Takes RP, Baatenburg de Jong RJ, Alles MJRC, Meeuwis CA, Marres HAM, Knegt PPM, de la Riviere GB, de Wilde PCM, Mooi WJ, Hermans J, van Krieken JHJM. Markers for Nodal Metastasis in Head and Neck Squamous Cell Cancer. Arch Otolaryngol Head Neck Surg. 2002;128(5):512-518. doi:10.1001/archotol.128.5.512
To identify markers that are relevant as predictors of lymph node metastasis in head and neck squamous cell cancer.
Expression of p53, Rb, cyclin D1, E-cadherin, and epithelial cell adhesion molecule was examined using immunohistochemical analysis and traditional histological parameters, and the correlation of these markers with the histologically verified presence of regional metastases was determined.
The study sample comprised 121 patients with head and neck squamous cell cancer from whom paraffin-embedded material of primary tumors was used.
Lymph node metastasis was correlated with the loss of expression of Rb (P = .04) and marginally correlated with the loss of expression of E-cadherin (P = .06). If the results are broken down to subsites, loss of E-cadherin expression in oral cancer (P = .04) and absence of eosinophilic infiltration in laryngeal cancer (P = .003) correlated with nodal metastasis. None of the other markers correlated. A combination of relevant parameters did not result in a much stronger correlation.
The expression of the investigated genetic markers and histopathological features of primary tumors can supply limited information on the metastatic behavior of tumors. Although the use of markers for regional metastasis would be a welcome additional tool, these results do not warrant the use of these parameters for clinical decision making concerning the treatment of the neck in patients with head and neck squamous cell cancer.
REGIONAL METASTASIS is an important factor in the prognosis and choice of treatment of patients with head and neck squamous cell cancer (HNSCC). The presence of nodal metastasis will significantly affect the survival of the patient.1 In most patients with HNSCC, a decision of whether to treat the lymph nodes of the neck has to be made. Diagnostic means to assess the lymph node status of the neck are not reliable. Owing to high false-negative rates, many patients with HNSCC will undergo elective neck treatment. For a considerable number of these patients, this means an unnecessary treatment for their neck with significant morbidity.2- 4
The techniques to assess the nodal status of the neck have improved in recent years. However, even ultrasound with ultrasound-guided fine-needle aspiration biopsy (presently, the most accurate technique to detect lymph node metastases) identifies clinically occult metastases with a sensitivity of no more than 48% to 76%.5,6 Moreover, because of the limited specificity in the range of 70% to 85%7- 9 of techniques such as computed tomography and magnetic resonance imaging, patients may receive unnecessary neck treatment based on false-positive findings. The limitation of all these techniques is that by palpation or imaging techniques small metastatic deposits will still be undetected, and uncertainty about the true lymph node status of the neck will remain.
The process of metastasis is a result of changes in properties of cells and of interaction between tumor cells and surrounding cells and structures. To metastasize, cells proliferate, lose contact with neighboring cells, migrate through the interstitial matrix, invade blood and lymph vessels, and grow out again in lymph nodes or distant organs. The metastatic cells, therefore, have to possess several properties to perform all these actions.10 These tumor cell properties will be based on changes in genes and their products. Based on the assumption that metastasis is mainly determined by properties of the primary tumor and its interaction with the surrounding structures,10 it is worthwhile to explore the possibility of predicting the presence of metastases based on features of the primary tumor. In that case it would be possible to obtain additional information concerning the chance of metastasis, irrespective of the size of the metastases, by studying features of the primary tumors themselves. In an earlier pilot study, it appeared to be feasible to assess the chance on metastasis in laryngeal cancers using a panel of relevant factors.11 If biological markers prove to be reliable diagnostic tools, they may reduce the need for elective neck treatment.12
Expression of p53, Rb, cyclin D1, E-cadherin, and epithelial cell adhesion molecule (Ep-CAM) and histological parameters (differentiation grade, growth pattern, tumor-associated eosinophilic infiltration, and inflammatory reaction) were examined in 121 primary tumors in patients with HNSCC. All patients were previously untreated. The correlation of these markers and parameters with the histologically verified presence of regional metastases was determined.
From the files of the departments of pathology of all participating institutions, 121 tissue blocks were retrieved from resection specimens of laryngeal, pharyngeal, and oral carcinomas, which were resected en bloc with the regional lymph nodes between the years 1990 and 1995. A portion of the patients had been enrolled in a multicenter study on the value of ultrasound with ultrasound-guided fine-needle aspiration biopsy in the assessment of the nodal status of the neck in patients with HNSCC.13 The population characteristics (ie, age; sex; site; and T, N, and pN stage) are summarized in Table 1. The proteins were analyzed using immunohistochemistry as previously described.11 The antibodies used are summarized in Table 2.
