What the clinician wants to know: surgical perspective and ultrasound for lymph node imaging of the neck.

Imaging of lymph node metastases in the neck can have two major indications: (1) prognosis and assisting with choice of treatment; (2) staging and detection of clinically occult metastases in different levels of the neck. Both indications are discussed. The role and limitations of US and US-guided fine-needle aspiration cytology are also reviewed. International Cancer Imaging Society.

Prognosis

Lymphatic metastasis is the most important mechanism in the spread of most head and neck carcinomas. The presence of lymph node metastases determines to a great extent the chances of locoregional cure or the development of distant metastases. The incidence of distant metastases in head and neck squamous cell carcinoma (HNSCC) ranges from 4% in clinical studies to over 50% in autopsy studies [1-3]. The lungs, the skeletal system, and the liver are the most frequent sites of distant metastases [3]. De Bree et al. [4] have shown that in patients with three or more nodal metastases, bilateral or low jugular lymph node metastases, large lymph node metastases (≥6 cm) or second primary malignancies, a search for distant metastases is warranted based on a high incidence. For screening of distant metastases, a chest computed tomography (CT) scan is the first choice modality, more effective than bone scans or ultrasound (US) of the liver [4].

Although lymph node level is used only in N-staging of nasopharyngeal carcinomas, several studies have shown its prognostic importance in other sites of the head and neck [5, 6]. Accurate depiction of the number and level of neck node metastases becomes important if selective neck dissections are considered or when radiotherapy is the primary treatment and no histopathology will become available [7, 8]. The accuracy of CT, US, and magnetic resonance imaging (MRI) for the assessment of the exact number of metastases or levels involved has not been studied. It is likely that imaging is not very accurate for this, as relatively large detectable metastases are very often accompanied by small undetectable micrometastases [9, 10]. However, CT and MRI may be helpful for the detection of retropharyngeal and possibly paratracheal and mediastinal lymph nodes. This may lead to a more expanded surgical treatment of the neck or extension of the radiotherapy fields. Furthermore, the presence of these lymph nodes is important for prognosis [11]. Unfortunately, paratracheal and retropharyngeal node metastases are often very small and difficult to detect at CT or US [12, 13].

Assessment of tumour volume has been shown to be an important prognosticator in laryngeal, and to a lesser extent, pharyngeal carcinoma [14-16]. Assessment of nodal volume [15, 17, 18] is studied less, but also has clinical importance in predicting outcome. A volume over 110 ml is a poor prognostic indicator in patients treated with radiotherapy [19].

Necrosis in lymph nodes, as depicted at CT or MRI can also be important to predict response to radiotherapy or chemoradiation. As nodal necrosis is a sign of tumour hypoxia, it can be anticipated that these lymph nodes respond less to radiotherapy. Indeed, Dietz has shown that diminished vascularity in lymph nodes, as shown with duplex Doppler, is a poor predictive sign for patients treated with chemoradiation [18]. In another study, it was shown that if the lymph node necrosis area at CT encompasses more than one-third of the total volume, the survival rate drops dramatically [19, 20]. Recently, it has been shown that with use of functional MRI, tumour hypoxia can be measured and chemoresistance predicted [21]. Another method of assessing tumour hypoxia is Tc-imidazole scintigraphy [22]. Recently, the standardized uptake value of fluorodeoxyglucose positron emission tomography (FDG-PET) was shown to be a predictive marker for patients treated with radiotherapy as well [23].

