pENE is commonly defined as extension of tumor cells outside the LN capsule into the perinodal soft tissue on histopathologic examination (10). It is often subcategorized as either microscopic/minor (≤2 mm in extent) or macroscopic/major (>2 mm in extent) or as a soft tissue metastasis (STM) ( Figure 1 ), as recommended by the AJCC for documentation purposes (9). The prognostic significance of the 2 mm extension threshold remains contentious, especially in the context of HPV-associated HNSCC (11–16).

Determining pENE can be complicated by the presence of an incomplete nodal capsule. This can occur as a result of sample preparation, or because the capsule has thinned and is difficult to identify, which is especially common at the hilum of the lymph node (17). When the capsule is deficient, pathologists sometimes elect to reconstruct it virtually based on the remaining evident portions, which can introduce heterogeneity and impact the reproducibility of results (12). Another clinical conundrum may occur where there is continuity of a primary tumor and an adjacent lymph node. Some pathologists may elect to diagnose ENE in this case, as it cannot be definitely excluded, while other pathologists may restrict the diagnosis of pENE to those cases where there is evidence of remnant capsule that is discontinuous with the primary tumor (10). A further challenge arises when grossly confluent LNs (also known as matted or coalescent nodes) are present. In this case, additional sections are recommended to exclude ENE, as confluent LNs may simply represent closely aggregated LNs with thickened capsules without actual microscopic evidence of ENE (10). Despite the aforementioned limitations, histopathology remains the gold standard for determining the presence of ENE, and in most cases, it can be determined and categorized as macroscopic (major) or microscopic (minor) (5).

ENE can also be visualized on pre-treatment morphologic imaging such as computed tomographic scans (CT) (18), magnetic resonance imaging (MRI) (19), ultrasound scans (US) (20) and positron emission tomography-computed tomographic scans (PET-CT) (21). Whilst the role of CT and MRI is well-established in the diagnosis of iENE, the value of PET-CT or US is questioned by many (22). There are several other issues that remain unresolved in defining or diagnosing iENE. Firstly, there is no clear consensus on the terminology, definitions or diagnostic criteria. iENE is often defined as an involved LN on imaging with an unequivocally ill-defined nodal border ( Figure 2 ), i.e., clearly discernible loss of the sharp plane between LN capsule and surrounding fat (23–26). Faraji et al. (27) performed a retrospective analysis of preoperative CT images of patients with HPV-associated HNSCC and concluded that irregular nodal margins and absence of perinodal fat plane were the most specific and sensitive features for iENE. The terms “conglomerate”, “matted” and “coalescent” have all been used to describe radiographically poorly delineated aggregates of two or more LNs, where iENE occurs between abutting nodes with loss of the intervening nodal planes ( Figure 3 , pattern 2; Figure 4A ) (5, 26, 30). However, there are limitations and some differences between these terms, even though many might use and interpret them interchangeably (31, 32). Another area of uncertainty in iENE reporting is the presence of central nodal necrosis. Some studies have recognized central nodal necrosis as a significant predictor of ENE, but others have dismissed it as an intranodal characteristic rather than a feature of ENE (33–35). It is possible that nodal necrosis is only an association with ENE since both might indicate an aggressive tumor phenotype).

Unfortunately, the use of imaging for ENE diagnosis prior to treatment is complicated by reports of low sensitivity, and poor negative predictive value ( Figure 5 ). Studies show that iENE demonstrates a sensitivity of 60-80% and a specificity of 72-96% to predict pENE (17, 36, 37). Maxwell et al. (38) assessed CT scans from 65 surgically treated HNSCC patients, with two radiologists scoring the likelihood of iENE using a 5-point scale. That method demonstrated high inter-rater variability and poor performance, with an area under the curve (AUC) of the receiver operating characteristic ranged from 0.65–0.69 (38). This has stimulated research into methods to improve iENE reporting, and as a result, various imaging features have been combined into grading or classification systems, Table 1 . Chin et al. (29, 39), Hoebers et al. (28) and Lu et al. (31) have all classified iENE into three patterns ( Figure 3 ; Table 1 ), while Ai et al. (40) categorized the presence of iENE into just two grades. All four systems identified grade 1 as iENE limited to perinodal fat only. Moreover, both Hoebers’ and Lu’s systems aligned well for grade 2 (coalescent or matted nodes) and grade 3 (ENE into adjacent structures like muscles, nerves, skin, etc.). Ai et al. omitted coalescent or matted nodes as feature from their classification (40).

