2. AJCC 7th Edition
The seventh edition of the AJCC staging manual (AJCC 7th edition), released in 2010, was the first to stratify patients based on nodal number, size, and laterality
[14]. The new N category and overall staging criteria was a replica of the mucosal staging from the same edition. Whilst a unified system promotes acceptance through familiarity
[15], it ignores the distinct biological differences between SCC of cutaneous and mucosal origin. This meant that T1N2M0 disease was grouped together with T4N3M1 disease as stage IV. In 2011, Brunner et al. evaluated the appropriateness of including N2 and N3 disease into stage IV in a multi-institutional cohort of 603 HNcSCC patients with regional metastasis
[16]. They found that patients with distant disease (M1) had a very poor, and markedly different prognosis to those with only regional disease, with five-year DSS for N2, N3, and M1 of 75%, 65%, and 11%, respectively. In addition, the 35 M1 patients from the study were all initially treated with curative intent but developed distant metastasis during follow-up suggesting that the true impact of distant metastases was likely to be worse and it was therefore recommended that regional disease be grouped independently from distant metastases regardless of extent.
2.1. AJCC 7th Edition vs. N1S3
In 2012, Clark et al. compared N1S3 directly with the new AJCC 7th edition nodal staging system in an Australian multi-centre cohort
[13]. Cox regression was used to adjust for treating institutional as well as clinicopathological variables such as immunosuppression, PORT, nodal margins, and ENE. The authors suggested that the complexity of the AJCC 7th edition was unhelpful as there was no prognostic distinctiveness for any N2 subgroupings, in particular N2c, which was rarely designated. The proportion of variation explained (PVE) was similar for both staging models indicating that the performance of the two models was similar, but N1S3 distributed patients better and stratified patients better according to risk of death.
2.2. AJCC 7th Edition Cutaneous vs. Oral
In 2014, Brunner et al. explored the justification for replicating the mucosal SCC N category in the AJCC 7th edition for HNcSCC, noting the different disease behaviour and patient populations
[1]. The study consisted of 672 patients with regional metastatic HNcSCC and 225 patients with regional metastatic oral cavity SCC treated with curative intent. The oral SCC N category stratified patients well according to survival, whereas HNcSCC stratification was poor. Within the HNcSCC N2 subgroups there was only 1% separation of DSS at 3 years. Adjusting for immunosuppression, age, gender, PORT, treatment institution, and ENE, the increase in risk by N category was neither clinically useful nor monotonic. In summary, the AJCC 7th edition worked well for mucosal SCC but performed poorly for HNcSCC, providing evidence for a standalone staging system for HNcSCC regional disease.
3. AJCC 8th Edition
HNcSCC was recognised as a separate entity for the first time in the AJCC 8th edition, published in 2017
[17]. Significant changes were made to the T category, but the N category remained aligned with mucosal SCC, with the introduction of ENE for the first time.
3.1. AJCC8 vs. AJCC7
In a single institution analysis of 382 patients, Liu et al. compared the performance of the AJCC 7th edition versus the AJCC 8th edition
[9]. In this cohort, 74.6% of patients were upstaged by the AJCC 8th edition due to the addition of ENE. Although risk stratification by N category was poor for the AJCC 7th edition, the results for the AJCC 8th edition were inferior, with no risk stratification observed between any nodal subgroups for either DSS or OSS. When translated to stage, 27.7% of patients were upstaged by the AJCC 8th edition, with 88% of patients with regional metastases designated stage IV versus 60% for AJCC 7th edition. Stage III and IV Kaplan–Meier curves overlapped for both DSS and OS. Similar results were reported in another study by Sood and colleagues
[18].
In 2020, Luk et al. evaluated the AJCC 8th edition and compared its prognostic utility in an Australian multicentre study of 1146 HNcSCC patients with regional metastasis
[19]. In this cohort, the N category was upstaged in 80.9% of patients by the AJCC 8th edition due to the presence of ENE, 29.9% were upstaged from stage III to IV, resulting in 90.6% of patients with nodal metastases being assigned to stage IV disease. The prognostic performance of the AJCC 8th edition N category was poor according to Harrell’s concordance index (C-index) (0.62 for DSS and 0.59 for OS) with no risk stratification observed between pN1, pN2a, pN2b, or pN2c for DSS. With regards to stage, the AJCC 8th edition performance was poor with C-indexes of 0.54 for DSS and 0.53 for OS; almost equivalent to “flipping a coin”. This also showed that the N3a nodal category (metastatic node >6 cm without ENE) is likely redundant, being reported in only 0.3% of patients and N3a disease carrying a worse prognosis than N3b.
