Journal of Surgical Oncology 48937-95 (1991)
Prognostic Factors in Laryngeal Cancer Patients Submitted to Surgical Treatment LUlZ P. KOWALSKI, MD, PhD, EDUARDO L. FRANCO, PhD, JOZIAS DE ANDRADE SOBRINHO, MD, P m , BENEDIT0 V. OLIVEIRA, MD, AND P A U L 0 L. PONTES, MD, PhD From the Epidemiology and Biostatistics Unit, Ludwig Institute for Cancer Research (L.P.K., E.L.F.), Head and Neck Service, Heli6polis Hospital (L.P.K., I.D.A.S.), and Head and Neck Surgery-Otolaryngology, Paulista Schooi of Medicine (P.L.P.), Sao Paulo, and Head and Neck Service, Erasto Caertner Hospital (B.V.O.), Curitiba, Brazil.
The authors studied some demographic, clinical, pathologic, and therapeutic variables in relation to the disease-free interval (DFI) and to the overall survival of laryngeal cancer patients. The patient population consisted of 278 consecutive cases of primary laryngeal cancer who had been admitted to four head and neck surgery services in Sao Paulo and Curitiba (Brazil) between 1973 and 1986. All patients had surgical treatment. Although several variables exhibited individual associations with survival, only a few were deemed to have independent prognostic value using multivariate regression techniques based on Cox’s proportional hazards model. The primary tumor stage (T index) represented the variable with the highest predictive strength with respect to survival. Hazard ratios for the risk of disease recurrence were 2.3 (Tz-3 vs. T,) and 4.3 (T4 vs. T,). Hazard ratios for the risk of death were 1.6 (Tzp3vs. T,) and 3.3 (T4 vs. TI). The following additional factors were also independently associated with the DFI: age, site of primary tumor (transglottic, glottic, or supraglottic), extracapsular spread of lymph node metastasis, surgical margins, type of paratracheal node dissection, and prior tracheostomy. Except for extracapsular spread, the same factors related to the DFI were also significantly related to the overall survival. Race, sex, N-stage, vascular embolization, and histological grade were other independent predictors of the risk of death. KEY WORDS:survival analysis, disease-free interval, head and neck surgery
INTRODUCTION An estimated 12,300 new cases of laryngeal cancer are diagnosed each year in the United States. The number of deaths due to this neoplasm is close to one-third of that annually [ 11. The prognosis of untreated cases of laryngeal cancer is very poor. Advanced disease is characterized by one or more of the following conditions: respiratory insufficiency, dysphagia, odinophagia, hemorrhage, and cachexia. Despite the severity of the clinical presentation of the advanced form of the disease, laryngeal carcinoma is a potentially curable disease, especially in cases diagnosed at early stages. The decision regarding treatment regimen has traditionally been dependent on 0 1991 Wiley-Liss, Inc.
site, local extent, nodal status, and on the general medical condition of the patient [2-4]. Posttreatment survival rates are the most important indices of therapeutic efficacy, but even using rigid criteria of patient eligibility for standardized protocols, results are not uniform. Accepted for publication May 20, 1991. Address reprint requests to Dr. Luiz P. Kowalski, Instituto Ludwig de Pesquisa sio Cancer, Rua Prof. Antonio Prudente, 109, 01509-Sao Paulo, Brazil. L.P.K.’s present address is Head and Neck Surgery Department, A.C. Camargo Hospital-Antonio Prudente Foundation, Sao Paulo, Brazil. E.L.F.’s present address is Department of Epidemiology, lnstitut Armand-Frappier, University of Quebec, Laval, Canada.
88
Kowalski et al.
Although stage has long been clinically accepted as the best single determinant of survival for larynx cancer patients, prognosis is complex and may depend on several other risk factors such as age and sex [ 5 ] ,prior tracheostomy [6], histological grade and margins of surgical resection [4], invasion of the thyroid [7], and lymph node involvement [8,9]. However, the importance of each one of these factors has not been unequivocally demonstrated. Many of these variables are, in fact, interrelated and their apparent association with prognosis may be due to secondary relationships with genuine correlates of survival, such as measures of disease extension at diagnosis. Our present knowledge about the relative importance of these risk factors has been mostly derived from studies based on univariate analyses of candidate prognostic factors. The present study was prompted by the need to evaluate the prognostic importance of several clinical, pathological, and treatment-related variables after controlling for the effect of staging information and other determinants of survival.
