The accuracy of head and neck carcinoma sentinel lymph node biopsy in the clinically N0 neck
Sentinel lymph node (SLN) biopsy originally was described as a means of identifying lymph node metastases in malignant melanoma and breast carcinoma. The use of SLN biopsy in patients with oral and oropharyngeal squamous cell carcinoma and clinically N0 necks was investigated to determine whether the pathology of the SLN reflected that of the neck.
Patients undergoing elective neck dissections for head and neck squamous cell carcinoma accessible to injection were enrolled into our study. Sentinel lymph node biopsy was performed after blue dye and radiocolloid injection. Preoperative lymphoscintigraphy and the perioperative use of a gamma probe identified radioactive SLNs; visualization of blue stained lymphatics identified blue SLNs. A neck dissection completed the surgical procedure, and the pathology of the SLN was compared with that of the remaining neck dissection.
Sentinel lymph node biopsy was performed on 40 cases with clinically N0 necks. Twenty were pathologically clear of tumor and 20 contained subclinical metastases. SLNs were found in 17 necks with pathologic disease and contained metastases in 16. The sentinel lymph node was the only lymph node containing tumor in 12 of 16.
The SLN, in head and neck carcinomas accessible to injection without anesthesia, is an accurate reflector of the status of the regional lymph nodes, when found in patients with early tumors. Sentinel lymph nodes may be found in clinically unpredictable sites, and SLN biopsy may aid in identifying the clinically N0 patient with early lymph node disease. If SLNs cannot be located in the neck, an elective lymph node dissection should be considered. Cancer 2001;91:2077–83. © 2001 American Cancer Society.
The sentinel lymph node concept states that a tumor spreads via lymphatics to the first echelon lymph node encountered in the lymph node basin, and this spread is embolic in nature.1 If the sentinel, or first echelon, lymph node can be identified and examined for the presence of tumor metastases, the need to perform an elective staging lymph node dissection is negated.2 The concept has been applied mainly to breast carcinoma3 and malignant melanoma,4 and in these carcinomas, sentinel lymph nodes free from tumor suggest a regional lymph node basin free from tumor with a high degree of accuracy.
In squamous cell carcinoma of the upper aerodigestive tract, the investigation and treatment of the patient with a clinically N0 neck remains controversial.5 Depending on T classification, approximately 30% of patients with head and neck squamous cell carcinoma have subclinical metastases in the neck,6 and knowledge of lymph node disease alters management. Currently, although computed tomography (CT) and magnetic resonance imaging (MRI) scans commonly are used to classify the neck, their overall accuracy is limited to approximately 70%, and the only highly accurate means of identifying lymph node disease is to perform a staging lymph node dissection.7 For early disease, clinicians are reluctant to perform a neck dissection, because up to 85% of patients will not benefit, yet adopting a wait-and-see policy to all necks will result in a high proportion of patients subsequently developing late stage regional failure.8
Initial results of the sentinel lymph node procedure in head and neck carcinoma have been reported with mixed success. In a series of 16 cases, Pitman et al. were unable to find any blue lymph nodes in patients injected with blue dye alone,9 and in a series of five cases using radiocolloid alone, Koch et al. remained unconvinced of its role in the management of patients with head and neck carcinoma.10 The first case report of a successful sentinel lymph node biopsy in head and neck carcinoma by using radiocolloid to trace the first echelon lymph node was performed in 1996 by Alex and Krag on a patient with a supraglottic carcinoma,11 and in 1998, Bilchik et al. reported the use of sentinel lymph node biopsy in a variety of neoplasms, including head and neck carcinoma.12 More recently, we described our method for successful sentinel lymph node biopsy using blue dye and radiocolloid.13 Werner et al. have had success with the procedure in the clinically N0 neck,14 and Alex et al. have published their experience using radiocolloid alone.15
This study was performed to investigate whether the sentinel lymph node concept is applicable to patients with clinically N0 necks in oral or oropharyngeal carcinoma and to map the spread of colloid and dye from primary site to sentinel lymph node.
