Prognostic Factors for Survival in Patients With T3 Laryngeal Carcinoma Johannes J. Manni, MD, PhD, Chris H. J. Terhaard, MD, Phi), Maarten F. de Boer, MD, Gerard A. Croll, MD, Frans J. M. Hilgers, vii), Phi), Albertus A. Annyas, it), Phi), Andel G. L. van der Meij, i n , Phn, Gert J. Hordijk, MD, PhD, Utrecht, The Netherlands

In a total of 511 patients with T3,N0-3,M0 laryngeal carcinoma, 24 possible prognostic factors were analyzed retrospectively. The factors were age, sex, mode of treatment, duration of several clinical symptoms, the presence of sore throat, otalgia, dyspnea, and dysphagia, previous traeheotomy, tumor extension, lymph node status (five items ), histolagie grading,smokinghabits, and alcohol intake. For 300 patients in whom surgery was part of the primary treatment, pathologic staging of the primary tumor and of lymph nodes in neck dissection specimens, cartilage invasion, radicality of the operation, differentiation grade, and subglottie extension ware also evaluated. In a univariate analysis for the whole group, tumor extension (limited to the glottie region), lymph node status (clinically palpable lymph nodes, cytologically confirmed positive lymph nodes), level of lymph node metastasis (high and midjugular site), histologie grading (poor differentiation grade), and treatment modality (planned combined therapy) were considered to be prognostic factors of corrected actuarial survival. In the group that underwent surgery, all factors derived from specimens of the larynx and neck dissections had prognostic significance. Multivariate analysis revealed that the glottic site of the tumor, the presence of cyto- and histopathologicaUy proven metastatic lymph nodes, pretreatment tracheotomy, positive resection mar~'ns, and planned combined treatment had a significant influence on corrected actuarial survival.

From the Dutch Head and Neck Oncology Cooperative Group, Utrecht, The Netherlands. Requests for reprints should be addressed to Gert J. Hordijk, MD, PhD, Dutch Head and Neck Oncoiogy Cooperative Group, Servaas Bolwerk 18, 3512 NK Utrecht, The Netherlands. Presented at the Third International Conference on Head and Neck Cancer, San Francisco, California, July 26-30, 1992.

682

THE AMERICAN JOURNAL OF SURGERY

" ost treatment strategies for cancer have been devel. oped based on the staging systems of the extent of M the disease. Researchers have attempted to identify and assess other factors that are presumed to influence the natural history of malignant disease, in order to separate patients with the same-stage disease who have a better chance of control by irradiation from those who have a better chance of control by surgery. This is of particular importance when surgical treatment has major functional and/or cosmetic consequences. Selection of the optimal management of T3 squamous cell carcinoma of the larynx remains a dilemma. Irradiation, as the first modality of treatment, with salvage by total laryngectomy in case of recurrence, is advocated by several centers [1-7]. Others prefer surgery to radiotherapy for patients with the same stage of disease because they have had a better survival rate with surgery when compared with radiotherapy [8-10]. It would be helpful for treatment planning if patient subgroups could be targeted for the different therapeutic approaches by means of a combination of prognostic parameters. Ideally, this would result in the development of a simple scoring system. Moreover, in patients with disease characteristics that predict a poor outcome with current treatment methods, new experimental approaches can be developed. It is apparent that only multivariate analysis of a large number of variables in a large group of patients with the same stage of disease will reveal those variables with a statistically significant independent prognostic value. In 1987, the Dutch Head and Neck Oncology Cooperative Group, in which the head and neck tumor groups of all university hospitals and the two cancer centers in The Netherlands have been united, initiated a retrospective study concerning the treatment of T3,N0-3,M0 laryngeal carcinoma. This study was coordinated by two of the authors (CHJT, GJH). This study attempts to establish the impact of 24 possible prognostic factors on the survival rates of patients with laryngeal cancer by univariate and multivariate analysis. P A T I E I ~ S AND METHODS In this retrospective analysis, the data of 511 patients with histologically confirmed T3,N0-3,M0 squamous cell carcinoma of the larynx treated between 1975 and 1984 were compiled. All patients were staged, treated, and followed up at the participating university institutions and the two cancer centers. Assignment of regions, sites, and TNM categories, and staging of all tumors were performed according the requirements of the Union Internationale Contra le Cancer (UICC) convention of

