Resection specimens from 121 patients who had undergone ND surgery for OSCC at the National Cancer Institute of Naples, ‘G. Pascale’, Italy, between the years 1993–2004, formed the basis of the present retrospective analysis. All patients underwent MRND for OSCC (11,19). Patients who had had previous surgery (other than diagnostic biopsy) were excluded. Throughout the time period of the study, the resection specimens were performed by the same surgical team and were assessed by the same pathological team. Pathological report and topography of the extent and location of the metastatic disease for each patient were reviewed and number, size and histological patterns of LNM-positive cases (pN+) were re-evaluated according to previous literature (14,15,20,21) in order to calculate frequency, distribution and other significant statistical correlations existing between primary tumour features and pN+.

The series comprised 80 males (mean age of 63.10±10.79 years; range, 30–83 years, median 63 years) and 41 females (mean age, 63.12 ±15.19 years; range, 25–86 years, median, 65 years). The cTNM staging was assessed according to the 6th edition AJCC (2) since data refer to the period between 1999–2004. Patient demographic and clinical characteristics of the 121 cases are summarized in Table I.

At the National Cancer Institute of Naples ‘G. Pascale’, Italy, the selection of OSCC cases requiring ND depends on the cTNM (2). cTNM parameters are accurately quantified after performing a series of clinical-instrumental examinations such as PET, total body CT scan, neck echo-color Doppler, fibrolaryngoscopy, esophageal-gastric-duodenoscopy, bronchoscopy and, if any doubts persist, fine needle aspiration biopsy (FNAB) and biopsy.

According to the literature, MRNDs are considered mandatory for the OSCCs clinically showing cervical LNMs (any cT/N+ and therapeutic ND), for locally advanced primary tumours with clinically undetectable LNMS (cT3/N0 and cT4/cN0, elective ND) and in the cases of small primary tumours clinically negative to node involvement (cT1-T2/N0) that during intraoperative assessment of their sentinel lymph node reveal a positivity to metastasis (SLN+) (8,9).

For the ND surgical approach, we adopted MRND (10), thus involving levels I–IV for all oral sub-sites, except for trigone, posterior tongue and anterior pillar where ND extended until level V; level IIB was always comprised.

After surgery and histopathological assessment of the margins of excision of the primary tumour and after the evaluation of the involvement of the surroundings structures, adjuvant therapy was performed; radiotherapy for T4 tumours with free-surgical margins and/or ≥N2, and both radiotherapy and chemotherapy both for tumours with any N+ plus ECS, tumours with any T and positive or close margins or perineural invasion and/or neoplastic vascular embolization tumours T1–3 with free margins and pN0/pN1 were not enrolled for adjuvant treatments (12,13).

Since data refer to patients recruited from 1999 to 2004, prior to the last Cancer Staging Atlas publication, we considered the previous edition (2); hence, T4 tumours were not distinguished in T4a and T4b, consequently, stage IV has not been subclassified into stages IVa, IVb and IVc (7).

All 121 lymph node specimens dissected were histopathologically re-evaluated for the present study. In order to detect micrometastases and to confirm the involvement of lymph nodes, the standardized sectioning protocol was performed. One H&E stained section was prepared from each block and examined for the presence of nodal involvement by tumour. If present, metastatic disease was reported. If node was negative or equivocal for metastatic disease, or positive for micrometastases, the removed lymph nodes were also serially sectioned at 25-μm intervals 4 μm thick and alternately stained with H&E and immunohistochemical staining (IHC) using anti-cytokeratin and EMA antibodies (Fig. 1) as described below. This pattern was continued throughout the entire block. Each immunostained component was always compared with adjacent sections stained by H&E.