Patients with oral or oropharyngeal carcinomas were invited to enter our study. Tumors were amenable to injection without the need for general anesthesia, and only patients whose planned primary treatment included a formal elective neck dissection were enrolled. This comprised patients considered at risk of occult metastases. Staging of the neck was performed at the time of clinical examination during general anesthesia and CT or MRI were not used. Ethical approval was obtained from the local ethics committee before commencing our study.
Although the method used for sentinel lymph node biopsy has been described previously,13 it is summarized briefly here. The day before surgery, patients attended the Nuclear Medicine Department, where up to 40 MBq of Tc99m-labeled colloidal human serum albumin—either Nanocoll or Albures (Nycomed Amersham, High Wycombe, Bucks, UK) in approximately 0.5–1 mL of saline—was injected at as many points as necessary in an attempt to completely surround the tumor on its deep and lateral aspects. All patients were offered topical local anesthetic immediately before their injection of radiocolloid. All patients with oral pain or trismus were strongly encouraged to accept the offer of local anesthetic, and those patients with no oral pain or trismus were informed that the procedure was of only mild discomfort. The needle used was a size 27F-gauge needle, which was permanently attached to the syringe to avoid inadvertent separation of the needle from the syringe. The needle was passed into one to four areas as many times as necessary in an attempt to completely surround the tumor with radiocolloid, both on the deep aspect as well as the lateral aspect to ensure radiocolloid was present at the “metastasizing edge” of the tumor.
Initially, Albures (Nycomed Amersham) was the colloid used in all cases, but after our initial learning curve, Albures, with a mean particle size of 500 nm, subsequently was only used for primaries of the tongue and floor of mouth, and Nanocoll (Nycomed Amersham), with a mean particle size of 80 nm, was used for tumors at other sites. Static lymphoscintigraphy was performed at 15 minutes, 30 minutes, and 60 minutes after injection, or until the first appearance of sentinel lymph nodes within the neck. During surgery, approximately 0.5–2 mL of Patent Blue V dye (Laboratoire Guerbet, Aulnay-Sous-Bois, France) was injected into the same site as radiocolloid, and skin flaps suitable for a neck dissection were raised. The sternomastoid muscle was retracted posteriorly to expose the deep chain of cervical lymphatics, and attempts were made at identifying sentinel lymph nodes. Blue stained lymphatics were followed to blue lymph nodes, and radioactive lymph nodes were identified with a Neoprobe 1500 hand-held gamma probe (Neoprobe Corp, Columbus, OH). Sentinel lymph nodes were identified in their lymph node level and were labeled according to color and presence of radioactivity. Radioactivity was confirmed within the sentinel lymph node ex vivo. An appropriate neck dissection completed the surgical procedure. At the time of surgery, lymph node levels were marked with metal disks. The sentinel lymph nodes were fixed in 10% neutral buffered formalin and after fixation were bisected through their longest axis. If the thickness of the halves was more than 2 mm, the slices were trimmed further to provide additional 2-mm-thick blocks. The remainder of the neck specimen was dissected after fixation, and all lymph nodes greater than approximately 2.5 mm in maximum dimension were identified in their anatomic groups. Each lymph node was bisected through its longest axis, and one-half was processed for histologic examination. One hematoxylin and eosin (H & E)–stained section was prepared from each block and was examined for the presence of lymph node involvement by tumor.
Between July 1998 and March 2000, 40 necks were explored for sentinel lymph nodes in 37 patients with biopsy proven squamous cell carcinoma. Each neck side was considered a single case. Ten cases have been reported previously in our pilot study and are included in the data.
Thirty-nine cases were classified N0, and one case was classified clinically as Nx, because of palpable cervical lymphadenopathy from longstanding non-Hodgkin lymphoma. The clinical classification of the primary tumor was T1 in 14, T2 in 14, T3 in 3, and T4 in 9. The site of the primary squamous carcinoma was the tongue in 21, floor of mouth in 10, soft palate in 4, retromolar trigone in 3, the buccal mucosa in 1, and lower alveolus in 1. The male to female ratio was 2:1. The mean age of patients was 59 years (range, 29–84).