PATIENTS AND METHODS Patients We reviewed the records of 278 consecutive patients with newly diagnosed laryngeal carcinoma referred to four Head and neck Surgery Services in Sao Paulo (Heliopolis Hospital, Servidor Publico Municipal Hospital, and Paulista School of Medicine) and Curitiba (Erasto Gaertner Hospital) during the period of July 1973-June 1986. Only patients submitted to surgical treatment were considered eligible for the study. Patients submitted to radiotherapy or surgery prior to the referral, patients electively treated by irradiation (stage I glottic carcinoma), or those who refused surgical treatment were not included in the investigation. All diagnoses were confirmed histopathologically prior to surgery, and anatomical site was ascertained postsurgically . Staging of disease was done by the Union Internationale Contre le Cancer’s tumor-nodes-metastasis (TNM) staging system [ 101 (Table I). Management policy for these patients varied relatively little within each institution during the study. Initial surgical treatments were total laryngectomy in 196 patients (70.5%), total laryngectomy and partial pharyngectomy in 3 1 (1 1.2%), partial vertical laryngectomies in 40 ( 14.4%), and horizontal supraglottic laryngectomy in 12 (4.3%). Dissection of paratracheal lymph nodes was performed either unilaterally or bilaterally in patients with confirmed or presumed involvement of the subglottis. Unilateral paratracheal lymph node dissection was the standard technique used at Erasto Gaernter Hospital, and the bilateral technique was used at Heliopolis and Servidor Publico Municipal Hospitals. Resection of the ipsilateral paratracheal lymph node chain with thyroid lobectomy was performed in 50 cases and bilateral
TABLE I. Distribution of Laryngeal Cancer Cases According to T and N Categories of the UICC TNM Classification Nodal stage
Tumor stage
No
TI,
27 14 45 122 24 232
Tib
T2 T3 T4
Total
NI
4 15 5 24
Nla
4 I 5
Nzb
N2c
1
1 5
4
3 9
2 7
N3
1
1
Total
27 14 52 150 35 278
paratracheal lymph node dissection with thyroid lobectomy was done in 93 cases. A total of 176 neck dissections (29 bilateral) was performed concomitantly with the primary resection in 147 patients (elective in 101 patients and therapeutic in 46). Patients were distributed according to the surgical technique for cervical lymph node dissection as follows: 94 radical classical neck dissections (15 bilateral), 33 functional neck dissections (7 bilateral), 7 extended supra-omohyoid dissections (4 bilateral), and 13 jugular lymph node chain dissections (3 bilateral). Postoperative radiotherapy was used as an adjuvant to surgery in 60 cases because of involved surgical margins and/or positive lymph nodes. The treatment consisted of a wide field technique including both sides of the neck with a dose range of 50-65 Gy. The number of metastatic nodes ranged from &20 in 191 cases with identifiable lymph nodes in the specimen. The diameter of metastatic lymph nodes varied from 1-8 cm. The level of lymph node involvement was evaluated in 147 patients and classified according to the presence or absence of metastasis in the jugular lymph node chain that was ipsilateral to the primary tumor. Lower metastatic lymph node involvement was considered either independently or associated with upper level involvement. The following additional characteristics of metastatic nodes were considered as candidate prognostic variables: contralateral lymph node involvement, preand/or paratracheal lymph node involvement, and extracapsular spread of tumor in lymph node metastasis.
Statistical Analysis Product-limit estimates of the survivorship function were used for the computation of the cumulative survival rates [ 11,121. Two survival time variables were computed on the basis of the dates and events abstracted from the original patients’ records. The overall survival time was defined as the interval between the date of surgery and the date of last consultation for censored observations, or date of death for uncensored observations. The disease-free interval (DFI) was measured on the basis of the dates of surgery and diagnosis of first recurrence, except for patients who never presented a recurrence, in which case the DFI coincided with the overall survival.
Prognostic Factors in Laryngeal Cancer
89
TABLE 11. Survival Rates and Adjusted Hazard Ratios According to Selected Demographic and Clinical Characteristics 3-year ratesa Variable (%) Age 65 (20.1) Sex Male (92.4) Female (7.6) Race White (96.0) Non-White (4.0) T stage (TNM) T l a - l b (14.7) Tz-3 (72.7) T4 (12.6) N stage (TNM) No (83.5) NI (8.6) N2-3 (7.9) Primary tumor site Transglottic (7 1.9) Glottic (19.8) Supraglottic (8.3)
Adjusted hazard ratiosb
DFI
0s
Relapse
Death
57.5 63.2 52.5
61.2 68.6 53.0
1.0 (ref) 1.12 (0.7, 1.9) 1.77 (1.0, 3.2)
1.0 (ref) 0.99 (0.6, 1.7) 2.06 ( 1 . 1 , 3.9)
59.7 60.4
62.9 75.6
1.0 (ref) 0.75 (0.4, 1.5)
1.0 (ref) 0.46 (0.2, 1.1)
59.5 72.9
62.9 100.0
1.0 (ref) 0.47 (0.1, 2.0)
80.7 58.6 41.5
79.7 63.9 42.6
1.0 (ref) 3.04 (1.4, 6.6) 5.77 (2.3, 14.2)
1.0 (ref) 2.56 (1.2, 5.6) 4.89 (1.9, 12.3)
62.4 39.9 55.9
66.2 44.3 59.5
1.0 (ref) 1.90 (1.1, 3.4) 1.56 (0.8, 2.9)
1.0 (ref) 1.79 (1.O, 3.3) 1.53 (0.8, 3.0)
57.3 77.7 37.5
61.2 78.6 51.4
1.0 (ref) 0.57 (0.3, 1.2) 1.62 (0.9, 2.9)
1.0 (ref) 0.51 (0.2, 1.1) 1.38 (0.7, 2.6)
1.0 (ref) 65 years) experienced significantly increased risks of recurrence and death, as compared with younger patients (65 years Primary tumor site Glottic Supraglottic Extracapsular spread (positive nodes) No Yes Not resected Surgical margins Involved Paratracheal dissection Unilateral Bilateral Prior tracheostomy Yes Age and paratracheal dissection interaction Unilateral and >65 years Bilateral and >65 years Model for the risk of death T stage T2-3
T4 Age group >65 years Surgical margins Invo1ved Paratracheal dissection Unilateral Bilateral Prior trachestomy Yes Race Non- White Sex Female N stage NI+ Vascular embolization Yes Histological grade I1 or I11
-15.433 -0.795
62 1.43
