Indian Journal of Pathology and Microbiology

: 2015  |  Volume : 58  |  Issue : 1  |  Page : 7--11

Warthin's tumor: An unknown pathogenesis: A neoplasm or a reactive hyperplasia?

Arpaci Bozdogan Rabia1, Linke Serınsoz Ebru1, Kara Tuba1, Gursoy Didar1, Orekici Gulhan2, Ozcan Cengiz3,  
1 Department of Pathology, Mersin University Medical School, Mersin, Turkey
2 Department of Biostatistics, Mersin University Medical School, Mersin, Turkey
3 Department of Otorhinolaryngology, Mersin University Medical School, Mersin, Turkey

Correspondence Address:
Dr. Arpaci Bozdogan Rabia
Department of Pathology, Mersin University Medical School, Mersin


Background and Aims: To examine the probable role of angiogenesis and lymphangiogenesis in the pathogenesis of Warthin俟Q製 tumor. Materials and Methods: Sixty-one patients with Warthin俟Q製 tumor (n = 40), branchial cysts (n = 6), thymic cysts (n = 3), or tonsillar lymphoepithelial cysts (n = 12) were included. Forty Warthin俟Q製 tumors were used as the lesion group, and 21 lymphoepithelial cysts were used as a control group. 29 lymph nodes around the Warthin俟Q製 tumor, four of which showed salivary duct inclusions, were also evaluated. Blood vessel density was defined as an indicator of angiogenesis by examining CD31 and FVIII Ag expression, and lymphatic vascular density was defined as an indicator of lymphangiogenesis by evaluating LYVE-1 and podoplanin expression by immunohistochemical analysis. Statistical Analysis Used: Data are expressed with descriptive statistics. Comparative analysis was performed using Shapiro-Wilks, Mann-Whitney U, and Kruskal-Wallis tests. A P < 0.005 was considered to indicate statistical significance. Statistical analysis was performed using the MedCalc ® v.10.3.0 software. Results: The lesion group had higher mean values of age (58 vs. 11 years, P = 0.001), smoking rate (92.3% vs. 19%, P < 0.001), stromal degeneration (100% vs. 42.9%, P < 0.001), lymph node involvement around the lesion (87.9% vs. 12.1%, P < 0.001), salivary duct inclusion (25% vs. 0%, P = 0.0001), than those of lymphoepithelial cysts. Blood vessel density (51.92 ± 25.64 vs. 8 ± 5.35, number/5 high power fields (HPF), P < 0.001) and lymphatic vascular density (68.95 ± 21.32 vs. 21.10 ± 4.05 number/5 HPF, P < 0.001) were higher in Warthin俟Q製 tumors than lymphoepithelial cysts. Warthin俟Q製 tumors, and lymph nodes with inclusions had similar levels of blood and lymphatic vascular density, which was higher than those of lymph nodes (P < 0.0001). Conclusions: Warthin俟Q製 tumor is a true neoplastic epithelial proliferation associated with increased angiogenesis and lymphangiogenesis and induces reactive lymph node hyperplasia.

How to cite this article:
Rabia AB, Ebru LS, Tuba K, Didar G, Gulhan O, Cengiz O. Warthin's tumor: An unknown pathogenesis: A neoplasm or a reactive hyperplasia?.Indian J Pathol Microbiol 2015;58:7-11

How to cite this URL:
Rabia AB, Ebru LS, Tuba K, Didar G, Gulhan O, Cengiz O. Warthin's tumor: An unknown pathogenesis: A neoplasm or a reactive hyperplasia?. Indian J Pathol Microbiol [serial online] 2015 [cited 2021 Apr 15 ];58:7-11
Available from:

Full Text


The pathogenesis of Warthin's tumor is controversial. [1] The theory that is more commonly accepted is the tumor development from the oncocytic epithelium and lymphoid tissue that is derived from salivary ductal tissue entrapped within parotid lymph nodes. [2],[3],[4] Lymphoepithelial cysts may develop in the head and neck region, oral cavity, pancreas, thyroid gland, and mediastinum with a decreasing order of frequency. [5],[6],[7] Angiogenesis and lymphangiogenesis are also important for tumor growth, metastasis, and chronic inflammatory diseases. [8],[9] The aim of the present study was to evaluate the role of angiogenesis and lymphangiogenesis in the pathogenesis of Warthin's tumors by quantifying lymphatics and blood vessels immunohistochemically.