Most proteins (except cyclin D1 and E-cadherin) were studied using the 3-step indirect method. In brief, 5-µm sections of paraffin-embedded tissue were dewaxed in xylol for 15 minutes and rehydrated with alcohol. Endogenous peroxidase was blocked with 0.3% hydrogen peroxidase. Subsequently, the sections were pretreated for antigen retrieval as follows: for p53, Rb, and E-cadherin, the sections were first boiled in citrate buffer (pH 6.0) for 10 minutes and cooled down for at least 2 hours; for Ep-CAM, the sections were pretreated with trypsin solution (0.1% trypsin with 0.1% calcium chloride) (pH 7.4) at 37°C for 20 minutes. After washing with phosphate-buffered saline (PBS), the primary antibody was applied for overnight incubation with 1% bovine serum albumin in PBS. After washing with PBS, the sections were incubated with the secondary antibody. For monoclonals, rabbit antimouse IgG (DakoP161, Dako Corp, Carpinteria, Calif) was applied for 45 minutes, then washed in PBS, and finally incubated with the tertiary antibody, swine antirabbit IgG (DakoP217, Dako Corp) for 45 minutes. For polyclonals, no tertiary antibody was used. For cyclin D1 and E-cadherin, the avidin-biotin-peroxidase complex staining method was applied.
After the final washing with PBS, staining was performed by means of 3-amino-9-ethylcarbazole in dimethylformamide with hydrogen peroxide followed by counterstaining with Mayer hematoxylin for 30 seconds. The sections were dyed blue in tap water and mounted with glycerin gelatin.
All cases were stained simultaneously for each protein with appropriate specimens as a positive and negative control. As a positive control, tumor specimens were used that showed positive results in former studies. As a negative control, the sections were processed without the primary antibody. Moreover, nonneoplastic cells in the section served as an internal negative control.
Two observers (R.P.T. and J.H.J.M.v.K.) evaluated the staining results. Differences in scoring were discussed during examination at a multiheaded microscope until agreement was reached. For each antibody, scoring categories were made. For convenience in reporting, before statistical analysis a dichotomy (positive vs negative) was made as previously described.14 The cutoff points were based on the distribution of staining results in the different scoring categories. For p53 and Rb, the cutoff point was 0% to 15% vs greater than 15%; for cyclin D1, 0% to 5% vs greater than 5%; and for E-cadherin and Ep-CAM, cases with no staining were compared with cases showing any staining. Some cases were not able to be evaluated owing to the absence of sufficient tumor in the specimens or inconclusive staining results. Traditional histological parameters were scored as described before.11
The correlation of the investigated parameters with the histologically verified presence of lymph node metastasis was tested using the χ2 test or Fisher exact test. For all analyses, P<.05 was considered significant.
The markers and histological parameters were examined in 121 cases of HNSCC and related to the development of nodal metastasis. The expression rates of the investigated markers are summarized in Table 3.
Lymph node metastases was correlated with the loss of expression of Rb (P = .04) and marginally correlated with the loss of expression of E-cadherin (P = .06). None of the other markers or histological features showed a correlation with nodal metastasis (Table 3). T stage (P = .26, data not shown) and T1 and T2 vs T3 and T4 (P = .83, data not shown) did not correlate with metastasis either.
The choice of cutoff points to separate negative from positive results usually lacks a fundamental basis and is often rather arbitrary. To see if the choice of cutoff points in the scoring categories would influence the results, several cutoff points for considering results positive or negative were examined for each marker. Most of the cutoff points used in an earlier study14 also resulted in the most significant correlations (15% for Rb and 0% vs greater than 0% for E-cadherin). For expression of p53 and cyclin D1, no correlation with lymph node metastasis was found using any of the alternative cutoff points. For p53, the best alternative cutoff point of 50% did not result in a significant correlation (P = .67) and neither did the cutoff point of 75% for cyclin D1 (P = .48). For Ep-CAM, only 2 scoring categories were made, so no alternative cutoff points could be explored.
To investigate if a combination of parameters would be more informative, the results of the markers that showed some relation with nodal metastasis were combined. The combination of Rb and E-cadherin did not result in a better correlation with nodal metastasis (P = .15, data not shown) and neither did the addition of inflammatory reaction (P = .37, data not shown).