65 years) had approximately twice the risk of younger patients, whereas unilateral paratracheal dissections tended to be equated to almost a doubling in risk as compared with no neck dissection. On the other hand, patients submitted to bilateral paratracheal dissections fared slightly better than those with no dissection. However, the contribution of interaction terms for these two variables indicated that the above effects due to paratracheal dissection were consid-
94
Kowalski et al. DISEASE-FEE I ) ( T E R I #
A
LOW
INTERMEDIATE
I
HIGH
I
96
120 T I E (rnHt)
144
B I LOW INTERMEDIATE
HIGH
iao
96
TI=
144
(rnHt)
Fig. 1. Cumulative survival curves according to risk groups defined by the multivariate models shown in Table VI. A: DFI (recurrence as endpoint). B: Overall survival (death as endpoint).
erably magnified among elderly patients. The tendency for a poorer prognosis due to unilateral dissections was somewhat more pronounced in the older age groups (more than a doubling in risk as compared with no paratracheal dissection). In addition, while bilateral paratracheal dissections conferred a 30% reduction in risk of recurrence with respect to no dissection among younger patients, the equivalent risk reduction approached 80% among those older than 65 years. The reason for this significant variation in the prognostic effect of paratracheal dissection across age categories could suggest differences in the surgical management of the patients as a function of age. Although such differences could not be imputed to treatment policy per se it is plausible to assume that some freedom of judgment was left for the surgeon with respect to how radical node dissections should be based on the patient’s age. Regardless of the actual mechanism governing the effect modification due to age, our results indicate that bilateral paratracheal node dissection should be the procedure of choice for selected cases, possibly with maximum benefit when performed in elderly patients. The prognostic importance of the involvement of surgical margins has been recorded in previous studies [3,18,19]. Postoperative irradiation has been reported as improving the survival among affected cases [20]. However, this has not been confirmed by Mirimanoff et al.
[21]. Our finding of a strong effect of surgical margins on survival highlights the importance of the recommendation [22] that frozen sections of surgical margins be mandatory, and that additional resection of pharynx mucosa be performed whenever necessary. Extracapsular spread was another independent prognostic indicator in the present series. Patients in whom this condition was identified experienced more than twice the risk of recurrence as compared with those with no lymph node involvement. It is noteworthy that the absence of extracapsular spread conferred a similar risk level to that experienced by node-negative patients. Extracapsular spread was also more frequent in cases with nodes greater than 1 cm in diameter and in cases with multiple nodes (data not shown). The implication of this characteristic on prognosis has been previously documented [8,9]. Such cases should be submitted to radical neck dissection, independent of node size. In most cases with small nodes extracapsular spread can only be ascertained by histological diagnosis. Therefore, it is necessary that the surgeon performing a partial neck dissection maintain a high degree of suspicion. Postoperative radiotherapy may improve regional control and increase survival in patients submitted to radical neck dissection whose metastatic nodes present extracapsular spread [23]. There is general agreement in the literature that
Prognostic Factors in Laryngeal Cancer
parameters related to lymph node metastasis have a major impact on survival for most head and neck cancers [8]. We have found that besides extracapsular spread (for the risk of recurrence) metastasis in N stage was also an independent prognostic variable emerging in the analysis subset for overall survival. The basic difficulty in assessing the prognostic value of lymph node characteristics stems from the fact that these variables are strongly associated with each other. This high degree of interdependence may lead to collinearity and other estimability problems in regression analysis. It is unlikely that these circumstances have affected our results materially. In preliminary analyses controlled for age, stage, and hospital, contralateral nodes was the single most important nodal status variable. Prior tracheostomy was identified as an important prognostic factor in the present study. It should be noted that both the statistical strength and the magnitude of the association with this variable were highest when considering death as the endpoint of interest. This could be attributed to the tendency for the occurrence of paraostomal recurrences among patients submitted for tracheostomy . Although parastomal recurrences were not the most prevalent type of local failure they were the most lethal (data not shown). Similar results have been obtained by Yuen et al. [6]. The poorer prognosis among patients submitted to tracheostomy should be construed as an important decision criterion for performing emergency laringectomies on a more frequent basis. This study has confirmed that there is much to gain in our ability to predict the survival of patients with larynx carcinoma by utilizing the information from a larger set of variables than that containing only the classical disease stage indices. The importance of better allocating resources and improving the efficiency of diagnostic surveillance during follow-up cannot be overemphasized. Since some of the prognostic factors represent clinical characteristics and tumor-related conditions they are known at diagnosis and thus, they can be used in estimating levels of risk expectancy for subsequent optimization of treatment and follow-up planning. These estimates of risk of treatment failure can be used not only in the management of individual cases but also for the stratification of patients according to comparable prognostic subsets in prospective controlled studies of treatment efficacy.