Woolgar’s classification criteria were considered in order to classify the type of metastasis (21,22). Woolgar distinguished metastatic lymph nodes into two groups with different prognosis and features: typical metastatic pattern as ‘orderly involvement of successive anatomical nodal levels, creating an inverted cone with maximum volume and maximum ECS at levels I or II and a gradual reduction in the volume/extent of metastasis at the numerically higher levels’ (21), vs. atypical metastatic pattern, termed ‘aberrant’ by Woolgar, and characterized by various features. The ones referring to the atypical pattern and considered in the present study were the involvement of ‘other’ anatomical groups of nodes (including parapharyngeal, facial, buccal, lingual and sublingual nodes), involvement of controlateral cervical lymph nodes, skipping of anatomical levels other than level I and the presence of a single micrometastasis (21).

Once defined, the histopathological features of node metastases were correlated with the clinical and histopathological features of the primary tumour in order to establish statistical and prognostic correlations.

Histological and immunohistochemical analyses were performed on formalin-fixed, paraffin-embedded tissue samples. Immunostaining was performed using the linked streptavidin-biotin horseradish peroxidase technique (LSAB-HRP). Antigen retrieval was performed by microwave heating, a first time for 3 min at 650 W, a second and a third time for 3 min at 350 W, the slides immersed in 10 mM citrate buffer pH 6.0. After heating, the sections were blocked for 60 min with 1.5% horse serum (Santa Cruz Biotechnology) diluted in PBS buffer before reaction with the primary antibody (Ab). The primary monoclonal antibodies anti CK AE1/AE3 (dilution 1:50, pH 6.0; Dako, Carpinteria, CA, USA) and EMA (dilution 1:75, with protein K; Dako) were incubated overnight. After two washes with PBS, the slides were treated with biotinylated species-specific secondary antibodies and streptavidin-biotin enzyme reagent (Dako, Glostrup, Denmark), and the colour developed by 3,3′-diaminobenzidine tetrahydrochloride. Sections were counterstained with Mayer’s hematoxylin and mounted using xylene-based mounting medium. In negative controls, the primary antibody was omitted. The results of the IHC were separately evaluated by two independent observers by carefully examining the entire section with an optical microscope (Olympus BX41). For each case, the presence and the extent of positive cells in all sections examined was determined. Isolated tumour cells (ITCs) are defined as tumour cell clusters that are not >0.2 mm in largest diameter and are denoted as lymph node negative (pN0[i+]). Micrometastases are defined as metastases that are >0.2 mm in diameter but ≤2 mm, denoted as lymph node positive (pN1mi). Carcinoma macrometastases measured >2 mm in maximum extent. Two investigators experienced in oral pathology blindly and independently examined the study sections initially, and then they evaluated together the histopathological and immunostained sections until they reached an agreement.

Data were analysed by the GraphPad Prism software version 5.0 for Windows (GraphPad Software, San Diego, CA, USA; www.graphpad.com) and Excel Microsoft Office. Differences among the groups were estimated using the one-way analysis of variance (ANOVA) and the Student-Newman-Keuls test. Only P-values <0.05 were considered significant. Overall survivals in the different groups were calculated by the Kaplan-Meier curves and log-rank (Mantel-Cox) test was applied for comparing survival probabilities. The pathological positivity of the ND (pN+) and the type of LNMs were correlated with the site of primary tumour. The percentages of positive node metastases (pN+) in elective, therapeutic and secondary to SLN+ NDs and the number of lymph nodes harvested were also evaluated.

Patient demographics and characteristics such as gender, OSCC site, primary tumour differentiation grading and cTNM staging according to the international guidelines (2) of the 121 OSCCs considered are shown in Table I.

At pre-surgical clinical-instrumental evaluation, 61/121 cases (50.4%) were considered positive (cN+) and 60/121 (49.6%) negative (cN0) to node metastases. The histopathological lymph node assessment subsequent to ND revealed at least one node metastasis in 56/121 NDs, thus considered pathologically positive lymph nodes (pN+) and no node metastasis in the remaining 65 cases, which were considered pathologically negative lymph nodes (pN0). The percentages of true positive (cN+ and pN+), true negative (cN0 and pN0), false positive (cN+ and pN0) and false negative (cN0 and pN+) were 87, 95, 13 and 5% respectively, as reported with details of accuracy, sensitivity, specificity, and positive and negative predictive values in Fig. 2.

