- Email: [email protected]
Multivariate predictors of occult neck metastasis in early oral tongue cancer
Multivariate predictors of occult neck metastasis in early oral tongue cancer ANTHONY SPARANO, MD, GREGORY WEINSTEIN, Philadelphia, Pennsylvania
MD,
ARA CHALIAN,
OBJECTIVES: The elective dissection of cervical lymph nodes from patients with early oral tongue cancer and a clinically negative neck (T1/T2N0), remains an unsettled issue that continues to be investigated. This study examines clinical and histopathologic factors through univariate and multivariate analysis to correlate the risk of neck micrometastasis in patients with T1/T2N0 squamous cell carcinoma of the oral tongue. STUDY DESIGN AND METHODS: The clinical files and histologic sections of tumor from 45 clinically determined N0 patients were retrospectively analyzed. The factors examined include degree of tumor cell differentiation, T1/T2 staging, presence of perineural invasion, presence of angiolymphatic invasion, type of invasion front, depth of muscle invasion, and tumor thickness. RESULTS: Independent correlates of positive occult neck metastasis included greater tumor thickness (P ⴝ 0.01), greater depth of muscle invasion (P ⴝ 0.01), T2 stage (P ⴝ 0.01), poorly differentiated tumors (P ⴝ 0.007), infiltrating-type invasion front (P ⴝ 0.03), presence of perineural invasion (P ⴝ 0.001), and presence of angiolymphatic invasion (P ⴝ 0.005). The final multivariate model for estimation of an increased probability of occult neck disease included greater tumor thickness, presence of perineural invasion, infiltrating-type invasion front, poorly differentiated tumors, and T2 stage. CONCLUSIONS: The clinical and histopathologic factors studied herein permit greater selectivity and more informed decision-making than does presurgical evaluation, when addressing elective neck treatment for early N0 oral tongue cancer. The multivariate model derived from this study appears to be a more reliable method for determining the pa-
From the Department of Otorhinolaryngology–Head and Neck Surgery, Hospital of the University of Pennsylvania, Philadelphia, PA. 1st place winner of 2003 Research Award, Clinical. Reprint requests: Anthony Sparano, University of Pennsylvania, Department of Otolaryngology–NHS, 1218 Walnut St., Box 302, Philadelphia, PA 19101; e-mail, [email protected] 0194-5998/$30.00 Copyright © 2004 by the American Academy of Otolaryngology–Head and Neck Surgery Foundation, Inc. doi:10.1016/j.otohns.2004.04.008
472
MD,
MIKE YODUL,
MD,
and RANDAL WEBER,
MD,
tients most likely to benefit from elective neck dissection. (Otolaryngol Head Neck Surg 2004;131: 472-6.)
T he most significant prognosticator of survival for patients with squamous cell carcinoma of the oral tongue has been the association of neck nodal metastasis.1-2 Surgical treatment for patients with oral tongue cancer, who have clinical evidence of metastasis, is well accepted. Elective neck dissection of patients with early oral tongue cancer with a clinically negative neck (T1/T2N0), however, has generated controversy for more than 3 decades. Although treatment of the clinically negative neck may benefit some patients with occult nodal metastasis, it also produces unnecessary morbidity in others without metastasis. Some studies have shown no survival benefit or other advantages for patients who underwent elective neck dissection with treatment of early tongue cancer,3,4 whereas other studies claim to show the percentage of patients who benefit from elective neck dissection is too small for the associated risks and morbidity.5 For some, the role of elective neck dissection is more as a staging procedure than as a predetermined therapeutic one.6 Others, however, have demonstrated benefits of reduced regional recurrence and prolonged survival when the clinically negative neck is treated with elective neck dissection in the setting of early oral tongue cancer.7-11 Whether the neck of patients with T1/T2N0 oral tongue cancer should be observed closely postoperatively, or dissected at the time of primary tumor resection, should depend on the ability to identify patients at high risk for nodal metastasis. Several studies have examined the value of certain clinical and histologic factors in predicting the likelihood of occult nodal metastasis.12-14 The use of any single prognosticator by itself, is neither sensitive nor specific enough to reliably select the patients most likely to benefit from elective neck dissection, as has been demonstrated.11,15 The purpose of this study was to examine various clinical and histopathologic factors together in a multivariate model used to predict the pathologic findings within regional neck nodes dissected electively in patients with T1/T2N0 squamous cell carcinoma of the oral tongue.
Otolaryngology– Head and Neck Surgery Volume 131 Number 4
SPARANO et al
473
Fig 1. Relationships of correlates retained in the multivariate model with each other, as well as with the presence of neck micrometastasis by univariate analysis.