ACKNOWLEDGMENTS The authors are indebted to Drs. Humberto Torloni and Lourdes A. Marques for valuable suggestions, and to Ms. Raimunda N. Pereira for assistance with data processing.
95
REFERENCES 1 Silverberg E, Lubera J: Cancer statistics. CA 39:3-39, 1989. 2 DeSanto LW: T, glottic cancer: options and consequences of the options. Laryngoscope 94:1311-1315, 1984. 3 So0 KC, Shah JP, Gopinath KS, Gerold FP, Jaques DP, Strong EW: Analysis of prognostic variables and results after supraglottic partial laryngectomy. Am J Surg 156:301-305, 1988. 4 So0 KC, Shah JP, Gopinath KS, Jaques DP, Gerold FP, Strong EW: Analysis of prognostic variables and results after vertical partial laryngectomy. Am J Surg 156:264-268, 1988. 5 Robbins KT: Prognostic and therapeutic implications of gender and menopausal status in laryngeal cancer. J Otolaryngol 1 7 9 185, 1988. 6 Yuen A, Medina JE, Goepfert H, Fletcher G: Management of stage T, and T, glottic carcinomas. Am J Surg 148;467472, 1984. 7. Gilbert RW, Cullen RJ, Van Nostrand AWP, Bryce DP, Hanvood AR: Prognostic significance of thyroid gland involvement in laryngeal carcinoma. Arch Otolaryngol Head Neck Surg 1 12:85& 859, 1986. 8 Grandi C, Alloisio M, Moglia D, et al: Prognostic significance of lymphatic spread in head and neck carcinomas: therapeutic implications. Head Neck Surg 8:67-73, 1985. 9. Carter RL, Bliss JM, So0 KC, O’Brien CJ: Radical neck dissections for squamous carcinomas: pathological findings and their clinical implications with particular reference to transcapsular spread. Int J Radiat Oncol Biol Phys 13:825-832, 1987. 10. Hermanek P, Sobin LH: “Classification of Malignant Tumours.” Berlin: Springer-Verlag, 1987. 11. Kaplan EL, Meier P: Nonparametric estimation from incomplete observations. J Am Stat Assoc 53:457-481, 1958. 12. Campos-Filho N, Franco EL: Microcomputer-assisted univariate survival data analysis using Kaplan-Meier life table estimators. Comp Methods Prog Biomed 27:223-228, 1988. 13. Cox DR: Regression models and life tables. J Roy Stat Soc B34~187-220, 1972. 14. Campos-Filho N , Franco EL. Microcomputer-assisted multivariate survival data analysis using Cox’s proportional hazards regression model. Comp Methods Prog Biomed 3 1 :8 1-87, 1990. 15. Jacobs J R , Spitznagel EL, Sessions DG: Staging parameters for cancers of the head and neck: a multi-factorial analysis. Laryngoscope 95:1378-1381, 1985. 16. Kokal WA, Gardine RL, Sheibani K, et al: Tumor DNA content as a prognostic indicator in squamous cell carcinoma of the head and neck region. Am J Surg 156:27&280, 1988. 17. Pera E, Moreno A, Galindo L: Prognostic factors in laryngeal carcinoma: a multifactorial study of 416 cases. Cancer 58:928934, 1986. 18. Chen TY, Emrich W, Driscoll DL: The clinical significance of pathological findings in surgically resected margins of the primary tumor in head and neck carcinoma. Int J Radiat Oncol Biol Phys 13:833-837, 1987. 19 Lam KH, Lau WF, Wei WI: Tumor clearance at resection margins in total laryngectomy : a clinicopathologic study. Cancer 61 :22602272. 1988. 20. Gkram B, Strong EW, Shah J, Spiro RH: Failure at the primary site following multimodality treatment in advanced head and neck cancer. Head Neck Surg 6:720-723, 1984. 21. Mirimanoff RO, Wang CC, Doppke KP: Combined surgery and postoperative radiation therapy for advanced laryngeal and hypopharyngeal carcinomas. Int J Radiat Oncol Biol Phys 1 1 :499504, 1985. 22. Byers RM, Bland KI, Borlase B, Luna M: The prognostic and therapeutic value of frozen section determinations in the surgical treatment of squamous carcinoma of the head and neck. Am J Surg 1361525-528, 1985. 23. Ali S , Tiwari RM, Snow GB: False-positive and false-negative neck nodes. Head Neck Surg 8:78-82, 1985.