Extranodal spread is radiologically characterized by ablation of fat planes and irregular nodal borders. For this feature, Close et al. have reported that CT could only identify extranodal spread in large nodes [24], and Som reported a sensitivity of 100% [25]. Yousem et al. reported an accuracy of CT of 90%, whereas in their study MRI had an accuracy of 78% [26]. On the other hand, Carvalho studied the value of CT in detecting extranodal spread and found a sensitivity of 63% and a specificity of 60% [27]. In a study by Woolgar, in 16% of the cases N0 at CT, extranodal spread (ENS) was present at pathology [28]. Recently, King et al. have shown that both CT and MRI have an accuracy in the order of 73%–80% to detect ENS [29]. We have previously looked at the inter- and intra-observer variation in diagnosing ENS histopathologically [30]. From these studies it became clear that even among pathologists there is no consensus on the criteria of ENS and that frequently it is a subtle feature not detectable radiologically. In our opinion, only major macroscopic extranodal spread (infiltration) can be detected with imaging techniques. In a recent study from the VU Medical Center, MRI characteristics of lymph node metastases were evaluated for their predictive value for the development of distant metastases. The mean lymph node metastases volume was 11.4 cm 3 (range 0.3–122 cm 3). Ipsilateral ENS was observed in 28% and central necrosis was observed in 61% of the entire group. In the multivariate analysis, ENS as diagnosed on MRI was the only independent predictor for the development of distant metastases, warranting further screening for distant metastases in these patients. These results confirm that macroscopic ENS is probably more important than microscopic ENS [31].

Assessment of invasion of vital structures can be both prognostically and therapeutically relevant. In this respect, invasion of the common or internal carotid artery is probably most important [32], although invasion of both internal jugular veins, the skull base, or thoracic inlet pose similar therapeutic challenges. The reported accuracy of CT, MRI, and US in detecting tumour invasion into the carotid artery varies widely [33-35]. Palpation simultaneously with real-time US can be helpful to detect carotid wall invasion [36]. In general, a tumour encircling the vessel over 270^ on CT or MRI, or a tumour that is immobile from the vessel using sono-palpation, indicates involvement of the vessel wall and often non-resectability.

Staging

The sensitivity and specificity of palpation for neck node metastases are in the range of 60%–70%. The resulting risk of occult neck metastases is to a large extent dependent on the size and site and other characteristics of the primary tumour. Because of this, it is common practice to treat the neck electively by either surgery or radiotherapy in most patients. The ‘acceptable’ risk in refraining from elective treatment is hard to define. In a published meta-analysis, a risk of occult metastases of over 20% was shown to warrant elective treatment [37]. Apart from this risk, a more important question is whether a wait-and-see policy has any prognostic impact. So far, this has not been proven unequivocally, although many retrospective and some prospective studies point towards a survival advantage of elective treatment [38-40]. The prognostic impact is probably related to the delay in treatment of the occult metastases [41]. Many imaging modalities have been used to assess the neck and improve detection of small metastases in head and neck cancer.

In a neck without palpable nodes, imaging can help in detecting occult metastases or in increasing the confidence that the neck is really tumour negative and can be observed [42]. Depiction of suspicious non-palpable lymph nodes can convert selective neck treatment or a wait-and-see policy to more secure comprehensive treatment of all levels of the neck. Negative imaging results, on the other hand, can be used as an argument to refrain from elective treatment of the neck if the risk of radiologically occult metastases is considered to be low enough and close follow-up using either US or US-guided fine-needle aspiration cytology (FNAC) is guaranteed. So far, several authors have shown the applicability and reasonable prognosis using this approach [41, 43–47].

Another aspect is the assessment of the exact number of lymph node metastases and the levels involved. This is becoming more important as selective neck dissection and limited image-guided radiotherapy gain popularity. Unfortunately so far no imaging studies have been published relating to this important subject.