MATERIALS AND METHODS This study was retrospectively performed using the records of 45 clinically determined N0 patients diagnosed with oral tongue cancer between 1995 and 2001 at the Hospital of the University of Pennsylvania. Institutional review board approval for this study was obtained. Patients with base of tongue cancer, recurrent oral tongue cancer, nonsquamous cell histopathology, or Tis/T3/T4 staging using American Joint Committee Cancer Staging were not included in the study. All patients must have been treated with at least a partial glossectomy and at least a unilateral selective neck dissection initially to be included in the study. Of patients eligible for the study, 28 (62%) were male and 17 (38%) were female. The age distribution at time of surgery ranged from 17 to 86 years, with a mean of 55 years. The designated tumor stage was clinically determined, and 27 (60%) T1 and 18 (40%) T2 oral tongue tumors were reviewed. Lymph node dissections were oriented by the surgeon in the presence of the pathologist at the surgical pathology bench prior to processing. The specimens were dissected by the pathologist, and lymph nodes were submitted according to level. If the lymph nodes were small, multiple nodes were submitted in a single tissue cassette. If the lymph node was too large for 1 cassette, it was sectioned and submitted in toto in multiple cassettes. All tissues were fixed in buffered formalin and submitted for histology after standard overnight machine
processing. After overnight processing, the tissue was embedded in paraffin wax. Paraffin-embedded tissue was then cut into 5- sections, placed on glass slides and stained as per department routine with hematoxylin and eosin. A single slide from each tissue cassette was cut, and the cases were reviewed by the pathologist. The pathology report associated with each case was reviewed for the presence of lymph node micrometastasis. Hematoxylin- and eosin-stained histologic sections of tumor from each patient were recalled and analyzed by a single pathologist, and measurements were concomitantly confirmed by a second observer using an adjoining microscope. Clinical and histopathologic factors assessed include degree of tumor cell differentiation (categorized as well-moderate versus poor), T1 versus T2 staging, presence or absence of perineural invasion, presence or absence of angiolymphatic invasion, type of invasion front (categorized as with pushing border versus with infiltrating border), depth of muscle invasion, and tumor thickness. Tumors with a well-defined invading border were categorized as having a pushing-type invasion front. Those with cords of cells and less well-defined invading borders were categorized as having an infiltrating-type invasion front.16 Tumor thickness was measured from the deepest point of penetration of the tumor to the mucosal surface. In cases in which the surface was ulcerated, the closest intact mucosal surface was
Otolaryngology– Head and Neck Surgery October 2004
474 SPARANO et al
Table 1. Sites of occult lymph node metastasis Levels of neck I IIa IIb III IV V
Presence of positive nodes 2/45 (4.4%) 8/45 (17.8%) 1/11 (9.1%) 3/45 (6.7%) 0/29 0/17
used, and the point of deepest penetration was extrapolated laterally to a point directly underneath the intact surface. Tumor depth of muscle invasion was calculated by subtracting the mucosal thickness (i.e., the measurement of the thickness of the adjacent normal mucosa) from the tumor thickness. The mucosal thickness was the distance from the mucosal surface to the tip of the dermal rete peg. All statistical analyses were performed using Stata version 6.0.17 Variables were independently tested for their relation to presence of occult neck nodal metastasis using the Chi-square test, Fisher’s exact test, or logistic regression. Multiple logistic regression was used to identify the multivariate correlates of occult neck metastasis. The Wald criterion of an independent P-value ⱕ 0.10 was used as the predetermined level of significance required for inclusion into the multivariable model.18 Independent correlates were statistically related to each other using the Chi-square test, Fisher’s exact test, or analysis of variance. RESULTS The number of lymph nodes examined from each patient ranged from 10 to 91, with a median number of 26. Occult metastatic nodes were found in 13 of the 45 patients; thus, the false-negative rate of neck palpation was 28.9%. The levels of lymph node metastasis are shown in Table 1. Independent Correlates of Occult Neck Metastasis Univariate analysis indicated that greater tumor thickness (P ⫽ 0.01), greater depth of muscle invasion (P ⫽ 0.01), T2 stage (P ⫽ 0.01 vs. T1 stage, RR ⫽ 3.38), a poor tumor cell differentiation (P ⫽ 0.007 vs. well-moderate differentiation, RR ⫽ 3.39), infiltratingtype invasion front (P ⫽ 0.03 vs. pushing-type, RR ⫽ 6.0), presence of perineural invasion (P ⫽ 0.001, RR ⫽ 4.50), and presence of angiolymphatic invasion (P ⫽ 0.005, RR ⫽ 4.56) correlated with positive occult neck node metastasis (Table 2). The mean difference between tumor thickness and depth of muscle invasion was only 0.18 ⫾ 0.11 mm