VOLUME 164 DECEMBER 1992

PROGNOSTIC FACTORS FOR LARYNGEAL CANCER

TABLE I

TABLE

Treatment Modalities and Number of Patients

Treatment modality*

No. of Patients

Primary surgery (S) Primary radiotherapy (RT) Planned combined therapy (S and RT) Selecting radiotherapy (40 Gy or 50 Gy) Total

137 113 113 146 509*

*Two patients could not be categorized into one of these treatment modalities.

ii

Univariate and Multivariate Analyses of Possible Prognostic Factors on the Actuarial Corrected Survival

Factor Sex Age Duration of clinical symptoms Sore throat Ota!gia

1978. The patients comprised 470 men and 41 women; their ages ranged from 29 to 91 years (median: 63.9 years). The lymph node status revealed 397 NO patients, 92 N1, 15 N2, and 7 N3. The patients were followed up for a minimum of 36 months or until death (mean followup: 75 months). The primary treatment modalities and the distribution of the patients are summarized in Table L All participating institutions used one of the aforementioned primary treatment protocols as their preferred standard treatment. The preferred standard treatment protocol was abandoned for medical reasons or because of patient refusal in 29% of the patients. Statistical analysis concerning the actuarial corrected survival involved 24 variables (Table II), of which 6 were obtained after surgery in a group of 300 patients (Table HI). Survival was measured from the day of the first treatment. For statistical analysis, the SPSS/PC version 2.0 (SPSS Inc., Chicago, IL) was used. Corrected survival curves were calculated according to the actuarial method of Kaplan and Mder. Statistical significance was calculated by the • test. To analyze the possible prognostic factors simultaneously, the stepwise proportional hazards regression analysis according to Cox was used. A significance level of 0.05 was set as the limit for inclusion in the model. The absence of 10% of the available data was arbitrarily taken as the upper limit for inclusion in this study. RESULTS The actuarial corrected survival for the total group of patients was 73% after 3 years and 70% after 5 years of follow-up (Figure I). General results of local and/or regional control and the results of salvage therapy for local and/or regional recurrent disease are presented elsewhere [11]. Univariate analysis: ,As expected, the sex ratio was weighted heavily to males, with 92% of patients being male. Between females and males, there was no significant difference for tumor extension and treatment groups. Gender was not a prognostic factor for actuarial corrected survival. However, as reported earlier in a subanalysis regarding radiotherapy, females showed better local-regional control and survival [7]. Age and dyspnea were not prognostic factors. However, deviation from the preferred standard treatment protocol occurred significantly more often in elderly pa-

Dyspnea Dysphagia Tracheotomy Smoking habits Alcohol intake (>_3 versus 3 U/day) Tumor extension Glottic with or without ventricle involvement versus others

Univariate p Value

Multivariate p Value

--

--

-

-

-

-

-

-

-

-

-

-

-

-

-

-

_

_

-

-

-

-

-

-

-

-

- - (0.07)

0.026

NE

NE

0.001 *

NE

0.02

Nodal stage Clinical N - versus N + Cytologic N - versus N + Number of lymph nodes Location of lymph nodes High, midjugular versus other locations Homolateral versus contralateral, bilateral

0.025 < 0.005 --

0.0005

-< 0.0005 --

0.0043 --

---

Tumor diilerentiat~on o1 biopsy G1/G3 G1/G2 G2/G3

0.025 0.O5 --

----

Treatment modality

0.005

0.001

N-, N+ = clinical absence or presence of lymph nodes and cytologic absence or presence of malignant cells, respective!y; G1, G2, G3 = well, moderately, and poody differentiatedtumor; NE = not evaluab|e, *Not recorded in 30% of patients.