 Materials and Methods

Study design and patients

This retrospective study consisted of 61 patients with Warthin's tumor (n = 40), branchial cysts (n = 6), thymic cysts (n = 3), and tonsillar lymphoepithelial cysts (n = 12) diagnosed at the Department of Pathology, Mersin University Medical School between the years 2006 and 2011. Forty cases with a diagnosis of Warthin tumor were defined as the lesion group, and 21 cases with lymphoepithelial cysts, including branchial cysts, thymic cysts, and tonsillar lymphoepithelial cysts, were defined as the control group.

This study was approved by the Ethical Committee of Mersin University Medical School, and an informed consent was obtained from the patients.

The age and sex distribution of the patients, smoking status, as well as histopathological features of the tumors, ; that is, size and presence of stromal degeneration, were retrieved from pathology reports.

Pathological evaluation

The following criteria were used for pathologic evaluation as by Cope et al. [10] with slight modifications:

An intra or periparotideal lymph node in a Warthin's tumor was defined as a collection of lymphoid tissue of any size with a discrete capsule and subcapsular sinus that was separated from the Warthin tumor by normal parotid tissue, skeletal muscle or large vessels;Salivary duct inclusions were defined as collections of ductal cells within a lymph node;Warthin tumor was defined as a tumor, composed of slits or cystic spaces with papillary infoldings lined with two layers of oncocytic epithelium;Degeneration was defined as epithelium replaced by squamous cells and extensive necrosis, fibrosis, inflammation, and granuloma formation;Thymic cysts were defined as cysts lined by squamous or cubic epithelium with prominent lymphoid tissue in the wall; Branchial cysts were defined as cysts lined with squamous epithelium with prominent lymphoid tissue in the wall; andTonsillar lymphoepithelial cysts were defined as encapsulated cystic lesions surrounded by squamous epithelium containing parakeratotic cells in the cystic space within a lymphoid stroma.

Tissue sections were stained with hematoxylin and eosin and evaluated under a light microscope (System Microscope BX53, Olympus Corporation, Olympus Microscope BX53).

For any of the following variables, presence was scored as "1" and absence was scored as "0": Any lymph node close to a Warthin's tumor and lymphoepithelial cysts, the presence of salivary duct inclusions within a lymph node, smoking history and stromal degeneration. For lesions with stromal degeneration, the degree of degeneration was scored as follows:

If the degeneration was <50% of the lesion;If the degeneration was 50% of the lesion; andIf the degeneration was >50% of the lesion.

Immunohistochemical analysis

The subcapsular region of Warthin's, thymic cysts, branchial cysts and tonsillar lymphoepithelial cysts as well as subcapsular site of the lymph nodes (with or without inclusions), were evaluated for the number of lymphatics and blood vessels to define the microvessel density (either lymphatic vascular or blood vessel density) by evaluating immunohistochemical staining patterns with Lyve-1, podoplanin, CD31, and FVIII RAg.

For imunohistochemical analysis, 4-μm-thick sections were obtained from formalin-fixed, paraffin embedded tissues. Sections were stained using a standard avidin-biotin-immunoperoxidase technique for LYVE-1 (diluted 1:50, neomarkers, Ab-7, 1467-P0), podoplanin (diluted 1:50, Santa Cruz, cat sc-79), CD31 (diluted 1:50, neomarkers, Ab, JC/70), and FVIII RAg (diluted 1:50, neomarkers, Ab, JC/70). Positive and negative controls were included for each respective antibody. Cytoplasmic staining of endothelial cells was considered positive for each marker. Differentiation of lymphatic and blood vessels was achieved using a combination of immunohistochemical markers (i.e., LYVE-1 and podoplanin for lymphatics and FVIII Ag and CD31 for blood vessels). Measurements of microvessel density for blood and lymphatic vessels were carried out by counting the number of vessels in "hotspot" areas in five different high power fields (HPF) under ×400 magnification. For lymphatic vessel density, Lyve-1 and podoplanin stained vessels were counted, and the mean value of the counted vessels was used. The same method was used for CD31 and FVIII RAg stained blood vessels.