microscopically. Tumors ⱖ4 mm thick were associated with a higher rate of occult nodal metastasis (P ⫽ 0.009). Similar results were seen for tumors with ⱖ4 mm of invasion depth (P ⫽ 0.004; Table 3). Multivariate Model for Occult Neck Metastasis The independent correlates of occult neck disease were assessed using stepwise logistic regression modeling, and the final multivariate model for estimation of the probability of neck nodal disease is: ln (p/1 ⫺ p) ⫽ ⫺12.20 ⫹ 0.43 ⫻ (millimeters of tumor thickness) ⫹ 3.33 (if perineural invasion present) ⫹ 6.26 (if an infiltrating invasion front present) ⫹ 2.23 (if poorly differentiated tumor cells present) ⫹ 2.13 (if T2 stage present) (n ⫽ 45, P ⬍ 0.0005, df ⫽ 5, 2 ⫽ 30.6, R2 ⫽ 0.56) (Table 4). A subsidiary data set to test the model’s predictive capacity was not available, so the model was applied to the patient sample used to derive it, and the results are shown in Table 5. Greater tumor thickness was independently associated with T2 -stage tumors (4.9 ⫾ 0.5-mm thickness for T1 stage, n ⫽ 27 vs. 10.2 ⫾ 1.4-mm thickness for T2 stage, n ⫽ 18, P ⫽ 0.0003). Greater tumor thickness was also independently associated with the presence of perineural invasion (5.7 ⫾ 2.9-mm thickness with no perineural invasion, n ⫽ 30 vs. 9.8 ⫾ 7.0-mm thickness with perineural invasion, n ⫽ 15, P ⫽ 0.008). The relationships among the multivariate correlates of occult neck metastasis are depicted in Figure 1. DISCUSSION The most frequent cause of treatment failure following surgical removal of oral tongue cancer is regional recurrence.4 Statistical proof demonstrating survival benefit with clinically N0 elective neck dissection has been inconsistent, however.3 Although it has been shown that patients with advanced oral tongue cancers staged T3/T4N0 benefit from elective neck dissection,14 appropriate treatment of T1/T2N0 tumors seems undetermined. Though palpation is clearly the most practical means of assessing the neck, this study showed that clinical examination of the selected patient sample had a falsenegative rate of 28.9%. Other studies have shown the reliability of the clinical examination to be between 50% and 80%.8 Byers et al14 and Yuen et al6 showed that preoperative use of CT and ultrasound imaging offer negligible added sensitivity in identifying occult nodal metastasis. Because the critical issue regarding treatment for patients with early oral tongue cancer centers on the ability to identify the 20% to 35% of patients at highest risk for occult neck metastasis, it seems clear one cannot depend on presurgical evalua-
Otolaryngology– Head and Neck Surgery Volume 131 Number 4
SPARANO et al
475
Table 2. Univariate correlates of occult neck metastasis Correlates
Neck nodes (ⴙ)
Neck nodes (ⴚ)
Total
T1 stage T2 stage Well-moderated differentiation Poor differentiation Pushing-type invasion front Infiltrating-type invasion front (⫺) Perineural invasion (⫹) Perineural invasion (⫺) Angiolymphatic invasion (⫹) Angiolymphatic invasion
4 (14.8%) 9 (50%) 8 (21.1%) 5 (71.4%) 1 (6.7%) 12 (40%) 4 (13.3%) 9 (60%) 9 (22.0%) 4 (100%)
23 (85.2%) 9 (50%) 30 (78.9%) 2 (28.6%) 14 (93.3%) 18 (60%) 26 (86.7%) 6 (40%) 32 (78.0) 0 (0%)
27 18 38 7 15 30 30 15 41 4
Table 3. Relationship of thickness and depth of muscle invasion to occult neck metastasis Correlates
Neck nodes (ⴙ)
Neck nodes (ⴚ)
Total
Thickness ⬍4 mm Thickness ⱖ4 mm Depth of invasion ⬍4 mm Depth of invasion ⱖ4 mm
0 (0%) 13 (40.6%) 0 (0%) 13 (41.9%)
13 (100%) 19 (59.4%) 14 (100%) 18 (58.1%)
13 32 14 31
Table 4. Multivariable model for occult neck metastasis Correlates
Coefficients
Tumor thickness (mm) Perineural invasion Infiltrating invasion front Poorly differentiated T2 stage
0.43 3.33 6.26 2.23 2.13
Table 5. Predicted probability of occult neck metastasis applied to patient sample Modelderived probability of occult (ⴙ) nodes 0 ⬍ P ⱕ 0.25 0.25 ⬍ P ⱕ 0.75 0.75 ⬍ P ⱕ 1.0
Number within predicted range
Number with (ⴙ) nodes
%
28 9
1 4
3.6 44.4
8
8
100
tion alone for accurate determination of who should avoid the unnecessary morbidity associated with elective neck dissection. This study identifies clinical and histopathologic factors that, in combination, can more reliably predict those most likely to benefit from elective neck treatment. Previous studies have independently shown that tumor thickness,13 depth of muscle invasion,12,14 tumor cell differentiation,9,14,18 and pattern of tumor inva-
Standard error
Odds ratio
P value
0.25 1.44 2.85 1.37 1.31
1.54 28.03 525.57 9.33 8.38
0.08 0.02 0.02 0.10 0.10
sion,12,19,20 are significant correlates of occult neck metastasis. This study confirms these variables as determinants of occult neck disease, and adds to the list T2 stage (versus T1 stage), presence of perineural invasion, and presence of angiolymphatic invasion. It is our belief, though, that any of these variables alone provides insufficient information upon which to base the decision to treat with elective neck dissection with any significant reliability. In this study, only tumors ⱖ4 mm thick were associated with occult neck metastasis. However, of patients with tumors ⱖ4 mm thick, the rate without neck metastasis was approximately 59%. If T2 staging were used as the sole criterion qualifying patients for neck dissection, 50% of those operated on would have no evidence of occult disease. Conversely, 21% of wellmoderately differentiated tumors and 22% without associated angiolymphatic invasion had occult disease. The statistics are similar for the type of invasion front as well as for the presence of perineural invasion, when considered as sole predictors of occult disease (Table 2). The fact that each of these variables is significantly