3.2. AJCC 8th Edition vs. N1S3 vs. ITEM
Ebrahimi et al. compared the prognostic performance of the AJCC 8th edition against the N1S3 and ITEM systems in a multicentre cohort of 990 patients
[20]. Model performance was measured using PVE, C-index, Akaike Information Criterion (AIC), and Bayesian Information Criterion (BIC). The AJCC 8th edition N category distribution was poor with 84% of patients classified as either N2a or N3b due to the inclusion of ENE, whereas N2c and N3a appeared redundant. For the AJCC 8th edition, the majority of patients (90.6%) were grouped as stage IV. In comparison N1S3 and ITEM both exhibited a reasonably balanced distribution (N1S3 stage I, II, and III was 39.3%, 44.9%, and 15.8%, respectively; ITEM low risk, moderate risk, and high risk was 18.1%, 35.5%, and 46.5%, respectively). However, only 34.3% of patients were allocated to the same risk level by both N1S3 and ITEM reflecting the different variables used in each system.
For DSS, N1S3 performed similarly to the AJCC 8th edition N category and the Kaplan–Meier curves for N1S3 appeared to stratify patients well. Although N1S3 was preferred over the AJCC 8th edition due to simplicity, the C-index of 0.62 and PVE of 10.9% was relatively poor compared with other cancer staging models. A further finding was that all staging systems were better predictors of DSS than OS. This reflects the competing risk of age-related comorbidities in the elderly HNcSCC population, with a median age of 73 years and relatively high disease control rates. They recommended the use of DSS as the preferred endpoint for HNcSCC nodal staging rather than OS.
3.3. AJCC 8th Edition Within-Stage Heterogeneity
In view of the poor performance of the AJCC 8th edition and the fact that most patients appear to be classified as pN2a, pN3b, and stage IV, Ebrahimi and colleagues hypothesized that patients with widely different prognoses were being classified into the same risk group. To identify sources of within-stage prognostic heterogeneity, a cohort of 1146 HNcSCC patients with regional metastases was analysed
[21]. Conclusions regarding within-group homogeneity were not possible for pN2c or pN3a groups due to insufficient numbers. However, pN2a and pN3b exhibited significant heterogeneity due to the effect of immunosuppression, nodal size, nodal number, and perineural invasion (PNI). Within pN2a, patients without PNI or immunosuppression had a 5-year DSS of 87.1%, compared with 69.9% in the presence of either feature. Regarding stage, TNM stage III contained only pN1 patients, and was relatively homogenous; however, stage IV contained over 90% of the entire cohort with marked heterogeneity observed. Ebrahimi et al. created low, moderate, and high-risk subgroups in an exploratory analysis of the stage IV patients, based on immunosuppression, PNI, ENE, and the number and size of the nodal metastases, weighted according to the hazard ratios observed in the multivariate analysis. The 5-year DSS for the low-risk stage IV patients was 90%, whereas the 5-year DSS for the high-risk stage IV patients was only 60%. These results revealed a staggering 400% difference in disease-specific mortality risk for patients ultimately staged together by the AJCC 8th edition, largely due to the incorporation of ENE, with resulting within-stage heterogeneity contributing significantly to the overall poor performance.
4. Additional Research of Relevance to Nodal Staging
4.1. Low-Risk
Most studies have focused on identifying high-risk patients requiring treatment escalation. In 2011, Ebrahimi et al. published on their efforts to identify a subset of patients who could benefit from treatment de-escalation
[22]. Using the N1S3 system, 168 “low-risk” metastatic HNcSCC patients (with a single lymph node measuring ≤3 cm) who had undergone surgery as their primary treatment were analysed, of which 20% did not receive PORT. Reasons included patient refusal, poor performance status, physician preference based on favourable histopathological features, and a prior history of RT to the surgical bed. Amongst this cohort, the regional control rate was 92% and the remaining 8% who recurred were successfully salvaged with further surgery and PORT. The 5-year DSS was 100% for patients without ENE versus 93% for those with ENE, suggesting that N1S3 stage I patients without ENE or other adverse clinicopathological features could be treated with surgery alone. However, the researchers emphasised the prerequisite for patients with regional disease to have undergone appropriate parotidectomy and neck dissection, based on the location of the primary tumour, before considering the omission of PORT.
4.2. Soft Tissue Metastases (STM)
The prognostic impact of STM has been demonstrated in melanoma and mucosal SCC of the HN
[23][24]. In 2012, Kelder et al. compared the prognostic difference between STM and ENE in 164 HNcSCC patients with regional metastasis
[25]. Adjusting for the effect of age, margin status, primary lesion size, number of nodes, and PORT, the presence of STM was associated with worse survival than ENE, suggesting that pathology reports for HNcSCC should include not just the presence of ENE, but also the presence, number, and size of STMs. Furthermore, a lower threshold for aggressive multimodal treatment in the presence of STM was recommended.