Prognostic Factors in Laryngeal Cancer Patients Submitted to Surgical Treatment LUlZ P. KOWALSKI, MD, PhD, EDUARDO L. FRANCO, PhD, JOZIAS DE ANDRADE SOBRINHO, MD, P m , BENEDIT0 V. OLIVEIRA, MD, AND P A U L 0 L. PONTES, MD, PhD From the Epidemiology and Biostatistics Unit, Ludwig Institute for Cancer Research (L.P.K., E.L.F.), Head and Neck Service, Heli6polis Hospital (L.P.K., I.D.A.S.), and Head and Neck Surgery-Otolaryngology, Paulista Schooi of Medicine (P.L.P.), Sao Paulo, and Head and Neck Service, Erasto Caertner Hospital (B.V.O.), Curitiba, Brazil.
The authors studied some demographic, clinical, pathologic, and therapeutic variables in relation to the disease-free interval (DFI) and to the overall survival of laryngeal cancer patients. The patient population consisted of 278 consecutive cases of primary laryngeal cancer who had been admitted to four head and neck surgery services in Sao Paulo and Curitiba (Brazil) between 1973 and 1986. All patients had surgical treatment. Although several variables exhibited individual associations with survival, only a few were deemed to have independent prognostic value using multivariate regression techniques based on Cox’s proportional hazards model. The primary tumor stage (T index) represented the variable with the highest predictive strength with respect to survival. Hazard ratios for the risk of disease recurrence were 2.3 (Tz-3 vs. T,) and 4.3 (T4 vs. T,). Hazard ratios for the risk of death were 1.6 (Tzp3vs. T,) and 3.3 (T4 vs. TI). The following additional factors were also independently associated with the DFI: age, site of primary tumor (transglottic, glottic, or supraglottic), extracapsular spread of lymph node metastasis, surgical margins, type of paratracheal node dissection, and prior tracheostomy. Except for extracapsular spread, the same factors related to the DFI were also significantly related to the overall survival. Race, sex, N-stage, vascular embolization, and histological grade were other independent predictors of the risk of death. KEY WORDS:survival analysis, disease-free interval, head and neck surgery
INTRODUCTION An estimated 12,300 new cases of laryngeal cancer are diagnosed each year in the United States. The number of deaths due to this neoplasm is close to one-third of that annually [ 11. The prognosis of untreated cases of laryngeal cancer is very poor. Advanced disease is characterized by one or more of the following conditions: respiratory insufficiency, dysphagia, odinophagia, hemorrhage, and cachexia. Despite the severity of the clinical presentation of the advanced form of the disease, laryngeal carcinoma is a potentially curable disease, especially in cases diagnosed at early stages. The decision regarding treatment regimen has traditionally been dependent on 0 1991 Wiley-Liss, Inc.
site, local extent, nodal status, and on the general medical condition of the patient [2-4]. Posttreatment survival rates are the most important indices of therapeutic efficacy, but even using rigid criteria of patient eligibility for standardized protocols, results are not uniform. Accepted for publication May 20, 1991. Address reprint requests to Dr. Luiz P. Kowalski, Instituto Ludwig de Pesquisa sio Cancer, Rua Prof. Antonio Prudente, 109, 01509-Sao Paulo, Brazil. L.P.K.’s present address is Head and Neck Surgery Department, A.C. Camargo Hospital-Antonio Prudente Foundation, Sao Paulo, Brazil. E.L.F.’s present address is Department of Epidemiology, lnstitut Armand-Frappier, University of Quebec, Laval, Canada.
88
Kowalski et al.
Although stage has long been clinically accepted as the best single determinant of survival for larynx cancer patients, prognosis is complex and may depend on several other risk factors such as age and sex [ 5 ] ,prior tracheostomy [6], histological grade and margins of surgical resection [4], invasion of the thyroid [7], and lymph node involvement [8,9]. However, the importance of each one of these factors has not been unequivocally demonstrated. Many of these variables are, in fact, interrelated and their apparent association with prognosis may be due to secondary relationships with genuine correlates of survival, such as measures of disease extension at diagnosis. Our present knowledge about the relative importance of these risk factors has been mostly derived from studies based on univariate analyses of candidate prognostic factors. The present study was prompted by the need to evaluate the prognostic importance of several clinical, pathological, and treatment-related variables after controlling for the effect of staging information and other determinants of survival.
