The impact of histology and molecular biology in refining the prognostic value of surgical margins in head and neck oncology – A narrative review

Cover figure: (A) Surgical specimen of a right hemiglossopelvectomy. The tongue is oriented using a suture to indicate the tip; (B) Surgical specimen after formalin fixation, with resection margins marked in different colours of India ink to indicate the midline raphe, the margin on the floor of the mouth, the posterior and the anterior margin (the tip). The margins were assessed on the surgical specimen (sample-driven), which was sectioned using the “radial and perpendicular” method; (C) A macroscopic section of the lingual body obtained by radial sectioning. (D) The corresponding whole slide image stained with haematoxylin and eosin.

The intraoperative evaluation of surgical margins (SM) using frozen section analysis is a widely adopted practice in surgery for head and neck cancer (HNC), offering real-time pathological assessment. However, its reliability is not without challenges, as inconsistencies in specimen processing, tissue shrinkage, and interobserver variability can impact its accuracy. This review explores key aspects of SM assessment in HNC, discussing different approaches, particularly dealing with intraoperative assessment and comparing sample-driven and tumour bed-driven approaches. It examines the accuracy of these techniques and their prognostic implications, especially in cases where a positive frozen section analysis leads to revision of margins. Another important factor discussed is the impact of tissue artefacts on margin interpretation, alongside practical recommendations for assessing bone margins. A crucial point emphasised throughout is the need for strong collaboration between surgeons and pathologists to enhance the reliability of intraoperative margin evaluation. 
Beyond traditional histopathological methods, recent advancements in imaging techniques are also considered. Technologies such as narrow band imaging, Raman spectroscopy, optical coherence tomography, hyperspectral imaging, computer-assisted surgery, and confocal microscopy are gaining attention for their potential to refine margin assessment and improve surgical precision. Looking ahead, the focus should be on refining intraoperative SM assessment techniques, integrating artificial intelligence-driven imaging solutions, and developing standardised guidelines through large-scale meta-analyses. A multidisciplinary approach that brings together pathology, radiology, and surgical expertise will be crucial in optimising tumour resection strategies and help to reduce the risk of local recurrence.

Introduction

Achieving an initially negative surgical margin predicts the optimal prognostic outcome for patients with head and neck squamous cell carcinoma (HNSCC), and surgical margins (SM) represent the only factor potentially controllable by the surgeon. Negative margins indicate that the tumour has been completely removed and serves as a guide for postoperative adjuvant treatments. However, the issue of margin adequacy, and the definition of “negative” margins, are objects of longstanding scientific debate. A previous American Head and Neck Society (AHNS) survey highlighted the spectrum of surgeons’ opinions regarding margin adequacy, with nearly half defining a clear margin as > 5 mm ¹. In contrast, SM free from tumour but measuring < 5 mm are defined as “close”.

Patient prognostic stratification according to conventional margin status is supported by robust and large literature data, as summarised by a recent meta-analysis on 8,435 patients from 26 studies, most of whom with oral squamous cell carcinomas (OSCC) (Cover figure and Figure 1). Hamman et al. reported a significant stepwise improvement in local recurrence (LR) and overall survival (OS) relative risks as margin status categorically improved from positive to close to clear². Stratification by margin status defined as negative, close, and positive according to the cutoff of 5 mm has been introduced in the AJCC Staging Manual (8^th edition) and in international guidelines. It is important to note that, while HNSCC included in the validation studies covered all head and neck subsites ³, albeit with a disproportionate prevalence of OSCC, the AJCC formally adopted the 5 mm cutoff for close tumours only for OSCC, and that most published literature exploring its strengths and limitations focuses on this tumour site ².

Recent studies have challenged the 5 mm cut-off to reach a more precise definition of safe margin distance and negativity threshold. A recent review summarised the results of the most relevant studies exploring optimal cutoff values for microscopic surgical distance in OSCC with reference to different outcome measures (2-, 3-, or 5-year local control [LC], OS, LR, loco-regional recurrence [LRR], loco-regional control [LRC], disease-specific survival [DSS], or loco-regional recurrence free survival [LRRFS]) ⁴. Significant cutoff values ranged between 1 and 7.6 mm, confirming that consensus on the extent of safe margins is far from reached. A meta-analysis in 2023 including 7 studies concluded that margins between 4 and 4.9 mm had a similar LR risk ratio compared with margins ≥ 5 mm, while margins < 4 mm had significantly higher risks ⁵. No pooled data are available for other outcome parameters, including survival, due to the high heterogeneity of the available studies. Interestingly, recent studies have found that treating SM as continuous, rather than categorical, variables improves the correlation with survival ^6,7, as well as with recurrence ⁸.

Controversies exist on how to treat positive (and possibly close) margins, be it by re-excision in case of intraoperative identification of positive margins, or adjuvant therapy, as originally proposed by the EORTC and RTOG studies ^9,10. This is mostly due to contradictory results obtained by works aimed at assessing the prognostic impact of such decisions. There are several technical reasons beyond intrinsic tumour behaviour that may explain the lack of robust evidence in this field; among them, the limited standardisation of sample handling by surgeons and pathologists and of tissue processing plays a crucial role: in particular, it is virtually impossible to quantify and standardise the amount of thermal damage and shrinkage suffered by the tissue in the pre-analytical phases. Moreover, the approach to the evaluation of SM is influenced by the practice of the surgical team and pathologists, with particular reference to the sampling of margins from the tumour bed or from the resection specimen and could benefit from the application of recently developed technologies, including artificial intelligence (AI).

We will cover the above topics in the following sections to provide an overview of the status of margin assessment in HNSCC.

Dysplastic margins

The impact of dysplasia at margins of HNSCC resection is not well understood. Most respondents to the AHNS survey ¹ considered carcinoma in situ as a positive margin, and 76% considered dysplasia at the margin as a negative SM. On the other hand, the College of American Pathologists guidelines recommend that moderate and severe dysplasia at SM should be considered as positive margins ¹¹. A recent large retrospective study on 565 resected OSCCs showed that dysplastic and close SM did not significantly affect LRR or OS in comparison with negative SM ¹²; however, patients with severe dysplasia at margins had a higher recurrence rate than those with mild and moderate dysplasia, even if more than half received adjuvant radiation therapy ¹². Other studies reported that dysplastic SM was not a significant risk factor in OSCC and that the severity of dysplasia had no impact ¹³. Senarath et al. ¹⁴ recently reviewed articles correlating outcomes with the presence of dysplastic margins and concluded that moderate to severe dysplasia at the margin required re-excision. Overall, due to the lack of definite evidence, surgeons tend to address SM with high-grade dysplasia on a case-by-case basis, highlighting the need for definite guidelines on this topic and the importance that pathologists precisely document the presence and grade of dysplasia on margins.

