Keywords

Introduction

The lymph node (LN) capsule is a natural barrier for tumor progression. When tumor breaches the LN capsule (termed extranodal extension [ENE] by the 8th edition TNM [1, 2]), it presents a challenge to disease clearance regionally but more importantly, also augments risk of distant metastasis (DM). Presence of ENE could be a consequence of long-growing ignored tumor, but more likely represents an aggressive tumor phenotype [3].

ENE can be detected on pathology specimens, inferred from imaging, or be indirectly evident via clinical examination. Extent of ENE reflects incremental tumor invasion. The initial stages of ENE can only be detected under the microscope (namely pathologic ENE, pENE). When ENE continues, it can eventually become visible on radiologic imaging (namely radiologic ENE, rENE). When ENE further progresses to invade skin with hallmark changes of skin ulceration or dermal edema (e.g., peau d’orange) and/or adjacent soft tissue structures (e.g., muscles, nerve, and vessels) causing fixation and neurovascular impairment, it will result in obvious clinical features consistent with ENE (namely clinical ENE, cENE).

ENE in head and neck cancer (HNC) was first described by Willis [4] in autopsy material from a head and neck epidermoid carcinoma in 1930. Its prognostic importance was confirmed in subsequent studies [5,6,7]. Convincing evidence demonstrates that ENE is one of the strongest prognostic factors for both viral-related and unrelated HNC. Unequivocal rENE carries prognostic significance beyond traditional cN classification and has the potential for risk stratification and future N classification [8,9,10,11,12,13,14,15]. Detection of ENE may also directly impact clinical care and treatment planning. If ENE can be identified before surgery, it can help predict the likelihood of needing more intense adjuvant approaches including triple modality treatment (postoperative chemoradiotherapy, postop-CRT) due to presence of pENE. Therefore, early recognition of its presence can triage appropriate treatment recommendations. This is especially relevant for many HPV-positive oropharyngeal cancer (OPC) patients where equipoise concerning disease control has emerged between primary (chemo-)radiotherapy (RT/CRT) and transoral surgery (TOS) due to an important focus on functional preservation. However, the sensitivity of rENE for pENE remains unsatisfactory.

In this chapter, we summarize pathological, radiological, and clinical signs of ENE and their relationships. Since rENE may have a broader implication in pre-treatment risk stratification and treatment selection, we propose a means to augment sensitivity and specificity of rENE for pENE detection. Finally, we review emerging data on biomarkers that are associated with ENE.

Pathological, Radiological, and Clinical Signs of ENE

ENE refers to tumor invasion through the nodal capsule into perinodal fat and beyond. It can invade through a single LN or involve 2 or more adjacent LNs to form a coalescent nodal mass. It can also destroy the entire nodal structure and manifest as a soft tissue deposit within nodal regions without associated clearly identifiable LN(s).

The extent of pENE has been categorized differently by various authors [2, 16,17,18,19]. Carter et al. [16] in 1985 classified ENE as “microscopic” (microscopic breaks in the lymph node capsule, only evident on histologic examination) versus “macroscopic” (spread of tumor into identifiable structures within the specimen) pENE. The latter may also be evident with clinical and radiological assessment. However, this classification is somewhat rudimentary since much “microscopic” ENE also carries prognostic significance. Yamada et al. [19] later classified pENE into three types: “Type A”—few tumor cells outside the LN capsule; “Type B”—microscopic invasion of the tumor cells into perinodal fat tissue, with capsular destruction, and “Type C”—macroscopic tumor invasion into perinodal fat or muscle tissue. However, this classification is ambiguous in practice since specific descriptions of the pathologic assessment of “microscopic” vs “macroscopic” were not provided. Lewis et al. [18] classified pENE into four grades: “Grade 1”—tumor reaching LN capsule with thickening of the overlying capsule; “Grade 2”—tumor extending ≤1 mm into perinodal issue; “Grade 3”—tumor extending >1 mm beyond nodal capsule; and “Grade 4”—soft tissue deposit without residual nodal architecture. The latter is probably related to the effacement of the entire nodal capsule by tumor or due to tumor foci escaping from the lymphovascular pathway. However, “Grade 1” pENE category in this classification does not truly reflect the essence of ENE. The 8th edition American Joint Committee on Cancer (AJCC) TNM (TNM-8) [1, 2] recommended directly measuring the distance from the breached nodal capsule to the farthest extent of tumor to quantify pENE extent as “microscopic ENE (micro-ENE)” (≤2 mm) versus “major-ENE” (>2 mm) (Fig. 7.1A–C). When tumor destroys the entire nodal architecture with only a soft tissue deposit in the neck tissue, it represents the most advanced form of pENE and should be considered as “major-ENE” (Fig. 7.1D).