If the correlation of the investigated parameters with nodal metastasis is examined for the 3 subsites of the head and neck (ie, larynx, pharynx, and oral cavity), the results are different compared with the entire population (Table 4). It should be noted, however, that the number of cases is limited. For the larynx, absence of eosinophilic infiltration correlated with the presence of nodal metastasis (P = .003). For pharyngeal cancers, no significant correlations were found, and for oral cancers, E-cadherin expression correlated with nodal metastasis (P = .04). All tumors with loss of expression of E-cadherin had lymph node metastases.
Markers for assessment of nodal metastasis are particularly useful in the N0 neck. Table 5 summarizes the results of the analysis performed in the subpopulation of 60 patients with no palpable masses in the neck. In this population, the loss of expression of Rb was marginally correlated with the presence of nodal metastases (P = .06). No relation between the other investigated parameters and nodal metastasis was found. In the subpopulation of patients (n = 49) with no detectable metastasis in the ultrasound-guided fine-needle aspiration biopsy findings, no correlations between marker expression and the presence of nodal metastasis could be established (Table 6).
In recent years, an increasing number of studies have focused on finding parameters to assess the lymph node status of the neck in patients with HNSCC. Diagnostic imaging techniques improve continuously but have the fundamental limitation that the metastases need to have a minimal size of at least several millimeters to be detected. Moreover, they have a low specificity if the imaging is not combined with fine-needle aspiration biopsy.7- 9 More recently, histological features and changes in gene expression of tumors have been explored for predictors of nodal metastasis. In earlier studies, the relation of individual markers with metastasis has only been studied in the context of detecting clinicopathological correlations in general. Now, studies are performed focussing on the issue of the management of the N0 neck. The feasibility of assessing the chance on metastasis more reliably using a set of tumor-related parameters is currently under investigation.11,15
In the present study, we found a correlation with nodal metastasis for the loss of expression of Rb and a near significant correlation for the loss of expression of E-cadherin. Combining the results of these 2 markers did not result in a better correlation with metastasis. If the correlation of the investigated parameters with nodal metastasis is examined for the 3 subsites of the head and neck, the results are different compared with those of the entire population. For the larynx and pharynx, this may be due to the low number of cases without metastasis because in this group the number of elective neck dissections was low. Moreover, the number of cases per group is lower, resulting in less statistically significant relations. The difference in results between the subsites may also underline the possible difference in intrinsic biological properties between tumors arising in the several subsites of the head and neck.14 The results in the clinically N0 group do not essentially differ from those of the entire group, except for E-cadherin. Probably due to the lower number of cases, statistical significance is not obtained for all parameters showing a correlation with metastasis in the entire group.
The loss of Rb expression correlated positively with the presence of nodal metastasis. In an earlier study of laryngeal carcinomas, however, we found an inverse correlation.11 We do not have a good explanation for this difference. It may be due to the smaller number of cases in our previous study or a difference in population characteristics (eg, some of the patients in the earlier study received prior radiotherapy, and in the present study, more higher-stage tumors were included because all patients underwent neck dissection). The expression of Rb has not been frequently studied in HNSCC.16,17 However, in 1 study a relation between the expression of Rb and nodal status has been described.17
The other marker that nearly significantly correlated with the presence of nodal metastasis in our study was E-cadherin. E-cadherin is an important molecule in cell-cell adhesion, and changes in its expression may, together with other factors, play a role in the process of metastasis.18 A relation between the loss of expression of E-cadherin and the development of metastases has been described in several studies as reviewed by Jiang.19 Also in HNSCC, relations between the loss of E-cadherin expression and the presence of nodal metastasis have been reported.20 Other studies, however, failed to find a statistically significant relation between the loss of expression of E-cadherin in the primary tumor and the occurrence of nodal metastases.21- 23 For the other markers (ie, p53, cyclin D1, and Ep-CAM), no correlation with the development of lymph node metastasis was found in the present study.