Sentinel lymph nodes were found in 36 of the 40 necks (90%). In the four necks in which sentinel lymph nodes were not identified, one was a patient with a T2 anterior floor of mouth tumor undergoing bilateral neck dissections with palpable lymphadenopathy from squamous cell carcinoma on one side of the neck, and no sentinel lymph node was identified in the uninvolved neck. Three were patients with well lateralized tongue carcinomas undergoing unilateral neck dissections. These three tongue carcinomas were classified as T2 in one case and T4 in two. In all four cases, the neck dissection specimens were examined for blue and radioactive lymph nodes ex vivo; however, none was found. In three of four cases in which sentinel lymph nodes were not detected, the neck was classified pN+. In two cases, the primary tumor was a T4 tongue carcinoma, and the third was a T2 tongue tumor. One was classified pathologically as pN2b and two were pN1. In the case classified N2b, two lymph nodes were involved with tumor, one of which was extensively replaced by squamous cells and showed early extracapsular spread. In the two cases classified as N1, early lymphatic spread was observed in one lymph node, and in the other case, one small tumor deposit of approximately 200 microns in maximum dimension was observed within a lymphatic channel in the lymph node capsule.
In 36 cases in which sentinel lymph nodes were identified, 90 sentinel lymph nodes were found in total. Forty-one lymph nodes were both radioactive and blue (“hot blue” lymph nodes), 35 were hot only, and 14 were blue only. The median number of sentinel lymph nodes per neck was 2 with a mean of 2.2 (range, 0–6).
Sentinel lymph nodes were found in lymph node levels I to V and are summarized in Table 1. Two sentinel lymph nodes were found in level IIb, in the triangle bordered by the skull base, sternomastoid muscle and the spinal accessory nerve. Two patients had areas of radioactivity, suggestive of sentinel lymph nodes, identified out with the neck during lymphoscintigraphy. One patient had a T1 left body of tongue squamous cell carcinoma, and the other had a T2 left soft palate carcinoma. In both cases, lymphoscintigraphy identified the left tonsillar region as the site, but tonsillectomies were not performed. The necks of both patients were classified as pN0.
Table 1. No. of Sentinel Lymph Nodes Found by Lymph Node Level and Involvement by Squamous Cell Carcinoma
Eight hundred forty-nine lymph nodes, including sentinel lymph nodes, from 40 necks were examined. A radical or modified radical neck dissection was performed in 22 cases, a supraomohyoid neck dissection in 3, and a selective neck dissection excising levels I–IV in 15. In two cases of floor of mouth carcinomas, the planned procedure of a supraomohyoid neck dissection was converted to one that included level IV, because the sentinel lymph nodes were located inferior to the omohyoid muscle, and in one of these cases the sentinel lymph node in level IV contained tumor. Table 2 summarizes the number of lymph nodes examined and containing tumor from the neck dissections.
Table 2. No. of Lymph Nodes Examined and Containing Tumor by Lymph Node Level
Twenty necks were classified pathologically as N0 and 20 as pathologically N positive (9 were pN1, 8 were pN2b, and 3 were pN2c). In the N positive group (pN+), sentinel lymph nodes were found in 17 and contained tumor with conventional histology in 16. In these 16 necks from which sentinel lymph nodes containing tumor were harvested, the sentinel lymph nodes were the only lymph nodes containing tumor in 12. Thus, in only four cases, nonsentinel lymph nodes contained tumor in the presence of pathologically positive sentinel lymph nodes. In three of these four necks, one additional lymph node was found to contain tumor, and in one case two nonsentinel lymph nodes contained tumor. The details of these four cases are shown in Table 3.
Table 3. Anatomic Location of Nonsentinel Lymph Nodes Containing Tumor, When Nonsentinel Lymph Nodes Were Found to Harbor Metastases
There was one false-negative sentinel lymph node in a patient with an extensive T4 squamous cell carcinoma of the floor of mouth in whom negative sentinel lymph nodes were found in levels I and II, and tumor was found within a nonsentinel level II lymph node.