Demographic, clinical and pathological findings

Compared with patients with lymphoepithelial cysts (control group), patients with Warthin's tumors (lesion group) had a significantly higher mean age at the time of surgery (58 vs. 11 years, P = 0.001), longer smoking history rate (92.3% vs. 19%, P < 0.001) and stromal degeneration (100% vs. 42.9%, P < 0.001). However, there were no statistically significant differences between the lesion and control groups in terms of gender and rate of stromal degeneration [Table 1]. Twenty-nine patients (87.9%) in the lesion group had periparotideal or intraparotideal lymph node tissue, whereas only four patients (12.1%) in the lymphoepithelial cysts group had lymph node tissue around the lesion (P < 0.001). In 10 (25%) cases of Warthin tumor, salivary duct inclusions were detected in lymph node tissue, whereas there were no lymph node inclusions in the lymphoepithelial cyst group (P = 0.0001) [Table 1].{Table 1}

Immunohistochemical findings

Blood vessel density was 51.92 ± 25.64 and 8 ± 5.35 number/5 HPF in the Warthin's tumors and lymphoepithelial cyst groups, respectively (P < 0.001). The lymphatic vascular density was also significantly higher in the lesion group than the control group ([68.95 ± 21.32 vs. 21.10 ± 4.05 number/5 HPF, P < 0.001] [Table 1]).

We also compared the mean vascular density between Warthin's tumors, lymph nodes and lymph nodes with inclusions. The mean lymphatic vascular density of the lymph nodes with salivary duct inclusions (n = 10) was 79.10 ± 11.02 number/5 HPF, whereas that of, lymph nodes without inclusions (n = 29) was 33.07 ± 8.41 number/5 HPF. Lymphatic vascular density in the Warthin' tumor group was significantly higher than that of lymph node group (P < 0.0001), however, the difference was not statistically significant between Warthin's tumors and lymph nodes with salivary duct inclusions (P = 0.1536). Moreover, the difference between lymph nodes without inclusions (n = 29) and those with inclusions (n = 10) was also significant (P < 0.0001).

With regards to blood vessels, the results were similar. The mean blood vessel density was 51.92 ± 25.64 number/5 HPF in the Warthin's tumor group, 36.83 ± 11.66 number/5 HPF in the lymph node group, and 70.90 ± 14.87 number/5 HPF in the lymph nodes with salivary duct inclusions. The difference was significant (P < 0.0043) between Warthin's tumors and lymph nodes, but not between Warthin's tumors and lymph nodes with inclusions (P = 0.030). The difference between lymph nodes with and without inclusions was also significant (P < 0.0001) [Table 2].{Table 2}

In all thymic cysts, branchial cysts, tonsillar lymphoepithelial cysts, and Warthin's tumor samples, the vessels were more commonly seen in the subcapsular area and were frequently similar in shape and size among the different groups [[Figure 1]a-d]. Blood vessel endothelial cells positive for FVIII RAg and CD31, those of lymphatic vessels did not. While lymphatic vessels showed positive staining with LYVE-1 and podoplanin, erythrocyte-filled blood vessels did not [[Figure 2]a-d]. Similar to LYVE-1, podoplanin expression was detected in a few stromal cells in lymphoid tissue and some of the oncocytic epithelial cells of Warthin's tumors, as well as in lymph nodes with salivary duct inclusion, and endothelial cells from lymphatic sinuses but not in the epithelium of lymphoepithelial cysts and stroma (data not shown).{Figure 1}{Figure 2}


The pathogenesis of Warthin's tumor is debated, thus immunohistochemical and molecular evidence is needed to clarify the nature of the tumor. The present study aimed to evaluate angiogenesis and lymphangiogenesis in Warthin's tumors, comparing them to that of lymphoepithelial cysts to shed light to pathogenesis of this tumor. We found that smoking history, stromal degeneration, lymph node involvement around the lesion, and salivary duct inclusions were more common in Warthin's tumors than lymphoepithelial cysts. We also showed that both blood vessel and lymphatic vascular density were higher in Warthin's tumors than in lymphoepithelial cysts, and demonstrated that the vascular density of Warthin's tumors and lymph nodes with inclusions were similar to each other but different from that of lymph nodes without inclusions.