Fig. 7.1
A model diagram illustrates four extents of pENE described in the literature.

Schematic depiction of various extent of pENE described in literature

Full size image

The cutoff of “micro-pENE” versus “major-pENE” varies in the literature. For example, the ECOG 3311 trial [20] used a 1 mm cutoff where ≤1 mm pENE is considered “intermediate” risk and eligible to receive postoperative RT alone with either 50 Gy or 60 Gy in HPV-positive OPC following TOS. Wreesmann et al. [21] used receiver operator curve (ROC) analysis at specific time points and identified a prognostic cutoff for ENE extent at 1.7 mm in oral cavity squamous cell carcinoma (OSCC). Similarly, Mamic et al. [22] reported a 1.9 mm cutoff by ROC analysis for prognostically important ENE in 174 cN0 OSCC who underwent surgery with elective neck dissection. Arun et al. [23] found that a 2 mm cutoff did not show prognostic significance in 212 OSCC patients, whereas a 5 mm cutoff demonstrated significant differences in overall survival (OS) and disease-free survival (DFS).

Similar to pENE, the grading of rENE definition is also evolving [13, 24,25,26,27]. It is now recognized that rENE can manifest in any individual LN or affect multiple adjacent LNs to form an inseparable nodal mass. Chin et al. [28] recently proposed clearly defining rENE extent into three grades: “Grade 1” rENE—tumor breaching the nodal capsule of an individual LN characterized by unambiguously ill-defined nodal border(s), but confined to perinodal fat; “Grade 2” rENE—tumor invasion through two or more inseparable adjoining LNs exhibiting unambiguous effacement of any component of their internodal plane(s) (implying replacement by tumor) [1], invariably resulting in a lobulated appearing nodal mass; “Grade 3” rENE—tumor invasion beyond perinodal fat to overtly invade or encase adjacent structures, e.g. skin, muscle, and neurovascular structures (Fig. 7.2). Interestingly, other terms, such as “conglomerate” and “matted”, have been used to describe an aggregation of multiple juxtaposed LNs, without necessarily adhering to or fusing into each other, to form a compact mass. We prefer the term “coalescent” to describe two or more adjoining LNs consuming each other into an inseparable mass; this is characterized by unequivocal effacement of internodal planes that forges multiple LNs into a single entity.

Fig. 7.2
A table reprsents definitions of grade 1, 2 and 3 rENE, their schematic depiction and radiologic examples.

Definition and extent of radiologic extranodal extension

Full size image

cENE has been introduced as a new N-classifier in the TNM-8 for non-viral related HNC [2]. It represents the most overt form of ENE and refers to detectable ENE by clinical examination. When ENE is advanced, clinical signs emerge. Peau d’orange is a clinical sign of dermal infiltration with edema and, along with ulceration, is indicative of tumor invading skin. “Fixation” of a nodal mass during palpation is a clinical sign of tumor infiltration of deeper fascial structures and musculature. Brachial plexopathy is often a sign of tumor invasion to neural structures, but like any clinical findings, cENE should be interpreted in context. Thus, cENE can also be subjective. Fixation of an upper neck mass can sometimes be caused by advanced primary tumor extension rather than a nodal mass [29]. Therefore, TNM-8 mandates cENE to be supported by rENE [2].