One of the most frequently studied markers in recent years is the tumor suppressor gene p53. Point mutations of p53, leading to nuclear accumulation of the protein, are among the most frequent genetic alterations in HNSCC. Clinicopathological studies of alterations in p53 in HNSCC show varying results as discussed in a review of Raybaud-Diogene et al.24 Some authors did not find any correlation of p53 expression with clinical parameters,25 metastasis,26 or survival.27,28 However, others found a correlation between nuclear p53 accumulation and survival, although reports are contradictory: some found a correlation with worse survival rates,29,30 and others, with better.31
Cyclin D1 is a potentially relevant prognostic marker and marker for the development of metastasis. Cyclin D1 was first described as a candidate oncogene in 1991 by Motokura et al32 as PRAD-1 and plays an important role in cell cycle regulation. Overexpression of cyclin D1 may cause deregulation of the cell cycle and thus contribute to tumorigenesis. Studies of HNSCC concerning amplification of the chromosome 11q13 region, where the cyclin D1 and EMS-1 genes are harbored, indicated a relation of this amplification with the development of nodal metastasis.11,33- 37 The relation between lymph node metastasis and the expression of cyclin D1 has not been studied frequently. In examining the expression of cyclin D1 with immunohistochemistry, Michalides et al38 found no correlation of cyclin D1 expression with N stage; however, other authors did.36,39 Therefore, although amplification of 11q13 genes seems to be correlated with the presence of metastasis, this correlation has not been definitively confirmed for the expression of these genes.
The expression of Ep-CAM has not been studied frequently in HNSCC. Increased expression of Ep-CAM appears to result in decreased cadherin-mediated cell-cell adhesion and may lead to segregation of Ep-CAM positive cells from the parental cell population in vitro.40 This phenomenon may facilitate the development of metastases in vivo. Other in vitro and animal studies, however, suggest that expression of the molecule reduces the metastatic potential.41 In an earlier study, we found a near significant relation between the loss of expression and the development of nodal metastasis.11 However, in the present study of a larger series of HNSCC, this relation could not be confirmed.
The studies on correlations of expression of markers with nodal metastasis or other clinical parameters often show conflicting results, which may be due to a number of factors. Besides differences in techniques and antibodies, many of these factors are related to the heterogeneity of tumors in which no tumor is exactly alike and no tumor consists of a population of identical cells.42,43 Chromosomal aberrations and protein expression are different in some parts of the tumor compared with others, which is usually thought to be a result of clonal evolution. In tumor progression, some cells of a clone may acquire additional or different chromosomal alterations, resulting in a subclone with different properties. If this subclone is a relatively small part of the primary tumor, the expression of an examined marker in this tumor will probably be scored negative. However, this subpopulation of tumor cells may still be responsible for a certain biological behavior such as metastasis. Indeed, the expression of markers in the primary tumor does not always match that of their metastases,44 which makes the choice of biologically relevant cutoff points arbitrary. What percentage of cells showing expression should be considered a positive result and what percentage a negative result? As a result, cutoff points are often not the same among studies. Another consequence of the heterogeneity of tumors is that biopsy material used for examination may not represent the entire tumor.44
In several studies, histological features of the tumor were found to correlate with survival and/or the development of lymph node metastases. Correlations between the presence of an inflammatory reaction and the absence of lymph node metastasis have been described by several authors.11,45,46 In some studies, a relation of eosinophilic infiltration surrounding the tumor and favorable prognosis has been described, but a relation with lymph node metastasis was not found frequently.47- 50 Other studies did not find any significant clinicopathological correlations.51 A relation between the development of lymph node metastasis and grade of differentiation45,52,53 or growth pattern45,54- 59 has also been described. Therefore, traditional histological parameters seem to be able to provide useful information.
Because the process of tumor development and metastasis is complex, it is unlikely that a single parameter will be able to predict the metastatic behavior of a tumor. However, in our study the number of single parameters showing a correlation with nodal metastasis was low, and a combination of the most relevant parameters was not more predictive for nodal metastasis.
If more relevant parameters can be identified in the future, the combination of the expression of markers and histopathological features of primary tumors may be able to supply information on the metastatic behavior of tumors and influence clinical decision making concerning the treatment of the neck in patients with HNSCC. The investigated parameters in this study, however, did not show correlations with nodal metastasis strong enough to be useful in clinical practice. Moreover, the inconsistent results between studies in the literature hamper the actual introduction of these markers for clinical purposes. Uniform standards are required to make the results of studies comparable.
Accepted for publication October 11, 2001.
This study was presented at the Fifth International Conference on Head and Neck Cancer, San Francisco, Calif, July 31, 2000.
Corresponding author: Robert P. Takes, MD, PhD, Department of Otorhinolaryngology and Head and Neck Surgery, University Medical Center Nijmegen, Greet Grooteplein 10, PO Box 9101, 6500 HB Nijmegen, the Netherlands (e-mail: firstname.lastname@example.org).