Apart from detection of lymph node metastases in untreated necks, imaging is often crucial for the detection of recurrences. Detection of recurrences is clinically most relevant if therapeutic options are still present. In patients treated for the primary tumour only (wait-and-see for the neck) or those treated exclusively with radiotherapy on the neck, chemoradiation or limited surgery, routine follow-up examinations of the neck using imaging to detect early recurrences seem warranted. Thus far, CT and also MRI have disappointed in the early detection of recurrent or residual disease in the neck. CT, MRI, and US have a poor specificity to distinguish radiation or postsurgical oedema and scarring from recurrent tumour [48, 49]. In patients who have a high likelihood of loco-regional recurrence, a baseline MRI or CT can be obtained 2–4 months after the initial treatment. Using the baseline scan, abnormalities that develop later can then be interpreted better with respect to tumour. With respect to the detection of residual neck disease, positron emission tomography (PET) is very likely to be the most accurate technique [50-53]. When the PET scan is negative, further investigations can be obviated according to most authors. The use of US, US-FNAC or duplex Doppler for the follow-up of the treated neck has been shown by several authors [46, 54, 55]. Westhofen showed that US FNAC was superior to CT in detecting neck recurrences after previous treatment [46]. In our opinion, routine US-FNAC follow-up for at least 1 year is warranted if the neck was not treated electively. We also use it to confirm complete response in patients treated with (chemo)radiation for an N+ neck, 6–8 weeks post treatment. In these cases, however, the cytology is much more difficult to interpret.

US-guided FNAC in experienced hands is a highly specific and quite sensitive technique in detecting palpably occult metastases, and the authors have adapted their policy of elective neck treatment in selected patients [41, 43, 45]. In selected patients who can be treated with transoral excision for T1 (and T2) oral carcinomas, laser excision of T1–2 supraglottic carcinomas, or selected patients who undergo laryngectomy for laryngeal carcinomas, one can rely on the US-guided FNAC findings and not routinely treat the neck electively. These patients should be followed very meticulously, using US-guided FNAC at 12-week intervals for at least 1 year.

Thyroid cancer

Apart from lymph node metastases from mucosal squamous cell carcinomas, imaging can play a similar role for thyroid cancer, skin cancer and salivary gland cancer. The rate of occult metastases from papillary thyroid carcinoma is reported to be as high as 60%–80% [56, 57]. In patients with follicular carcinomas, lymph node metastases are less common. The most important echelons are the paratracheal and level 4 lymph nodes. Although neck node metastases are a risk factor for developing locoregional recurrences, the influence on survival is probably limited [58]. There is little literature on the accuracy of imaging on the detection of paratracheal metastases. Because radioactive iodine can cure small metastases after thyroidectomy, imaging before treatment has relatively few implications in papillary carcinomas without palpable neck nodes. However, to detect metastases early during follow-up, US-guided FNAC is the most reliable technique routinely used for follow-up [59-61]. As no iodine contrast agents should be used, CT is less useful than US-guided FNAC. In medullary carcinomas, the rate of metastases to the neck is also high. Regional lymph nodes metastases are present in over 75% of cases at the time of diagnosis [62, 63]. Because of the prognostic significance, in medullary carcinomas elective neck dissection is often recommended but still controversial and imaging can play a pivotal role in decision making. As imaging of the paratracheal nodes is not very reliable, a routine paratracheal dissection is always recommended.

Salivary gland carcinoma

Lymph node metastases are an important prognostic factor in salivary gland cancer [64, 65]. The incidence of lymph node metastases from salivary gland cancer is dependent on the size of the primary tumour and the histologic subtype. Overall, some 20% of all parotid carcinomas are pN+, whereas lymph node metastases are rare in low grade acinic cell carcinomas and relatively common in high grade mucoepidermoid cancer [38, 65]. However, in a recent study from Stennert et al. [66], the reported incidence of (occult) metastases was much higher. Because the incidence of neck node metastases is in general reported to be below 20%, elective neck dissection is controversial [38, 66]. A common policy is to perform frozen section of the first echelon nodes in level 2. If these are positive, the parotidectomy will be followed by a neck dissection. This policy has the disadvantage that surgery time is difficult to plan. Therefore, preoperative assessment of the neck, using either MRI or US-FNAC is a logical approach [67]. As the treatment of most parotid carcinomas is surgery with postoperative radiotherapy, there is a tendency to treat the primary with surgery and postoperative radiotherapy, and the neck with elective radiotherapy if staged N0 preoperatively and at frozen section of level 2 nodes.