The role of salivary ductal inclusions in the pathogenesis of Warthin's tumors is uncertain. Inclusions were shown to be higher in compared to nonsmokers among pleomorphic adenoma patients in a study conducted by Cope et al. [10] Smoking may trigger ductal hyperplasia and oncocytic metaplasia in preexisting intraparotid lymph nodes. [10] Oncocytic metaplasia, cyst formation and papillary hyperplasia are also seen in oncocytic papillary cystadenomas of the larynx, which is more common among smokers. [11] In the current study, the presence of oncocytic inclusions was a significant predictor of Warthin's tumor versus lymphoepithelial cysts (most of which are congenital lesions), as were smoking and aging likely affect the formation and neoplastic transformation of this lesion.

Immunological interactions between tumor cells and lymphocytic infiltrations with heterotopia is another hypothesis for the pathogenesis of Warthin's tumors. [12] Cells of the immune system might play a role in inflammatory and neoplastic angiogenesis/lymphangiogenesis through the expression of several angiogenic factors and their receptors and co-receptors. [8] Any such reaction could also lead to neoplastic proliferation of the epithelium due to tumor growth. However, the numbers of lymph nodes around Warthin's tumors and lymphoepithelial cysts in our study suggest that oncocytic epithelium in Warthin's tumors induces lymphoid hyperplasia through an alternative pathway different from that of lymphoepithelial cysts.

According to the World Health Organization (2005) classification of salivary gland tumors, Warthin tumors are benign epithelial neoplasms. [13] However, according to some authors, they can be classified in the group of tumor-like lesions, as the epithelial and lymphoid components are polyclonal in origin. [13] Additionally, the low incidences of recurrence and malignant transformation of this tumor, and the pathogenesis similar to lymphocyte-induced lesions (i.e., multilocular thymic cysts, branchial cleft cysts, and multiple cysts in Hashimoto thyroiditis, support the "tumor-like lesion" concept. [14],[15]

Although some molecular studies have shown that Warthin's tumor may have a clonal origin. [16],[17] Data from O'Neil [16] demonstrated that a chromosomal translocation of a novel fusion-oncogene is a remarkable feature of some cases of Warthin's tumors which favors that Warthin's tumor is a true clonal and a neoplastic process. However, Nakamura et al. [17] reported that various growth factors and their receptors, extracellular components, and cell adhesion molecules interact and regulate the proliferation and cell attachment of both the epithelial and lymphoid components of this tumor. This supports the theory of an epithelial neoplasm inducing a lymphocytic response in the pathogenesis of Warthin's tumors. [17]

Angiogenesis and lymphangiogenesis are important steps in tumor growth. [8] Therefore, it is important to assess the increased lymphangiogenesis and angiogenesis in this tumor. We showed that the vascular density was higher in Warthin's tumors than in lymphoepithelial cysts and the vascular densities of Warthin's tumors and lymph nodes with inclusions were similar to each other but different than that of lymph nodes without inclusions. Hence, we can speculate that oncocytic epithelium induces a neoplastic proliferation and results in a reactive lymphoid hyperplasia.

LYVE-1 is a specific marker for lymphatic vessels. [18],[19] However, its expression has also been detected in normal liver blood sinusoids, spleen endothelium and activated tissue macrophages. [20] LYVE-1 positive endothelial cells have been in lymph node sinuses and some reticular cells in lymph nodes, as described previously. [21] podoplanin has been demonstrated to specifically stain lymphatic endothelial cells, but not blood vessels. [22],[23] However, podoplanin stains not only lymphatics but also myofibroblasts and the myoepithelium. [22],[24] The expression of different cytokines and growth factors suggests a close immunological interaction between epithelial tumor cells and lymphoid stroma in Warthin tumors. [25],[26] Ogawa et al. [27] found similarities between lymphoid components of the tumor andlymphoid tissue of the intestinal mucosa. Soma et al. [28] showed numerous myofibroblasts between the oncocytic epithelium and lymphatic stroma. Schwerer et al. [29] detected podoplanin staining at the inner layer of the epithelium and speculated that these were myofibroblastic and myoepithelial cells.