TABLE III

Univariate and Multivariate Analyses of Possible Prognostic Pretreatment Variables After Primary Surgery (n = 300) and Actuarial Corrected S u r v i v a l

Variable pT3 versus pT4

Univariate p Value 0.0025

Multivariate p Value --

Cartilage invasion

< 0.001

--

pN Status

< 0.001

0.019

Differentiation grade G1,2 versus G3 Subglottic extension Irradical resection

0.009

*

0.1~.05

--

0.025

0.002

*Borderline significance (0.08).

THE AMERICAN JOURNAL OF SURGERY

VOLUME 164

DECEMBER 1992

683

MANNI ET AL

% CORRECTED

SURVIVAL

% CORRECTED

SURVIVAL

8O 70

-

~,~,

xf

70

I

I

I

60 50 F 40 30H

-+-

PRIMARY SURQERY('I87)

30 F

/I

|1 + 2 0 [ _It, - B /

PRIMARY RIg, (11~) SELECTIVE RT. (,4e)

20}

I -~

i

GLOTTIG ONLY.tMS (42)

~

OTHERS (469)

I 72

'

101

o

~ 0

12

24

38

48

80

72

84

98

108

0~_t_ 0

120

J

I 12

'

I 24

J

I 36

MONTHS

SURVIVAL

8O 70

3O 20 10~-

~

aN- 151)

" - ~ pN*.EN- 128) II

t

+

pN*,EN* 1181

I

I

'

O'l 0

I 12

'

I 24

'

I 36

'

J 48

'

I 60

'

f 72

'

[ 84

'

I 96

'

I ' I 108 120

MONTHS Figure 3. Actuarial corrected survival related to histopathologically confirmed absence (pN--) or presence (pN+) without extranodal (EN-) or with exlranodal (EN+) spread of metastatic lymph nodes in the neck dissection specimens. In four patients, the EN pattern was not reported.

tients and in patients with dyspnea. Older patients underwent radiotherapy significantly more often, and dyspnea resulted in a significant increase in surgical treatment. The durations of pain in the throat and dysphagia were not found to be significantly related to prognosis. Otalgia was significantly related to the clinical presence of lymph nodes. Pretreatment tracheotomy had a borderline significant impact on the prognosis of survival but none on the incidence of stomal recurrences. Patients who had undergone a pretreatment tracheotomy underwent significantly more primary surgical treatment. Unfortunately, smoking habits could not be extensively evaluated since the number of years patients were smoking was unknown in the majority of cases. The alcoholic intake was not recorded in 30% of the patients. Analysis of the evaluable patients, however, revealed that alcohol intake significantly decreased corrected survival, which was strongly correlated to the amount of alcohol consumption. 684

I 48

'

I 60

'

I 84

'

I 96

'

r ~ t 108 120

MONTHS

Figure 1. Actuarial corrected survival for the four treaVnent modairies and the total group of patients. Two patients could not be categorized into one of the mentioned treatment modalities. TH = therapy; RT = radiotherapy.

% CORRECTED

'

THE AMERICAN JOURNAL OF SURGERY

Figure 2. Actuarial corrected survival for glottic tumors, with o r without involvement of the Morgagni sinus (4- MS), versus other laryngeal tumor sites.

Tumor extension, nodal stage, tumor differentiation of the biopsy, and treatment modality all had a significant prognostic value on corrected survival. Patients with glottic tumors presgnted with significantly fewer clinically suspicious lymph nodes than those with supraglottic tumors. Patients with glottic tumors, with or without involvement of the ventricle, but without ingrowth of the false cords, experienced the best survival when data were corrected for clinical nodal stage (Figure 2). Division of the supraglottic region into subsites (epiglottis, false cord, aryepiglottic fold) with respect to tumor extension did not demonstrate prognostic value for survival. The number of patients with subglottic extension was too small for valuable statistical evaluation. Cytologically proven metastatic lymph nodes had a significant prognostic value on survival. The clinical presence of high or midjugnlar lymph nodes had a significantly better prognosis than lymph nodes at other locations. The number of involved lymph nodes did not have a prognostic significance. The level of the nodes appeared to be of more importance than the lateralization. The survival of patients with poorly differentiated and anaplastic tumors that were oonfirmed by biopsy appeared worse compared with those with well-differentiated tumors. In 41 (8%) patients, the differentiation grade of tumors as revealed by biopsy was not reported. Three hundred sixty-two of 511 patients received treatment according to the standard protocol of each participating institution. Patients treated according to the preferred standard protocol of the institution were revealed to have significantly better 3- and 5-year survival rates when compared with patients not treated according to the standard protocol. Analysis of the influence of the treatment modalities on the actuarial corrected survival revealed that primary surgery and primary radiotherapy had equivalent effects. Planned combined treatment significantly increased the corrected survival, and radiotherapy alone significantly decreased the corrected survival. In patients who underwent laryngectomy as part of their treatment, the prog-