Skin cancer

Metastatic patterns from skin carcinomas and melanomas differ and are more variable than metastases from mucosal carcinomas. For all skin carcinomas, lymph node metastases are a dismal prognostic feature [68-72]. The parotid gland is a major nodal echelon for all skin tumours anterior to a vertical plane through the ear. Tumours behind this line mainly spread to the posterior neck nodes and occipital nodes. Metastases to superficial nodes, e.g. along the external jugular vein, occur more frequently than in mucosal squamous cancers. Whereas basal cell carcinomas very rarely give rise to neck metastases, squamous cell carcinomas, especially when infiltrating deeply, do so in 2%–15% of cases [73-75]. Melanomas give rise to lymph node metastases more often, although the patterns of metastases are less predictable than in squamous cell carcinomas [76]. The incidence of lymph node metastases is in the range of 20% for intermediate thickness melanomas. Because of that, the sentinel node procedure has gained widespread acceptance although it has not yet been clarified with certainty whether early detection of lymph node metastases (and early treatment) has prognostic importance in skin melanoma. For the head and neck area, the accuracy of the sentinel node procedure is less than for other parts of the body, and in over 10% of patients, the sentinel node cannot be identified or renders false negative results [77-79]. To assess the neck non-invasively, several authors have shown that US or US-guided FNAC is the modality of first choice, more reliable than palpation or CT [80-84]. Thus, US-FNAC can be used to stage advanced skin squamous cancers and melanomas. When US-guided FNAC is negative, a sentinel node procedure should be considered in intermediate thickness melanomas. Apart from initial assessment, US-guided FNAC can be used during follow-up [83, 85].

Ultrasound

In general, US is reported to be superior to palpation in detecting lymph node metastases [86-88]. Whereas some authors report it to be superior to contrast-enhanced CT and MRI [89], others have found similar accuracies [90, 91]. The advantages of US over other imaging techniques are its price and low patient burden. Furthermore, US is the only available imaging technique that can be used for frequent routine follow-up.

Because irregular echogeneity as a sign of metastatic involvement is often not present in small lymph node metastasis, the size of lymph nodes plays an important role in assessing their nature [92]. It is clear that size and shape criteria are not very accurate for the clinically N0 neck. The criteria used in the literature vary between 8 and 30 mm [9, 25, 93–95]. Several studies have tried to define criteria by evaluating nodal size and the histopathological outcome in neck dissection specimens [9, 91, 96–98]. Friedman [98] found a maximal axial diameter of 1 cm optimal, whereas Giancarlo [91] found a minimal diameter of 1 cm. By comparing three lymph node diameters we previously found that the minimal axial diameter is a better criterion than the more widely used maximal axial diameter or the longitudinal diameter [9]. Don et al. [96] found that 68 of 102 (67%) metastatic nodes had a longitudinal diameter smaller than 1 cm, whereas in our study we found that 102 of 144 (71%) were smaller than 1 cm. As a consequence, the current size criterion of 1 cm or larger misinterprets the majority of all metastases. This is especially the case in clinically N0 patients. In an US study in clinically node-negative patients [99], we found that for level 2 a criterion of 7 mm for the minimal diameter renders the best compromise, whereas for the rest of the neck, lymph nodes with a minimal diameter of 6 mm should be considered suspicious. During follow-up, an increase in size is a strong argument for metastasis [44].

As lymph nodes with metastases tend to become a rounder shape, shape is used as a criterion by several authors. In general, a round shape is considered more suspicious than an oval or flat shape [95]. In reactive nodes, the ratio of the longest diameter over the shortest diameter is 2 or higher in 86% of cases [93]. In stead of diameter or shape, the axial surface might be a better criterion. Umeda et al. showed that a surface area of 45 mm 2 correlated better with histopathology than using a minimal or maximal axial diameter [92].