Margins at extra-oral sites

There is a lack of conclusive evidence regarding SM beyond the oral cavity. The anatomical differences among subsites in the head and neck region make it difficult to directly compare the clinical significance of a margin greater than 5 mm across these areas ¹⁵. A key question remains: is achieving a margin greater than 5 mm feasible in the oropharynx, larynx, and hypopharynx? Several studies have indicated that the anatomical constraints of the head and neck region limit SM due to the thinness of various tissue layers ¹⁶. In the literature, only a limited number of studies have specifically addressed SM in total laryngectomy/laryngopharyngectomy procedures ^17-23 (Fig. 2). Among them, only one adhered to the Royal College of Pathologists (RCP) guidelines ²⁴. Two studies classified close margins as being less than 5 mm and considered margins as positive when tumour cells were present at the resection surface ^22-25. The remaining studies described margin status using qualitative terms rather than precise numerical values, employing classifications such as ‘positive’, ‘microscopically positive’, ‘tumour at the resection surface’, ‘negative’, ‘safe margins’, or ‘no invasive tumour at the resection surface’. Due to these inconsistencies, determining the clinical relevance of SM in the larynx and hypopharynx remains challenging. Achieving consensus on the prognostic significance of SM at these sites would require meta-analyses of studies that share standardised definitions, methodologies, and postoperative treatment strategies ²⁶. From a practical point of view, for glottic cancers a smaller margin of 1-2 mm is considered sufficient with particular regards to transoral laser resections ^27,28. As part of the Department of Veterans Affairs Laryngeal Cancer Study, one paper reported 50% survival outcomes for clear margins, 57% for close (less than 5 mm) margins, and 27% for involved margins in laryngeal SCC ²⁹. In transoral endoscopic and robotic approaches for oropharyngeal cancers, margins of 1.5-2 mm are considered acceptable⁴. In hypopharyngeal tumour, submucosal patterns of spread could predispose to inadequate margins ³⁰. A study has shown that the propensity for submucosal tumour extension was least in the superior direction (3-10 mm) and highest in the medial (2-37 mm), lateral (2-37 mm), and inferior direction (3-35 mm) ³¹. The same group has also emphasised adequate radial clearance and suggested that a radial clearance of less than 1 mm is an independent prognostic factor for OS, DFS, and regional RFS ³². Hence, greater margins up to 30 mm have been recommended in hypopharyngeal tumours. To integrate the minimal free margin proposed by the Royal College of Pathologists guidelines and establish it as a standard of care for laryngeal and hypopharyngeal cancer, further investigation into its clinical relevance is essential.

Intraoperative evaluation of tumour margins

The intraoperative assessment of SM by frozen section analysis is a widely accepted and commonly utilised approach in HNSCC surgery. Its reliability and clinical value have been well demonstrated. The 2024 NCCN guidelines report: ‘Frozen section margin assessment is always at the discretion of the surgeon and should be considered when it will facilitate complete tumour removal. Achieving adequately wide margins may require resection of an adjacent structure in the oral cavity or laryngopharynx such as the base of the tongue and/or anterior tongue, mandible, larynx, or portions of the cervical oesophagus. This technique enables rapid pathological evaluation during surgery ³³. Ideally, margin assessment should follow a uniform, standardised approach for all cases. However, specific standards for head and neck procedures have not yet been established. Findings from the survey “Critical evaluation of frozen section margins in head and neck cancer resections” indicate that there is no universally standardised approach for frozen section margin analysis in head and neck surgeries. According to the responses, 32% of participants reported using multiple processes for intraoperative margin assessment, while 18% stated that more than one method was applied for final margin evaluation, although this step was determined by the pathologist. The approach often varies by institution and tends to be internally developed to fit specific institutional practices ³⁴. Additionally, many pathologists face uncertainty regarding the exact areas sampled for frozen section analysis and the margins that require further assessment. To address these challenges, various intraoperative techniques and strategies for SM evaluation have been developed ³⁵ (Tab. I).

Defect-driven vs sample-driven

There is currently no evidence-based data confirming that intraoperative margin assessment improves local control. Conducting research is challenging due to the variability in margin sampling, evaluation, and reporting; however, it is clear that margins can be assessed from the specimen, referred to as specimen margins or sample-driven margins (SDM), or from the tumour bed, known as tumour bed margins (TBMs) or defect-driven margins ^1,34,36. In the SDM approach, the entire resected specimen is sent to the pathologist for margin assessment. Conversely, in the defect-driven approach, the surgeon samples the tumour bed, collecting margins from the wound, or cavity, after completing the resection. In the latter method, the selection of TBMs, including their size, location, and quantity, is not directly guided by the formal margin assessment of the resected specimen. The debate extends beyond the method of margin sampling to the reliability of peripheral margin interpretation. In fact, there is a weak correlation between margins taken from the surgical bed and those directly sampled from the resected specimen. Most studies on margin assessment overlook the distinction between these two approaches and fail to consider their relationship. Research suggests that prognostically significant margins are those derived from the actual resection specimen, while tumour bed sampling alone has been associated with poorer local control ^37,38. In a retrospective multicentre cohort study, Maxwell et al. found that among 95 patients relying solely on TBMs, the sensitivity was just 24.2% (95% CI: 16-34%), whereas the specificity reached 92% (95% CI: 85-97%) when SM were used as the reference standard ³⁹. Similarly, Buchakjian et al. reported that TBMs had a 35% sensitivity in detecting positive specimen margins ⁴⁰. Another study by Chang et al. found that TBMs failed to identify positive glossectomy margins in 67% of cases ³⁷.

In evaluating the two types of margins, an economic aspect also emerged. The assessment of intraoperative margins has been associated with prolonged surgical times and increased costs ⁴¹. Moreover, these additional expenses can extend beyond the operating room, as false negative results may necessitate further therapeutic escalation ⁴². The current 2024 NCCN guidelines do not specify which type of margin between SDM and TBM is best to use intraoperatively with frozen sections ³³.

TBM sampling remains widely used because of its efficiency and ease of execution, enabling surgeons to influence the pathology report. In a 2021 AHNS survey conducted among head and neck surgeons, 45% still favoured the tumour bed sampling method ⁴³. This preference may stem from its technical advantages, as it is a faster approach that allows surgeons greater control over the processing of pathological specimens. However, a survey conducted among Canadian surgeons highlights that a slightly higher number of surgeons preferred a tumour-driven approach ⁴⁴.

Over the years, efforts have been made to refine this technique by incorporating TBM orientation. Van Lanschot et al. explored the feasibility of a similar approach using paired markers placed on both the specimen and the surgical bed. The primary specimen underwent intraoperative analysis by both surgeon and pathologist. Based on these assessments, additional resections were carried out. As a result, margin status improved in 28 of 31 cases, and surgeons reported that identifying the positive SDM from the specimen to the wound bed became easier ⁴⁵.

Villemure-Poliquin et al. evaluated the accuracy of an oriented TBM technique in comparison to an SDM technique ⁴⁶. Their findings indicated that the use of an oriented TBM approach did not negatively impact LC, LRC, or DFS when compared to SDM. Moreover, the oriented TBM method demonstrated higher sensitivity than the SDM technique, with rates of 50% and 30%, respectively. The effectiveness of this approach improved when performed by the surgeon who originally developed the technique, suggesting that its application may vary depending on the operator’s experience and consistency. These findings are favourable compared to previous studies, which reported lower sensitivity values, ranging between 15% and 32% ^39-47 (Tab. II).

