Incidence of Occult Lymph Node Metastasis in Primary Larynx Squamous Cell Carcinoma, by Subsite, T Classification and Neck Level: A Systematic Review.

BACKGROUND
Larynx cancer is a common site for tumors of the upper aerodigestive tract. In cases with a clinically negative neck, the indications for an elective neck treatment are still debated. The objective is to define the prevalence of occult metastasis based on the subsite of the primary tumor, T classification and neck node levels involved.


METHODS
All studies included provided the rate of occult metastases in cN0 larynx squamous cell carcinoma patients. The main outcome was the incidence of occult metastasis. The pooled incidence was calculated with random effects analysis.


RESULTS
36 studies with 3803 patients fulfilled the criteria. The incidence of lymph node metastases for supraglottic and glottic tumors was 19.9% (95% CI 16.4-23.4) and 8.0% (95% CI 2.7-13.3), respectively. The incidence of occult metastasis for level I, level IV and level V was 2.4% (95% CI 0-6.1%), 2.0% (95% CI 0.9-3.1) and 0.4% (95% CI 0-1.0%), respectively. For all tumors, the incidence for sublevel IIB was 0.5% (95% CI 0-1.3).


CONCLUSIONS
The incidence of occult lymph node metastasis is higher in supraglottic and T3-4 tumors. Level I and V and sublevel IIB should not be routinely included in the elective neck treatment of cN0 laryngeal cancer and, in addition, level IV should not be routinely included in cases of supraglottic tumors.


Introduction
The larynx is the site for the second most common cancer of the upper aerodigestive tract. While in some countries the incidence of larynx cancer has risen in the last forty years, due to smoking and drinking, in the USA and Northern Europe it has decreased dramatically [1,2].
Adequate treatment of larynx cancer centers around local control of the primary tumor, as well as control of clinical or occult neck lymph node metastasis. If the patient presents with positive lymph nodes in the lateral compartment of the neck and the planned treatment option for the primary tumor is surgery, there is no doubt that the preferred treatment is a neck dissection most likely followed by radiotherapy alone or in combination with chemotherapy [3]. The lymph node levels to be included in the dissection and the radiotherapy fields depend on the level of positive nodes found by palpation and radiological examination, as well as the laryngeal subsite(s) involved by cancer and T classification of the tumor. However, there still is an ongoing discussion about which neck levels should be included [4][5][6][7]. On the other hand, in cases with a clinically negative neck, the indications for and the extent of an elective neck dissection or elective neck irradiation are still debated vigorously in the literature, the scientific community and multidisciplinary tumor board meetings. Although, depending mainly on the subsite and T classification of the primary tumor, the final decision making should take into account the expected survival advantages as a result of elective treatment of the neck, as well as the functional and esthetic morbidity that may result from it.
The rationale to offer elective treatment of the neck in patients with larynx cancer is the potential presence of occult lymph node metastases, which can result in recurrence after successful initial treatment of the primary tumor. However, it has been reported that the results with elective neck dissection or irradiation are not superior to those of "watchful" surveillance without neck surgery or radiotherapy in early stage laryngeal cancer [8]. It is also known that the social and psychological make-up of some patients with larynx cancer is such that when they present with a recurrence in the neck after a "watchful" waiting policy, it is often at an advanced stage and salvage treatment may not be successful. Consequently, many clinicians prefer performing an elective neck dissection/irradiation in spite of the fact that tumor recurrence owing to adverse biological characteristics of the cancer may develop even after a neck dissection and radiation therapy. Therefore, the procedure may not offer any advantage in survival and the morbidity associated with an elective neck dissection may decrease the quality of life [9,10].
Central to the discussion of elective treatment of the neck is the ability to predict correctly the likelihood of occult metastasis in a given patient and the neck node levels that are at highest risk of containing metastases, which, therefore, need to be removed or irradiated electively. Unfortunately, neither modern imaging techniques nor molecular marker analysis of the primary tumor, nor sentinel lymph node biopsy [5,11], are generally felt to be accurate enough to be considered the standard of care.
The objective of this review is to assess the available literature in order to define the prevalence of occult metastasis in larynx cancer based on three key factors: subsite of the primary tumor (glottic or supraglottic), T classification and neck node levels involved. Information about bilateral metastasis was not included. The data are intended to provide a rationale for the indications and extent of elective neck treatment based on the specific factors analyzed.