As the size, shape and necrosis criteria are hampered by the fact that they are not very adequate for the clinically N0 neck, researchers keep looking for better criteria. Morphological criteria, such as focal cortical widening or depiction of small tumour areas inside a lymph node, will become more important as the contrast and spatial resolution of imaging techniques increases. Thus far, however, these are not shown to be reliable in lymph nodes measuring less than 1 cm. The potential value of Doppler US criteria (avascular pattern, scattered pattern, peripheral vascularity) as an adjunct to differentiate between benign and metastatic lymph nodes has been the topic of various reports. This technique enables the visualization of small irregularities in vascularization [100, 101]; however, these irregularities are seldom visible in lymph nodes smaller than 1 cm. Because of that, it is our opinion that lymph nodes should be aspirated to obtain cells for cytological assessment if management consequences are attached to these radiological findings.

Ultrasound-guided aspiration cytology

Because many authors have found that borderline lymph nodes cannot be reliably characterized on US, CT, and MRI, and because radiological criteria are not as reliable as cytology, US-guided FNAC has gained popularity since its introduction some 20 years ago [102]. In the United States this technique has received less acceptance because it is operator dependent. Although the technique is not difficult, considerable training is required to aspirate from lymph nodes as small as 4–5 mm and still obtain sufficient cells [67], and to select the most suspicious lymph nodes from which to aspirate. For this it is necessary to have clinical information on the primary tumour and knowledge about the patterns of lymphatic spread from this tumour.

It has been shown that US-guided FNAC has a very high specificity, approaching 100% as epithelial cells in lymph nodes are seldom diagnosed falsely. To obtain a high enough sensitivity, lymph nodes as small as 4–5 mm in the first two echelons should be aspirated. Although aspirating smaller nodes will probably increase the sensitivity, it is difficult to obtain a diagnostic aspirate from nodes of 3 mm or smaller. In a previous report, we found that with use of this US-guided FNAC we obtained a sensitivity of 73% with a specificity of 100% in N0 necks [90, 103]. This was significantly better than CT or MRI. Only two other studies have compared US-guided FNAC to CT and MRI and found it to be superior as well [104, 105]. Also for melanoma metastasis it was found to be the most accurate technique. Recently, however, in a multicentre study using US-guided aspiration, Takes et al. reported a sensitivity of only 42% for the N0 neck [106]. Righi et al. found a sensitivity of 50%, which was inferior to the 60% for CT [107]; however, in Righi’s study, most false negatives were found at the beginning of the study and some of these were irradiated patients or non-squamous cell carcinoma patients.

False-negative US-guided FNAC results may be the result of aspirating the wrong node or the wrong part of the correct node (=sampling error). Furthermore, the cytopathologist may overlook single tumour cells. A technique which was supposed to increase the accuracy of US-guided aspiration is better selection of the node to aspirate by the sentinel node procedure. The concept of the sentinel node approach is based on the knowledge that nodal metastases progress in an orderly manner with the first site of metastases occurring in the sentinel node. Initial reports on sentinel node biopsy in oral cancer have shown promising results. However, it remains an invasive technique and lymph node metastases close to the primary tumour, e.g. level 1 nodes in oral cancer, can be difficult to detect using scintigraphy [84, 108]. The sentinel node detection technique involves injecting around the primary tumour site with Tc-99m-labelled colloid. The localization of the sentinel node is then performed by planar scintigraphy and the use of a hand-held gamma camera. We have tried to combine the non-invasive US-guided FNAC procedure with lymphoscintigraphic detection of the sentinel node [109]. Unfortunately, this combination of the sentinel node procedure and US-guided FNAC has not improved our results obtained without sentinel node scintigraphy [43, 110]. In these studies we could also show that the sensitivity of US-guided FNAC for the clinically N0 cases varied widely in relation to the patient population studied. In patients treated with elective neck dissection, the sensitivity was 71%, similar to our previous studies [90]. However, in the group of patients treated with transoral excision only and follow-up of the neck, the sensitivity was only 25%. The reasons for this lower sensitivity might be the unreliability of histopathological examination in the electively treated group. Probably more important is the fact that in the transoral excision group the primary tumours and thus the metastases were smaller and thus more difficult to detect.