Presented orally at the 3rd East-West Symposium on Nasopharyngeal Carcinoma, Twin Waters Resort, Sunshine Coast, Queensland, Australia, July 30–August 2, 2007.
Retropharyngeal lymph node (RLN) metastasis was not included in the current American Joint Committee on Cancer (AJCC) staging system (6th edition) for nasopharyngeal carcinoma (NPC).The object of the current study was to investigate the prognostic value and staging categories of RLN metastasis in NPC detected by magnetic resonance imaging (MRI).
All 924 consecutive patients with newly diagnosed NPC were examined with MRI before treatment with definitive intent radiotherapy.
The incidence of RLN metastasis was 73.5%. On multivariate analysis, RLN metastasis was found to be an independent prognostic factor for distant metastasis-free survival (DMFS) in all patients (P = .040). In patients with N0 disease, significant differences were observed between patients with and those without RLN metastasis after adjusting for T classification (P = .046). With regard to laterality, no significant differences were observed in DMFS between patients with unilateral and bilateral RLN metastasis in N0 disease (P = .734). No significant difference in the hazards ratios for either DMFS or disease-free survival (DFS) was found between patients with N0 disease with RLN metastasis and patients with N1 disease (P = .092 and P = .149, respectively). When RLN was classified as N1 disease, there was a better segregation of different N classifications in terms of DFS and DMFS curves, whereas the difference in hazards ratios for N0 and N1 disease was more obvious in DMFS (from 0.461 vs 0.785 to 0.317 vs 0.690).
Although nasopharyngeal carcinoma (NPC) is rare in other parts of world, it is endemic in certain regions, especially in Southeast Asia. The incidence is approximately 30 to 80 cases per 100,000 people per year in Southern China.1 Newly diagnosed, nonmetastatic nasopharyngeal carcinoma (NPC) typically is treated with nonsurgical intervention, and has a higher incidence of cervical lymph node metastasis compared with other head and neck cancers.2 There is a well-developed network of lymph nodes in the nasopharynx. The retropharyngeal lymph node (RLN), regarded as 1 of the key lymph nodes in NPC,3 is not amenable to clinical palpation (Fig. 1) and is particularly difficult to diagnose histopathologically because of its deep anatomic location.4
To our knowledge, the significance of RLN metastasis in both prognosis and clinical staging systems has not been defined yet for NPC. RLN metastasis also was not included in the current American Joint Committee on Cancer (AJCC) staging system (6th edition),5 which results in varied classifications of RLN among different centers.6 One reason RLN metastasis is not included is that the data supporting the current staging system are based mainly on computed tomography (CT) imaging, which detects RLN metastases at a low rate (29.1–51.5%).7, 8 In addition, to our knowledge there is no common set of diagnostic criteria for RLN metastases. Very rare reports have investigated the prognostic value of RLN metastasis, and the results were controversial.7–9
The development of diagnostic and therapeutic techniques revolutionized the management of NPC. The increased sensitivity of magnetic resonance imaging (MRI) allows a more accurate evaluation of the extent of the primary tumor and the lymph node spread pattern compared with CT, thereby shifting the classification and changing treatment strategies for NPC.10, 11 MRI is the optimal method for diagnosing RLN metastases, with an accuracy rate of >38% compared with CT.7, 10 Because MRI is such an important tool for pretreatment staging and determination of the treatment regimen for NPC,10 we investigated how using MRI to diagnose RLN metastasis could improve the prognostic value and staging categories for NPC.
MATERIALS AND METHODS
Between January 2003 and December 2004, 924 consecutive patients with newly diagnosed, untreated, and nonmetastatic NPC were included in the current study. There were 591 male patients and 333 female patients, with a male:female ratio of 1.8:1. The median age of the patients was 44 years (range, 11–78 years). Histologically, 99% of the patients had World Health Organization (WHO) type II or III disease, 0.8% had WHO type I disease, and the remaining patients (0.2%) had adenocarcinoma. All patients underwent pretreatment evaluation, including complete history, physical and neurologic examinations, hematology and biochemistry profiles, MRI scans of the neck and nasopharynx, chest radiography, and abdominal ultrasonography. All patients with N2 through N3 disease underwent emission CT, and 56 patients (6.1%) underwent positron emission tomography (PET)-CT. Medical records and imaging studies were analyzed retrospectively, and all patients were staged according to the 6th edition of the AJCC staging system.5 When RLN metastasis was disregarded, the distribution of T classification, N classification, and overall stage was as follows: T1: 183 patients (19.8%); T2: 227 patients (24.6%); T3: 315 patients (34.1%); T4: 199 patients (21.5%); N0: 323 patients (35.0%); N1: 327 patients (35.4%); N2: 190 patients (20.6%); N3: 84 patients (9.1%); Stage I: 79 patients (8.5%); Stage II: 216 patients (23.4%); Stage III: 363 patients (39.3%); and Stage IV: 266 patients (28.8%).