Sensitivity and Specificity of rENE for pENE

Since pENE is identified microscopically, it is regarded as the most sensitive and objective way of identifying ENE; comparatively, subjectivity exists for rENE and cENE detection. Hence, pENE often serves as a gold standard to examine the accuracy of rENE and cENE. It is understandable that not all pENE would have rENE since the method of assessment differs significantly, using the “naked eye” on one hand compared to the microscopic in the other. Studies in HPV-negative HNC showed that only about 50% pENE may have signs of rENE [10] and only about 50% rENE could have evidence of cENE [29]. Despite low sensitivity, when more stringent criteria are used for rENE declaration, the specificity is generally high [8].

The sensitivity and specificity of rENE for pENE depends on pENE extent, rENE grade and the level of certainty a radiologist has adopted for declaration. Blasco et al. [30] showed that the sensitivity of rENE is higher in identifying major pENE compared to micro-pENE. Chang et al. [31] found that rENE carried a higher hazard ratio (HR) for OS compared to major-pENE or micro-pENE (2.27 vs 1.06 vs 0.49). Hu et al. [24] showed that the higher the certainty of rENE declaration, the higher the inter-rater concordance; and a higher grade of rENE is associated with less ambiguity for a radiologist to declare rENE. Of importance, “suspicious” rENE did not carry prognostic difference vs no rENE while those with a high certainty of rENE patients had lower distant control (DC).

Studies have consistently shown that rENE is one of the strongest prognostic factors for survival in HPV-negative HNC (HR 1.3–3.3), nasopharyngeal cancer (HR 1.6–1.9), and HPV-positive oropharyngeal cancer (OPC) (HR 3.9). A recent meta-analysis showed that rENE-positive HPV-positive OPC patients suffered a much higher risk of death (HR 2.6) versus rENE-negative HPV-positive OPC, mainly related to increased risk of DM (HR 3.8), and the HR of rENE for survival was even higher than that of pENE (HR 2.6 vs. 1.9). This is likely because, on average, unequivocal rENE recognizes a worse version of extranodal disease than pENE which includes cases with minimal extent of ENE lacking the same prognostic significance under contemporary treatment. Therefore, rENE can serve an important role in clinical care and risk stratification in HNC and has been used as an exclusion factor in some de-intensification trials.

The priority of sensitivity over specificity and vice versa depends on the clinical scenario. For clinical care, such as triaging cases of T1-2N1 OPC to surgery vs radiotherapy, a high sensitivity of rENE in identifying pENE would be important to avoid potential triple modality treatment which is ordinarily recommended if pENE is detected; therefore, a relatively modest level of certainty (>50%) may be used for declaration of rENE before treatment assignment. For staging, the preservation of prognostic importance to minimize dilution due to the inclusion of uncertain cases (or those with less important minimal disease) is needed, and a high level of certainty (>90%) should be maintained for rENE declaration.

Radiology assessment has always played an essential role in clinical decision making and staging of HNC patients. Standardization of taxonomy describing nodal features and the certainty for their declaration will help facilitate clear communication and interpretation of radiology reports. Chin et al. [28] studied interrater concordance by two radiologists assessing 7 nodal features frequently highlighted in the literature in 413 HPV-positive OPC patients and found that variation existed in interpretation regarding radiologic nodal features. Clearly defined nomenclature results in improved interrater reliability when assessing radiologic nodal features, especially for coalescent adenopathy and ENE. A multicenter study by Hoebers, et al. [32] showed that a learning curve exists for rENE assessment. Reliability of rENE assessment across institutions improved after consolidation of rENE operational definitions. Higher levels of certainty were associated with higher inter-rater agreement. The authors propose a strategy to augment the reliability of rENE ascertainment including: high certainty for declaration, consolidating operating definitions, and sharing experience among the radiology community [32].