476 SPARANO et al
correlated with occult neck metastasis, makes clear that each patient has a composite set of prognosticators together establishing some relatively specific magnitude of risk amidst an estimated spectrum. Thus, the multivariate model constructed from the univariate correlates, both corrects for colinearities between variables, and estimates the overall risk with consideration of each independent predictor weighted against the others. As shown in Table 5, patients with a model-derived predicted probability of occult neck metastasis ⱕ0.25, indeed had neck disease only 3.6% (1/28) of cases. Patients with a predicted probability of neck metastasis ⬎0.75, had neck disease in 100% (8/8) of cases. This model was able to explain 56% (R2) of the variation accounting for the presence of neck micrometastasis, and appears to be a more useful estimation of probability of occult neck disease than any single variable alone. It is recognized that the major limitations to this study include a relatively small patient sample size, and the inability to test the model’s predictive capacity using an independent patient sample. It is also of note, that in this study of 45 patients, none of the neck dissection specimens demonstrated neck micrometastasis in level IV or V nodes. These considerations remain for future study. However, the status of neck lymph nodes remains the most important prognosticator of the clinical course of oral tongue cancer. Assessment of the clinical and histopathologic factors studied here enables a more informed decision addressing elective neck treatment than does presurgical evaluation of early N0 oral tongue cancers. Although the multivariate model derived from this study needs to be validated across an independent patient sample, the results do suggest that its use may be a more reliable method for selecting the patients most likely to benefit from elective neck dissection. REFERENCES 1. Di Troia JF. Nodal metastasis and prognosis in carcinoma of the oral cavity. Otolaryngol Clin North Am 1972;5:333-42.
Otolaryngology– Head and Neck Surgery October 2004
2. Teichgraeber JF, Clairmont AA. The incidence of occult metastases for cancer of the oral tongue and floor of the mouth: treatment rationale. Head Neck Surg 1984;7:15-21. 3. Yii NW, Patel SG, Rhys-Evans PH, et al. Management of the N0 neck in early cancer of the oral tongue. Clin Otolaryngol 1999;24:75-9. 4. Yuen APW, Wei WI, Wong YM, et al. Elective neck dissection versus observation in the surgical treatment of early oral tongue carcinoma. Head Neck 1997;19:583-8. 5. Jesse RH, Barkley Jr, HT Lindberg RD, et al. Cancer of the oral cavity. Is elective neck dissection effective? Am J Surg 1970;102:505-8. 6. Yuen APW, Lam KY, Chan AC, et al. Clinicopathologic analysis of elective neck dissection for N0 neck of early oral tongue carcinoma. Am J Surg 1999;177:90-2. 7. Spiro RH, Strong EW. Surgical treatment of cancer of the tongue. Surg Clin North Am 1974;54:757-65. 8. Kaya S, Yilmaz T, Gursel B, et al. The value of elective neck dissection in treatment of cancer of the tongue. Am J Otolaryngol 2001;22:59-64. 9. Mendelson BC, Wood J, Beahrs O. Neck dissection in the treatment of carcinoma of the anterior two-thirds of the tongue. Surg Gynecol Obstet 1976;143:75-80. 10. Lee J, Litton W. Occult regional metastases: carcinoma of the oral tongue. Laryngoscope 1972;82:1273-80. 11. Whitehurst J, Droulias C. Surgical treatment of squamous cell carcinoma of the oral tongue. Arch Otolaryngol 1977;103:212-5. 12. Fukano H, Matsuura H, Hasegawa Y, et al. Depth of invasion as a predictive factor for cervical lymph node metastasis in tongue carcinoma. Head Neck 1997;19:205-10. 13. Spiro R, Huvos A, Wong G, et al. Predictive value of tumor thickness in squamous carcinoma confined to the tongue and floor of the mouth. Am J Surg 1986;152:345-50. 14. Byers R, El-Naggar A, Lee YY, et al. Can we detect or predict the presence of occult nodal metastases in patients with squamous carcinoma of the oral tongue? Head Neck 1998;20:138-44. 15. Johnson JT, Leipzig B, Cummings CW. Management of T1 carcinoma of the anterior aspect of the tongue. Arch Otolaryngol 1980;106:249-51. 16. Eneroth CM, Morberger G. Histologic malignancy grading of squamous cell carcinoma of the palate. Acta Otolaryngol 1973;75:293-5. 17. Stata Corporation. Intercooled Stata: Release 6.0. College Station, Texas; 1999. 18. Jekel JF. Multivariable analysis. In: Jekel JF, Elmore JG, Katz DL, editors: Epidemiology, biostatistics, and preventive medicine. Philadelphia: W.B. Saunders Co.; 1996. pp. 172-9. 19. Rasgon B, Cruz R, Hilsinger R. Relation of lymph-node metastasis to histopathologic appearance in oral cavity and oropharyngeal carcinoma: a case series and literature review. Laryngoscope 1989;99:1103-10. 20. Crissman JD, Liu WY, Gluckman JL. Prognostic value of histopathologic parameters in squamous cell carcinoma of the oropharynx. Cancer 1984;54:2995-3001.
