N1S3: A revised staging system for head and neck cutaneous squamous cell carcinoma with lymph node metastases

Results of 2 Australian Cancer Centers

Authors

  • Veronique-Isabelle Forest MD, MSc,

    1. Sydney Head and Neck Cancer Institute, Sydney Cancer Center, Royal Prince Alfred Hospital, University of Sydney, Sydney, Australia
    Search for more papers by this author
  • Jonathan J. Clark MBBS,

    Corresponding author
    1. Sydney Head and Neck Cancer Institute, Sydney Cancer Center, Royal Prince Alfred Hospital, University of Sydney, Sydney, Australia
    • Sydney Head and Neck Cancer Institute, Royal Prince Alfred and Liverpool Hospitals, Sydney, Australia
    Search for more papers by this author
    • Fax: (011) 61 2 9601 6459

  • Michael J. Veness MBBS, MMed, MD,

    1. Head and Neck Cancer Service, Westmead Hospital, University of Sydney, Sydney, Australia
    Search for more papers by this author
  • Chris Milross MBBS, MD

    1. Sydney Head and Neck Cancer Institute, Sydney Cancer Center, Royal Prince Alfred Hospital, University of Sydney, Sydney, Australia
    Search for more papers by this author

Abstract

BACKGROUND.

A staging system was designed for metastatic cutaneous squamous cell carcinoma (SCC) that would incorporate the parotid as a regional level and facilitate a better prognostic discrimination between subgroups.

METHODS.

A retrospective review of clinical and pathological information of patients treated for metastatic cutaneous SCC to the parotid and/or neck was conducted. Potential prognostic factors were analyzed using univariate and multivariate analyses. A staging system was elaborated and externally validated.

RESULTS.

Two hundred fifteen patients were included. All patients had surgery as their primary treatment; 148 had parotidectomy with neck dissection, 50 parotidectomy alone, and 18 neck dissection alone. One hundred seventy-five patients received postoperative radiotherapy. On univariate analysis, the number of involved lymph nodes (P < .001), maximal size (P = .01), and extracapsular spread (P = .003) were found to be significant predictors of survival. On Cox regression, the number of involved lymph nodes as single or multiple (P = .006) was significant. The N1S3 staging system incorporates involved lymph nodes from parotid and neck (single or multiple) and the size (< or >3 cm). This system demonstrates significant predictive capacity for locoregional control (P < .001), disease-specific survival (P<.0001), and overall survival (P<.0001). N1S3 was tested on a different cohort of 250 patients, and the results confirmed those obtained from our primary analyses.

CONCLUSIONS.

The N1S3 system stages patients according to the number of involved lymph nodes and size, and incorporates parotid as 1 of the regional levels. These 2 predictors are easily applied on both clinical and pathological data. Cancer 2010. © 2010 American Cancer Society.

In Australia, 374,000 cases of nonmelanoma skin cancers are diagnosed every year, and cutaneous squamous cell carcinomas (SCCs) make up 20% to 25% of these. The majority of these skin cancers are low risk, with infrequent local recurrence and a low incidence of regional metastases (≤5%); consequently, local excision is considered to be sufficient treatment.1 Despite the relatively low risk, O'Brien et al demonstrated that the parotid gland is an important “metastatic basin” for head and neck (HN) cutaneous SCC, and that >50% of patients will have concurrent cervical lymph node involvement (occult or clinical) when there is clinical evidence of parotid metastases.2

Although surgery and radiotherapy are accepted as standard treatment for HN cutaneous SCC, the extent of parotidectomy, neck dissection, clinical target volume of radiotherapy, and use of concurrent chemotherapy are contentious. One of the present limitations is the lack of a validated staging system to guide the management of these patients.3-5 The current clinical TNM staging system for skin cancer recognized by the American Joint Committee on Cancer offers a simplistic classification dividing patients into 2 categories (N0/N1), in contrast to mucosal SCC, where there are 5 categories (N1, N2a, N2b, N2c, N3). As such, it has not proved helpful in the clinic nor in stratifying patients in clinical research.