PATIENTS AND METHODS Patients We reviewed the records of 278 consecutive patients with newly diagnosed laryngeal carcinoma referred to four Head and neck Surgery Services in Sao Paulo (Heliopolis Hospital, Servidor Publico Municipal Hospital, and Paulista School of Medicine) and Curitiba (Erasto Gaertner Hospital) during the period of July 1973-June 1986. Only patients submitted to surgical treatment were considered eligible for the study. Patients submitted to radiotherapy or surgery prior to the referral, patients electively treated by irradiation (stage I glottic carcinoma), or those who refused surgical treatment were not included in the investigation. All diagnoses were confirmed histopathologically prior to surgery, and anatomical site was ascertained postsurgically . Staging of disease was done by the Union Internationale Contre le Cancer’s tumor-nodes-metastasis (TNM) staging system [ 101 (Table I). Management policy for these patients varied relatively little within each institution during the study. Initial surgical treatments were total laryngectomy in 196 patients (70.5%), total laryngectomy and partial pharyngectomy in 3 1 (1 1.2%), partial vertical laryngectomies in 40 ( 14.4%), and horizontal supraglottic laryngectomy in 12 (4.3%). Dissection of paratracheal lymph nodes was performed either unilaterally or bilaterally in patients with confirmed or presumed involvement of the subglottis. Unilateral paratracheal lymph node dissection was the standard technique used at Erasto Gaernter Hospital, and the bilateral technique was used at Heliopolis and Servidor Publico Municipal Hospitals. Resection of the ipsilateral paratracheal lymph node chain with thyroid lobectomy was performed in 50 cases and bilateral
TABLE I. Distribution of Laryngeal Cancer Cases According to T and N Categories of the UICC TNM Classification Nodal stage
Tumor stage
No
TI,
27 14 45 122 24 232
Tib
T2 T3 T4
Total
NI
4 15 5 24
Nla
4 I 5
Nzb
N2c
1
1 5
4
3 9
2 7
N3
1
1
Total
27 14 52 150 35 278
paratracheal lymph node dissection with thyroid lobectomy was done in 93 cases. A total of 176 neck dissections (29 bilateral) was performed concomitantly with the primary resection in 147 patients (elective in 101 patients and therapeutic in 46). Patients were distributed according to the surgical technique for cervical lymph node dissection as follows: 94 radical classical neck dissections (15 bilateral), 33 functional neck dissections (7 bilateral), 7 extended supra-omohyoid dissections (4 bilateral), and 13 jugular lymph node chain dissections (3 bilateral). Postoperative radiotherapy was used as an adjuvant to surgery in 60 cases because of involved surgical margins and/or positive lymph nodes. The treatment consisted of a wide field technique including both sides of the neck with a dose range of 50-65 Gy. The number of metastatic nodes ranged from &20 in 191 cases with identifiable lymph nodes in the specimen. The diameter of metastatic lymph nodes varied from 1-8 cm. The level of lymph node involvement was evaluated in 147 patients and classified according to the presence or absence of metastasis in the jugular lymph node chain that was ipsilateral to the primary tumor. Lower metastatic lymph node involvement was considered either independently or associated with upper level involvement. The following additional characteristics of metastatic nodes were considered as candidate prognostic variables: contralateral lymph node involvement, preand/or paratracheal lymph node involvement, and extracapsular spread of tumor in lymph node metastasis.
Statistical Analysis Product-limit estimates of the survivorship function were used for the computation of the cumulative survival rates [ 11,121. Two survival time variables were computed on the basis of the dates and events abstracted from the original patients’ records. The overall survival time was defined as the interval between the date of surgery and the date of last consultation for censored observations, or date of death for uncensored observations. The disease-free interval (DFI) was measured on the basis of the dates of surgery and diagnosis of first recurrence, except for patients who never presented a recurrence, in which case the DFI coincided with the overall survival.
Prognostic Factors in Laryngeal Cancer
89
TABLE 11. Survival Rates and Adjusted Hazard Ratios According to Selected Demographic and Clinical Characteristics 3-year ratesa Variable (%) Age 65 (20.1) Sex Male (92.4) Female (7.6) Race White (96.0) Non-White (4.0) T stage (TNM) T l a - l b (14.7) Tz-3 (72.7) T4 (12.6) N stage (TNM) No (83.5) NI (8.6) N2-3 (7.9) Primary tumor site Transglottic (7 1.9) Glottic (19.8) Supraglottic (8.3)
Adjusted hazard ratiosb
DFI
0s
Relapse
Death
57.5 63.2 52.5
61.2 68.6 53.0
1.0 (ref) 1.12 (0.7, 1.9) 1.77 (1.0, 3.2)
1.0 (ref) 0.99 (0.6, 1.7) 2.06 ( 1 . 1 , 3.9)
59.7 60.4
62.9 75.6
1.0 (ref) 0.75 (0.4, 1.5)
1.0 (ref) 0.46 (0.2, 1.1)
59.5 72.9
62.9 100.0
1.0 (ref) 0.47 (0.1, 2.0)
80.7 58.6 41.5
79.7 63.9 42.6
1.0 (ref) 3.04 (1.4, 6.6) 5.77 (2.3, 14.2)
1.0 (ref) 2.56 (1.2, 5.6) 4.89 (1.9, 12.3)
62.4 39.9 55.9
66.2 44.3 59.5
1.0 (ref) 1.90 (1.1, 3.4) 1.56 (0.8, 2.9)
1.0 (ref) 1.79 (1.O, 3.3) 1.53 (0.8, 3.0)
57.3 77.7 37.5
61.2 78.6 51.4
1.0 (ref) 0.57 (0.3, 1.2) 1.62 (0.9, 2.9)
1.0 (ref) 0.51 (0.2, 1.1) 1.38 (0.7, 2.6)
1.0 (ref) 65 years) experienced significantly increased risks of recurrence and death, as compared with younger patients (65 years Primary tumor site Glottic Supraglottic Extracapsular spread (positive nodes) No Yes Not resected Surgical margins Involved Paratracheal dissection Unilateral Bilateral Prior tracheostomy Yes Age and paratracheal dissection interaction Unilateral and >65 years Bilateral and >65 years Model for the risk of death T stage T2-3
T4 Age group >65 years Surgical margins Invo1ved Paratracheal dissection Unilateral Bilateral Prior trachestomy Yes Race Non- White Sex Female N stage NI+ Vascular embolization Yes Histological grade I1 or I11
-15.433 -0.795
62 1.43