Results
The primary search found 2025 studies. Only 36 studies with 3803 patients fulfilled the inclusion criteria for the review and are presented in Table 1 [10,. Most exclusions were based on indeterminate outcomes or because the primary treatment was with radiotherapy/chemoradiotherapy. Ten studies were excluded because cN+ patients were also included in the sample [47][48][49][50][51][52][53][54][55][56] and one because it was not possible to get full text [57]. Data from Deganello et al. [10], Pinilla et al. [23] and Zhang et al. [31] reported individual data about subsites and were divided in order to include their cases in different categories [glottic, supraglottic or larynx, not otherwise specified (LNOS)] (Table 1). Three studies from the same center had overlapping patients but they were analyzed independently because the numbers, outcomes and objectives were different [32,33,35].

Methodological Quality
The methodological quality of the included studies was moderate to good. The mean score was 8.3 ± 1.4 over a maximum score of 10. Only 4/36 (11%) of the studies scored lower than 7. The item with the lowest score was "Did the case series have complete inclusion of participants?", where 21/36 (58%) studies had a no/unclear result, followed by "Did the case series have consecutive inclusion of participants?", where 14/36 (39%) had a no/unclear result ( Figure 1).

Methodological Quality
The methodological quality of the included studies was moderate to good. The mean score was 8.3 ± 1.4 over a maximum score of 10. Only 4/36 (11%) of the studies scored lower than 7. The item with the lowest score was "Did the case series have complete inclusion of participants?", where 21/36 (58%) studies had a no/unclear result, followed by "Did the case series have consecutive inclusion of participants?", where 14/36 (39%) had a no/unclear result ( Figure 1).

Rate of Occult Metastasis
All studies, except two [34,39], reported an overall rate of occult metastasis of 18.7% (711/3803 patients). The pooled incidence was not calculated due to the expected clinical and statistical heterogeneity by subsites.
Was the condition measured in a standard, reliable way for all participants included in the case series?
Were valid methods used for identification of the condition for all participants included in the caseseries?
Did the case series have consecutive inclusion of participants?
Did the case series have complete inclusion of participants?
Was there clear reporting of the demographics of the participants in the study?
Was there clear reporting of clinical information ofthe participants?
Were the outcomes or follow up results of cases clearly reported?
Was there clear reporting of the presenting site(s)/clinic(s) demographic information?
Was statistical analysis appropriate?
Methodological quality of included studies Yes No Unclear

Rate of Occult Metastasis
All studies, except two [34,39], reported an overall rate of occult metastasis of 18.7% (711/3803 patients). The pooled incidence was not calculated due to the expected clinical and statistical heterogeneity by subsites.

Occult Metastasis by T Classification
Five studies reported data on the T1 stage [16,21,23,29,31]. Five of 105 patients (4.8%) had occult metastasis. It was impossible to calculate a pooled incidence as Elo et al. [21], Pinilla et al. [23] and Lim et al. [29] did not observe any event in this category.

Occult Metastasis by T Classification
Five studies reported data on the T1 stage [16,21,23,29,31]. Five of 105 patients (4.8%) had occult metastasis. It was impossible to calculate a pooled incidence as Elo et al. [21], Pinilla et al. [23] and Lim et al. [29] did not observe any event in this category.