VOLUME 164

DECEMBER 1992

PROGNOSTIC FACTORS FOR LARYNGEAL CANCER

therapy) have been studied [12-27]. However, the possible interrelations of these factors have only recently received attention, due to the prior lack of appropriate statistical assessment for which multivariate analysis is necessary in order to identify the independent significance of the factors under consideration. The present study was retrospective; thus, some factors currently known or suggested as influencing prognosis were not examined, i.e., performance status, blood transfusion during operation, immunologic status, DNA ploidy, and continuation of smoking after therapy [24,25,27-30]. Among the prognostic factors found to be significant in the univariate analysis of our study, pretreatment tracheotomy, tumor extension, clinical and histopathologic nodal status, tumor differentiation, cartilage invasion, and irradical resection have already been reported for a wide range of laryngeal tumors. More specifically, in patients with T3 laryngeal carcinoma treated by radiotherapy with surgery used for salvage, univafiate analysis demonstrated that the clinical extent of the disease did not significantly influence survival, whereas female gender, age older than 60 years, and irradiation dose greater than 1,700 rets had a significant impact on survival. However, the number of females was small [31,32]. Eiband et al [33], evaluating I2 prognostic factors in 152 patients with laryngeal cancer in all stages of disease, unlike our results, found that patients with glottic tumors had a significantly poorer prognosis than those with supraglottic lesions. The prognostic impact on survival of deviation from the preferred standard treatment oftbe institution has not been tested by univariate or multivariate analysis, since these patients apparently are not included in the analyses. In the present study, this factor demonstrated a significantly worse effect on corrected survival. In this study, multivariate analyses revealed that only 6 of 24 variables were significantly related to poor prognosis. Clearly, this does not mean that other variCOMMENTS ables have no prognostic impact, but, in the context of a Since controversies still exist in the choice of treat- large number of different variables, pretreatment trachement for patients with T3 laryngeal carcinoma, identifi- otomy, glottic site of the tumor, presence of cytologically cation of prognostic factors could enable more precise and histopathologically positive lymph nodes, tumor-posassessment of which treatment strategy should be fol- itive resection margins, and planned combined treatment lowed. Furthermore, identification of favorable and unfa- were the only important variables for the actuarial corvorable prognostic factors enables an accurate definition rected survival. of areas in which new treatment protocols should be deThe analysis of other variables revealed only additionsigned. Moreover, the recognition of prognostic factors al significant prognostic information. For example, otalcould assist in the comparison of studies performed at gia was significantly related to the presence of lymph other centers by better characterization of the patient nodes, whereas the pre,scnce of lymph nodes was signifipopulation. Controversial results reported by the differ- cantly related to supraglottic extension of tumor and poor ent groups with the same treatment regimen are possibly differentiation grade. Pretreatment tracheotomy was sigdue to lack of comparability in the prognostic factors of nificantly related to dyspnea, which resulted in signifithe patients involved in the trials. Some of these factors cantly more surgery and deviation from the preferred are known, whereas others may as yet be unrecognized. standard treatment protocol of the institution. In the search for reliable prognostic factors, a variety of In a subanalysis, pretreatment tracheotomy was sigpatient factors (i.e., age, sex, immunologic status, and nificantly related to histopathologieally proven invasion status of performance), tumor factors (i.e., site of pri- of the cartilage framework of the larynx specimen. Premary tumor, tumor size, possible metastatic spread, de- treatment trachcotomy had no influence on the corrected gree of tumor differentiation, and histology, morphology, survival when surgery was included in the treatment. The and DNA content of tumor cells), and treatment factors younger age group underwent significantly more planned (i.e., the response to surgery, radiotherapy, and chemo- combined surgical and radiation treatments. Older pa-