Andruchow et al published in 2006 a clinical staging system where parotid and neck metastases were classified separately.6 This study was designed to validate O'Brien's hypothesis that parotid (P stage) and neck (N stage) involvement may have independent prognostic significance, and to address the deficiencies in the current TNM system. Although O'Brien's system represented a substantial improvement on the existing TNM system, several problems remained. First, the system did not adequately stratify all groups (ie, P1 vs P2 vs P3 and N0 vs N1 vs N2). Second, the system did not perform well when applied to pathological data. Third, separating the parotid from the neck increased the complexity of the staging system without substantial gain. Recently Hinerman et al, in a large series of metastatic HN cutaneous SCC from Florida, affirmed the prognostic significance of P stage, but again failed to demonstrate stratification of risk by N stage (N1 vs N2) or to validate the separation of P and N stage on regression analysis.7

The aim of this study was to identify and combine independent pathological variables to modify O'Brien's existing staging system for HN cutaneous SCC, incorporating the parotid as 1 of the regional levels. This novel staging system will then be applied to a patient cohort from another major Australian cancer center to assess its performance.

MATERIALS AND METHODS

Clinicopathological data from patients treated at the Sydney Head and Neck Cancer Institute, Royal Prince Alfred Hospital, Australia have been recorded prospectively in a computerized database since 1987. The database records demographic data, clinical and pathological staging, treatment details, histopathological features, the duration of follow-up, and status at last visit.

All patients treated with curative intent between 1987 and 2007 for metastatic HN cutaneous SCC to the parotid and/or neck were identified. The pathology details recorded in the database were cross-checked with the original pathology reports, and the database was updated where necessary. This information was then used for statistical analysis, identification of prognostic factors, and elaboration of the staging systems. All patients included had undergone parotidectomy and/or neck dissections as their primary treatment modality.

Patients

Two hundred fifteen patients were included after detailed evaluation of their records for completeness of pathological information. There were 188 men and 27 women, with a median age of 73 years (range, 25-94 years). Median follow-up was 34 months. Five patients were immunosuppressed, with a diagnosis of chronic lymphocytic leukemia (2 patients), non-Hodgkin lymphoma (1 patient), and human immunodeficiency virus (2 patients). Use of immunosuppressive medication was not recorded in the database.

Surgery and Radiotherapy

All patients had surgery as their primary treatment: 148 patients had a parotidectomy with neck dissection, 50 patients had a parotidectomy alone, and 18 patients underwent neck dissection only. One patient had a bilateral neck dissection for a metastatic SCC of the scalp.

One hundred seventy-five patients received adjuvant radiotherapy. Among them, 17 patients received concomitant chemotherapy. The median dose of adjuvant radiotherapy given was 54 grays (Gy) to the parotid bed and 50 Gy to the neck (range, 32-66 Gy). Of the 40 patients who did not receive radiotherapy, reasons included patient refusal, early recurrence or death, favorable histopathological features, and previous radiotherapy to the neck for lymphoma/leukemia (2 patients).

Pathology of Metastatic Disease

The median number of nodes or foci of metastatic SCC in the parotid was 1 (range, 1-6), and the median size of the metastatic deposit was 22 mm (range, 5-75 mm). In the neck, the median number of positive lymph nodes was 1 (range, 0-67), and the median size was 19 mm (range, 6-65 mm). Extracapsular spread was present in 58% of patients. The margins of resection were involved or close in 47% of patients.

External Validation

The proposed novel staging system was applied to a different patient cohort of 250 patients, from Westmead Hospital, Sydney, Australia. Patients were eligible if they had had no previous treatment and had undergone surgery (± adjuvant radiotherapy) with curative intent. Patients treated exclusively with radiotherapy and those where pathological data could not reliably be determined were excluded. A minimum follow-up of 2 years was required. Relevant data were extracted from a prospective database.