Discussion
The role of elective neck treatment for larynx cancer patients with cN0 disease has been a subject of debate for many years [3,6,7]. This has been further complicated by the advent and effectiveness of non-surgical organ preservation protocols, as well as the introduction of endoscopic surgery as primary tumor treatment [5]. Moreover, since elective neck dissection may be a risk factor for pharyngocutaneous fistula after total laryngectomy, unnecessary neck dissection should be avoided [58]. The same holds true when radiotherapy is chosen as treatment, where extending the radiation fields unnecessarily can lead to serious complications that can adversely affect the quality of the patient's life [59]. Given these issues, the present study was undertaken in an attempt to provide more homogeneous data from the available literature by quantitating the incidence of occult nodal disease based on mucosal site and T classification. Furthermore, an attempt was made to subdivide the presence of nodal disease based on neck level in order to elucidate the application of the various types of selective neck dissections.
Lymphatic drainage patterns have been widely studied since the 1960s. The studies of Welsh et al. [60][61][62] demonstrated the differential drainage patterns of the larynx according to subsites. These studies demonstrate that the supraglottis is rich in lymphatic networks while for instance vocal cords do not have lymphatic vessels in their free margin. Liu et al. [63] demonstrated that the inferior surface of the vocal cord does have an important lymphatic network, with a similar drainage as the subglottis. These anatomical factors explain the pattern of distribution of lymph node metastasis, with levels II and III [6,20,39] being the most frequently involved. Metastasis to level VI is not usually reported, except in specific studies on primary subglottic tumors or advanced tumors involving this area.
In this review, the evaluation of occult lymph node metastasis by subsite demonstrated a higher risk for supraglottic (19%) compared to glottic tumors (8%). However, there exists a wide variation in the reported results. Unfortunately, the most common clinical and histopathological characteristics of the primary tumor provided by the studies analyzed do not help to elucidate the cause of such heterogeneity. For example, for larynx tumors NOS, Deganello et al. [10] reported a low incidence of