Accuracy assessment

The interpretation of a negative TBM remains unclear ⁴⁸. Unlike SDM, which can be categorised as positive, close, or negative, TBM is assessed in a binary manner, either positive or negative. Additionally, frozen sections taken from the tumour bed do not allow precise distance measurements as it would be on a resected specimen, and for this reason interpreting frozen margins remains challenging. If an SDM is negative and the corresponding TBM is also negative, the margin is unequivocally classified as negative. However, ambiguity arises in other situations. For instance, if a specimen has a close margin (ranging from 2 to 5 mm) but a negative TBM, should it still be considered a close margin? Similarly, if a TBM is initially positive but turns negative after multiple reassessments, should that margin be classified as negative or close in the final pathology report, provided that the latter scenario could result in escalated therapy? Currently, no standardised guidelines exist to categorise such cases, and adjuvant treatment decisions vary across institutions.

Prognostic impact of margin revision after positivity on frozen section analysis

A highly debated topic concerns the impact and management of SM, either SDM or TBM, that were initially positive at intraoperative frozen sections but later revised to negative. Although identifying positive margins intraoperatively and immediately resecting occult disease may seem logical, a definitive prognostic benefit and therapeutic significance of these revised margins continues to be a subject of discussion ⁴⁹. Therefore, although many surgeons attempt further tissue resection when encountering a positive intraoperative margin, it remains unclear whether this additional resection significantly affects prognosis ^39,48,49. Previous evidence suggests that survival rates and local tumour control among patients who undergo additional resections to achieve negative margins are comparable to those who achieve negative margins on the first attempt ^50-52. Zhang et al. found that immediate further resection guided by frozen section analysis could enhance oncologic outcomes in patients with positive main specimen margins ⁵¹. Additionally, they confirmed the poor sensitivity (17%) of TBMs in predicting the status of the margins of the main specimen ⁵¹. However, the study included patients with diverse clinical characteristics, such as those who did or did not undergo adjuvant therapy and neck dissection, those at different tumour stages, and those with an uncertain SM status.

On the other hand, some researchers have suggested that re-resections provide limited benefit in reducing LRs and that tumours with an initial cut-through SM often exhibit more aggressive biological behaviour compared to those with an initially negative SM ⁵³. Patients with initially positive margins who undergo additional resections still tend to have a worse prognosis compared to those with negative margins from the outset. A meta-analysis of retrospective studies assessing LRFS in early-stage OSCC found that revising initially positive SM to negative margins using frozen section guidance does not yield the same outcome as having clear margins from the outset and does not significantly enhance LC ⁵³. Additionally, Bulbul et al. demonstrated that patients requiring additional resection were 2.9 times more likely to experience recurrence than those with initially negative margins, even if the final resection resulted in what was considered a negative margin ⁵³. Buchakjian et al. compared LC in patients with positive main specimen margins who underwent additional tissue resection and found that once a positive margin was encountered on the main specimen, additional tissue resection did not impact LC. The risk of LR was lowest when SM were initially negative, higher when positive margins were revised to negative, and highest when final margins remained positive ⁵⁴. Guillemaud et al. conducted a study on 65 patients, showing that the presence of tumour cut-through with positive frozen margins was associated with poorer LC, regardless of the final margin status after additional resection and revision ⁵⁵. Similarly, Patel et al. found that microscopic tumour with positive frozen margins remained a negative prognostic factor even when revised to negative margins ⁴⁹. Coutu et al. reported a cohort study indicating that an initial positive margin was an independent predictor of reduced survival and poorer LC, irrespective of the outcomes of additional resection ⁵⁶. Recently, Kim et al. investigated the oncologic impact of revised negative SM (R1-R0) in a cohort of 441 patients with early-stage (pT1-2/N0) OSCC who underwent surgery without adjuvant therapy. In their retrospective analysis, LR was significantly higher (p = 0.045) in the R1-R0 group (13.1%) than in the R0 group (5.5%), and LRFS was better in the R0 group; however, there were no significant differences in overall recurrence or OS between these two groups. In their multivariate analysis, the authors identified an initial positive margin as an independent risk factor for LR (HR = 2.2; 95% CI, 1-4.9; p = 0.043). A revised clear SM after an initial cut-through margin increased the risk of LR in early-stage OSCC, emphasizing the need for caution in managing these patients ⁵⁷.

Overall, these studies reinforce the idea that the presence of a positive margin during the initial resection negatively affects prognosis and may not be fully corrected by further resection to achieve control rates comparable to those of patients with initially negative margins. Several hypotheses have been proposed to explain why additional tissue resection may not improve survival or LC. If a positive margin is not the result of a surgical error but rather due to diffuse tumour infiltration and the absence of a clearly defined tumour edge, this could indicate an underlying tumour biology that is less responsive to primary surgical treatment ⁵⁶. Moreover, studies have shown that the prognostic significance of positive margins during the initial resection is influenced by the presence of regional metastases. This further reinforces the idea that individual tumour biology plays a role in modifying the impact of a positive margin ⁴⁹. Secondly, identifying the precise location of a positive margin after tumour resection can be challenging. Given the low sensitivity of conventional tumour bed sampling techniques, it is difficult to ensure that an adequate revision has been achieved or that the re-resection has been performed in the appropriate location. The spatial relationship between the resected specimen and the revised margin plays a crucial role in ensuring proper margin revision. Studies have indicated that, when a frozen section is positive, surgeons often struggle to precisely identify the corresponding site for additional sampling ^39,47-49. Previous research has shown that accurately re-orienting the surgical specimen is difficult, often resulting in an error exceeding 10 mm ^37,58,59. Another challenge in margin assessment is tissue shrinkage, which varies depending on the resection site (e.g., dorsal tongue or hard palate vs floor of the mouth, buccal mucosa, or soft palate) ^40,60 (Tab. III).

Are tissue artefacts on margins a real issue?

In 1997, Johnson et al. investigated the potential shrinkage of tumour-free margins caused by tissue fixation. By simulating margins on the benign oral mucosa of dogs, they observed a reduction of 30-47% when comparing in vivo, post-fixation, and slide preparation measurements. However, the use of non-human subjects and non-malignant tissue raises concerns about the applicability of these findings to human OSCC and surrounding mucosa ⁵⁴. Mistry et al. examined margin shrinkage in the oral cavity 30 minutes after tumour resection, reporting an average reduction of 23%, but did not account for changes following formalin fixation ⁶¹. Cheng et al. retrospectively analysed pre-excision margins in oral cavity specimens and compared them to the final histopathologic margins, finding an average reduction of 59%, and showed that this shrinkage occurs immediately post-resection and, to a lesser degree, following formalin fixation ⁶². Several other factors may also influence the extent of tissue shrinkage, including the resection technique used, the time elapsed before transferring the specimen to a formalin container (cold ischaemia), and delays in tissue processing ⁶³, but also intrinsic contractility of mucosal tissues. Umstattd et al. analysed shrinkage in OSCC along the pre-analytical phase, evaluating tumour and margin measurements at different stages: in vivo, post-resection, and post-formalin fixation. Their findings indicate that the most significant shrinkage occurs between pre- and post-excision measurements, while no significant changes were observed between post-excision and post-formalin fixation. Given that the primary alterations in measurement take place before fixation, it is suggested that tissue shrinkage is predominantly due to inherent tissue properties rather than the effects of formalin processing ⁶⁴. Additionally, it appears to vary across different regions within the same specimen and to affect less neoplastic than normal tissue; consequently, SM tend to move closer to the tumour mass ⁶⁵. This difference is likely due to the infiltrative nature of tumour tissue. The variation in shrinkage can be explained by the distinct structural properties of the tissues. The intra-neoplastic stroma is often characterised by dense collagenisation, which creates a compact and rigid framework that resists formalin-induced shrinkage more effectively than the more flexible and loosely structured normal tissues. Studies have shown that shrinkage is more pronounced in the tongue (average shrinkage of 23.5%) compared to buccal mucosa (21.2%) ⁶¹. In a recent study, Michcik et al. introduced the term macroscopic resection margins (MRM) to eliminate clinical ambiguities in their research. MRM refers to the visibly unchanged tissue surrounding the resected OSCC, extending from the tumour’s outer edge to the surgical cut line. Their findings suggest that variations in MRM size are influenced by the specific tumour location. Notably, tongue tumours exhibited the most significant reduction in margins, with a median shrinkage of 1.5 mm. In comparison, tumours located in the floor of the mouth and maxilla showed slightly less shrinkage, with median values of 0.8 and 0.9 mm, respectively ⁶⁶. Additionally, tumours at earlier stages (T1/T2) tend to exhibit greater shrinkage than those at more advanced stages (T3/T4) ⁶¹.

