When Neck Dissection is Not Indicated in the Treatment of the Clinically Node-Negative Head and Neck Squamous Cell Carcinoma

Editorial

Neck lymph node metastases (LNM) in patients with head and neck squamous cell carcinoma (HNSCC) are a key prognostic indicator of survival, with their presence taken to indicate decreased survival outcomes by up to approximately 50% [1]. As inappropriate neck management often results in regional failure, accurate detection and treatment of cervical LNM are critical. However, given the current limitations of palpation and diagnostic imaging work-up for the assessment of LNM, patients with a clinically and radiologically negative (cN0) neck still have a substantial risk of having occult metastases [2]. This leads to the question “Should all patients be treated prophylactically with an elective neck dissection (END), or can the neck be left untreated with a ‘watchful waiting’ follow-up policy in selected patients?” As HNSCC is a heterogeneous disease anatomically and biologically, having a uniform guideline regarding END indications is challenging. This question has been an ongoing debate for years, especially for surgically treated patients who do not require surgical access through the neck for primary tumor resection or reconstruction. This is particularly true for early oral squamous cell carcinoma (SCC) [3]. Furthermore, during the last decade, the emergence of transoral robotic and non-robotic surgery for early pharyngeal and laryngeal tumor resection has led to a reappraisal of the indications for END also in these subsites.

The rationale for END in patients with cN0 neck is based on the likelihood of clinically occult LNM, which varies according to tumor site and stage. In 1994, based on an extensive literature review, Weiss et al. constructed a decision tree using a computer model to compare the different management strategies: END, elective neck radiotherapy (ENR) and observation [4]. A sensitivity analysis was then performed to determine the optimal threshold for treatment of the neck. These authors recommended elective treatment of the neck in patients with primary HNSCC clinically staged as node negative but having a ≥ 20% probability of occult LNM [4]. Contingent on the assumption that END and ENR are equally effective in controlling cN0 neck, the choice between the two procedures thus generally depends on the treatment modality selected for the primary tumor, which in turn mainly depends on institutional policy. The basic rule driving the choice between surgery and radiotherapy is to favor the use of a single-modality therapy to avoid overtreatment. However, this 20% cut-off is quite arbitrary and might vary depending on specific circumstances, such as lack of access to close follow-up. Importantly, in their publication, Weiss et al. emphasized that the conclusions of their analysis are only as good as the data that are fed into the model [4]. This qualification is quite important as the parameters that were used in their analysis may have changed in the last 30 years. Weiss et al. expected that salvage therapy would be successful in only 50% of the patients in the observation group, with LNM appearing in the neck later [4]. However, it can now be anticipated that improved follow-up strategies, such as watchful waiting and wait and scan, will increase the success rate of salvage neck dissection. For example, the use of ultrasound-guided fine needle aspiration cytology during follow-up can improve the salvage rate [5]. In the study of Weiss et al., the utility (desirability) ratings for outcome were established by a panel of three experienced head and neck physicians and by the use of the time trade-off method in which the question to be answered was how much time of their life would a person give up for a better quality of life [4]. However, a patient’s preferences may differ substantially from the expectations of treating physicians. Probably even more important is the improvement in detection of occult LNM in the last decades [2]. The risk of occult LNM in cN0 neck may have changed in this period. Typically, sentinel node biopsy (SNB) may decrease this risk from 40% to 8% in patients with early-stage oral SCC [6].

There are several clinical situations when END is not indicated and can be excluded as a treatment option. One of these is early glottic carcinoma limited to the vocal cord, without supraglottic and subglottic extension; based on a very low probability of occult LNM, there is a general agreement to spare END in these cases. A recent systematic review confirmed this policy, i.e. that END was not justified in early glottic tumors, but that END including sublevel IIa and level III was recommended in T1–T4 cN0 supraglottic tumors [7]. In addition, END was also recommended in T1–T2 oral SCC because the incidence of occult LNM is around 30%. In these tumors, END has become the standard of care based on evidence provided by two large randomized controlled trials [8, 9]. However, the price to be paid for a survival benefit is significant as most patients are true N0 and consequently overtreated with this approach, resulting in higher morbidity, particularly shoulder morbidity (at least temporarily) caused by spinal accessory nerve dysfunction, in a significant number of patients [9]. Moreover, there seems to be substantial differences in pretreatment work-up and follow-up between head and neck centers around the world, limiting the generality and applicability of the results of these studies [10].

The need for non-invasive approaches to rule out the presence of occult LNM and avoid an unnecessary END has resulted in the development of novel diagnostic tools, including advanced imaging techniques, gene signatures and histopathological prediction models. However, these tools are hampered by their low sensitivity and negative predictive value [11–16]. In a trend promoting minimally-invasive approaches to reduce surgical morbidity, SNB has been widely investigated in patients with early-stage cN0 oral SCC as an prognostic factor to avoid neck dissection in SNB-negative patients. SNB has become a reliable diagnostic procedure for staging the neck in this setting [17, 18]. More recently, two large phase III randomized trials demonstrated that the oncologic outcome of the patients included in the SNB arm was equivalent to those included in the END arm, but that a shorter hospital stay and a lower morbidity were observed in the SNB arm, supporting SNB as a standard of care in patients with T1–T2 cN0 oral SCC [19, 20]. These findings were confirmed in systematic reviews [21, 22]. In addition to a reduced complication rate and a lower postoperative morbidity, SNB has also other advantages over END, including increased patient preference and cost-effectiveness [23].