Discussion
The role of elective neck treatment for larynx cancer patients with cN0 disease has been a subject of debate for many years [3,6,7]. This has been further complicated by the advent and effectiveness of non-surgical organ preservation protocols, as well as the introduction of endoscopic surgery as primary tumor treatment [5]. Moreover, since elective neck dissection may be a risk factor for pharyngocutaneous fistula after total laryngectomy, unnecessary neck dissection should be avoided [58]. The same holds true when radiotherapy is chosen as treatment, where extending the radiation fields unnecessarily can lead to serious complications that can adversely affect the quality of the patient's life [59]. Given these issues, the present study was undertaken in an attempt to provide more homogeneous data from the available literature by quantitating the incidence of occult nodal disease based on mucosal site and T classification. Furthermore, an attempt was made to subdivide the presence of nodal disease based on neck level in order to elucidate the application of the various types of selective neck dissections.
Lymphatic drainage patterns have been widely studied since the 1960s. The studies of Welsh et al. [60][61][62] demonstrated the differential drainage patterns of the larynx according to subsites. These studies demonstrate that the supraglottis is rich in lymphatic networks while for instance vocal cords do not have lymphatic vessels in their free margin. Liu et al. [63] demonstrated that the inferior surface of the vocal cord does have an important lymphatic network, with a similar drainage as the subglottis. These anatomical factors explain the pattern of distribution of lymph node metastasis, with levels II and III [6,20,39] being the most frequently involved. Metastasis to level VI is not usually reported, except in specific studies on primary subglottic tumors or advanced tumors involving this area.
In this review, the evaluation of occult lymph node metastasis by subsite demonstrated a higher risk for supraglottic (19%) compared to glottic tumors (8%). However, there exists a wide variation in the reported results. Unfortunately, the most common clinical and histopathological characteristics of the primary tumor provided by the studies analyzed do not help to elucidate the cause of such heterogeneity. For example, for larynx tumors NOS, Deganello et al. [10] reported a low incidence of occult metastasis of 12% while Tsushima et al. [46], Xu et al. [40] and Candela et al. [13] reported incidences higher than 32%. Heterogeneity was also found in supraglottic tumors. For example, Lawson et al. [36] found a 32.5% incidence of occult metastasis while Tu et al. [19] found 8.3%, but we are unable to explain these dramatic differences with the available data. Factors related to the indication for neck dissection, the policy of wait and see without providing the selection criteria used and the number of patients observed may explain these differences. Other factors responsible for this heterogeneity may be differences in diagnostic work-up and in the depth of histopathological examination of the neck dissection specimen. Routine histopathological examination of the neck dissection specimen can miss micrometastases in up to 15.2% and therefore the real incidence of occult metastases may be underestimated [64]. Serial sectioning and immunohistochemistry, as used in the examination of a sentinel lymph node biopsy, will detect micrometastases better, but is costly and can introduce a verification bias due to the higher number of sections. Finally, the use of more accurate diagnostic imaging techniques will result in lower reported incidence rates of occult metastases than the use of palpation only.
Analysis by T classification helps to reduce the variability among the patients studied and provides an improved understanding of the risk of occult nodal disease for each subsite. The assessment of occult lymph node metastases according to T classification confirms the increasing risk with increasing T stage. In fact, at the same T classification, supraglottic tumors always had a higher incidence in comparison to glottic tumors (16.5% vs. 4.7% in T2 tumors and 23.8% vs. 14.4% in T3 tumors). We found that early supraglottic tumors have a lower rate of occult metastasis than those with more advanced tumors. However, these differences were not significantly different. A remarkable result was found regarding glottic T3 tumors, which had a lower incidence than -T2 supraglottic tumors.
The data from this analysis support the notion that neck levels I and V are at low risk for occult nodal disease in patients with cN0 laryngeal cancer. For level I, the incidence derived from the pooled data from two studies is 2.4%. However, it should be pointed out, as Candela et al. [13] state in their paper, that Level I is rarely involved and that the involvement they observe occurred usually in cases with neck node metastases in levels II, III or IV (75% of the time). It should also be noted that Level I involvement in their study occurred in T3 or T4 primary tumors exhibiting histologic extra laryngeal spread. While these two studies did not provide information about sublevel (IA or IB) involvement, others suggest that the risk of involvement of sublevel IA in cN0 larynx tumors is insignificant [4]. Similarly, the rate of occult metastases in level V was only 0.4%. Thus, routine elective treatment of these levels is not indicated in patients undergoing elective lymphadenectomy to remove potential occult disease among patients with laryngeal cancer and a cN0 neck. This recommendation is also supported by the results of a prospective trial that compared selective neck dissection of levels II to IV vs. modified radical neck dissection [65]. With regard to the inclusion of Level IV in elective treatment, our study found a very low rate of involvement of the nodes in this level in patients with supraglottic cancer, thus supporting the practice of not including them routinely in those cases. Similarly, our results regarding sublevel IIB indicated a 0.5% rate of occult disease supporting the practice of not including this sublevel routinely when performing a neck dissection for occult disease [66,67]. Some of these studies follow a uniform pattern regarding the lymphatic levels resected, while others use intraoperative findings or even frozen section results to guide the extent of the dissection. There is also still a lack of standardization in the nomenclature and in the execution of neck dissections, despite several initiatives to standardize them [68].
Even though there is a large number of publications dealing with the surgical treatment of the neck in larynx cancer, the quality of many studies is moderate, as shown in Figure 1. This is due, in part, to a selection bias of surgical studies, which may lead to a selection of more favorable primary tumors for surgery and those unfavorable for chemo/radiotherapy. Another factor is that most of these studies are retrospective case series, which are again prone to selection bias that can affect the pooled results. Other weaknesses are related to patient selection in individual studies. Furthermore, one study included a small number of patients with hypopharynx tumors, but the corresponding effect on the calculation of the pooled incidence is minimal [34]. Finally, during the more than 50-year period covered by our analysis, pre-therapeutic diagnostic procedures and protocols have changed dramatically, and thus the accuracy of the assessment of the neck stage prior to treatment.
Obviously, the few factors analyzed in this review (T classification and laryngeal subsite) are clearly insufficient to accurately predict the risk of lymph node metastasis. Other tumor features also have been reported to be relevant in the development of lymph node metastases, such as the infiltrating pattern of growth of the tumor, "tumor budding" and lymphovascular and perineural invasion [69][70][71][72]; molecular changes, such as loss of N33, STK11 and TP53 [73]; as well as immune alteration, such as HLA-E overexpression [74]. However, the means to assess these characteristics are not readily available in all institutions. Furthermore, most of these can only be evaluated after resection of the primary tumor, which precludes its clinical use to make decisions about elective neck dissection. The use of an SLN biopsy could be explored in some of these cases [5,36].