Following surgical excision and fixation, the specimen undergoes visual assessment during the grossing process. In HNSCC, the approach to sectioning remains a topic of debate due to the absence of standardised protocols, unlike in organs such as the colon or breast. Typically, specimens are sectioned using either the “radial or perpendicular” method, which involves cutting at a right angle to the margin (Cover figure), or the “parallel or en-face” (shave margin) technique. Notably, only the first method enables precise measurement of the microscopic distance between the (inked) margin and the tumour ⁶⁷. Although the grossing step may appear straightforward and systematic, it is essential to acknowledge the possibility of errors during the initial sampling stages. Additionally, individual pathologist interpretation can introduce a degree of subjectivity in the subsequent histological assessment ⁶⁸. Regardless of the grossing method used for tumour processing, each subsequent histological step, including dehydration, clearing and infiltration, microtomy, haematoxylin and eosin (H&E) staining, and slide preparation, has the potential to induce tissue alterations that may affect microscopic evaluation ⁶⁹.

As a result of all of the above, pathological margins are often significantly smaller than in-situ margins, which can influence the assessment of marginal status.

This brings up an important question: Can specimen shrinkage be reliably predicted? Moreover, how can this knowledge be applied to improve margin evaluation? A noteworthy study by Burns et al. explores this issue in greater detail ⁷⁰. The study identified an average shrinkage of 26% following resection and processing, leading to the suggestion that margin measurements should be adjusted accordingly. Implementing this adjustment could reduce the need for overly extensive resections, particularly in cases where tumour location limits surgical options due to functional or aesthetic concerns. Additionally, it may help prevent unnecessary re-resections or adjuvant therapy in cases with close or positive margins. It is important to emphasise that this is currently just a proposal. As suggested by Sengupta et al., variations in tissue shrinkage across different institutions and laboratories should be thoroughly assessed before determining a universally optimal margin distance. Each institute and laboratory should measure and quantify tissue shrinkage while accounting for all possible influencing factors ⁷¹. If a strong inter-institutional correlation in the data is identified, the recommended margin distance could be applied universally. However, if significant discrepancies arise, tissue shrinkage should be reassessed based on influencing factors to determine an appropriate margin distance for each institution and possibly anatomical area. Additionally, the authors recommend that pathology reports on margin assessment include the percentage of shrinkage, enabling surgeons to interpret the findings with both aspects in mind. These institutional measurements could contribute to improved prognosis for HNSCC. Therefore, future studies should focus on the correlation between tissue shrinkage and the optimal margin distance required for favourable prognostic outcomes ⁷². Recently Michcik et al., using modern 3D imaging techniques, developed a protocol for creating virtual images of OSCC that can be utilised in various ways, including enhancing surgeon-pathologist communication, mapping and describing tumours, and presenting cases at multidisciplinary board meetings. The creation of virtual images provides nearly unlimited possibilities for conducting and repeating measurements at countless points. Compared to traditional methodologies described in the literature, the use of virtual OSCC images represents a significant breakthrough, unlocking new opportunities that were previously unavailable ⁶⁶ (Tab. IV).

Suggestions on how to evaluate the bone margin using frozen sections

While frozen sections provide intraoperative histological assessment of soft tissue margins, microscopic evaluation of bone margins is generally not available within the surgical timeframe, provided that calcified bone prevents the preparation of frozen sections. As a result, the surgical planning for bone-infiltrating carcinomas relies primarily on preoperative imaging techniques, such as MRI and CT. Despite advancements in imaging technology, accurately defining the extent of bone involvement before surgery remains a challenge. Additionally, preoperative imaging data cannot be directly applied intraoperatively to distinguish malignant from healthy tissue in real time ³. As a result, uncertainty persists regarding bone infiltration, which may lead to misjudgment of the extent of carcinoma involvement, potentially causing either over- or under-treatment. Multivariate analyses have demonstrated that only medullary invasion has clinical significance, whereas cortical invasion alone does not ^73-75. According to the latest NCCN guidelines, segmental resection is advised in cases of medullary space invasion. Frozen section analysis of bone marrow can be utilised intraoperatively to assist in guiding resection decisions ²⁷. Given the preoperative challenges in assessing bone invasion and the intraoperative limitation of frozen sections to soft tissue evaluation, Nieberler et al. implemented intraoperative cytological assessment of bone (ICAB) SM as an alternative approach ^3,76-78. In their latest research, ICAB showed a sensitivity of 88% and a specificity of 78.6% in determining adequate resection ³. However, its clinical implementation revealed technical and logistical challenges, including desiccation of bone margins, insufficient or excessive trabecular material, and interference from blood cells, leading to cytological samples with poor quality. To improve diagnostic accuracy, bone margins were rinsed to remove excess blood, bone dust, and floating carcinoma cells, which do not indicate true bone infiltration. It is also crucial to avoid contamination from instruments, gloves, or residual tumour tissue during transport to prevent false positive results. The quality of ICAB relies heavily on expert cytological assessment, as prior radiotherapy may alter cells morphology, and immature haematopoietic precursor cells can be misidentified as malignant. Additionally, complex margins, infiltrative carcinoma growth, and prior radiotherapy pose challenges to both frozen sections and ICAB.

Collaboration with the surgical team

At the institutional level, several measures can enhance the accuracy of SM assessment, such as dedicated head and neck pathologists, standardised synoptic reporting, uniform laboratory processing protocols, and external quality control through routine audits. To improve specimen orientation and facilitate interpretation, the Canadian surgeons survey reported employing 3D specimen orientation techniques, using various marking methods to indicate the closest representative margin ⁴⁴. Another promising area of research is the integration of 3D modelling to improve communication between surgeons and pathologists. Saturno et al. ⁷⁹ and Sharif et al. ⁸⁰ have demonstrated the feasibility of scanning surgically resected specimens to generate virtual 3D models, which can be annotated by pathologists for real-time intra- and postoperative consultations. While this technology remains in the experimental phase and was not directly examined in this survey, it holds potential as a valuable tool for enhancing SM assessment in the future.