In our study, the shape and size of the vasculature were similar, whereas concentrations of these vessels were slightly higher in the subcapsular area in all groups. In addition, podoplanin and LYVE-1 stained not only lymphatics, but also a few cells (nature unknown) particularly those in close proximity to the epithelial and lymphoid elements of Warthin's tumors and lymph nodes with inclusions as well as some oncocytic cells that were not observed in lymphoepithelial cysts (data not shown). These cells may play a role via smoking, aging and stromal degeneration due to lymphoid stroma and oncocytic epithelium come into contact with a pathway that is at present unknown. Hence, oncocytic epithelium may be gaining a neoplastic character and also lead to a reactive lymphoid hyperplasia. It may be thought that stromal degeneration, aging and smoking may play a role in the relationship between stroma and oncocytic epithelium. These findings suggest that the lesion may be an epithelial neoplasm rather than a heterotopia or embryological abnormality.

There has been an ongoing debate in the literature concerning the nature of Warthin's tumors. Some molecular studies suggest that the epithelium in Warthin's tumor has a clonal origin while others report it to be of polyclonal origin. However, these studies have reported controversial results, with some showing clonal cytogenetic abnormalities and others demonstrating polyclonal deoxyribonucleic acid (DNA). [13],[30],[31],[32],[33] The lymphoid infiltrate has been shown to be polyclonal and nonneoplastic. [34]

In our study, the presence of oncocytic inclusions was a significant predictor of Warthin tumors versus lymphoepithelial cysts. In addition to the existence of clearly increased lymphangiogenesis compared to lymphoepithelial cysts which do not include oncocytic epithelium, the existence of increased lymphangiogenesis in lymph nodes with inclusions compared with lymph nodes suggest that oncocytic epithelium in a neoplastic nature may play a role in the progress of Warthin' tumor.

In conclusion, we suggest that Warthin's tumors are true neoplastic epithelial proliferations which induce a reactive lymphoid hyperplasia supported by the findings of a greater degree of (lymph)angiogenesis in Warthin's tumors and lymph nodes containing oncocytic inclusions compared to lymph nodes without inclusions and lymphoepithelial cysts. However, further ultrastructural studies in larger series are needed to provide further information on the nature of Warthin's tumors.