Despite evidence of the benefits of SNB for patients, routine implementation of SNB remains a challenge worldwide, even in countries with abundant resources [24]. The need for infrastructure, experienced staff and multidisciplinary collaboration between nuclear medicine physicians, surgeons and pathologists, as well as equipment costs, are the most frequently reported reasons for this limited implementation [25]. SNB has a learning curve, and the diagnostic accuracy varies with the volume of SNB performed at the center and the experience of the multidisciplinary team. The associated labor-intensive pathology protocol also represents a challenge to the feasibility of its implementation, not only at a low-volume center due to absence of multidisciplinary expertise but also at a high-volume center as it entails considerable additional workload in pathology departments [25]. Moreover, SNB remains an invasive surgical procedure, and approximately 25% of patients require a subsequent therapeutic neck dissection, which must be performed promptly in a second operation, presenting potential logistical challenges. Typically, performing a subsequent neck dissection is more complex than END, and carries a higher risk of complications [3]. In addition, a neck dissection as a second surgical procedure is associated with a greater burden on patients and the healthcare system [23, 26]. Therefore, it is important to further improve patient selection for SNB by investigating other (less invasive) diagnostic techniques. As suggested a few years ago, to reduce the need for SNB, the scoring criteria of standardly used imaging techniques, such as fluor-18-fluorodeoxyglucose ([18F]-FDG) in conjunction with positron emission tomography/computed tomography (PET/CT), should focus on a high positive predictive value instead of sensitivity, since it is unlikely that these techniques will reliably detect small micro-metastasis, which can be detected by SNB. If a LNM is then present with a very high likelihood, a neck dissection can be performed in one stage with the primary tumor resection, avoiding the need for SNB, whereas in other situations a SNB can be performed [27].

To enable SNB to be widely applicable and feasible worldwide, efforts should be directed towards finding innovative solutions and to adapt the lessons learned from research in other sites. The role of artificial intelligence (AI) and deep learning (DL) in the detection of metastasis in breast cancers following SNB is being researched with promising results [28]. DL algorithms are being explored to identify LNM in HNSCC [29].

Although SNB is a minimally-invasive surgical procedure, it is nonetheless an invasive approach requiring an additional neck incision. Yang et al. developed a three-stage DL model based on magnetic resonance imaging (MRI) with three-dimensional sequences that can serve as a practical and non-invasive tool for differentiating LNM in patients with oral SCC [30]. This model can detect LNM and decrease the occult metastasis rate from 15.30% to 1.64% in the primary cohort (area under the curve [AUC] 0.99), and from 20.44% to 5.69% in the external validation (AUC 0.81) [30]. Accordingly, this DL model can rightly classify the LNM in patients with cN0 neck by reducing the occult metastasis rate by up to 89.50%, subsequently resulting in a more appropriate treatment plan. This model achieved consistent and robust performance across three cohorts and even surpassed the performance of specialized radiologists. This strategy can assist clinicians in reducing occult LN metastasis rates for appropriate neck management. Another approach to reduce the need for SNB may be the use of AI to predict with a high positive predictive value the presence of LNM based on MRI characteristics of the primary tumor.

At the present time, END is not justified to treat early glottic larynx carcinoma in patients with HNSCC with clinically cN0 neck. In early oral SCC, growing evidence suggests that END should be substituted by SNB. A randomized phase II/III trial (NRG-HN006 study) is ongoing with enrollment of 686 patients with early-stage oral SCC and has, with the primary objective to determine if patient-reported neck and shoulder function and related quality of life at 6 months after surgery are superior with SNB compared to END [31]. The results of this trial are expected for 2031 and should provide a definitive answer. Although END is still recommended when SNB is not available, surveillance may be an option in thin oral cancer of < 5 mm thickness or in older and fragile patients [32]. In pharyngeal and supraglottic larynx SCC, END is still recommended even in early tumors. However, when a transoral approach is planned, SNB is a valuable option in centers with an extensive expertise. A systematic review has shown that SNB has also a very high accuracy in detecting occult metastasis in patients with pharyngeal and laryngeal SCC [33].

It is to be hoped that advances in AI will in the very near future lead to the detection of the absence of occult lymph node metastases with high accuracy, allowing many patients to avoid unnecessary surgery.

Author Contributions

Conceptualization and design: Marc Hamoir, Alfio Ferlito. Data acquisition: Marc Hamoir. Data analysis: Marc Hamoir. Drafting of the manuscript: Marc Hamoir. Critical revision of the manuscript: Marc Hamoir, Remco de Bree, Primoz Strojan, Nabil F Saba and Alfio Ferlito. Final approval of the version to be published: Marc Hamoir, Remco de Bree, Primoz Strojan, Nabil F Saba and Alfio Ferlito.

Funding

No funding or sponsorship was received for this Editorial or its publication.

Declarations

Conflict of Interest

Alfio Ferlito is an editorial board member of Oncology and Therapy. Alfio Ferlito was not involved in the selection of peer reviewers for the manuscript nor any of the subsequent editorial decisions. Marc Hamoir, Remco de Bree, Primoz Strojan and Nabil F. Saba have no other conflicts of interest to declare.

Ethical Approval

This article is based on previously conducted studies and does not contain any new studies with human participants or animals performed by any of the authors.