The use of tissue biomarkers in the evaluation of tumour margins

The rationales supporting the search for biomarkers that could indicate either absence or presence of neoplastic transformation on SM are: (1) the observation that tumour tissue and non-neoplastic mucosa bear inherent molecular differences; (2) the theory of field cancerisation in the upper aerodigestive tract mucosa, and (3) the possibility that the number of tumour cells left beyond resection margins is too small to be detected by routine histopathology (minimal residual cancer, MRC) ⁸¹. Several findings have shown that microscopically normal mucosa surrounding SCC may host molecular abnormalities shared with tumour tissue, suggesting preneoplastic transformation ⁸². They also supported the hypothesis that the presence of these abnormalities on SM can negatively affect prognosis and possibly explain the limited association between conventional margin stratification and patient outcomes discussed in the above sections. Recurrence rates at the primary tumour site after R0 resection range between 12% and 30% ^83,84. For these reasons, morphological margin examination may be supported using ancillary techniques. These are roughly distinguished in two main groups, although significant overlap exists: molecular and immunohistochemical analyses. Both may integrate morphological examinations and contribute to clarifying SM status. Molecular techniques provide a more comprehensive description of tissues as well as of their genomic status. However, they are expensive and time-consuming, requiring many days to be performed. Immunohistochemical analyses, instead, are less expensive and require less time, representing a valid support for routine practice (Fig. 1). More recently, protein biomarkers have also been proposed for fluorescence-guided surgical identification of negative mucosal margins ^85,86.

Not surprisingly, the number of biomarkers that are expressed differentially in HNSCC and normal mucosa, and that may signal the presence of tumour cells along SM, is very high, including p53, microsatellite instability, DNA ploidy, CCDN1, p16, Chk2, laminin-5, glycosylated oncofetal fibronectin, galectin-9, hypermethylated gene promoters ^87-89, and overall genomic imbalance ⁹⁰. A relevant number of studies have aimed to identify if the presence of any of these biomarkers on SM could predict LR (Tab. V) ^{76,79,82,86,88,92-122}. The list of those that have been proven to significantly predict LR in small case-control series is relatively limited. Among these, the status of p53 has been most widely studied with increasingly accurate methods over time. Other promising biomarkers include overexpression of eIF4E, loss of PERP and cornulin, overall loss of heterozygosity (LOH), p16, cyclin D1, beta-catenin, MMP9 and PTHLH RNA levels, chromosome instability, and profile of genes with promoter methylation.

It is worth noting that the bulk of studies on so-called “molecular margins” was published in the first 2 decades of this century; moreover, there are no prospective studies validating the prognostic value of biomarker expression analysis on traditionally sampled HNSCC margins, and biomarkers have not yet been applied as a support to surgical or therapeutic decisions. In reviewing the literature and discussing the multitude of molecular markers that have been evaluated for use, one factor that limits their application is the low coverage of these markers within the SM, and the relatively low specificity for tumour. For the widely studied p53, the myriad of genetic variants makes it difficult to develop simple assays that can be integrated into the clinical setting and limits the utility of mutant p53 assessment. Other examples, which include the early genetic event involving LOH at 9p21 to 22, observed in approximately 70% of HNSCC cases, dysregulation of p16, either via homozygous deletion (67%) or increased promoter methylation (21%), or microsatellite instability (50-60%), illustrate that no marker is expressed homogeneously (Tab. V). Definite evidence of practical benefits in treatment decisions stemming from molecular evaluation of SM is still lacking. Possibly the most significant concern that has so far prevented the use of molecular alterations to assess the SM in clinical practice is determining a particular biomarker’s predictive value for malignant disease.

Moreover, most of the methods reported so far are not compatible with the rapid response required for margin status by clinicians. However, Goldenberg and colleagues used quantitative methylation-specific PCR (QMSP) to determine the hypermethylation pattern of p16 and MGMT promoters in the SM. Although the number of patients in this study was small, the investigators showed that QMSP analysis could be performed within a period compatible with intraoperative assessment ¹²². One further limitation is related to the complexity of complete tissue sampling over a large surface of marginal tissue. The concept of margin imprinting presented by Hayashi et al. ¹²³ is interesting. Their strategy may ease the difficulty of collecting relevant SM samples that represent the entire area of irregular surgical resection.

More recently, whole genome sequences of HNSCCs available in public databases offer the opportunity for a different approach to molecular profiling of cancer, and possibly to overcome the limitations of single biomarker analysis. One recent study ¹²⁴ provided a gene signature associated with positive SM in tongue SCC, possibly predicting tumour resectability and informed surgical decisions. Furthermore, a pioneering microRNA signature in peripheral blood has recently been associated with disease recurrence in OSCC ¹²⁵, paving the way for the application of liquid biopsy in the future.

Non-histopathological approaches to the evaluation of tumour margins: back to the surgical bed

Evaluating tumour margins intraoperatively is crucial to ensure complete tumour resection and minimise recurrence. In recent years, significant efforts have been directed toward developing non-invasive optical imaging techniques to support clinical diagnosis and treatment. These imaging methods provide a way to analyse the structural and compositional characteristics of tissues and cells in vivo, by detecting variations in their optical properties, resulting in highly detailed images of organs and biological structures ¹²⁶. Various optical imaging modalities integrating surgery have been documented in the literature, including narrow band imaging (NBI), Raman spectroscopy (RS), optical coherence tomography (OCT), hyperspectral imaging (HSI), computer-assisted surgery (CAS), and fluorescence imaging. These techniques leverage anatomical and metabolic alterations, as well as the chemical composition of tissues ¹²⁷. Due to its ability to minimise patient exposure to harmful radiation while offering valuable insights into the properties of soft tissue, optical imaging has become an essential tool in clinical practice. It is widely used to identify abnormal lesions, enabling early disease detection, providing real-time visualisation of SM, assisting in tumour removal, and tracking treatment response.

NBI is a fast, noninvasive, and safe endoscopic technique that employs two specific wavelength filters, blue and green, to improve the visualisation of mucosal structures and superficial blood vessels, thus allowing the distinction of patterns of distribution between tumour and normal mucosa. With central wavelengths of 415 nm for blue light and 540 nm for green light, NBI takes advantage of haemoglobin’s strong absorption properties, causing blood vessels to appear in darker shades, such as brown or cyan, in contrast to the surrounding mucosa ¹²⁸. As a result, surgeons utilise NBI to assess characteristics of intra-epithelial papillary capillary loops for early lesion detection and differentiation between benign and malignant conditions. Key features examined include the extent of vessel dilation, curvature, variations in vessel calibre and morphology, as well as colour alterations in surrounding mucosal tissue ¹²⁹. Research indicates that defining the incisional margin using NBI can lower the likelihood of positive tumour margins, enhance patient survival, and decrease recurrence rates compared to standard white-light endoscopy. In a study by Tirelli et al., NBI demonstrated a sensitivity of 100% and a specificity of 88.9% in evaluating incisional margins ¹³⁰.

NBI has notable limitations when evaluating deep incisional margins, as it is unable to detect submucosal tumour extension. Consequently, NBI is most effective for assessing the margins of superficial and early-stage tumours. For locally-advanced tumours, where precise evaluation of deep margins is required, fluorescence molecular imaging offers a more accurate depiction of tumour boundaries ¹³¹.