All patients underwent MRI with a 1.5-Tesla system (Signa CV/i; General Electric Healthcare, Chalfont St. Giles, U.K.). The region from the suprasellar cistern to the inferior margin at the sternal end of the clavicle was examined with a head-and-neck combined coil. T1-weighted fast spin-echo images in the axial, coronal, and sagittal planes (repetition time of 500–600 milliseconds [ms] and an echo time of 10–20 ms) and T2-weighted fast spin-echo images in the axial plane (repetition time of 4000–6000 ms and an echo time of 95–110 ms) were obtained before the contrast material was injected. After the intravenous administration of gadopentetate dimeglumine (Gd-DTPA) (Magnevist; Schering, Berlin, Germany) at a dose of 0.1 mmol/kg body weight, spin-echo T1-weighted axial and sagittal sequences and spin-echo T1-weighted fat-suppressed coronal sequences were performed sequentially, with parameters that were used before Gd-DTPA injection. We used a section thickness of 5 mm, and a matrix size of 512 × 512.
Determination of Lymph Node Metastasis
Two radiologists specializing in head and neck cancers evaluated all scans separately. Any disagreements were resolved by consensus. Diagnostic MRI criteria for a metastatic lymphadenopathy included: 1) lateral RLNs with a minimal axial dimension (MID) of ≥5 mm and any lymph node observed in the median retropharyngeal group or lymph nodes with a MID ≥11 mm in the digastric region, and 10 mm for all other cervical lymph nodes except the retropharyngeal group; 2) lymph nodes of any size with central necrosis or a contrast-enhancing rim; 3) lymph node grouping (the presence of ≥3 contiguous and confluent lymph nodes, each of which should have a MID of 8–10 mm).12, 13
All patients were treated with definitive intent radiotherapy. The majority of the patients (83.7%) were treated with conventional techniques, but 12.7% were treated with intensity-modulated radiotherapy (IMRT) and 3.6% were treated with 3-dimensional conformal radiotherapy (3-DCRT). Details regarding the radiotherapy technique performed at the Cancer Center of Sun Yat-Sen University have been reported previously.8, 14, 15
The majority of the patients (517 of 629 patients [82.2%]) with stage III or IV disease (classified as T3-T4 and/or N2-N3) received neoadjuvant, concomitant, or adjuvant chemotherapy in conjunction with a platinum-based therapeutic clinical trial. When possible, salvage treatments (including afterloading, surgery, and chemotherapy) were provided in the event of documented disease recurrence or when the disease persisted.
The follow-up duration was calculated from the first day of treatment to either the date of death or last examination. Patients were examined at least every 3 months during the first 2 years; thereafter, a follow-up examination was performed every 5 months for 3 years or until death. The median follow-up period for the entire group was 40 months (range, 2–56 months).
SPSS software (version 11.0; (SPSS Inc, Chicago, Ill) was used. All events were measured from the date of the initiation of treatment. The actuarial rates were calculated using the Kaplan-Meier method,16 and the differences were compared with the log-rank test. The following endpoints (time to the first defining event) were assessed: overall survival (OS), disease-free survival (DFS), distant metastasis-free survival (DMFS), and the locoregional recurrence-free survival (LRRFS).
Multivariate analyses with the Cox proportional hazards model were used to test the independent significance17 and the hazard consistency and the hazard discrimination. Host factors (age and sex) were included as covariates in all tests. In addition, the T classification was included as a covariate in analysis of the N classification.
A 2-tailed P value <.05 was considered to be statistically significant.