Materials and Methods
A search was conducted using the MESH and free-text terms "cancer", "larynx" and "neck dissection" in the PubMed, Embase and LILACS databases for studies published between 1966 and March 2019. An expanded search was conducted, and references were explored to identify additional articles. We did not restrict the publication language. All selected studies provided the rate of occult metastases in cN0 larynx squamous cell carcinoma patients primarily treated. All abstracts were reviewed by the authors. Those related with the subject were selected for further analysis. Those studies in which it was not possible to isolate specific data about larynx tumors were excluded. A differential analysis was made when studies did not report outcomes by subsites. The main outcome measure was the incidence of occult metastasis in neck dissection specimens based on the mucosal subsite, T classification and neck levels involved. Studies dealing with organ-preserving approaches were not included.
Data were collected based on sample size, patient characteristics and outcomes. Data from each study were extracted and recorded in an Excel spreadsheet. The unit of analysis was the study and statistical analysis was performed with Excel (Microsoft). The pooled incidence (95% confidence interval (CI)) was calculated for each outcome with random effects analysis because this method is a conservative summary estimate and incorporates between-and within-study variance [75]. For data with 0 events, a correction with 0.1 was made in order to obtain calculations. Studies that reported the frequency of occult metastases but did not discriminate it by subsite were classified as larynx tumors not otherwise specified (LNOS). The others were classified as having glottic and supraglottic tumors. The rate of positive nodes by neck level was calculated from the number of positive lymph nodes in each level divided by the total number of patients. Statistical heterogeneity was calculated with the Higgins I2 statistic. This represents the amount of variation in incidence of the included studies and when found it is recommended to look for an explanation based on the clinical factors, methods or analysis. Results of the intervention effects are presented with a forest plot graph.
Methodological quality of the included studies was evaluated with the Checklist for Case Series of the Joanna Briggs Institute Critical Appraisal tools for use in systematic reviews [76].

Conclusions
In summary, the incidence of occult lymph node metastasis in cN0 larynx cancer depends on the subsite and T classification, where supraglottic and T3-4 tumors have the higher percentages. The final recommendations for the patient with larynx cancer and a clinically negative neck are the following: For cN0 T1/2 glottic tumors, neck dissection is not recommended; for cN0 T1/2 supraglottic tumors, level I, level IV, level V and sublevel IIB should not be routinely included in the elective neck treatment; and for cN0 T3/4 supraglottic tumors and cN0 T3/4 glottic tumors, level I, level V and sublevel IIB should not be routinely included in the elective neck treatment. This study did not allow us to make a recommendation about bilateral neck dissection in cN0 T1/2 supraglottic tumors.
Supplementary Materials: The following are available online at http://www.mdpi.com/2072-6694/12/4/1059/s1, Figure S1: Pooled incidence of occult lymph metastasis in larynx cancer by subsite, Figure S2: Pooled incidence of occult lymph metastasis in supraglottic larynx tumor, T1-2 stage, Figure S3: Pooled incidence of occult lymph metastasis in larynx cancer, subsite NOS (not otherwise specified), T3-4 stages, Figure S4: Pooled incidence of occult lymph metastasis by levels for larynx tumor, subsite NOS (not otherwise specified), Figure S5: Pooled incidence of occult lymph metastasis by levels for supraglottic tumor, Figure S6: Pooled incidence of occult lymph metastasis for glottic tumor by stage, Figure S7: Pooled incidence of occult lymph metastasis for supraglottic tumor by stage, Table S1: Data on involvement of lymph node levels by occult neck metastasis in cN0 larynx cancer.