RS is a light scattering technique based on the Raman effect and is capable of generating unique spectral fingerprints that reveal the molecular composition and biochemical characteristics of tumour tissue. By reflecting histopathological changes, RS provides valuable insights for clinical applications, including early cancer detection, tumour margin assessment, identification of metastatic lymph nodes, and identification of recurrence. The method is fast, non-invasive, requires minimal sample material, does not rely on fluorescence properties, and eliminates the need for special labelling, making it a promising tool in cancer diagnostics ¹³². RS has certain limitations. One challenge is the difficulty in fully distinguishing tumour from normal tissue. Due to biochemical similarities between SCC and the squamous epithelium, along with changes in nucleic acid and protein levels in inflamed connective tissue, there is a risk of misdiagnosing normal epithelium or inflammatory tissue as SCC ^133,134. However, studies suggest that false positive results do not impact surgical outcomes as long as a 5-10 mm margin around the tumour is excised ¹³⁴. Another limitation is the lack of a standardised diagnostic threshold to differentiate cancerous from non-cancerous tissue using spectral analysis. Yan et al. in their research introduced a novel approach combining Raman imaging technology with an AI generative model to enhance intraoperative margin assessment. In their research, Raman images were converted into H&E-stained histopathological representations using an AI generative model. This approach facilitated the transformation of RS images into visual histopathological images, eliminating the need for traditional tissue section preparation and staining processes. The AI-generated H&E images provided a clear visualisation of the pathological characteristics of the tissue and demonstrated superior performance compared to conventional clinical practice, particularly in cases of HNSCC with metastatic lymph nodes or moderate differentiation ¹³⁵.

HSI is a remote optical imaging technique that captures a single image across potentially hundreds of distinct wavelengths. This method provides detailed information about tissue composition, assisting surgeons in identifying tumour margins more precisely. Fabelo et al. highlighted the importance and effectiveness of HSI in detecting brain cancer in vivo and designed a visualisation system that leverages a machine learning-based classification algorithm for near-real-time guidance ¹³⁶. Additionally, research on deep learning approaches applied to the same dataset revealed superior performance compared to traditional machine learning methods. Boehm et al. conducted a proof-of-concept study to evaluate whether HSI, combined with AI through a Convolutional Neural Network (CNN) and a Graph Neural Network (GNN), can improve the differentiation between healthy and malignant tissue in ex vivo head and neck tumour specimens. The goal was to determine its potential for effective margin assessment. Their study included 32 patients diagnosed with HNSCC. Following surgical resection, their specimens underwent ex vivo HSI imaging, with annotated regions used to train the CNN and GNN models. Imaging parameters were carefully optimised to enhance efficiency, allowing for a more objective and rapid evaluation of SM. Their findings suggest that integrating HSI with neural networks provides additional benefits in distinguishing cancerous from healthy tissue. Despite the relatively small dataset (32 patients), the AI models demonstrated high accuracy ¹³⁷.

CAS plays a crucial role in improving margin assessment during HNC surgery. By integrating imaging techniques such as CT, MRI, and PET, CAS enables precise preoperative visualisation and delineation of tumour margins, reducing the risk of leaving residual cancerous tissue ^138,139. In OSCC with mandibular involvement, CAS has shown effectiveness in controlling SM, with outcomes comparable to those of traditional surgical techniques. Some studies suggest that CAS enhances both surgical precision and postoperative functional and aesthetic results due to its accuracy in planning and executing resections ¹⁴⁰. However, there is no conclusive evidence that cutting guides significantly improve oncological margin control in mandibular SCC resections. Research into mixed reality systems, which integrate real-time 3D imaging with CAS, is exploring ways to further improve intraoperative margin assessment, though these technologies remain in experimental phases ^141,142.

In recent years, there has been growing interest in the use of fluorescent markers to enhance tumour detection and delineation during surgery, providing real-time intraoperative guidance. These markers aim to provide a clear distinction between malignant and healthy tissues, ensuring precise excision margins. An ideal fluorescent marker should selectively bind to cancer cells, accumulate sufficiently to produce fluorescence, and effectively differentiate malignant from healthy tissue. Several fluorescent markers have been studied for their potential use in HNSCC detection, including toluidine blue, ICG, 5-ALA, and IRDye800CW-conjugated anti-EGFR antibodies. Toluidine blue has seen limited recent research ¹⁴³ due to concerns over specificity. While ICG is FDA- and EMA-approved ¹⁴⁴, its effectiveness in HNSCC remains uncertain ¹⁴⁵, although newer ICG-based markers like ONM-100 may offer better tumour visualisation. 5-ALA, already used in glioma surgery ¹⁴⁶, has shown promise for HNSCC in recent studies ¹⁴⁷. Meanwhile, fluorescently labeled cetuximab and panitumumab are gaining attention, with multiple ongoing clinical trials. Panitumumab-IRDye800CW has demonstrated potential for improving surgical decision-making ¹⁴⁸. Srinivasan et al. conducted a systematic review to identify and compare fluorescent markers used for the detection and delineation of HNSCC. Among the most promising candidates, they identified 5-ALA, cetuximab-IRDye800CW, panitumumab-IRDye800CW, and targeted ICG derivatives such as ONM-100 ¹⁴⁹. Recent studies have explored ex vivo microscopy for evaluating excision margins in HNSCC. Shavlokhova et al. ¹⁵⁰ reported a sensitivity and specificity of 99% and 95% using ex vivo imaging of oral cavity lesions. A systematic review of 6 confocal laser endomicroscopy studies ¹⁵¹ showed 95% sensitivity and 93% specificity, while Zanoni et al. ¹⁵² demonstrated 90% sensitivity and 98.3% specificity for in vivo reflectance confocal microscopy (Tab. VI).

Conclusions

In HNC surgery, evaluation of margin status represents an important part of tumour staging and prognostic assessment. This may be obtained in frozen and definitive sections, sometimes with the use of ancillary techniques. It should be noted, however, that the risk of LR is affected not only by the surgical excision margin. Alongside this, T category, degree of differentiation, pN, lympho-vascular, perineural, depth, and pattern of invasion are other independent risk factors for LR.

Conflict of interest statement

The authors declare no conflict of interest.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Author contributions

All authors contribuited equally to the work.

Ethical consideration

Not applicable.

History

Received: March 15, 2025

Accepted: March 21, 2025

Figures and tables

Figure 1. (A) A neoplastic lesion on the left lingual border in a male patient affected by graft-versus-host disease; (B) Tumour bed after lesion resection; (C) Whole slide image of the lesion stained with haematoxylin; (D) Immunohistochemical analysis for p53 shows a pattern of basal and suprabasal overexpression, while the resection margins are negative.