MD,
ARA CHALIAN,
OBJECTIVES: The elective dissection of cervical lymph nodes from patients with early oral tongue cancer and a clinically negative neck (T1/T2N0), remains an unsettled issue that continues to be investigated. This study examines clinical and histopathologic factors through univariate and multivariate analysis to correlate the risk of neck micrometastasis in patients with T1/T2N0 squamous cell carcinoma of the oral tongue. STUDY DESIGN AND METHODS: The clinical files and histologic sections of tumor from 45 clinically determined N0 patients were retrospectively analyzed. The factors examined include degree of tumor cell differentiation, T1/T2 staging, presence of perineural invasion, presence of angiolymphatic invasion, type of invasion front, depth of muscle invasion, and tumor thickness. RESULTS: Independent correlates of positive occult neck metastasis included greater tumor thickness (P ⴝ 0.01), greater depth of muscle invasion (P ⴝ 0.01), T2 stage (P ⴝ 0.01), poorly differentiated tumors (P ⴝ 0.007), infiltrating-type invasion front (P ⴝ 0.03), presence of perineural invasion (P ⴝ 0.001), and presence of angiolymphatic invasion (P ⴝ 0.005). The final multivariate model for estimation of an increased probability of occult neck disease included greater tumor thickness, presence of perineural invasion, infiltrating-type invasion front, poorly differentiated tumors, and T2 stage. CONCLUSIONS: The clinical and histopathologic factors studied herein permit greater selectivity and more informed decision-making than does presurgical evaluation, when addressing elective neck treatment for early N0 oral tongue cancer. The multivariate model derived from this study appears to be a more reliable method for determining the pa-
From the Department of Otorhinolaryngology–Head and Neck Surgery, Hospital of the University of Pennsylvania, Philadelphia, PA. 1st place winner of 2003 Research Award, Clinical. Reprint requests: Anthony Sparano, University of Pennsylvania, Department of Otolaryngology–NHS, 1218 Walnut St., Box 302, Philadelphia, PA 19101; e-mail, [email protected] 0194-5998/$30.00 Copyright © 2004 by the American Academy of Otolaryngology–Head and Neck Surgery Foundation, Inc. doi:10.1016/j.otohns.2004.04.008
472
MD,
MIKE YODUL,
MD,
and RANDAL WEBER,
MD,
tients most likely to benefit from elective neck dissection. (Otolaryngol Head Neck Surg 2004;131: 472-6.)
T he most significant prognosticator of survival for patients with squamous cell carcinoma of the oral tongue has been the association of neck nodal metastasis.1-2 Surgical treatment for patients with oral tongue cancer, who have clinical evidence of metastasis, is well accepted. Elective neck dissection of patients with early oral tongue cancer with a clinically negative neck (T1/T2N0), however, has generated controversy for more than 3 decades. Although treatment of the clinically negative neck may benefit some patients with occult nodal metastasis, it also produces unnecessary morbidity in others without metastasis. Some studies have shown no survival benefit or other advantages for patients who underwent elective neck dissection with treatment of early tongue cancer,3,4 whereas other studies claim to show the percentage of patients who benefit from elective neck dissection is too small for the associated risks and morbidity.5 For some, the role of elective neck dissection is more as a staging procedure than as a predetermined therapeutic one.6 Others, however, have demonstrated benefits of reduced regional recurrence and prolonged survival when the clinically negative neck is treated with elective neck dissection in the setting of early oral tongue cancer.7-11 Whether the neck of patients with T1/T2N0 oral tongue cancer should be observed closely postoperatively, or dissected at the time of primary tumor resection, should depend on the ability to identify patients at high risk for nodal metastasis. Several studies have examined the value of certain clinical and histologic factors in predicting the likelihood of occult nodal metastasis.12-14 The use of any single prognosticator by itself, is neither sensitive nor specific enough to reliably select the patients most likely to benefit from elective neck dissection, as has been demonstrated.11,15 The purpose of this study was to examine various clinical and histopathologic factors together in a multivariate model used to predict the pathologic findings within regional neck nodes dissected electively in patients with T1/T2N0 squamous cell carcinoma of the oral tongue.
Otolaryngology– Head and Neck Surgery Volume 131 Number 4
SPARANO et al
473
Fig 1. Relationships of correlates retained in the multivariate model with each other, as well as with the presence of neck micrometastasis by univariate analysis.