Artificial Intelligence and Machine-Learning in Identifying rENE

Recognizing the limitations of imaging interpretation by human eyes, some researchers have turned to artificial intelligence and machine-learning with automated detection algorithms to improve interrater concordance [33]. Kann et al. [33] trained on a dataset of 2875 CT-segmented LN samples with corresponding pENE data and derived an algorithm which predicted pENE with area under the receiver operating characteristic curve (AUC) of 0.91 (95%CI: 0.85–0.97). The subsequent validation study from two different datasets (one from an external institution and the other from The Cancer Genome Atlas (TCGA) Head and Neck Squamous Cell Carcinoma imaging data repository) and showed improved AUC compared to radiologists’ assessment. Moreover, the diagnostic accuracy of the radiologists improved when receiving assistance from the detection algorithm.

Although Kann’s work shows promise of artificial intelligence in enhancing sensitivity and objectivity in recognizing rENE and improving rENE-pENE correlation, it is not ready for routine clinical practice. In part this is because it relies on modelling processes for prediction of a status among a group of patients, rather than declaring its presence in an individual case. For deep-learning performance evaluation, the authors used the primary endpoint of AUC. Whether AUC is the optimal endpoint for developing models to guide clinicians on treatment recommendation is uncertain since AUC measures the overall “goodness-of-fit” of the model. In clinical situations, specific requirements often dictate the priority for either high sensitivity or high specificity and AUC rarely provides adequate information in this regard. If the objective is to avoid tri-modality treatment in TOS-eligible patients, identifying ENE before treatment with high sensitivity is important. Conversely, a false-positive declaration of ENE may prompt chemoradiotherapy when surgery-alone may have been sufficient. One can argue for a high sensitivity test to identify any ENE to avoid tri-modality treatment. In contrast, one could also argue for a high specificity test to identify only those cases bearing prognostically important major ENE for staging and treatment recommendation. This fits the quintessential staging rule of the UICC and AJCC: when there is doubt, a lower stage (i.e. less ominous prognostic level) should be assigned [34].

Biomarkers Associated with ENE

ENE has been proven to be associated with aggressive phenotypes in many cancers including HNC. Several biomarkers (whether protein, RNA, DNA, or epigenetic markers) have been reported to be associated with presence of ENE [3] and mostly in OSCC population. Podoplanin is a small mucin-type transmembrane glycoprotein that promotes local invasion and metastasis through the regulation of tumor cell migration and epithelial–mesenchymal transition [35, 36]. Lee et al. [37] recently found that almost all (93%) ENE-positive OSCC patients had podoplanin expression in the peri-nodal stroma of metastatic LNs compared to 47% in ENE-negative patients, and the intensity of podoplanin was also higher in ENE-positive patients. Noda et al. [38] examined clinical features associated with the tumor microenvironment in 186 surgically treated OSCC patients and 83 matched biopsy specimens, and found that ENE-positive patients had a high tumor budding pattern, low tumor-infiltrating lymphocytes, and an immature desmoplastic reaction in the primary site. Gieber-Netto et al. [39] observed that tumors carrying high-risk TP53 mutations had a significantly increased risk of developing ENE in OSCC. Similarly, Sandulache et al. [40] analyzed TCGA OSCC dataset and found that pENE-positive patients had the highest proportion of high-risk TP53 mutations while wild-type TP53 was highly representative in pN0 patients. Dhanda et al. [41] performed microarray and immunohistochemistry staining of primary tumors in 102 OSCC patients and found that high or intermediate expression of both SERPINE1 and SMA at the primary site was strongly associated with the presence of ENE. In addition, expression of both SERPINE1 and SMA was associated with poor OS.

It is important to point out that all these studies are retrospective and conducted on OSCC sites. It is unclear if these observations can be replicated in more diverse disease sites.