Incidence of RLN Metastasis
No patient had a metastatic median RLN, and the incidence of lateral RLN metastasis in the current study was 73.5% (679 of 924 patients). Of these 679 patients, 58.8% (399 patients) had unilateral disease, whereas 41.2% (280 patients) had bilateral involvement. The mean values of the MID and the maximal axial dimension (MAD) of the positive RLNs were 10.05 ± 4.57 mm (range, 4–40 mm) and 13.55 ± 5.86 mm (range, 5–40 mm), respectively.
The incidence of RLN metastases in patients with unilateral cervical NPC was lower than in patients with bilateral cervical involvement (78.0% vs 88.6%; chi-square = 10.99 [P = .001]). A higher incidence of RLN involvement was found in the 6th edition AJCC staging system5 categories: T2 to T4 disease compared with T1 disease, N1 to N3 disease compared with N0 disease, and in stage II through IV compared with Stage I disease. These differences were statistically significant (chi-square = 39.19 [P < .001], chi-square = 66.96 [P < .001], and chi-square = 40.10 [P < .001], respectively) (Table 1).
Table 1. Correlation Between Retropharyngeal Lymph Node Metastasis and the T and N Classifications and Overall Stage
P < .001 for RLN vs T1 and T2.; P = .218 for RLN vs T2 and T3.; P = .993 for RLN vs T3 and T4.; P < .001 for RLN vs T1 and T2–4.; P < .001 for RLN vs N0 and N1; P = .002 for RLN vs N1 and N2; P = .126 for RLN vs N2 and N3; P < .001 for RLN vs N0 and N1–3; P < .001 for RLN vs stage I and stage II–IV.
No. of Patients (Column %)
No. of Patients (Column %)
185 (57. 3)
No. of Patients (Column %)
Prognostic Value of RLN
A total of 215 patients (23.2%) had experienced locoregional failure or distant metastases, and 148 patients (16.0%) had died. The 3-year result for the entire series was as follows: the OS rate was 86.1%, the DFS rate was 76.5%, the DMFS rate was 85.0% and the LRRFS rate was 91.6%. A significant difference in patient survival was observed with regard to OS, DMFS, and DFS (P = .006, P < .001, and P = .002, respectively). No significant difference was observed with regard to LRRFS (P = .238).
Multivariate analysis was performed to adjust for various prognostic factors. The following parameters were included in the Cox proportional hazards model: age (≥50 years vs <50 years), sex, T classification, N classification, chemotherapy, radiation technique, and RLN metastasis. RLN metastasis was found to be an independent prognostic factor for DMFS (P = .040) (Table 2), but not for DFS, whereas both N classification and T classification were found to be independent prognostic factors for both DMFS and DFS.
Table 2. Summary of Multivariate Analysis of Prognostic Factors in Nasopharyngeal Carcinoma
95% CI for exp (B)
95% CI indicates 95% confidence interval; RLN, retropharyngeal lymph node metastasis; NS, not significant.
In patients with N0 disease, a statistically significant difference was observed for DMFS (P = .019) and a marginally significant difference was observed for DFS (P = .059), but no statistically significant differences were observed for LRRFS and OS (P = .548 and P = .379, respectively). After adjusting for T classification, RLN metastasis was still found to be a prognostic factor for DMFS (P = .046). With regard to laterality, no statistically significant differences were observed in OS, DFS, DMFS, and LRRFS between patients with unilateral and those with bilateral RLN metastasis (P = .423, P = .759, P = .734, and P = .264, respectively).
Hazard Consistency and Hazard Discrimination
To evaluate the relative severity of disease, the hazards ratios for each N classification subgroup were calculated. The patients were divided into 5 groups: N0 disease without RLN metastasis, N0 disease with RLN metastasis (N0 with RLN), N1 disease, N2 disease, and N3 disease. Host factors (age and sex) and T classification were included as covariates. No significant difference in the hazards ratios of DMFS and DFS were found between patients with N0 disease with RLN metastasis and those with N1 disease (DMFS: 0.628 vs 1 [P = .092] and DFS: 0.739 vs 1 [P = .149]) (Table 3). The survival curves also demonstrated no significant difference for DMFS and DFS between patients with N0 disease with RLN and those with N1 disease (P = .101 and P = .175, respectively). The difference in DMFS and DFS between patients with N0 disease without RLN and those with N0 disease with RLN was also found to be statistically significant (P = .019 and P = .048, respectively) (Fig. 2).