MATERIALS AND METHODS This study was retrospectively performed using the records of 45 clinically determined N0 patients diagnosed with oral tongue cancer between 1995 and 2001 at the Hospital of the University of Pennsylvania. Institutional review board approval for this study was obtained. Patients with base of tongue cancer, recurrent oral tongue cancer, nonsquamous cell histopathology, or Tis/T3/T4 staging using American Joint Committee Cancer Staging were not included in the study. All patients must have been treated with at least a partial glossectomy and at least a unilateral selective neck dissection initially to be included in the study. Of patients eligible for the study, 28 (62%) were male and 17 (38%) were female. The age distribution at time of surgery ranged from 17 to 86 years, with a mean of 55 years. The designated tumor stage was clinically determined, and 27 (60%) T1 and 18 (40%) T2 oral tongue tumors were reviewed. Lymph node dissections were oriented by the surgeon in the presence of the pathologist at the surgical pathology bench prior to processing. The specimens were dissected by the pathologist, and lymph nodes were submitted according to level. If the lymph nodes were small, multiple nodes were submitted in a single tissue cassette. If the lymph node was too large for 1 cassette, it was sectioned and submitted in toto in multiple cassettes. All tissues were fixed in buffered formalin and submitted for histology after standard overnight machine
processing. After overnight processing, the tissue was embedded in paraffin wax. Paraffin-embedded tissue was then cut into 5- sections, placed on glass slides and stained as per department routine with hematoxylin and eosin. A single slide from each tissue cassette was cut, and the cases were reviewed by the pathologist. The pathology report associated with each case was reviewed for the presence of lymph node micrometastasis. Hematoxylin- and eosin-stained histologic sections of tumor from each patient were recalled and analyzed by a single pathologist, and measurements were concomitantly confirmed by a second observer using an adjoining microscope. Clinical and histopathologic factors assessed include degree of tumor cell differentiation (categorized as well-moderate versus poor), T1 versus T2 staging, presence or absence of perineural invasion, presence or absence of angiolymphatic invasion, type of invasion front (categorized as with pushing border versus with infiltrating border), depth of muscle invasion, and tumor thickness. Tumors with a well-defined invading border were categorized as having a pushing-type invasion front. Those with cords of cells and less well-defined invading borders were categorized as having an infiltrating-type invasion front.16 Tumor thickness was measured from the deepest point of penetration of the tumor to the mucosal surface. In cases in which the surface was ulcerated, the closest intact mucosal surface was
Otolaryngology– Head and Neck Surgery October 2004
474 SPARANO et al
Table 1. Sites of occult lymph node metastasis Levels of neck I IIa IIb III IV V
Presence of positive nodes 2/45 (4.4%) 8/45 (17.8%) 1/11 (9.1%) 3/45 (6.7%) 0/29 0/17
used, and the point of deepest penetration was extrapolated laterally to a point directly underneath the intact surface. Tumor depth of muscle invasion was calculated by subtracting the mucosal thickness (i.e., the measurement of the thickness of the adjacent normal mucosa) from the tumor thickness. The mucosal thickness was the distance from the mucosal surface to the tip of the dermal rete peg. All statistical analyses were performed using Stata version 6.0.17 Variables were independently tested for their relation to presence of occult neck nodal metastasis using the Chi-square test, Fisher’s exact test, or logistic regression. Multiple logistic regression was used to identify the multivariate correlates of occult neck metastasis. The Wald criterion of an independent P-value ⱕ 0.10 was used as the predetermined level of significance required for inclusion into the multivariable model.18 Independent correlates were statistically related to each other using the Chi-square test, Fisher’s exact test, or analysis of variance. RESULTS The number of lymph nodes examined from each patient ranged from 10 to 91, with a median number of 26. Occult metastatic nodes were found in 13 of the 45 patients; thus, the false-negative rate of neck palpation was 28.9%. The levels of lymph node metastasis are shown in Table 1. Independent Correlates of Occult Neck Metastasis Univariate analysis indicated that greater tumor thickness (P ⫽ 0.01), greater depth of muscle invasion (P ⫽ 0.01), T2 stage (P ⫽ 0.01 vs. T1 stage, RR ⫽ 3.38), a poor tumor cell differentiation (P ⫽ 0.007 vs. well-moderate differentiation, RR ⫽ 3.39), infiltratingtype invasion front (P ⫽ 0.03 vs. pushing-type, RR ⫽ 6.0), presence of perineural invasion (P ⫽ 0.001, RR ⫽ 4.50), and presence of angiolymphatic invasion (P ⫽ 0.005, RR ⫽ 4.56) correlated with positive occult neck node metastasis (Table 2). The mean difference between tumor thickness and depth of muscle invasion was only 0.18 ⫾ 0.11 mm