Figure 2. (A) Total laryngectomy specimen. It is crucial to assess all resection margins. Intraoperatively, frozen sections must always be performed, especially on the superior margin at the level of the vallecula/base of the tongue and the inferior margin at the level of the piriform sinuses; (B) A whole slide image (WSI) of a cross-section obtained at the tumour site in the corresponding total laryngectomy specimen; (C) Open partial horizontal laryngectomy (OPHL) Type II specimen. Proper orientation of the specimen is essential. It is important to evaluate the margins and assess thyroid cartilage infiltration, as this can influence the indication for adjuvant treatment; (D) A WSI of a cross-section obtained at the tumour site in the corresponding OPHL Type II specimen; (E) Type VI cordectomy. Close collaboration with the surgeon is essential, particularly in orienting the specimen in the operating room. In this case, the margins to be evaluated include: the posterior vocal cord margin, the inferior subglottic margin, and the superior margin towards the ventricle. The anterior and lateral margins, adjacent to the perichondrium of the thyroid cartilage, may also be positive. However, in such cases, achieving negative margins is challenging. If the posterior, inferior, or superior margins are positive, a wider endoscopic resection may be performed. However, if the anterior or lateral margins are involved and further excision is necessary, an open surgical approach is required. In general, an indication for wider endoscopic resection is given only in cases of multiple positive superficial margins or a positive deep margin. However, depending on the surgeon’s experience, an endoscopic follow-up approach may also be considered. This is because margins can appear negative intraoperatively, but due to shrinkage from formalin fixation or the effects of the carbon dioxide laser, may be positive in the final histological examination; (F) A WSI of a cross-section obtained from the cordectomy specimen.

Table I. Intraoperative evaluation of tumour margins.

Key points	Description
Frozen section analysis	Widely used in head and neck cancer surgery for intraoperative margin assessment
NCCN guidelines	Frozen section margin assessment is at the surgeon’s discretion; wide margins may require additional tissue resection
Proximal and distal nerve margins	Recommended for frozen section analysis to ensure complete tumour removal
Reliability of frozen sections	Demonstrated to be a valuable diagnostic tool for rapid intraoperative pathological evaluation
Lack of standardisation	There is no universally accepted standard for frozen section margin evaluation in head and neck surgery
Institutional variability	Different institutions apply varying approaches, often developed internally
Survey findings	32% of respondents use multiple intraoperative margin assessment methods; 18% apply more than one method for final evaluation

Table II. Defect-driven vs sample-driven margin assessment.

Key points	Sample-driven margins (SDM)	Defect driven or tumour-bed margins (TBM)
Definition	Margins assessed on the resected specimen	Margins taken from the surgical bed after resection
Guidance	Pathologist-driven margin assessment	Surgeon-guided margin selection
Correlation	Stronger correlation with true SM	Weaker correlation with actual resection specimen margins
Local control	Better local control	Lower local control, higher false negative rate
Sensitivity	Higher sensitivity in detecting positive margins	Lower sensitivity (reported between 15% and 35%)
Surgical time	Potentially increases operative time for thorough assessment	Faster, as it avoids waiting for detailed pathological processing
Economic aspect	Higher costs due to extended surgery and pathology processing	Lower initial costs, but false negatives may lead to additional treatment expenses
Surgeon preference	Less commonly preferred due to logistical challenges	More widely used due to efficiency and ease of execution
Advancement	Standardised in most pathology departments	Some methods attempt to improve accuracy by orienting TBM samples

Table III. Prognostic impact of margin revision after positivity on frozen section analysis.

Key points	Description
Negative margins	Clear margins improve prognosis in head and neck cancer
Surgical challenges	Obtaining negative margins in the first attempt can be difficult
Margin revision debate	Uncertainty exists on whether re-resection improves outcomes
Recurrence risk	Higher recurrence rates in patients with revised margins compared to initially negative ones
Tumour biology role	Patients with aggressive tumours may not benefit from additional resection
Re-resection challenges	Difficult to identify and accurately resect positive margins

Table IV. Are tissue artefacts on margins a real issue?

Key points	Description
Specimen processing	After excision, tissue is fixed and sectioned, but standardisation in head and neck cancers is lacking
Tissue shrinkage	Margins shrink due to processing, with reductions up to 59%, depending on tissue type and location
Impact on margin assessment	Shrinkage leads to margins appearing smaller, affecting classification of negative, close, or positive margins
Regional variability	Tongue tissue shrinks more (23.5%) than buccal mucosa (21.2%), impacting interpretation of margins
Predicting shrinkage	Studies suggest measurements of margins should be adjusted to compensate for shrinkage effects
Institutional differences	Margin shrinkage varies across institutions; standardised reporting could improve prognostic evaluation

Table V. Summary of studies evaluating the prognostic impact of biomarker expression on surgical margins.