nostic factors related to the surgically removed specimen were analyzed (Table III). Two variables, vascular invasion and perineural spread, could not be analyzed due to the lack of data. Histopathologically proven invasion of the thyroid cartilage resulted in a significant decrease in the corrected survival rate. Subglottic extension (greater than 5 mm) of the tumor had no significant influence on loco-regional or stomal recurrences but had a marginally worse influence on survival. A poor differentiation grade had no influence on loco-regional or stoma recurrent disease but caused a significant increase in the incidence and the number of palpable nodes as well as a decrease in survival. Positive resection margins of the specimen and histopathologically confirmed metastatic lymph nodes (Figure 3) resulted in a significant decrease in survival. In 18 patients with histopathologically proven lymph node metastasis, extranodal spread was reported. Extranodal spread had no additional significant impact on corrected survival, but the small number of patients does not allow for valid conclusions. The results of the univariate analysis are summarized in Tables II and III. Multivariate analysis: Since some of the possible prognostic factors were unequally distributed in the different treatment groups, the multivariate analysis was performed with and without the variable "treatment modality." Multivariate analysis without the covariable treatment modality revealed that age, tumor extension, and cytologically positive lymph nodes were prognostic factors, with a significant decrease in survival. Multivariate analysis that included treatment modality as a variable showed that pretreatment tracheotomy, tumor extension, cytologically and histopathologically positive lymph nodes, and tumor-positive resection margins were prognostic factors, with a significant decrease in survival. Patients with tumors graded as poorly differentiated had a borderline (0.08) significance for survival.

THE AMERICAN JOURNAL OF SURGERY

VOLUME 164

DECEMBER 1992

685

MANNI ET AL

tients underwent significantly more radiotherapy. In the multivariate analysis in which treatment modality was not considered, older age appeared to be an unfavorable prognostic factor. However, since treatment modality was related to age, multivariate analysis was also performed including therapy as a prognostic factor. As a result of this analysis, age showed no independent correlation with the corrected actuarial survival. From the number of patients evaluated for alcohol intake in this study, it can be concluded that a careful recording and analysis is mandatory, since alcohol intake had a significant prognostic impact. Continuation of smoking after radiotherapy for laryngeal carcinoma was a negative prognostic factor for local control [7]. The lack of data related to alcohol intake and smoking habits in this study should highlight the need for the compilation of an appropriate medical history. The most important findings of this study come from the multivariate analysis. The only multivariate studies on prognostic factors in laryngeal cancer are those of Wolfensberger [34], Pera et al [35], and Stell [24]. Using a multifactorial analysis, Pera et al [35] evaluated pretreatment and posttreatment parameters in laryngeal and hypopharyngeal cancer in 416 patients treated with surgery followed by radiotherapy (5,000 to 6,000 rads) between 1969 and 1981. In their study, 14 of 20 variables were found to be related to survival. Only 5 of 14 were found to influence survival independently: T classification, histologic grading, positive lymph nodes, ulceration, and location. Information on lymph nodes, ulceration, and location was obtained from the gross pathologic description, and, therefore, these factors were con~sidered as posttreatment prognostic factors. In a personal series of 765 previously untreated patients with laryngeal carcinoma, Stell [24] evaluated 4 tumor factors: site, T stage, N stage, and histologic grade and found only N status to be a significant prognostic factor. In another study, the host factors age, sex, and performance status were evaluated; only the latter was found to have significant prognostic impact on survival [24]. The results of our analysis of prognostic factors in T3,N0-3,M0 laryngeal cancer indicate that, in relation to therapeutic approaches, there are three groups of patients: one group with a better chance of control by radiotherapy, one group with a better chance of control by surgery, and a group for which additional research is indicated for identification of the optimal treatment. Primary radiotherapy seems indicated for T3 glottic tumors with or without involvement of the Morgagni sinus in the absence of regional metastasis and for females. In the present study, this would apply to about 15% of the total number of patients. Primary surgical treatment seems to be the treatment of choice in patients with a preoperative tracheotomy or cytologically proven lymph node metastasis. In the present study, this group would also represent about 15% of the patients. For accurate diagnosis, it is imperative to use such valuable diagnostic tools as nuclear magnetic resonance