1Hunt LJ. Warthin tumors do not have microsatellite ýnstability and express normal DNA mismatch repair proteins. Arch Pathol Lab Med 2006;130:52-6.
2Simpson RHW, Eveson JW. Warthin Tumour. In: Barnes L, Eveson JW. Reichart P, Sidronsky D, editors. Pathology and Genetics Head&Neck Tumours. Lyon: IARC Press; 2005. p.263-65.
3de Ru JA, Plantinga RF, Majoor MH, van Benthem PP, Slootweg PJ, Peeters PH, et al. Warthin's tumour and smoking. B-ENT 2005;1:63-6.
4Bernier JL, Bhaskar SN. Lymphoepithelial lesions of salivary glands; histogenesis and classification based on 186 cases. Cancer 1958;11:1156-79.
5Zizzi A, Aspriello SD, Ferrante L, Stramazzotti D, Colella G, Balercia P, et al. Immunohistochemical correlation between microvessel density and lymphoid infiltrate in radicular cysts. Oral Dis 2013;19:92-9.
6Ozcan I, Ozcan M, Dere H. Lenfoepitelyal kist: Olgu sunumu. KBB Forum 2005;4:4.
7Chaves FN, Feitosa SG, Pereira KM, Costa FW. Multiple lymphoepithelial cysts from oral mucosa of a healthy woman: Case report and literature review. Indian J Pathol Microbiol 2013;56:473-4.
8Marone G, Granata F. Angiogenesis, lymphangiogenesis and clinical implications. Preface. Chem Immunol Allergy 2014;99:XI-XII.
9Sharma B, Sriram G, Saraswathi TR, Sivapathasundharam B. Immunohistochemical evaluation of mast cells and angiogenesis in oral squamous cell carcinoma. Indian J Dent Res 2010;21:260-5.
10Cope W, Naugler C, Taylor SM, Trites J, Hart RD, Bullock MJ. The association of warthin tumor with salivary ductal inclusions in intra and periparotid lymph nodes. Head Neck Pathol 2014;8:73-6.
11Brandwein M, Huvos A. Laryngeal oncocytic cystadenomas. Eight cases and a literature review. Arch Otolaryngol Head Neck Surg 1995;121:1302-5.
12Teymoortash A, Werner JA. Tissue that has lost its track: Warthin's tumour. Virchows Arch 2005;446:585-8.
13Honda K, Kashima K, Daa T, Yokoyama S, Nakayama I. Clonal analysis of the epithelial component of Warthin's tumor. Hum Pathol 2000;31:1377-80.
14Louis DN, Vickery AL Jr, Rosai J, Wang CA. Multiple branchial cleft-like cysts in Hashimoto's thyroiditis. Am J Surg Pathol 1989;13:45-9.
15Suster S, Rosai J. Multilocular thymic cyst: An acquired reactive process. Study of 18 cases. Am J Surg Pathol 1991;15:388-98.
16O'Neill ID. New insights into the nature of Warthin's tumour. J Oral Pathol Med 2009;38:145-9.
17Nakamura Y, Yamamoto M, Sakamoto K, Ohta K, Umeda A, Tsukamoto T, et al. Growth factors, extracellular matrix components and cell adhesion molecules Warthin's tumor. J Oral Pathol Med 2001;30:290-5.
18Akishima Y, Ito K, Zhang L, Ishikawa Y, Orikasa H, Kiguchi H, et al. Immunohistochemical detection of human small lymphatic vessels under normal and pathological conditions using the LYVE-1 antibody. Virchows Arch 2004;444:153-7.
19Giorgadze TA, Zhang PJ, Pasha T, Coogan PS, Acs G, Elder DE, et al. Lymphatic vessel density is significantly increased in melanoma. J Cutan Pathol 2004;31:672-7.
20Mouta Carreira C, Nasser SM, di Tomaso E, Padera TP, Boucher SI, Tomarev SI, et al. LYVE-1 is not restricted to the lymph vessels: Expression in normal liver blood sinusoids and downregulation in human liver cancer and cirrhosis. Cancer Res 2001;61:8079-84.
21Jackson DG, Prevo R, Clasper S, Banerji S. LYVE-1, the lymphatic system and tumor lymphangiogenesis. Trends Immunol 2001;22:317-21.
22Kaiserling E. Immunohistochemical identification of lymph vessels with D2-40 in diagnostic pathology. Pathologe 2004;25:362-74.
23Hansen T, Kirkpatrick CJ. Expression of podoplanin in Warthin tumours. Oral Maxillofac Surg 2010;14:223-6.
24Kanner WA, Galgano MT, Atkins KA. Podoplanin expression in basal and myoepithelial cells: Utility and potential pitfalls. Appl Immunohistochem Mol Morphol 2010;18:226-30.
25Takahashi H, Tsuda N, Tezuka F, Okabe H. An immunoperoxidase investigation of S-100 protein in the epithelial component of Warthin's tumor. Oral Surg Oral Med Oral Pathol 1986;62:57-62.
26Korsrud FR, Brandtzaeg P. Immunohistochemical characterization of cellular immunoglobulins and epithelial marker antigens in Warthin's tumor. Hum Pathol 1984;15:361-7.
27Ogawa Y, Hong SS, Toyosawa S, Chang CK, Yagi T. Expression of major histocompatibility complex class II antigens and interleukin-1 by epithelial cells of Warthin's tumor. Cancer 1990;66:2111-7.
28Soma L, LiVolsi VA, Baloch ZW. Dendritic interstitial and myofibroblastic cells at the border of salivary gland tumors. Arch Pathol Lab Med 2001;125:232-6.
29Schwerer MJ, Kraft K, Baczako K, Maier H. Cytokeratin expression and epithelial differentiation in Warthin's tumour and its metaplastic (infarcted) variant. Histopathology 2001;39:347-52.
30Martins C, Fonseca I, Roque L, Soares J. Cytogenetic characterisation of Warthin's tumour. Oral Oncol 1997;33:344-7.
31Mark J, Dahlenfors R, Stenman G, Nordquist A. Chromosomal patterns in Warthin's tumor. A second type of human benign salivary gland neoplasm. Cancer Genet Cytogenet 1990;46:35-9.
32Arida M, Barnes EL, Hunt JL. Molecular assessment of allelic loss in Warthin tumors. Mod Pathol 2005;18:964-8.
33Bañez EI, Krishnan B, Ansari MQ, Carraway NP, McBride RA. False aneuploidy in benign tumors with a high lymphocyte content: A study of Warthin's tumor and benign thymoma. Hum Pathol 1992;23:1244-51.
34Takezawa K, Jackson C, Gnepp DR, King TC. Molecular characterization of Warthin tumor. Oral Surg Oral Med Oral Pathol Oral Radiol Endod 1998;85:569-75.