Clinical ENE (cENE) was defined by the AJCC 8^th edition of the TNM staging manual as: 1) unambiguous evidence of gross ENE on clinical examination, i.e. invasion of skin, infiltration of musculature or dense tethering to adjacent structures or cranial nerve, brachial plexus, sympathetic trunk or phrenic nerve invasion with dysfunction, and 2) strong radiographic evidence of ENE to support the clinical examination (9) ( Figure 4B ). Clinical ENE will typically correlate with grade 3 in the Hoebers and Lu classification systems (28, 31, 41). As such, this designation is reserved for only the most obvious, and relatively uncommon, cases of ENE.

pENE typically indicates a poorer prognosis (42, 43). While some of the older studies concluded that there was no relationship between pENE and survival (44, 45), these studies exhibited significant weakness due to insufficient statistical power, heterogeneity in adjuvant treatment strategies, and variable pathological interpretation of pENE (46, 47). More recent studies examining the prognostic importance of pENE in HNSCC have reported up to 50% lower relative overall and disease-specific survival (DSS) for patients with pENE (43, 48–55). Several publications (43, 46, 52, 56) have demonstrated that pENE is a poor prognostic indicator for distant metastasis (DM) (pooled OR 2.18, 95% CI 1.23–3.87), and loco-regional recurrence (LRR) (pooled OR 1.33, 95% CI 0.86–2.07). Other studies have concluded that pENE is a better predictor of OS than either resection margins (50, 57, 58) or TNM staging (43, 57, 58).

Based on its prognostic importance, pENE is considered an indication for intensification in treatment protocols for HNSCC patients. A pooled subset analysis of two landmark trials [EORTC 22931 (7) and RTOG 9501 (8)] cemented the paradigm of treatment intensification for patients with pENE (6). The addition of concomitant high-dose cisplatin to RT in patients with pENE and/or positive margins reduced the risk of LRR and death by 42% and 30% respectively compared to adjuvant RT alone (6). Although these trials did not analyze pENE and positive margins separately, they still support the role of intensified treatment in pENE cases as over 50% of enrolled patients had pENE (5, 6). Importantly, this adjuvant therapy for cases with pENE attenuates the reported negative impacts of pENE on prognosis (5, 59). It should be noted however that these studies consisted mainly of non-oropharyngeal cancer and so recruited cases were highly likely to be mainly HPV-negative.

The significance of the extent of pENE remains unclear. This controversy potentially stems from the lack of a universally accepted pENE definition and is further confounded by variations in sample processing and interpretation (5, 47, 49). While macroscopic pENE consistently indicates poor prognosis in HNSCC (60–62), the prognostic significance of microscopic pENE has not been widely proven (49, 50, 60, 63). Carter et al. (63) and Brasilino de Carvalho (60) found that macroscopic pENE increased the risk of recurrence (RR 3.5, 95% CI 1.7–7.0), and worsened recurrence-free survival (RFS), but they found that microscopic pENE had no impact (RR 1.3, 95% CI 0.6–3.0). Similarly, Clark et al. (62) studied a mixed cohort of HNSCC with advanced nodal stage and reported that while microscopic pENE had a similar risk of regional recurrence as no pENE, those with macroscopic pENE fared significantly worse (RR 4.3; 95% CI 1.87–9.89). However, the authors noted that patients with microscopic pENE had intermediate DSS outcomes between those with no pENE and those with macroscopic pENE, but the authors did not report actual values (62). Moreover, Jose et al. (49) analysed survival outcomes in a mixed cohort of HNSCC patients (71% had laryngeal and hypopharyngeal cancers), and demonstrated no statistical difference in RFS between microscopic and macroscopic pENE. It is apparent that there is differential impact of pENE by its extent. However, optimal cutoff of pENE extent remains debatable. The 8^th edition AJCC/UICC (TNM8) staging manual did not include a cut-off for the extent of pENE in its definition, but they recommended documenting pENE grade as minor (≤2mm) or major (>2mm) ENE (64). Wreesmann et al. (65) and Mamic et al. (13) used ROC curve analysis to define discriminatory thresholds of 1.7 and 1.9 mm respectively in patients with oral cavity cancers, supporting the AJCC recommendation of 2 mm cutoff threshold. de Almeida et al. also showed prognostic difference in minor vs major pENE using a 2 mm cutoff (66). Similarly, Kwon et al. (11) and Arun et al. (16) found that empirically defined pENE extension thresholds of 2 mm (in a mixed cohort of HNSCC) and 5 mm (in oral cavity cancers) respectively, were able to produce significant survival differences (HR 3.8, 95% CI 2.0-7.2, and HR 95% CI). However, other studies did not support the arbitrary 2 mm extension threshold (15, 16). At the least, future efforts to better define and standardize pENE criteria would be well-served by agreeing upon a consistent terminology, i.e., major/minor or macroscopic/microscopic. Since ENE is most often diagnosed by a pathologist using a microscope and very rarely by the naked eye, even when greater than 2mm, we suggest the use of the terms major/minor as opposed to microscopic/macroscopic.