Table 3. Hazard Consistency of Different N Classification Subsets by the Risk of Distant Failure and Disease Failure
RLN Metastasis Should Be Included in N Classification for NPC Staging
When RLN metastasis was classified as N1, the survival curves for DMFS and DFS were found to well separated between N0 and N1 disease (Figs. 3 and 4), and the difference between the hazards ratios for N0 and N1 disease was more obvious for DMFS (from 0.461 vs 0.785 to 0.317 vs 0.690) (Table 4).
Table 4. Hazard Discrimination of Different N Classification Subsets by the Risk of Distant Failure and Disease Failure
If RLN were included as a criterion for N1 staging, 185 patients with N0 disease (20.0%) would be upgraded to N1 disease, but only 34 patients with stage I disease (3.7%) would be upgraded to stage II. The distribution of patients according to this N classification method was as follows: 138 patients with N0 disease (14.9%), 512 patients with N1 disease (55.4%), 190 patients with N2 disease (20.6%), and 84 patients with N3 disease (9.1%). The distribution of patients in each stage by this method was as follows: 45 patients with stage I disease (4.9%), 250 patients with stage II disease (27.1%), 363 patients with stage III disease (39.3%), and 266 patients with stage IV disease (28.8%).
Review of the historic record demonstrated that the 5-year relative survival rates for NPC have increased from approximately 50% to 75% over the past 10 years.18, 19 Advances in technology and the increasing availability of facilities (both diagnostic and therapeutic) have made obvious contributions to this achievement.20 At the current time, MRI is preferred over CT in the evaluation of NPC. In addition, IMRT was found to reduce the damage to surrounding tissue by improving the targeting of tumor tissue, and the increasing use of chemotherapy improved OS in the treatment of NPC.21 The introduction of new therapeutic interventions or new technologies may require modifications of the classification and staging systems.5
Incidence of RLN Metastasis
A MAD of ≥10 mm was the criteria for radiographic involvement of RLN by CT in early studies.9, 22 However, the measurement of MID may be more accurate for diagnosing tumor-positive cervical lymph nodes.13 To our knowledge, 2 MRI studies to date have reported that normal RLNs measured <4 mm or 4.5 mm in MID. Hence, an MID of ≥5 mm was considered an abnormal, or metastatic, lymph node.4, 11 Using both the criteria described above and lymph node necrosis, the results of the current study demonstrated that the incidence of RLN metastasis was 73.5%, which is similar to the results of other studies which used MRI,4, 11 but much higher than the incidence found using CT data (range, 29.1–51.5%).6, 7 More metastatic lymph nodes are found using MRI because the superior soft-tissue contrast allows for the identification of smaller lymph nodes, as well as the discrimination of individual lymph nodes from direct tumor extension and adjacent normal structure.23 In addition, metastatic RLNs have the same density as nonmetastatic RLNs without an enhanced ring, and are contiguous with the primary tumor, which may not be identified as a separate mass on a CT scan but was identified by MRI.