Clinical Implication of ENE

Compelling evidence demonstrates that the presence of pENE carries prognostic significance in both HPV-negative and HPV-positive HNC, and mainly affects DM [3, 9, 37]. As mentioned earlier, recent data shows that micro-pENE and major-pENE have different prognostic importance [17, 20, 21]. Wreesmann et al. [21] showed that OSCC patients with ≤1.7 mm pENE had similar 5-year DFS versus no ENE while patient with >1.7 mm pENE had much lower DFS. De Almeida et al. [17] analyzed 348 OSCC patients and found that patients with micro-pENE (≤2 mm from the nodal capsule) had no difference in locoregional control (LRC, 72 vs. 74%, p = 0.86) compared to patients with no ENE, while DC was only slightly lower (80 vs. 86%, p = 0.17). In contrast, patient with major-pENE had significantly worse 5-year OS (16 vs. 45%, p = 0·002), DFS (15 vs. 42%, p = 0·004), and DC (58 vs. 80%, p = 0·005) compared to patients with minor ENE, although LRC was only marginally worse (61 vs. 72%, p = 0·07). More importantly, the addition of cisplatin chemotherapy improved DFS for patients with major-pENE but had no impact in patients with micro-pENE. The effect of cisplatin chemotherapy on DFS was mainly due to enhanced LRC but not DC. The ECOG 3311 trial [20] result shows that HPV-positive OPC with ≤1 mm pENE can be safely treated with reduced dose postop-RT without chemotherapy. Emerging data suggest that differentiating micro-pENE from major-pENE might impact treatment choice in the future, although more robust trial data are warranted.

Data on the prognostic value of rENE are also emerging. Almulla et al. [8] showed that within pENE-positive patients, rENE-negative status had less prognostic importance than rENE positive status. The meta-analysis of HPV-positive OPC by Benchetrit et al. [9] showed that both pENE and rENE were prognostic and mainly impacted on DM rather than LRC. Emerging data has consistently show that ENE is one of the most powerful prognostic factor for all HNC including OSCC [8], HPV-negative OPC [29], HPV-positive OPC [10, 27], and nasopharyngeal cancer (NPC) [12, 13, 24]. In viewing the prognostic importance of rENE, many authors have now proposed to include rENE in pre-treatment risk stratification and future staging [12, 13, 15, 27, 29]. Some clinical trials have also included rENE as an exclusion criterion when designing studies addressing treatment deintensification [42, 43].

As mentioned earlier, ENE affects DFS, mainly through increased risk of DM [17]. Effective systemic agents to address DM are needed. However, cisplatin, the most commonly used chemotherapeutic agent, seems insufficient to address this risk. Felon et al. analyzed a NCDB dataset of 14,071 HPV-positive OPC undergoing TOS treatment and found that both micro-pENE and major-pENE patients had inferior OS, but addition of cisplatin did not improve OS of pENE-positive patients although disease specific endpoints (locoregional or distant failure) were not reported. Huang et al. [27] also showed that cisplatin chemotherapy could improve LRC in HPV-positive OPC but could not fully negate the high DM risk of being rENE-positive.

Conclusion

Convincing evidence exists that ENE is one of the strongest prognostic factors for both viral related and unrelated HNC. It is mainly associated with a higher risk of DM with some additional influence on risk of locoregional recurrence. cENE and pENE are new N-category modifiers for non-viral HNC. Minor and major pENE may have different clinical importance. Similar to pENE, rENE should have a promising role in risk stratification of HNC. However, more work is needed to improve reliability of rENE assessment. Radiology reporting rENE needs to consider both sensitivity and specificity. To avoid inadvertent intense treatment combinations, including triple modality treatment, and to optimize treatment recommendations upfront, high sensitivity is important; to avoid falsely up-staging to the detriment of prognostication, high specificity is important. A standardized radiologic taxonomy and reporting template is warranted. Cisplatin appears to have insufficient effect in negating DM risk associated with ENE, novel systemic agents are needed to better address risk of DM in patients with ENE.