The impact of pENE on prognosis of HPV-associated OPSCC is widely questioned (67–69). Multiple retrospective single-centre studies (67, 69–79), two multi-centric studies (80, 81), one small national cancer database (NCDB) study (82), and two systematic reviews (46, 59) failed to demonstrate a negative prognostic impact for pENE in HPV-associated OPSCC treated with surgery and post-operative adjuvant therapy and suggest that the addition of chemotherapy to adjuvant radiation may not be necessary in such cases. Consequently the AJCC/UICC staging system excluded pENE for those patients (5, 64, 83) However, it must be noted that most of these studies were small and arguably inadequately powered to detect statistical significance for pENE in HPV-associated OPSCC, especially in this group of patients with significantly better survival and fewer events (5, 83, 84). Furthermore, the use of adjuvant chemoradiotherapy may have mitigated the negative prognostic impact of the pENE.

More recently, increasing evidence has suggested that pENE (especially >1 mm) in HPV-associated OPSCC does indeed negatively impact survival and regional control (83–90). Multivariate analysis of 92 p16-positive patients found pENE was an independent risk factor for overall survival (OS) and disease progression (87). Several other retrospective analyses of patients with HPV-associated OPSCC show that pENE is associated with worse survival, albeit with moderate effect size (5-11%) (84, 88–90). Importantly, a recent systematic review by Benchitrit et al. (83) pooling 1349 patients from 6 studies, concluded that pENE is associated with a relative reduction in OS of 89% (HR 1.89, 95% CI 1.15-3.13) in HPV-associated OPSCC. Moreover, the phase II ECOG 3311 trial (91) demonstrated that arm D, in which 88% of low risk HPV-associated OPSCC patients had pENE >1 mm, showed significantly poorer 2-year progression-free survival (PFS) of 90.7% (90% CI, 86.2 to 95.4) compared to arm A patients (2 year PFS= 96.9% (90% CI, 91.9 to 100%), who had no pENE, and arm B and C patients who had <1 mm pENE (2year PFS 94.9% (90% CI, 91.3 to 98.6]) and 96.0% (90% CI, 92.8 to 99.3) respectively. These differences in OS are further underscored by the fact that arm D patients received significantly more intensive tri-modal therapy (surgery, 66Gy of RT and cisplatin), compared to arm A patients (surgery only), and arm B and C (surgery and RT only at 50 or 60 Gy) (91).

Since the evidence on pENE only applies to surgically treated patients, there is great interest in the prognostic value of iENE, which could be applied to a wider cohort of HNSCC patients (5, 59, 83). In a systematic review, Benchetrit et al. (83) pooled iENE data from 1468 patients with HPV-associated OPSCC and demonstrated that iENE led to worse OS (HR 2.64, 95% CI 1.46-4.78), with a greater contribution by increased risk of distant failure (HR 3.83, 95% CI 1.88-7.80) than locoregional failure (HR, 2.03, 95% CI 0.86-4.79). A similar trend is also seen in the nasopharyngeal (30, 31, 39, 92) and HPV-negative HNSCC (41) literature: one large systematic review (93) that pooled data from 7532 patients with nasopharyngeal cancer found that iENE was associated with worse OS (HR 1.85-2.62) and distant metastasis-free survival (HR 2.07 -3.14). iENE has also been shown to be an independent factor associated with worse survival and distant control in patients with HPV-negative tumours in the oropharynx (41), oral cavity (23) and in mixed cohorts of HNSCC (94–96). Together, this literature points to a higher prognostic impact for iENE than pENE, possibly indicating that unequivocal iENE correlates with more advanced pENE (23, 83).