65 years) had approximately twice the risk of younger patients, whereas unilateral paratracheal dissections tended to be equated to almost a doubling in risk as compared with no neck dissection. On the other hand, patients submitted to bilateral paratracheal dissections fared slightly better than those with no dissection. However, the contribution of interaction terms for these two variables indicated that the above effects due to paratracheal dissection were consid-
94
Kowalski et al. DISEASE-FEE I ) ( T E R I #
A
LOW
INTERMEDIATE
I
HIGH
I
96
120 T I E (rnHt)
144
B I LOW INTERMEDIATE
HIGH
iao
96
TI=
144
(rnHt)
Fig. 1. Cumulative survival curves according to risk groups defined by the multivariate models shown in Table VI. A: DFI (recurrence as endpoint). B: Overall survival (death as endpoint).
erably magnified among elderly patients. The tendency for a poorer prognosis due to unilateral dissections was somewhat more pronounced in the older age groups (more than a doubling in risk as compared with no paratracheal dissection). In addition, while bilateral paratracheal dissections conferred a 30% reduction in risk of recurrence with respect to no dissection among younger patients, the equivalent risk reduction approached 80% among those older than 65 years. The reason for this significant variation in the prognostic effect of paratracheal dissection across age categories could suggest differences in the surgical management of the patients as a function of age. Although such differences could not be imputed to treatment policy per se it is plausible to assume that some freedom of judgment was left for the surgeon with respect to how radical node dissections should be based on the patient’s age. Regardless of the actual mechanism governing the effect modification due to age, our results indicate that bilateral paratracheal node dissection should be the procedure of choice for selected cases, possibly with maximum benefit when performed in elderly patients. The prognostic importance of the involvement of surgical margins has been recorded in previous studies [3,18,19]. Postoperative irradiation has been reported as improving the survival among affected cases [20]. However, this has not been confirmed by Mirimanoff et al.
[21]. Our finding of a strong effect of surgical margins on survival highlights the importance of the recommendation [22] that frozen sections of surgical margins be mandatory, and that additional resection of pharynx mucosa be performed whenever necessary. Extracapsular spread was another independent prognostic indicator in the present series. Patients in whom this condition was identified experienced more than twice the risk of recurrence as compared with those with no lymph node involvement. It is noteworthy that the absence of extracapsular spread conferred a similar risk level to that experienced by node-negative patients. Extracapsular spread was also more frequent in cases with nodes greater than 1 cm in diameter and in cases with multiple nodes (data not shown). The implication of this characteristic on prognosis has been previously documented [8,9]. Such cases should be submitted to radical neck dissection, independent of node size. In most cases with small nodes extracapsular spread can only be ascertained by histological diagnosis. Therefore, it is necessary that the surgeon performing a partial neck dissection maintain a high degree of suspicion. Postoperative radiotherapy may improve regional control and increase survival in patients submitted to radical neck dissection whose metastatic nodes present extracapsular spread [23]. There is general agreement in the literature that
Prognostic Factors in Laryngeal Cancer
parameters related to lymph node metastasis have a major impact on survival for most head and neck cancers [8]. We have found that besides extracapsular spread (for the risk of recurrence) metastasis in N stage was also an independent prognostic variable emerging in the analysis subset for overall survival. The basic difficulty in assessing the prognostic value of lymph node characteristics stems from the fact that these variables are strongly associated with each other. This high degree of interdependence may lead to collinearity and other estimability problems in regression analysis. It is unlikely that these circumstances have affected our results materially. In preliminary analyses controlled for age, stage, and hospital, contralateral nodes was the single most important nodal status variable. Prior tracheostomy was identified as an important prognostic factor in the present study. It should be noted that both the statistical strength and the magnitude of the association with this variable were highest when considering death as the endpoint of interest. This could be attributed to the tendency for the occurrence of paraostomal recurrences among patients submitted for tracheostomy . Although parastomal recurrences were not the most prevalent type of local failure they were the most lethal (data not shown). Similar results have been obtained by Yuen et al. [6]. The poorer prognosis among patients submitted to tracheostomy should be construed as an important decision criterion for performing emergency laringectomies on a more frequent basis. This study has confirmed that there is much to gain in our ability to predict the survival of patients with larynx carcinoma by utilizing the information from a larger set of variables than that containing only the classical disease stage indices. The importance of better allocating resources and improving the efficiency of diagnostic surveillance during follow-up cannot be overemphasized. Since some of the prognostic factors represent clinical characteristics and tumor-related conditions they are known at diagnosis and thus, they can be used in estimating levels of risk expectancy for subsequent optimization of treatment and follow-up planning. These estimates of risk of treatment failure can be used not only in the management of individual cases but also for the stratification of patients according to comparable prognostic subsets in prospective controlled studies of treatment efficacy.