Biomarker	Method	Site	Tumour-associated defect	Cases/controls	Prevalence	Specificity (%)	Sensitivity (%)	Analysis	Significance	Notes	Ref.
TP53 gene	SequencingSouthern blot	Head neck	Mutation	5/20	13/25	5/5(100)	12/20(60)	KM/log-rank	Yes		82
EIF4E protein	IICH	Head neck	Overexpression	22/43	36/65	20/22(91)	27/43(63)	KM/log-rank	Yes		92
EIF4E proteinP53 protein	IICHIICH	Larynx	OverexpressionOverexpression	25/29	32/546/54	21/25(84)6/23(26)	18/29(82)31/31(100)	KM/log-rankMultivariateKM/log-rankMultivariate	YesYesYesNo		93
TP53 geneP53 protein	Phage assayIICH	Oral cavity	MutationOverexpression	5/6	6/11	4/5(80)	4/6(67)	ND			94
Microsatellites	PCR, Gel electrophoresis	Multiple	Instability	8/17	11/25	7/8(88)	13/17(76)	KM/log-rankMultivariate	YesYes		95
P53 proteinMMP9 protein4E protein	IICHIICHIICH	Head neck	Overexpression	14/38	24/5228/5227/52	9/14(64)11/14(78)12/14(86)	23/38(60)21/38(55)22/38(58)	KM/Cox proportional hazard/multivariate	NoNoYes/Yes		96
TP53 gene	Plaque assay	Multiple	Mutation	15/61	50/76	14/15(93)	25/61(41)	KM/log-rank	Yes		76
Microsatellites	PCR, Gel electrophoresis	Multiple	Instability	5/21	7/26	5/5(100)	19/21(90)	KM/log-rank	Yes		97
DNA	DNA-image cytometry	Oral cavity	Aneuploidy	20/20	16/40	14/20(70)	18/20(90)	Fisher exact test	Yes		98
TP53 gene	PCR, Plaque assay	ND	Mutation	13/12	16/25	11/13(85)	7/12(58)	KM/log-rank	Yes	Deep margins also evaluated	99
CDKN2A (p16), CCNA1, DCC genes	QMSP	Multiple	Promoter methylation	5/22	11/27	5/5(100)	16/22(73)	KM/log-rank	Yes	At least 1 marker positive	100
BIRC5 proteinCD44v6 protein	IICHIICH	Larynx	OverexpressionOverexpression	41/71	44/11235/112	26/41(63)20/41(49)	53/71(75)56/71(79)	Cox proportional hazard/multivariate	Yes/YesYes/Yes		101
DNA	Interphase-FISH	Oral cavity	Polyploidy	4/15	8/19	4/4(100)	11/15(73)	Fisher exact test	Yes		102
P53 protein	IICH	Oral cavity	Overexpression	4/15	8/19	3/4(75)	10/15(67)	Fisher exact test	No		103
Cytokeratin 4 proteinCornulin protein	IICHIICH	Oral cavity and oropharynx	HypoexpressionHypoexpression	23/23	23/4632/46	17/23(74)16/23(70)	17/23(74)16/23(70)	KM/log-rank	YesYes		104
CDKN2A gene	QMSP	Tongue	Promoter methylation	6/24	13/30	5/6(67)	16/24(67)	KM/log-rank	Yes		105
TP53 gene	LigAmp	Head neck	Mutations	15/80	35/95	9/15(60)	54/80(67)	Log-rank	No		106
Chromosome 3pChromosome 9pChromosome 17pP53 protein	SequenceSequenceSequenceIICH	Oral cavity and oropharynx	LOHLOHLOHOverexpression > 5%	16/19	ND	NDND (62)NDND (62)	NDND (74)NDND (84)	KM/Cox proportional hazard	NoYesNoYes		107
TP53 geneLy-6D mRNA	LigAmpqRT-PCR	Oral cavity	MutationExpression	43/59	51/102ND	NDND	NDND	KM/log-rank/multivariate	Yes/YesNo	Combined markers significant	108
PTHLH mRNAEPCAM mRNAMMP9 mRNA	qRT-PCRqRT-PCRqRT-PCR	Multiple	Overexpression 5xOverexpression 10xOverexpression 10x	4/51	5/556/5513/55	0/4 (0)0/4 (0)4/4(100)	46/51(90)45/51(88)42/51(82)	KM/log-rankCox proportional hazard	NoNoYes		79
DCC, EDNRB, p16, KIF1A genes	QMSP	Head neck	Promoter methylation profile	6/11	8/17	5/6(83)	8/11(72)	N/A	NA	Tissue imprints	109
DSPP proteinBSP proteinOPN proteinMMP-2 proteinMMP-3 proteinMMP-9 protein	IICHIICHIICHIICHIICHIICH	Oral cavity	Expression	9/11	11/2013/2011/2012/2013/207/20	8/9(89)6/9(67)7/9(78)5/9(55)7/9(78)6/9(67)	6/11(54)4/11(36)7/11 ((64)4/11(36)5/11(45)10/11(91)	KM/Cox proportional hazard	YesNoNoNoNoYes		110
P53 proteinCyclin D1 protein	IICHIICH	Multiple	OverexpressionOverexpression	35/81	34/11646/116	ND	ND	KM/Cox proportional hazard/ multivariate	NoYes/No	No association with OS	111
DCC, EDNRB, HOXA9, KIF1A, NID2, NR2B genes	QMSP	Head neck	Promoter methylation profile	Total 65	ND	ND	ND	Cox proportional hazard	Significant combined markers; no single	Significant on margin imprint, negative histological	112
PAX5 gene	QMSP; ddQMSP	Head neck	Promoter methylation	35/25	16/60	ND	ND	Cox proportional hazard	Yes	Margin imprint and tissue	86
Microsatellite	Sequence	Oral cavity	InstabilityLOH	24/58	40/8276/82	NDND	NDND	KM/log-rank	YesNo		113
P53 proteineIF4E Protein	IICHIICH	Oral cavity and oropharynx	OverexpressionOverexpression	7/17	11/2416/24	4/7(57)1/7(14)	10/17(59)15/17(88)	KM/log-rank	NoNo		114
Chromosome 9pChromosome 17pP53 proteinP14 proteinP15 proteinP16 protein	PCR/gel electrophoresisIICHIICHIICH	Oral cavity	LOHLOHOverexpressionOverexpressionOverexpressionOverexpression	15/56	14/6517/6722/7139/7130/7130/71	6/13(46)6/15(40)10/15(67)11/15(73)9/15(60)15/15(100)	44/52(85)41/52(79)44/56(78)28/56(50)35/56(62)38/56(68)	KM/Cox proportional hazard/multivariate Cox regression	Yes/YesYes/NoYes/NoNo/NoYes/NoYes/No	TP16+1 9p microsatellite significant in multivariate analysis	115
Chromosome 1 & 7P53	FISHIICH	Oral cavity	Copy number variationOverexpression	11/2411/31	18/3512/42	9/11(82)4/11(36)	15/24(62)16/31(52)	KM/log rank test/multivariate Cox regression	Yes/YesNo		116
βcatenin proteinβcatenin proteinβcatenin mRNA	Western blotIICHqRT-PCR	Oral cavity	OverexpressionNuclear expressionOverexpression	26/54	47/80	21/26(81)	28/54(52)	KM/log-rank	Yes	Pooled results	117
Microsatellite	Sequence	Oral cavity	Instability	26/29	41/55	22/26(85)	10/29(34)	Logistic regression	Yes	Pre-selected MSI+ tumour	118
P53 protein	IICH	Oral cavity	Overexpression	16/20	32/72	14/16(87)	38/40(95)	KM/Cox proportional hazard	Yes	Dysplastic margins	119
PERP protein	IICH	Multiple	Loss of expression	NA	19/46	NA	NA	KM/log-rank	Yes		120
Conulin proteinKI67/MIB1 proteinISG15 protein	IICHIICHIICH	Oral cavity	IRS-score (hig/low)% positive (hig/low)IRS-score (hig/low)	15/19	10/34	7/15(47)2/15(13)3/15(20)	16/19(84)18/19(95)18/19(95)	Pearson Chi-square/binary logistic regression	YesNoNo		121
TP53 geneP53 protein	RT-PCRIICH	Oral cavity	AmplificationOverexpression	21/19	38/40	16/21(76)19/21(90)	14/19(74)17/19(89)	KM/Cox proportional hazard/Multivariate COX regression	NoNo	Significant association with OS	122
TP53 geneCirculating miRNA	NGS + ddPCR	Multiple	VAFsLevel quantification	4749 (28 both tests)	30/4726/49	27/42(64)25/47(53)	2/5(40)0	KMKM	YesYes	Significant biomarker combination	88
KM: Kaplan-Meier; IICH: immunohistochemistry; ND: not detailed; PCR: polymerase chain reaction; QMSP: quantitative methylation-specific PCR; FISH: fluorescent in situ hybridisation; LOH: loss of heterozygosity; qRT-PCR: quantitative retro-transcribed polymerase chain reaction; NA: not assessed; OS: overall survival; MSI: microsatellite instability; IRS: immune response score; RT-PCR: real time PCR; NGS: next generation sequencing; ddPCR: digital droplet polymerase chain reaction.

Table VI. Non-histopathological approaches to the evaluation of tumour margins.

Key points	Description
Importance of margin assessment	Intraoperative tumour margin assessment is crucial for complete tumour resection and reduction of recurrence risk
Optical imaging techniques	NBI, RS, OCT, HSI, CAS, and fluorescence imaging improve tumour margin visualisation
Narrow band imaging (NBI)	Enhances superficial tumour margin assessment but is ineffective for deep margins
Raman spectroscopy (RS)	Provides biochemical insights but has limitations in differentiating SCC from normal tissue
HyperSpectral Imaging (HSI)	Effective with AI assistance but limited by long acquisition times
Computer-sssisted surgery (CAS)	Improves preoperative planning but lacks conclusive oncological margin control evidence
Fluorescence imaging	Several markers (5-ALA, cetuximab-IRDye800CW, etc.) show promise for intraoperative tumour delineation
Ex vivo microscopy	Studies demonstrate high sensitivity and specificity in SCC margin assessment