The recent changes to the AJCC/UICC TNM system (64) were well received by the head and neck oncologic community worldwide (104), but the criteria for “unambiguous”, clinically-overt ENE (105) remain vague. Moreover, there is significant uncertainty regarding the diagnostic criteria of pENE, leading to heterogeneity when making a diagnosis. This may be contributing to the conflicting evidence commonly encountered in pENE research. There is still no consensus among pathologists on a preferred terminology for pENE, and a constellation of terms like extranodal extension, extracapsular spread, or extranodal spread are used interchangeably. Furthermore, while most pathologists will agree that extension of tumor cells in the perinodal fat and soft tissue is diagnostic for pENE, there is still significant uncertainty around determining pENE in challenging cases with matted nodes, nodal hilar involvement or in cases with direct extension of primary tumor into a node. Moreover, there is still no agreement between pathologists whether HPV status should be taken into consideration when interpreting pENE features. These challenging issues are best addressed first by consensus, to standardize the definitions, terminologies and synoptic reporting used for pENE, especially for microscopic versus macroscopic, in both HPV-associated and HPV-negative HNSCC. The pathology community should also come to agreement on a standardized lymph node processing and sampling methodology for pENE (5, 10). This may help facilitate much needed research on the prognostic power of pENE in both HPV-associated and -negative HNSCC.

The importance of a common language for defining iENE at the time of diagnosis has been poorly addressed so far, and the lack of standardized definitions and nomenclature has contributed to conflicting evidence, both in clinical trials and real-world data (5). However, a rational roadmap to develop a standardized nomenclature for iENE faces some challenges. Widespread agreement in the radiology community regarding the diagnostic criteria for ENE on imaging is still lacking. Features like nodal size, central nodal necrosis and capsular thickening are still being debated as criteria for iENE. Moreover, there is still no consensus regarding interpreting findings like matted/coalescent nodes, or the impact of HPV status on iENE features. Furthermore, there is still no conclusive evidence regarding the best imaging modality for iENE identification. There are several published classification systems for iENE in head and neck cancer as shown in Table 1 , but none of these systems have been widely adopted in routine clinical practice. Research into assessing, improving and validating these systems is needed, to enable wide adoption into clinical practice and there is a need for better appreciation of the impact of ENE, at least in its worst form, on the outcome of patients rather than disregarding it.

Thus, there is a pressing need for standardized diagnostic criteria for iENE, which could improve reproducibility and facilitate research and widespread clinical implementation. In our view, an international consensus process aiming to standardize the iENE criteria and address these gaps in the literature on iENE is needed.

Robust and large-scale studies are needed to quantify the prognostic impact of pENE and iENE in HPV- associated and HPV-negative HNSCC. Such studies need to be adequately powered to definitively address the prognostic significance of the different grades of pENE, and to validate the commonly used 2 mm threshold, especially in HPV-associated tumors. This could then be integrated into the TNM system, de-escalation trials, and everyday practice.

Recent advances in artificial intelligence (AI) may hold promise for use in the diagnosis of iENE and outcome prediction. Kann et al. (106) trained a 3-dimensional convolutional neural network using 2875 CT-segmented LN samples, correlated with pathology samples to act as a ground truth. They demonstrated an improvement in the AUC to 0.91 with a sensitivity of 88% (false negative rate: 12%), and specificity of 85% (false positive rate: 15%) (106). They later validated this approach using two external cohorts, consisting of a total of 200 LNs (107). The algorithm achieved an AUC of 0.84 (83.1% accuracy) and 0.90 (88.6% accuracy) in the two cohorts, outperforming two independent radiologists’ AUCs of 0.70 and 0.71 in the first cohort, and 0.60 and 0.82 in the second cohort respectively. The diagnostic accuracy and inter-rater variability of both radiologists improved when they were supported with deep learning assistance. Ariji et al. (108) also developed a deep learning algorithm and compared performance to radiologists. Once again, the deep learning system achieved high accuracy (84.0%) for diagnosing iENE, using a set of AI-determined features. In comparison, the radiologists’ accuracies based on a set of radiological criteria - minor axis ≥ 11 mm, central necrosis, and irregular borders- were 55.7%, 51.1% and 62.6% respectively (108). These efforts are still in early stages of development and will need to undergo wider external validation before routine implementation in clinical practice.

In the last decade, advances in biomarker technologies have led to multiple discoveries in HNSCC diagnosis and prognosis. There are currently several promising molecular biomarkers that could potentially be used for predicting pENE before commencement of treatment. However, these are still in early stages of development, with a high rate of false discoveries (46, 109). External validation in larger cohorts, and in some cases, better biomarkers are needed to confirm these associations and their clinical impact before being incorporated into clinical treatment paradigms.