Statistical Analysis

The staging system was designed using covariates demonstrating independent prognostic significance for disease-specific survival and subsequently reanalyzed to determine their capacity to stratify patients according to risk of regional failure and mortality. Locoregional control, disease-specific survival, and overall survival were calculated using the Kaplan-Meier method, and curves were compared using the log-rank test. Regression analysis of independent variables was carried out using the Cox proportional hazards model. Data were collated and filtered using Excel (Microsoft, Redmond, Wash), and the statistical analysis was performed using SPSS version 12.0 (SPSS Inc., Chicago, Ill).

RESULTS

Survival Analysis

In this cohort, the cumulative overall survival at 2 and 5 years was 82% and 69%. Disease-specific survival at 2 and 5 years was 87% and 77%. The locoregional control rate was 81% at 2 years and 73% at 5 years.

Univariate analysis

For the purpose of analysis, the number of nodes was divided into 3 groups (1 positive lymph node, 2-3 positive nodes, 4 or more positive nodes) and into 2 groups (single lymph node vs multiple nodes) according to the existing mucosal staging system. Both demonstrated a significant increase in mortality with greater number of metastatic lymph nodes (P < .001). The maximum size of the positive lymph node was also found to be a predictor of survival, with larger nodes associated with poorer survival. Positive nodes were classified according to the existing mucosal staging system into 3 groups (0-3 cm, 3-6 cm, and >6 cm). Although this classification discriminated between patients with nodes <3 cm and 3 to 6 cm (P = .04), patients with nodes >6 cm performed similarly to those with intermediate-sized nodes. For the purpose of inclusion in the model, the latter 2 groups were combined (0-3 cm, >3 cm; P = .01).

The presence of extracapsular spread (ECS) was also an adverse predictor of survival (P = .003), whereas receiving adjuvant radiotherapy did not prove to be statistically significant (P = .55).

Multivariate analysis

On Cox regression when including number of lymph nodes, lymph node size, presence of ECS, and radiotherapy, only the number of involved nodes as single or multiple (hazard ratio [HR], 3.8; 95% confidence interval [CI], 1.5-10.0; P = .006) was an independent predictor of disease-specific survival. There was a trend toward ECS (HR, 3.9; 95% CI, 0.9-17.8; P = .076) and radiotherapy (HR, 0.4; 95% CI, 0.2-1.2; P = .10) being independent prognostic variables. There was substantial correlation between ECS and size, because most metastases with ECS are large. This correlation may interfere with the ability of the regression to detect a difference for both ECS and size. Given that ECS cannot be applied to a clinical staging system, we elected to remove ECS from the model, despite evidence to suggest that ECS may exert a greater effect than lymph node size. With ECS removed, both number of nodes (P = .001) and size (P = .025) were significant.

Revised (N1S3) Staging System

On the basis of the analysis of this clinicopathological data set, a novel staging system was devised and entitled N1S3. This incorporates the number of involved nodes from the parotid and the neck as being single or multiple and the maximum size as being < or >3 cm. The N1S3 staging system classifies patients into 3 groups: I) single lymph node ≤3 cm; II) single lymph node >3 cm or multiple nodes ≤3 cm; and III) multiple nodes maximum diameter >3 cm (Table 1). The N1S3 system has significant predictive capacity for locoregional control (P < .001) (Fig. 1), disease-specific survival (P < .0001) (Fig. 2), and overall survival (P < .0001) (Fig. 3). Cox regression (Table 2) incorporating the new staging system, radiotherapy, and ECS demonstrated that each stage was significant with a HR of 3.3 (P = .04) for stage II and 7.7 (P = .001) for stage III compared with stage I. Radiotherapy (P = .08) and presence of ECS (P = .07) exhibited a trend toward being independent predictors of disease-free survival when analyzed with the N1S3 system, and radiotherapy was a significant independent predictor of improved overall survival (P = .035).