When a sentinel lymph node was found, the sensitivity of the procedure was 94% (95% confidence interval, 82–100%). All sentinel lymph nodes in patients with pN0 necks were free of tumor (specificity 100%), and the sentinel lymph node correctly predicted the positivity or negativity of the neck in 35 of 36 cases (overall accuracy, 97%). Forty-one sentinel lymph nodes were harvested in total from the 20 pN+ necks, and tumor was found in 18 lymph nodes. Table 4 summarizes the presence of tumor, radioactivity and blue dye within all sentinel lymph nodes harvested from pN+ necks.
Table 4. Sentinel Lymph Nodes in the pN+ Neck According to Presence of Tumor, Dye, and Radiocolloid
Table 5 details the site of the primary tumor, the number of sentinel lymph nodes found, the pathologic classification of the neck, and the number of sentinel lymph nodes containing tumor for each case.
Table 5. Position of Tumor, No. of Lymph Nodes per Case, pN Classification of Neck and No. of Sentinel-Lymph Nodes Containing Tumor for All Patients
This study was performed to determine if the sentinel lymph node concept was valid for head and neck squamous cell carcinoma in necks with impalpable metastases undergoing elective neck dissections. We conclude that the sentinel lymph node, when identified using a combination of the hand-held gamma probe and blue dye visualization, reflects the positivity or negativity of the neck with a high degree of accuracy. We found that sentinel lymph node pathology using conventional H & E stains reflected that of the neck dissection in 94% of cases with impalpable disease, but only in those cases when the SLN was found. Frozen section analysis was not used in this study, thus avoiding a further variable and ensuring that all material was available for pathologic analysis. The accuracy of frozen section has been questioned in melanoma and breast carcinoma sentinel lymph node biopsy.16, 17
Sentinel lymph node biopsy is technically challenging and difficulties with identifying sentinel lymph nodes may be encountered during lymphoscintigraphy and surgery. Sentinel lymph nodes may be too close to the primary injection site to be discernible by the gamma camera, and this is particularly troublesome for sentinel lymph nodes in level I for a primary tumor located in the floor of mouth. Although the use of lead shields and software masking may highlight level I sentinel lymph nodes, if the procedure is to find a role in the management of the clinically N0 neck for floor of mouth lesions, we would recommend exploration of level I in all cases in which the primary tumor is located in the floor of mouth. During surgical exploration of level I, the hand-held gamma probe will detect scatter and shine-through from the primary site; the use of sterilized lead plates will aid in isolating radioactivity from lymph nodes. Blue dye visualization may be the primary means of identifying lymph nodes in level I, with the hand-held probe being used to confirm the presence of radiocolloid within the lymph node ex vivo. Last, removal of the primary tumor does not remove all radioactivity from the injection site, despite adequate tumor resection margins, although the reduction in radioactivity within the primary site can aid in the subsequent identification of hot lymph nodes close to the primary tumor.
Both blue dye and radiocolloid are required for sentinel lymph node biopsy to be successful. Two sentinel lymph nodes containing tumor were blue but contained no radiocolloid, and six sentinel lymph nodes containing tumor were hot but had no blue dye within them. In finding the blue lymph nodes, preoperative lymphoscintigraphy aided in localizing the presence of blue dye. Because the pharmacodynamics of blue dye will be different from radiocolloid, it is not surprising that some sentinel lymph nodes are identified with one modality only.
Sentinel lymph nodes have been identified in clinically unpredictable sites. Traditional anatomic teaching from cadaveric studies has suggested that lymph drainage may be sequential from level I to the upper jugular chain and thence to the lower jugular chain. This has given rise to the misconception of “skip metastases,” in which tumor metastases bypass upper lymph nodes in levels I and II and are found in lower levels. If the anatomic location of sentinel lymph nodes is a reflection of direct lymphatic drainage, the findings that sentinel lymph nodes may be located at various levels in the neck indicates that tumor drainage is dependent on individual anatomic pathways in some cases.
In most cases, sentinel lymph nodes have been the only lymph nodes with tumor in the involved neck (12 of 16 cases). When nonsentinel lymph nodes also are involved with tumor, these occasionally have been identified at higher anatomic levels in the neck. This phenomenon may be a reflection of tumor spread from the sentinel lymph node to a second echelon lymph node at a higher anatomic level.