It should be emphasized that radiographic criteria were based on a normal group with noninfective and nonmalignant conditions of the head or neck. Using this size criterion, it was difficult to distinguish benign reactive adenopathy with an MID ≤5 mm from malignant RLNs.4 Morrissey et al. reported that the sensitivity of the radiologist reading was 50%, the specificity was 70%, the positive predictive value was 33%, and the negative predictive value was 82% in squamous cell carcinoma of the head and neck.24 Unlike other head and neck malignancies, NPC without distant metastasis is typically treated with radiotherapy, and therefore biopsy of RLNs in the diagnosed NPC breaches medical ethics. Furthermore, RLNs are located deep within the neck and are very close to the primary tumor. It is difficult to access these lymph nodes by imaging-guided fine-needle aspiration biopsy. It is particularly difficult to define the radiologic criteria of RLN metastasis by radiologic-histopathologic correlations based on a large sample. Therefore, the current study used an MID ≥5 mm as the size criteria for metastatic RLNs based on the recommendation of published reports.12
In the current study, a higher incidence of RLN metastases was also found more frequently in patients with advanced T (T2–4) and N (N1–3) classified disease compared with those with T1 disease and N0 disease. The pharyngobasilar fascia, an effective barrier to tumor invasion, may account for these results. Anatomically, the pharyngobasilar fascia attaches to the skull base, anterior to the medial plate of the pterygoid process separating the levator from the tensor palatini muscles, and superior to the inferior part of the petrous apex. Posteriorly, the pharyngobasilar fascia attaches to the occipital pharyngeal tubercle and the prevertebral muscles.25 Tumor invasion beyond the confinement of the pharyngobasilar fascia may be correlated with an increased risk of more advanced primary disease and RLN metastasis.10
Prognostic Value of RLN
The prognosis for patients with RLN metastasis is controversial.7–9, 22 With CT, using a MAD of 10 mm as the size criterion for metastatic RLNs, RLN was found to be a significant factor for DMFS on univariate analysis in small samples.9, 22 In contrast, the study by Chua et al. demonstrated that there was no significant difference after adjusting for the T and N classifications.7 In addition, including RLN metastases as a factor was found to affect the DMFS rates and had a negative impact on the prognosis of patients with N0 disease on multivariate analysis, using a MID of 5 mm as the criterion.8 Data from the current study indicate that RLN metastasis detected by MRI was an independent prognostic factor for DMFS in the entire group of patients. Different imaging methods, diagnosing criteria, and sample sizes could explain the difference noted in incidence, and therefore may account for this finding.
RLNs are located in the retropharyngeal space, which is a part of the lymphatic system.3 According to the general principles of the current AJCC staging system,5 RLNs are classified as N1 if unilateral and N2 if bilateral. However, the results of the current study demonstrated no significant differences in any of the endpoints between patients with unilateral and those with bilateral RLN metastasis. There is no evidence to upgrade patients with N0 to N1 disease with bilateral RLN metastasis to N2 disease.
Comparing Staging Categories of RLN Metastasis
A good staging classification system should meet the following criteria: 1) survival rates should differ among the groups (hazard discrimination); 2) the subsets defined by the T, N, and M classifications that comprise a given group should have similar survival rates (hazard consistency); 3) the patient distribution across the groups should be balanced; and 4) the cure prediction should be high (outcome prediction).26 Judging from the hazards ratios and survival curves of the different N classification subsets, the prognosis for patients with N0 disease with RLN metastasis was similar to that for patients with N1 disease, and distinguishable from that of patients with N0 disease without RLN metastasis. Including N0 disease with RLN metastases in N1 disease would improve the hazard consistency of this subset of patients. In addition, this inclusion improves the hazard discrimination among N classification subsets, as noted by the similarities noted in the hazards ratios.
If the N classification system is altered to include patients with N0 disease with RLN metastases within the N1 subgroup, 55.4% of patients are classified as N1, which unfortunately is an unbalanced distribution. However, in the current study 185 patients (20.0%) with negative cervical lymph nodes were found to have RLN metastasis. Many (151 patients [16.3%]) of these 185 patients were also found to have an advanced T classification (T2–4), and therefore the overall stage was unchanged regardless of whether RLN was classified as N1. Only 34 patients (3.7%) were reclassified from stage I to stage II disease, and the pattern of distribution for overall clinical stage was still quite balanced, including similar group sizes for patients with stage II to stage IV disease.
The current analysis, with a large number of patients at a single institution and a systemic staging workup performed using more sensitive and specific imaging technologies as well as a stratified treatment modality, offers valuable information for evaluating the prognostic value of RLN metastasis. The current study was based on solid data with full statistical justification, and all analyses were duly adjusted for related parameters to minimize potential biases. More importantly, it meets the requirement of diagnostic development, introducing RLN metastasis, an independent prognostic factor, into the current staging system for NPC.
However, it should be noted that 118 patients (12.8%) were treated with IMRT, but only 33 (3.6%) were treated with 3-DCRT in the current series. Although excellent local control has been achieved with IMRT, patients unfortunate continue to fail distantly.27, 28 Data from the current study also demonstrated that radiation techniques were not found to be an independent factor on multivariate analysis. This suggests that RLN was a significant factor for DMFS in patients with NPC. Thus, inaccuracy resulting from different radiotherapy techniques was minimal in our investigation.
The results of the current MRI-based study demonstrate that RLN metastasis affects the DMFS rates of patients with NPC. We propose that RLN metastasis be classified as N1 disease, regardless of its laterality.