in imaging techniques, direct visualization of smaller nerves and proximal nerve branches have become possible. MRI allows contrast between the nerve fibers and the surrounding cerebrospinal fluid, perineural vascular plexus, and fat pads. Involvement of cranial nerves may be direct or perineural; proposed MRI criteria of cranial nerve involvement in our study include both possibilities. MRI, including the contrastenhanced fat suppression technique, is particularly helpful in detection of direct and perineural invasion of the nerve in patients with NPC, so we often identified cranial nerve involvement on gadoliniumenhanced T1-weighted images. Perineural spread, in which pathologic conditions spread along the connective tissues of the perineurium, is often associated with malignant disease of the head and neck. The imaging hallmark of this

Without contrast between the nerve and the surrounding structures, CNs in the parapharyngeal space could not be identified on MRI; therefore, parapharyngeal space invasion was not defined as evidence of CN involvement in our study. A small number of symptomatic patients with only carotid sheath or retropharyngeal lymph node invasion along the course of ipsilaterally paralyzed CNs were regarded as MRI negative but clinically

Because NPC is a nonsurgical disease, the limitation of our study is that there was no pathologic correlation of CN involvement with MRI findings. Another limitation is the slice thickness of 5 mm used in our study, which may limit the sensitivity of detecting cranial nerve involvement. Ideally, slice thickness should be 3 to 4 mm in future studies; this may

We observed that most CN palsies had 1 or more MRI findings of intracranial or orbital CN involvement; this may be a reason for the adverse prognostic significance of CN palsies, and

Several investigators have reported that the perineural tumor spread in patients with nonnasopharyngeal carcinoma of the head and neck is associated with an increased incidence of recurrence [11-14]. However, in our series, MRI-detected CN involvement was not associated with the 3-year LRFS rate in NPC patients with local advanced disease. The primary treatment modality for NPC was radiation therapy, whereas that of other carcinomas of the head and neck was surgery. Lawrence and Cottel reported that postoperative radiotherapy of squamous cell carcinoma with perineural invasion resulted in a much improved survival probability when compared with that observed after

**4.1 Difference between clinical CN palsy and MRI-detected CN involvement** 

process has been discussed in detail in a previous articles [1-3].

**4.2 Prognosis of clinical CN palsy in MRI-detected CN involvement** 

patients with CN palsies are determined to be at stage T4 [4].

lead to earlier identification of CN involvement.

**4. Discussion** 

NPC patients.

positive in our series.

respectively), but no significant differences were observed between Groups 2 and 3 (*P* = 0.322 and *P* = 0.809, respectively). No significant differences were observed in the LRFS rates between Groups 1 and 2 (*P* = 0.750), and no significant difference was observed between Groups 2 and 3 (*P* = 0.079).

Multivariate analysis was performed to adjust for various prognostic factors in T3 and T4 disease. Parameters were included in the Cox proportional hazards model by backward elimination of insignificant explanatory variables: age (50<=years vs >50 years), sex, skull base extension, paranasal sinus extension, intracranial extension, infratemporal extension, orbital extension hypopharynx extension, MRI-detected CN involvement, CN palsy, N classification, radiotherapy technique, and chemotherapy. Both clinical CN palsy and MRIdetected CN involvement were significant predictive factors for the DMFS and OS rates in local advanced disease (Table 2).


OS, overall survival; LRFS, local relapse-free survival; DMFS, metastasis-free survival.

Table 2. Summary of Multivariate Analyses of Prognostic Factors in T3-4 Disease.

#### **3.4 Difference in the prognostic implications between clinical CN palsy and MRI-detected CN involvement**

In 336 patients with clinical and/or MRI-detected CN involvement, no significant difference was observed in the 3-year DMFS (74.6% vs 84.6%, *P* = 0.094) or LRFS (86.7% vs 87.9%, *P* = 0.899) rates between patients with and without clinically detected CN palsies. A marginally significant difference was observed in the 3-year OS (74.2% vs 80.1%, *P* = 0.067) rate; however, after adjusting for N classification, there was no significant difference (*P* = 0.102).

#### **3.5 Prognosis of lesion localization in patients with MRI-detected CN involvement**

Of the 333 patients with MRI-detected CN involvement, no significant differences were observed with regard to the 3-year OS (78.3% vs 72.9%, *P* = 0.120), LRFS (89.7% vs 84.1%, *P* = 0.154) or DMFS (79.6% vs 74.8%, *P* = 0.466) rates between patients with and without intracranial or orbital MRI-detected CN involvement.