microscopically. Tumors ⱖ4 mm thick were associated with a higher rate of occult nodal metastasis (P ⫽ 0.009). Similar results were seen for tumors with ⱖ4 mm of invasion depth (P ⫽ 0.004; Table 3). Multivariate Model for Occult Neck Metastasis The independent correlates of occult neck disease were assessed using stepwise logistic regression modeling, and the final multivariate model for estimation of the probability of neck nodal disease is: ln (p/1 ⫺ p) ⫽ ⫺12.20 ⫹ 0.43 ⫻ (millimeters of tumor thickness) ⫹ 3.33 (if perineural invasion present) ⫹ 6.26 (if an infiltrating invasion front present) ⫹ 2.23 (if poorly differentiated tumor cells present) ⫹ 2.13 (if T2 stage present) (n ⫽ 45, P ⬍ 0.0005, df ⫽ 5, 2 ⫽ 30.6, R2 ⫽ 0.56) (Table 4). A subsidiary data set to test the model’s predictive capacity was not available, so the model was applied to the patient sample used to derive it, and the results are shown in Table 5. Greater tumor thickness was independently associated with T2 -stage tumors (4.9 ⫾ 0.5-mm thickness for T1 stage, n ⫽ 27 vs. 10.2 ⫾ 1.4-mm thickness for T2 stage, n ⫽ 18, P ⫽ 0.0003). Greater tumor thickness was also independently associated with the presence of perineural invasion (5.7 ⫾ 2.9-mm thickness with no perineural invasion, n ⫽ 30 vs. 9.8 ⫾ 7.0-mm thickness with perineural invasion, n ⫽ 15, P ⫽ 0.008). The relationships among the multivariate correlates of occult neck metastasis are depicted in Figure 1. DISCUSSION The most frequent cause of treatment failure following surgical removal of oral tongue cancer is regional recurrence.4 Statistical proof demonstrating survival benefit with clinically N0 elective neck dissection has been inconsistent, however.3 Although it has been shown that patients with advanced oral tongue cancers staged T3/T4N0 benefit from elective neck dissection,14 appropriate treatment of T1/T2N0 tumors seems undetermined. Though palpation is clearly the most practical means of assessing the neck, this study showed that clinical examination of the selected patient sample had a falsenegative rate of 28.9%. Other studies have shown the reliability of the clinical examination to be between 50% and 80%.8 Byers et al14 and Yuen et al6 showed that preoperative use of CT and ultrasound imaging offer negligible added sensitivity in identifying occult nodal metastasis. Because the critical issue regarding treatment for patients with early oral tongue cancer centers on the ability to identify the 20% to 35% of patients at highest risk for occult neck metastasis, it seems clear one cannot depend on presurgical evalua-
Otolaryngology– Head and Neck Surgery Volume 131 Number 4
SPARANO et al
475
Table 2. Univariate correlates of occult neck metastasis Correlates
Neck nodes (ⴙ)
Neck nodes (ⴚ)
Total
T1 stage T2 stage Well-moderated differentiation Poor differentiation Pushing-type invasion front Infiltrating-type invasion front (⫺) Perineural invasion (⫹) Perineural invasion (⫺) Angiolymphatic invasion (⫹) Angiolymphatic invasion
4 (14.8%) 9 (50%) 8 (21.1%) 5 (71.4%) 1 (6.7%) 12 (40%) 4 (13.3%) 9 (60%) 9 (22.0%) 4 (100%)
23 (85.2%) 9 (50%) 30 (78.9%) 2 (28.6%) 14 (93.3%) 18 (60%) 26 (86.7%) 6 (40%) 32 (78.0) 0 (0%)
27 18 38 7 15 30 30 15 41 4
Table 3. Relationship of thickness and depth of muscle invasion to occult neck metastasis Correlates
Neck nodes (ⴙ)
Neck nodes (ⴚ)
Total
Thickness ⬍4 mm Thickness ⱖ4 mm Depth of invasion ⬍4 mm Depth of invasion ⱖ4 mm
0 (0%) 13 (40.6%) 0 (0%) 13 (41.9%)
13 (100%) 19 (59.4%) 14 (100%) 18 (58.1%)
13 32 14 31
Table 4. Multivariable model for occult neck metastasis Correlates
Coefficients
Tumor thickness (mm) Perineural invasion Infiltrating invasion front Poorly differentiated T2 stage
0.43 3.33 6.26 2.23 2.13
Table 5. Predicted probability of occult neck metastasis applied to patient sample Modelderived probability of occult (ⴙ) nodes 0 ⬍ P ⱕ 0.25 0.25 ⬍ P ⱕ 0.75 0.75 ⬍ P ⱕ 1.0
Number within predicted range
Number with (ⴙ) nodes
%
28 9
1 4
3.6 44.4
8
8
100
tion alone for accurate determination of who should avoid the unnecessary morbidity associated with elective neck dissection. This study identifies clinical and histopathologic factors that, in combination, can more reliably predict those most likely to benefit from elective neck treatment. Previous studies have independently shown that tumor thickness,13 depth of muscle invasion,12,14 tumor cell differentiation,9,14,18 and pattern of tumor inva-
Standard error
Odds ratio
P value
0.25 1.44 2.85 1.37 1.31
1.54 28.03 525.57 9.33 8.38
0.08 0.02 0.02 0.10 0.10
sion,12,19,20 are significant correlates of occult neck metastasis. This study confirms these variables as determinants of occult neck disease, and adds to the list T2 stage (versus T1 stage), presence of perineural invasion, and presence of angiolymphatic invasion. It is our belief, though, that any of these variables alone provides insufficient information upon which to base the decision to treat with elective neck dissection with any significant reliability. In this study, only tumors ⱖ4 mm thick were associated with occult neck metastasis. However, of patients with tumors ⱖ4 mm thick, the rate without neck metastasis was approximately 59%. If T2 staging were used as the sole criterion qualifying patients for neck dissection, 50% of those operated on would have no evidence of occult disease. Conversely, 21% of wellmoderately differentiated tumors and 22% without associated angiolymphatic invasion had occult disease. The statistics are similar for the type of invasion front as well as for the presence of perineural invasion, when considered as sole predictors of occult disease (Table 2). The fact that each of these variables is significantly