ACKNOWLEDGMENTS The authors are indebted to Drs. Humberto Torloni and Lourdes A. Marques for valuable suggestions, and to Ms. Raimunda N. Pereira for assistance with data processing.
95
REFERENCES 1 Silverberg E, Lubera J: Cancer statistics. CA 39:3-39, 1989. 2 DeSanto LW: T, glottic cancer: options and consequences of the options. Laryngoscope 94:1311-1315, 1984. 3 So0 KC, Shah JP, Gopinath KS, Gerold FP, Jaques DP, Strong EW: Analysis of prognostic variables and results after supraglottic partial laryngectomy. Am J Surg 156:301-305, 1988. 4 So0 KC, Shah JP, Gopinath KS, Jaques DP, Gerold FP, Strong EW: Analysis of prognostic variables and results after vertical partial laryngectomy. Am J Surg 156:264-268, 1988. 5 Robbins KT: Prognostic and therapeutic implications of gender and menopausal status in laryngeal cancer. J Otolaryngol 1 7 9 185, 1988. 6 Yuen A, Medina JE, Goepfert H, Fletcher G: Management of stage T, and T, glottic carcinomas. Am J Surg 148;467472, 1984. 7. Gilbert RW, Cullen RJ, Van Nostrand AWP, Bryce DP, Hanvood AR: Prognostic significance of thyroid gland involvement in laryngeal carcinoma. Arch Otolaryngol Head Neck Surg 1 12:85& 859, 1986. 8 Grandi C, Alloisio M, Moglia D, et al: Prognostic significance of lymphatic spread in head and neck carcinomas: therapeutic implications. Head Neck Surg 8:67-73, 1985. 9. Carter RL, Bliss JM, So0 KC, O’Brien CJ: Radical neck dissections for squamous carcinomas: pathological findings and their clinical implications with particular reference to transcapsular spread. Int J Radiat Oncol Biol Phys 13:825-832, 1987. 10. Hermanek P, Sobin LH: “Classification of Malignant Tumours.” Berlin: Springer-Verlag, 1987. 11. Kaplan EL, Meier P: Nonparametric estimation from incomplete observations. J Am Stat Assoc 53:457-481, 1958. 12. Campos-Filho N, Franco EL: Microcomputer-assisted univariate survival data analysis using Kaplan-Meier life table estimators. Comp Methods Prog Biomed 27:223-228, 1988. 13. Cox DR: Regression models and life tables. J Roy Stat Soc B34~187-220, 1972. 14. Campos-Filho N , Franco EL. Microcomputer-assisted multivariate survival data analysis using Cox’s proportional hazards regression model. Comp Methods Prog Biomed 3 1 :8 1-87, 1990. 15. Jacobs J R , Spitznagel EL, Sessions DG: Staging parameters for cancers of the head and neck: a multi-factorial analysis. Laryngoscope 95:1378-1381, 1985. 16. Kokal WA, Gardine RL, Sheibani K, et al: Tumor DNA content as a prognostic indicator in squamous cell carcinoma of the head and neck region. Am J Surg 156:27&280, 1988. 17. Pera E, Moreno A, Galindo L: Prognostic factors in laryngeal carcinoma: a multifactorial study of 416 cases. Cancer 58:928934, 1986. 18. Chen TY, Emrich W, Driscoll DL: The clinical significance of pathological findings in surgically resected margins of the primary tumor in head and neck carcinoma. Int J Radiat Oncol Biol Phys 13:833-837, 1987. 19 Lam KH, Lau WF, Wei WI: Tumor clearance at resection margins in total laryngectomy : a clinicopathologic study. Cancer 61 :22602272. 1988. 20. Gkram B, Strong EW, Shah J, Spiro RH: Failure at the primary site following multimodality treatment in advanced head and neck cancer. Head Neck Surg 6:720-723, 1984. 21. Mirimanoff RO, Wang CC, Doppke KP: Combined surgery and postoperative radiation therapy for advanced laryngeal and hypopharyngeal carcinomas. Int J Radiat Oncol Biol Phys 1 1 :499504, 1985. 22. Byers RM, Bland KI, Borlase B, Luna M: The prognostic and therapeutic value of frozen section determinations in the surgical treatment of squamous carcinoma of the head and neck. Am J Surg 1361525-528, 1985. 23. Ali S , Tiwari RM, Snow GB: False-positive and false-negative neck nodes. Head Neck Surg 8:78-82, 1985.