686

imaging and ultrasound-guided fine needle aspiration to exclude cartilage invasion and nodal spread, respectively [36,37]. The results of the present study will guide a multiinstitutional prospective clinical trial to address the needs of the remaining group of patients for whom the treatment policy has not yet been defined. ~ C E S 1. Bryce D, Rider WD. Preoperativeirradiation in the treatment of advanced laryngeal carcinoma. Laryngoscope1971; 81: 1481-90. 2. Harwood AR, BtTceDP, Rider WD. Management of T3 glottic cancer. Arch Otolaryngol 1980; 160: 697-9. 3. SteU PM, Dalby JE, Singh SD, Ramadan MF, Bainton R. The management of gtottic T3 carcinoma. Clin Otolaryngol 1982; 7: 175-80. 4. Wang CC, Suit HD, Phil D, Blitzer PH. Twice-a-dayradiation therapy for supragiottic carcinoma. Int J Radiat Oncol Biol Phys 1986; 12: 3-7. 5. Karim ABMF, Kralendonk JH, Njo KH, Tierie AH, Hasman A. Radiation therapy for advanced (T3-4 N0-3 M0) laryngeal carcinoma: the need for a change of strategy: a radiotherapeutic viewpoint. Int J Radiat Oncol Biol Phys 1987; 13: 1625-33. 6. Parsons JT, MencienhallWM, MancusoA, Cassisi N J, Stringer SP, Million RR. Twice-a-dayradiotherapy for T3 squamous cell carcinoma of the giottic larynx. Head Neck 1989; 11: 123-8. 7. Terhaard CHJ, Karim ABMF, Hoogenraad WJ, et al. Lo~l control in T3 laryngeal cancer treated with radical radiotherapy, time dose relationship: the concept of nominal standard dose and linear quadratic model. Int J Radiat Oncel Biol Plays 1991; 20: 1207-14. 8. Skolnik EM, Martin L, Yee KF, Wheatley MA. Radiation failures in cancer of the larynx. Ann Otot 1975; 84:804-11. 9. Alajmo E, Fini-Storchi O, Polli G. Five-year results of 100 patients operated on for cancer of the larynx. Acta Otolaryngol 1976; 82: 437-9. 10. De Santo LW. T3 glottic cancer: options and consequencesof the options. Laryngoscope 1984; 94:1311-5. 11. Terhaard CHJ, Hordijk GJ, van den Brock P, et al. A retrospectivestudy of the Dutch Head and Neck OncologyCooperative Group: study designand general. Clin Otolaryngol 1992 (in press). 12. Lauerma S. Treatment of laryngeal cancer. A study of 638 cases. Acta Otolaryngol (Suppl) 1967; 225: 1-67. 13. Eneroth C-M, Moberger G. Histologicalmalignancy grading of squamous cell carcinoma of the palate. Aeta Otolaryngol (Stockh) 1973; 75: 293-5. 14. Jakobsson PA, Eneroth C-M, Kilander D, Moberger G, Martensson B. Histologicclassificationand grading of malignancy in carcinomaof the larynx.Acta Radiol Ther Phys Bio11973;12: 1-8. 15. Till JE, Bruce WR, Elwan A, et al. A preliminaryanalysis of end results for cancer of the larynx. Laryngoscope 1975; 85: 259-75. 16. Sala O, Ferlito A. Morphologicalobservationof immunology of laryngeal cancer. Acta Otolaryngol 1976; 81: 353-63. 17. Huygen PLM, van den BrockP, Kazem I. Age and mortalityin laryngeal cancer. Clin Otolaryngol 1980; 5: 129-37. 18. Schuller DE, McGuirt WF, McCabe FB, Young D. The prognostic significanceof metastatic lymph nodes. Laryngoscope1980; 90: 557-67. 19. Holm L-E. Cellular DNA amounts of squamous cell carcinomas of the head and neck region in relation to prognosis.Laryngoscope 1982; 92: 1064-9. 2 0 . Snow GB, Annyas AA, Slooten van EA, Bartelink H, Hart AAM. Prognosticfactors of neck nodemetastasis. Clin Otolaryngol