476 SPARANO et al
correlated with occult neck metastasis, makes clear that each patient has a composite set of prognosticators together establishing some relatively specific magnitude of risk amidst an estimated spectrum. Thus, the multivariate model constructed from the univariate correlates, both corrects for colinearities between variables, and estimates the overall risk with consideration of each independent predictor weighted against the others. As shown in Table 5, patients with a model-derived predicted probability of occult neck metastasis ⱕ0.25, indeed had neck disease only 3.6% (1/28) of cases. Patients with a predicted probability of neck metastasis ⬎0.75, had neck disease in 100% (8/8) of cases. This model was able to explain 56% (R2) of the variation accounting for the presence of neck micrometastasis, and appears to be a more useful estimation of probability of occult neck disease than any single variable alone. It is recognized that the major limitations to this study include a relatively small patient sample size, and the inability to test the model’s predictive capacity using an independent patient sample. It is also of note, that in this study of 45 patients, none of the neck dissection specimens demonstrated neck micrometastasis in level IV or V nodes. These considerations remain for future study. However, the status of neck lymph nodes remains the most important prognosticator of the clinical course of oral tongue cancer. Assessment of the clinical and histopathologic factors studied here enables a more informed decision addressing elective neck treatment than does presurgical evaluation of early N0 oral tongue cancers. Although the multivariate model derived from this study needs to be validated across an independent patient sample, the results do suggest that its use may be a more reliable method for selecting the patients most likely to benefit from elective neck dissection. REFERENCES 1. Di Troia JF. Nodal metastasis and prognosis in carcinoma of the oral cavity. Otolaryngol Clin North Am 1972;5:333-42.
Otolaryngology– Head and Neck Surgery October 2004
2. Teichgraeber JF, Clairmont AA. The incidence of occult metastases for cancer of the oral tongue and floor of the mouth: treatment rationale. Head Neck Surg 1984;7:15-21. 3. Yii NW, Patel SG, Rhys-Evans PH, et al. Management of the N0 neck in early cancer of the oral tongue. Clin Otolaryngol 1999;24:75-9. 4. Yuen APW, Wei WI, Wong YM, et al. Elective neck dissection versus observation in the surgical treatment of early oral tongue carcinoma. Head Neck 1997;19:583-8. 5. Jesse RH, Barkley Jr, HT Lindberg RD, et al. Cancer of the oral cavity. Is elective neck dissection effective? Am J Surg 1970;102:505-8. 6. Yuen APW, Lam KY, Chan AC, et al. Clinicopathologic analysis of elective neck dissection for N0 neck of early oral tongue carcinoma. Am J Surg 1999;177:90-2. 7. Spiro RH, Strong EW. Surgical treatment of cancer of the tongue. Surg Clin North Am 1974;54:757-65. 8. Kaya S, Yilmaz T, Gursel B, et al. The value of elective neck dissection in treatment of cancer of the tongue. Am J Otolaryngol 2001;22:59-64. 9. Mendelson BC, Wood J, Beahrs O. Neck dissection in the treatment of carcinoma of the anterior two-thirds of the tongue. Surg Gynecol Obstet 1976;143:75-80. 10. Lee J, Litton W. Occult regional metastases: carcinoma of the oral tongue. Laryngoscope 1972;82:1273-80. 11. Whitehurst J, Droulias C. Surgical treatment of squamous cell carcinoma of the oral tongue. Arch Otolaryngol 1977;103:212-5. 12. Fukano H, Matsuura H, Hasegawa Y, et al. Depth of invasion as a predictive factor for cervical lymph node metastasis in tongue carcinoma. Head Neck 1997;19:205-10. 13. Spiro R, Huvos A, Wong G, et al. Predictive value of tumor thickness in squamous carcinoma confined to the tongue and floor of the mouth. Am J Surg 1986;152:345-50. 14. Byers R, El-Naggar A, Lee YY, et al. Can we detect or predict the presence of occult nodal metastases in patients with squamous carcinoma of the oral tongue? Head Neck 1998;20:138-44. 15. Johnson JT, Leipzig B, Cummings CW. Management of T1 carcinoma of the anterior aspect of the tongue. Arch Otolaryngol 1980;106:249-51. 16. Eneroth CM, Morberger G. Histologic malignancy grading of squamous cell carcinoma of the palate. Acta Otolaryngol 1973;75:293-5. 17. Stata Corporation. Intercooled Stata: Release 6.0. College Station, Texas; 1999. 18. Jekel JF. Multivariable analysis. In: Jekel JF, Elmore JG, Katz DL, editors: Epidemiology, biostatistics, and preventive medicine. Philadelphia: W.B. Saunders Co.; 1996. pp. 172-9. 19. Rasgon B, Cruz R, Hilsinger R. Relation of lymph-node metastasis to histopathologic appearance in oral cavity and oropharyngeal carcinoma: a case series and literature review. Laryngoscope 1989;99:1103-10. 20. Crissman JD, Liu WY, Gluckman JL. Prognostic value of histopathologic parameters in squamous cell carcinoma of the oropharynx. Cancer 1984;54:2995-3001.