Figure 1.

Locoregional control using N1S3 staging system is shown.

Figure 2.

Disease-specific survival is shown when the N1S3 staging system is applied.

Figure 3.

Overall survival when applying the N1S3 staging system is shown.

Table 1. N1S3 Staging System Applied to the Cohort of 215 Patientsa With Disease-Specific Survival and Overall Survival at 2 Years and 5 Years
StageNo.DSSOS
% at 2 Years% at 5 Years% at 2 Years% at 5 Years
  • DSS indicates disease-specific survival; OS, overall survival.

  • a

    The total number of cases was 216 because 1 patient underwent a bilateral neck dissection.

I: Single lymph node measuring ≤3 cm10295908878
II: Single lymph node measuring >3 cm or multiple lymph nodes measuring ≤3 cm8888758269
III: Multiple lymph nodes measuring >3 cm2666426441
Table 2. Cox Regression Analysis of Disease-Specific Survival Using the N1S3 Staging System, RT, and ECS for Both Cohorts
 Royal Prince Alfred HospitalWestmead Hospital
PHazard Ratio95% CI for HRPHazard Ratio95% CI for HR
  1. RT indicates radiotherapy; ECS, extracapsular spread; CI, confidence interval; HR, hazard ratio.

N1S3      
 Stage I.0031.00 .0191.00 
 Stage II.0443.301.03-10.53.0263.371.15-9.85
 Stage III.0017.702.31-25.67.0055.151.65-16.09
RT.0810.440.17-1.11.0700.670.43-1.03
ECS.0733.920.88-17.42.0005.632.20-14.38

External Validation

The model/staging system developed from the Sydney Head and Neck Cancer Institute data was applied on an external data set from Westmead hospital. This cohort of patients included 207 men and 43 women with a median age at diagnosis of 67 years (range, 34-95 years). The median follow-up was 54 months. All patients underwent surgery, either alone (11%) or followed by external beam adjuvant radiotherapy (89%). Twenty-seven patients had only a parotidectomy, 125 patients underwent a parotidectomy and neck dissection, and 98 patients had a neck dissection alone. The median number of metastatic nodes was 2 (range, 1-29), and the median size was 26 mm (range, 10-92 mm). Extracapsular spread was present in 82% of patients. Seventy of 250 (28%) patients recurred; 51 had a regional recurrence, and 9 developed distant metastasis. Median time to recurrence was 8 months, and the majority of patients (73%) who developed regional metastasis died of disease. Twenty patients were removed from analysis because of incomplete pathological data for the staging system because they underwent parotidectomy alone and had a single lymph node involved, where the addition of neck dissection may have reclassified them from stage I to stage II.

When the N1S3 staging system was applied to this independent external cohort, the results from the primary analysis were confirmed. N1S3 offered discrimination between the 3 different groups for locoregional control (log-rank P = .01) (Fig. 4) and disease-specific survival (P = .004) (Fig. 5). Cox regression (Table 2) also demonstrated comparable results in terms of HRs for the staging system and radiotherapy in predicting disease-specific survival. Figure 6 presents the results of disease-specific survival when adjusted for radiotherapy. However, extracapsular spread was found to be a highly significant predictor of survival (P < .001), although HRs were not markedly different.

Figure 4.

Locoregional control is shown when the N1S3 staging system is applied to the external cohort.

Figure 5.

Disease-specific survival is shown when the N1S3 staging system is applied to the external cohort.

Figure 6.

Disease-specific survival adjusted for radiotherapy when applying the N1S3 staging system to the external cohort is shown.

DISCUSSION

The aim of this study was to elaborate a staging system for HN cutaneous SCC that would facilitate improved prognostic discrimination between subgroups. It was imperative that the model be simple, usable, and applicable in all clinical settings. According to our analysis, the number of nodes is the most important predictor of survival (HR, 3.8), followed by the maximum size of the involved nodes and presence of ECS.