THE AMERICAN JOURNAL OF SURGERY VOLUME164 DECEMBER 1992

PROGNOSTIC FACTORS FOR LARYNGEAL CANCER

1982; 7: 185-92. 21. Katz AE. Immunobiologic staging of patients with carcinoma of the head and neck. Laryngoscoge 1983; 93: 445-63. 22. Davis RK. Prognostic variables in head and neck cancer--tumor site, stage, nodal status, differentiation, and immune status. Otolaryngol Clin North Am 1985; 18: 411-9. 23. Gavilfin J, Gavilfin C, Manos-Pujol M, Herranz J. Discriminant analysis in predicting survival of patients with cancer of the larynx or hypopharynx. Clin Otolaryngol 1987; 12: 331-5. 24. Stell PM. Prognosis in laryngeal carcinoma: host factors. Clin Otolaryngol 1990; 15: 111-9. 25. Stell PM. Prognosis in laryngeal carcinoma: tumor factors. Clin Otolaryngol 1990; 15: 69-81. 26. Stephenson WT, Barnes DF, Holmes FF, Norris ChW. Gender influences subsitr of origin of laryngeal cancer. Arch Otolaryngol 1991; 117: 774-8. 27. Z.r~tterstrrm UK, Wennerbcrg J, Sven-Brrje E, Willrn R, Attewell R. Prognostic factors in head and neck cancer: histologic grading, DNA ploidy, and nodal status. Head Neck 1991; 13: 6: 477-88. 28. Merkel DE, McGuire WL. Ploidy, proliferative activity and prognosis--DNA flow cytometry of solid tumors. Cancer 1990; 65: 1194-205. 29. Alun-Jones T, Clark P J, Morrissey S, Mill J. Blood-transfusion and laryngeal cancer. Clin Otolaryngol 1991; 16: 240-5.

30. MacRae JD, Lampe H, Banerjee D. Blood transfusions and phenotypic immune prof'de in head and neck cancer patient undergoing surgical resection. J Otolaryngol 1991; 20: 5: 310-4. 31. Harwood AR, DeBoer G, Kazim F. Prognostic factors in Ta giottic cancer. Cancer 1981; 47: 367-72. 32. Lundgren JAV, Gilbert RW, van Nostrand AWP, Harwoord AR, Keane T J, Brant TDR. T3NoMo glottic carcinoma a failure analysis. Clin Otolaryngol 1988; 13: 455-65. 33. Eiband JD, Elias EG, Suter CM, Gray WC, Didolkar MS. Prognostic factors in squamous cell carcinoma of the larynx. Am J Surg 1989; 158: 314-7. 34,. Wolfensberger M. Multifactorial analysis of prognostic factors in head and neck cancer using Cox's proportional hazards model [abstract]. Second International Conference on Head and Neck Cancer 1982: 56. 35. Pera E, Moreno A, Galindo L. Prognostic factors in laryngeal carcinoma. Cancer 1986; 58: 928-34. 36. Castelijns JA, Golding RP, van Schaik C, Valk J, Snow GB. MR findings of cartilage invasion by laryngeal cancer: value in predicting outcome of radiation therapy. Radiology 1990; 174: 669-73. 37. Baatenburg de Jong R J, Rongen R J, Verwoerd CDA, Overhagen van H, Lameris JS, Knegt P. Ultra sound guided f'me needle aspiration biopsy of neck nodes. Arch Otolaryngol 1991; 117: 402-4.

THE AMERICAN JOURNAL OF SURGERY VOLUME 164 DECEMBER 1992 687