After adopting their own recommendations, the same group published an updated analysis on ENE and STM in 2020 on a series of 535 HNcSCC patients with regional metastasis
[26]. Interestingly, only 22.4% had nodal disease without either ENE or STM versus 35.4% in 2012 and 26.2% had STM alone versus 13.4% in 2012. The finding of STM in the absence of ENE had doubled and STM in the presence of ENE had reduced, suggesting that the increased recognition of STM had led to both increased identification and the reclassification of cases previously labelled as ENE. Whilst ENE alone, STM alone, and ENE with STM were all associated with reduced OS, the impact of STM was now similar to that of ENE. The diminished survival impact of STM was considered a stage migration effect associated with greater recognition and more aggressive treatment of STM when identified compared with the results of the 2012 study. Based on this evidence it was recommended that STM be managed equivalent to ENE from a prognostic and treatment perspective, but that these findings should be individually documented in pathology reports to facilitate further study.
4.3. Disease-Free Interval (DFI)
Time to development of lymph node metastases in melanoma patients has been shown to be prognostic
[27]. Consequently, Ebrahimi et al. in 2012 investigated whether the time interval between treatment of an index primary HNcSCC and subsequent presentation of regional metastatic disease impacted survival outcomes
[28]. With a cohort of 229 HNcSCC patients with regional metastasis, only 20% of patients presented with concurrent primaries and regional metastatic disease. In the remaining patients, the mean and median DFI were 11.7 and 7.2 months, respectively. On multivariate analysis, DFI (≤9 vs. >9 months) was associated with locoregional control and DSS. The researchers concluded that DFI was likely to be an inverse surrogate for aggressive disease biology, and unlike many other variables proposed for staging, DFI is available to clinicians before definitive treatment is delivered.
4.4. Lymph Node Ratio (LNR)
Several studies have suggested that the LNR may be a superior prognosticator compared to the number of involved nodes in HNcSCC
[29][30][31]. In 2017, Vasan et al. examined the utility of LNR in a cohort of 326 HNcSCC patients with regional metastasis
[32]. For both DFS and OS, a significant decrease in survival was found for patients with a LNR >6% with almost double the risk of death after adjusting for the effects of other covariates. Surprisingly, unlike oral SCC
[33], nodal yield was not an independent predictor of survival. However, when 18 or more nodes were removed, Vasan et al. determined that a LNR >6% was a suitable threshold to stratify patients into low-risk and high-risk categories based on a 3-year OS difference of 17.5%.
4.5. Number of Nodal Metastases
The AJCC staging manual has only considered the nodal number as single versus multiple for HNcSCC. In 2019, Sood et al. found a cumulative increase in risk with each additional involved node
[34]. In a large multicentre study involving 1128 HNcSCC patients with regional metastasis, Ebrahimi et al. confirmed that number of nodal metastases was an independent predictor of survival and provided additional prognostic information compared with AJCC 8th edition
[35]. Model performance was compared against the AJCC 8th edition using C-index, PVE, AIC, and BIC. Exploratory analyses determined the optimal cut-offs to be 1–2 versus 3–4 versus ≥5 nodes with good distribution amongst these groups. Using this categorisation, the risk of disease-specific death was 1.6 times higher for the 3–4 group, and 2.9 times higher for the ≥5 group. Interestingly, the hazard ratios remained consistent despite adding the N category and overall stage in the multivariate model, suggesting that incorporating the number of nodes provides additional prognostic information to the AJCC 8th edition. When the prognostic performance of this simple categorical variable was compared to the AJCC 8th edition staging using objective measures and adjusting for institution and treatment, the categorical number of nodes was superior in to the AJCC 8th edition TNM stage and equivalent to the AJCC 8th edition N category, but with superior distribution and parsimony providing strong support for incorporating the nodal number into future HNcSCC staging systems.
4.6. Location—Parotid vs. Neck
In 2018, the results of the Trans-Tasman Radiation Oncology Group (TROG) randomized phase III trial were published, where the addition of concurrent chemotherapy (weekly carboplatinum) to PORT failed to improve locoregional control or survival in high-risk HNcSCC
[36]. The presence of parotid metastasis was considered a high-risk criterion satisfying inclusion. Mooney et al. in 2021 aimed to re-evaluate the impact of the location of nodal metastasis on survival in patients with HNcSCC
[37]. Of the 535 patients, 235 had parotid metastases in isolation, 96 had neck metastases alone, and 204 had both. On univariate analysis, disease in the neck resulted in worse DSS and OS compared to isolated parotid disease. On multivariate analysis, this finding persisted and further analysis demonstrated that patients with multiple parotid nodes behaved similarly to those with a single involved neck node and that patients with multiple cervical nodal deposits had the lowest survival.
4.7. A Hierarchical Approach to Nodal Staging
Clinicopathologic characteristics considered in the AJCC 8th edition are proven prognosticators, but the hierarchy or order in which these variables should be considered is not well understood. Recently, Hurrell et al. utilised recursive partitioning analysis (RPA), a statistical method for prioritising variables and creating homogenous groups, to determine the impact and hierarchy of these variables on a cohort of 366 HNcSCC patients with regional metastasis
[38]. The RPA for DSS revealed that the number of metastatic deposits (including STM) was of primary significance, followed by deposit size and the presence of immunosuppression together. ENE was significant at the third level of hierarchy.