Other authors have identified increasing lymph node size (>3 cm) and multiple lymph nodes as independent adverse prognostic factors.8, 9 The N1S3 staging system combines these 2 predictors in a simple fashion that can be applied in both the preoperative and postoperative settings. We have externally applied this staging system on an additional cohort of 250 patients, where it has also been shown to offer excellent discrimination between groups. Apart from the prognostic application, these data suggest that disease-specific survival at 5 years is approximately 90% for stage I disease. Given this, it may be reasonable to treat selected patients confirmed to have a single small lymph node with parotidectomy and selective neck dissection alone and omit the additional morbidity of adjuvant radiotherapy, assuming adequate margins are achieved and ECS is not present.10

Recognition of the prior work by O'Brien11 is fundamental to understand the evolution of this staging system. His research culminated in a multicenter study published in 20066 confirming the hypothesis that extent of disease is prognostic in metastatic HN cutaneous SCC, in a similar fashion to mucosal SCC, and highlighted the limitations of the current American Joint Committee on Cancer system. However, on critical analysis, the P and N staging system did not adequately stratify all subgroups of patients, particularly when applied to pathological data. Furthermore, the separation of P and N was quite complex, and the relative contribution of each in predicting outcome was not clearly defined. The parotid gland contains the first echelon lymph nodes for many primary lesions located on the face and scalp.12 However, there are no anatomical or functional barriers between the parotid and neck. This is evidenced by the finding that metastatic cutaneous SCC within the parotid is associated with a 26% rate of clinical cervical metastases, and an additional 35% have occult neck disease.2 The present study suggests that it is reasonable to incorporate the parotid gland as 1 of the regional cervical levels, at least for the purpose of predicting disease-related outcomes in HN cutaneous SCC. The odds of death because of HN cutaneous SCC in each stage have been clearly described and confirmed in an external cohort. Although most staging systems are primarily clinical, we believe that decisions regarding adjuvant therapy are preferentially based on pathological parameters when available.

The importance of ECS is an issue that warrants further comment. The rate of ECS in our cohort is 58%, which is lower than the 82% reported in the external cohort. This may reflect less reliable reporting of extracapsular spread in histopathological reports dating before 1990. Alternatively, metastatic cutaneous SCC may be present within the parotid bed as a soft tissue deposit with no lymph node remnant identifiable, and thus may not always be classified as ECS. The prognostic significance of soft tissue deposit versus lymph node metastasis or the definition of what is considered ECS in the parotid are unanswered questions at the moment. This may explain the wide range of ECS reported in the literature. Various studies have demonstrated the importance of ECS in locoregional control and survival. ECS was previously identified as a predictor for survival on univariate analysis by Veness et al, although on multivariate analysis, only number of lymph nodes and treatment modality remained significant.13 Similarly, Ch'ng et al did not find ECS to be an independent prognostic factor.14 There is a positive correlation between the presence of extracapsular spread and the size of the involved lymph node. Puri et al evaluated this association in noncutaneous SCC. ECS was found in 60% to 100% of lymph nodes >3 cm, in 39% to 59% of lymph nodes <3 cm, and in only 23% of lymph nodes <1 cm.15 Thus, we can hypothesize that size of the lymph node and ECS are surrogates for each other, and that collinearity prevents both factors from contributing to the regression model.

Conclusions

To design the proposed staging system, we have elected to rely on 2 predictors of survival that have been consistently reported in the literature: the number of lymph nodes and the maximal size of the involved lymph nodes. These 2 predictors, when combined, offer excellent discrimination by stage and have been externally validated. We believe that other clinicopathological factors such as ECS, immunosuppression, and treatment modality are equally important but less readily applied as a clinical and pathological staging system. The widespread application of an acceptable staging system is long overdue for HN cutaneous SCC in both the clinical and research arenas.

CONFLICT OF INTEREST DISCLOSURES

The authors made no disclosures.

Ancillary