The pN+ distribution according to the grading, staging MRND intent and, conversely, the pN+ frequency in each group are reported in Table II.

Among the 121 NDs considered, 61/121 (50%), presenting at least a clinical node involvement (cN+), were therapeutic NDs; 19/121 (16%), presenting a cT3/cT4 and cN0 were elective NDs; the remaining 41/121 (34%), presenting primary tumours small in size (cT1/cT2) and no evident clinical node metastases (cN0) but positive to sentinel lymph node metastases (SLN+) during the intraoperative assessment of sentinel lymph node, were NDs performed to establish the presence of other node metastases in addition to SLN+ (Table II). The distribution of 56 pN+ according to clinical-instrumental indications and, conversely, the pN+ frequency in each indication are reported in Table II.

After excluding 19 pN+ cases due to lack of data useful to classify the pN+ histopathological pattern, the remaining 37 pN+ NDs were distinguished into 17/37 with typical (46%) and 20/37 with atypical (54%) node metastases according to Woolgar’s classification (21). The atypical pattern was largely represented by skip metastases [10/20 (50%)], followed by the frequency of bilateral metastases [5/20 (25%)], micrometastases [2/20 (10%)], involvement of other lymph nodes alone [2/20 (10%)], and 1/20 (5%) showing both micrometastases and other lymph nodes co-interested (Fig. 3A). According to primary tumour site, skip metastases indiscriminately involved any site of the oral cavity, except cheek and fornix, which, on the contrary, preferentially showed micrometastases; bilateral metastases were frequently detected in OSCCs affecting tongue as a single site or in association with floor of the mouth or trigone; ‘other different lymph nodes’ were involved in floor of the mouth tumours (Fig. 3B and C).

Among 56 pN+ cases, 15/56 pN+ (26.8%) involved levels I–III, 11/56 pN+ (19.6%) involved levels IV–V, and 30/56 (53.6%) pN+ were censored due to level not reported.

ND anatomical levels I–III were typically involved by node metastases while levels IV–V showed pN+ only for primary tumours of trigone, floor of the mouth plus tongue, tongue alone and multiple sites (data not shown).

A total of 3,390 lymph nodes were harvested in 121 NDs, mean 18.5±22.7 per ND, range 1–110. One hundred and sixty-two lymph nodes were pN+ in 56 NDs, mean 2.9±0.03, range 1–10.

Recurrence was observed in 8 cases, whose clinical, histological and lymph node features are shown in Table III.

Overall survival statistical significance and percentages of subjects alive at 1, 2 and 5 years from the diagnosis are reported in Table IV.

Overall survival was statistically different according to early AJCC vs. late AJCC stages (P=0.0004) with a 5-year survival of 78.34 and 46.16%, respectively, and according to absence/presence of nodal metastases (pN0 vs. pN+, P=0.0003) with a 1–5 year overall survival ranging between 97–74% in pN0 cases and between 69–54% in pN+ cases.

Overall survival according to histological grade and typical vs. atypical lymph nodal metastatic pattern did not reach statistical significance when independently considered or when together. When comparing pN0 subjects, pN+ with macrometastases or ECS and pN+ with micrometastases, patients reporting pN+ with micrometastases showed an intermediate probability of survival after 5 years and these data were statistically significant (P=0.004). With regard to the number of lymph nodes harvested, statistically significant different survivals (P=0.005) were found in cases with >30 lymph nodes harvested, whose 5-year survival was >80% compared to 46.93% in cases with <20 lymph nodes and 66.67% in the 20–30 lymph node group, independently of the pN status.

With regard to intent to surgery, overall survival among subjects who had undergone elective, therapeutic and after SLN+ NDs were always statistically significant with the worst prognosis in patients who had undergone therapeutic ND and the most positive one in patients who had undergone ND after SLN+. pN status did not appear to be responsible for differences in survival since patients with pN0/therapeutic MRND showed a 15% 5-year survival compared to pN+/therapeutic MRND patients with a 45.72% 5-year survival.