In one case of a patient with long standing cervical lymphadenopathy from a low grade non-Hodgkin lymphoma, the neck was classified clinically Nx. The primary tumor was a T1 lateral tongue carcinoma, and the neck was found to contain squamous cell carcinoma in three lymph nodes in levels II, III, and IV. Three sentinel lymph nodes were found in levels III and IV, with one of the two “cold blue” lymph nodes from level IV containing tumor. Despite the presence of palpable lymphadenopathy from a different pathologic process, sentinel lymph node biopsy correctly identified lymph node disease from the squamous cell carcinoma. In such cases, lymph nodes with increased hydrostatic pressure may divert lymphatic flow, so that sentinel lymph nodes are found at clinically aberrant locations. In these cases, the sentinel lymph node location may reflect the true anatomic drainage pathway of the primary site. Although this article addresses the use of sentinel lymph node biopsy in the clinically N0 neck, the use of sentinel lymph node biopsy in this patient led to noteworthy and significant findings. If sentinel lymph node biopsy finds a role in the management of patients with head and neck carcinoma, it will be in the group of patients in which the presence of tumor metastases within the lymph nodes is uncertain as in this patient.
On average, more than one sentinel lymph node will be found in a single neck. Because the oral cavity has a rich plexus of lymphatic vessels, these findings may reflect the diverse lymphatic drainage pathways to the neck.
Sentinel lymph nodes were not found in 4 cases (10%), and this is similar to cutaneous melanoma of the head and neck in which the rate of sentinel lymph node identification varies between 90% and 96%.18–21 We cannot comment on the learning curve for head and neck carcinoma sentinel lymph node biopsy, but in our continuing experience of more than 100 cases, the rate of failure to identify sentinel lymph nodes remains much the same.
In this series, sentinel lymph nodes were not identified in four cases. Three cases were tongue tumors: in all cases, the tumors were indurated and ulcerated. Blue dye leaked out of the injection site into the oral cavity at the time of surgery. We presume the colorless injected colloid followed the same pathway, preventing lymphatic uptake of both dye and radioactivity.
The remaining case was a neck from a patient undergoing bilateral neck dissections for a tumor close to the midline. In this case, lymphoscintigraphy identified drainage to one side of the neck only, and the lack of sentinel lymph nodes on the contralateral, clinically clear neck was most likely a reflection of the true drainage pattern of the tumor. The decision to perform bilateral neck dissections on patients with tumors close to the midline can be difficult. Because the role of sentinel lymph node biopsy has yet to be defined in patients with head and neck carcinoma, it seems reasonable to perform the technique on a variety of patient types that are undergoing a neck dissection to address this question, at least in initial studies. For this reason, we chose to include such patients in the group studied.
Sentinel lymph node pathology did not reflect that of the remaining neck dissection in one case. There thus was one false-negative sentinel lymph node. In this case, the patient had an extensive T4 tumor of the floor of mouth, invading into the tongue, mandible, and skin. In this case, the injection site of colloid and dye was not at the metastasizing edge of the tumor, because the small volume of injection was insufficient to completely surround the tumor on its deep and lateral aspects. If sentinel lymph node biopsy is to be used as a staging tool for head and neck carcinomas, it is important to inject around the whole tumor periphery.
If sentinel lymph node biopsy finds a role in the management of the clinically N0 neck as a staging technique, then in cases in which a sentinel lymph node is not identified a formal elective neck dissection should be considered for staging purposes. The technique is only reliable when a sentinel lymph node is identified.
We propose a multicenter study to investigate sentinel lymph node biopsy in patients with T1 or T2 lesions of the oral cavity and oropharynx and clinically N0 necks. If the sentinel lymph node is free of tumor, no elective treatment of the neck will be undertaken. Multiple serial sections of the sentinel lymph node will be examined using routine histology and immunohistochemistry, and if viable tumor is found within the sentinel lymph node, patients will undergo a neck dissection and other treatment to the neck according to local protocols. Sentinel lymph node biopsy can be a technically challenging procedure, but we believe it should be investigated further by those clinicians competent at the procedure, in the context of a multicenter trial. We have commenced such a trial in Europe, details of which are published on our web site (http://www.canniesburn.org).