Prognostic Factors in Mobile Tongue and Floor of Mouth Carcinoma BERT BROWN, MD,* LEON BARNES, MD,t JUAN MAZARIEGOS, MD,t FLOYD TAYLOR, ScD,$ JONAS JOHNSON, MD,* AND ROBIN L. WAGNER, BS'

This study identifies significant prognostic factors in squamous cell carcinomas of the anterior tongue and floor of mouth. It is clear that the TMN staging system does not account for other important variables that affect tumor prognosis. Tumor thickness and the presence of perineural invasion and intralymphatic tumor emboli should be examined in all resected tumors. Tumor thickness, tumor size, and perineural invasion all have an impact on survival and must be considered in treatment plans. Tumors measuring between 2 mm and 3 mm may or may not have metastases and further evaluation of this group needs to be done. Most importantly, the data in this study supports a multiinstitutional prospective evaluation of pathology specimens. Precise guidelines must be established for handling of the specimen, which must then be evaluated for the variables mentioned above. In this way, more definitive conclusions can be reached in the management of tumors of the anterior tongue and floor of mouth. Cancer 64:1195-1202,1989.

T

is intended to provide standard terminology to facilitate patient counseling, therapeutic decision-making, and comparison of results of treatment. To improve the usefulness of the staging system, a retrospective study was undertaken to evaluate the effect of tumor thickness, perineural invasion, and intravascular and intralymphatic tumor emboli on subsequent outcome after treatment. Pathologic specimens were reviewed without the knowledge of clinical course. All patients were followed a minimum of 2 years after completion of therapy. Appropriate pathologic material and clinical follow-up were available on 87 patients with squamous cell carcinoma of the mobile tongue and anterior floor of mouth. Increases in tumor thickness were correlated with decreased survival for death from all causes as well as for death from disease. Patients with tumor thickness less than or equal to 3 mm had improved survival (P= 0.0006) when compared to patients with tumor thickness between 3 mm and 7 mm and those with tumor thickness greater HE TNM STAGING SYSTEM

From the *Department of Otolaryngology, University of Pittsburgh School of Medicine, Eye and Ear Hospital of Pittsburgh, Pittsburgh; the ?Department of Pathology, University of Pittsburgh School of Medicine, Presbyterian-University Hospital, Pittsburgh; and the $Department of Community Medicine, University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania. Address for reprints: Jonas T. Johnson, MD, Professor and Vice Chairman, Department of Otolaryngology, Eye and Ear Institute, Suite 500, 203 Lothrop Street, Pittsburgh, PA 152 13. Accepted for publication March 13, 1989.

than 7 mm. Multivanant analysis revealed that perineural invasion (P = O.OOl), intralymphatic tumor emboli (P = 0.006), and tumor stage ( P = 0.012) were associated with the development of cervical metastasis. These data are in keeping with observations made by other investigators supporting the need for the integration of pathologic information into therapeutic decision-making. Close cooperation between surgeons and pathologists is critical to facilitate consistently reliable pathologic data. The TNM staging system provides standard terminology for patient counseling, therapeutic decision-making, and reporting end results. The clinical stage, however, does not reliably predict prognosis. Previous studies have suggested that there are other variables, not included in the TNM system, that correlate with tumor behavior. In particular, tumor thickness and perineural invasion of tumor were shown to correlate statistically with both survival and locoregional recurrence. Other variables such as intralymphatic and intravascular tumor emboli have also been considered as potential prognostic factors.6The goal of this study was to further evaluate tumor thickness, perineural invasion, and intravascular and intralymphatic tumor emboli to determine if they could be correlated with tumor recurrence and survival. A better predictor of metastasis and prognosis would allow improved planning of therapy and avoid unnecessary neck dissections or radiotherapy. A retrospective review of both clinical and pathologic data on surgically treated anterior tongue and floor of mouth squamous cell carcinomas was, therefore, undertaken.

1195

CANCERSeptember 15 1989

1196

FIG. 1.

Diagram of methods of measurement of tumor thickness.

Materials and Methods The hospital charts of all patients who had surgery for squamous cell carcinoma of the anterior tongue or floor

Vol. 64

of mouth between 1974 and 1984 were reviewed. Only those patients who had surgery as their first treatment modality were included. In addition, only those patients who had documented follow-up for 2 years or more and had all pathologic slides (including the original biopsy) available for review were eligible for the study. All pathology slides were reviewed with the pathologist. An ocular micrometer was used to measure the maximal thickness of tumor to the nearest millimeter, measuring vertically from the surface of the tumor. All measurements were taken from the most perpendicular cuts available. Surface keratin, parakeratin, and inflammatory exudate were not included in the measurement. In those cases where the tumor was exophytic, a measurement was taken from the tip of the papilla to maximal depth in a radial fashion (Fig. 1, diagram 1). In the case of an ulcerative tumor, a measurement was made from the ulcer base in a radial fashion to the deepest penetration of tumor (Fig. 1, diagram 2, measurement B). In addition, attempts were made to draw an imaginary mucosal line over the ulcer to the outer cellular limits of the adjacent intact mucosa, and then take measurements from this line to the deepest point of invasion (Fig. 1, diagram 2, measurement C). Comparisons were made between the depths obtained using an imaginary mucosal line versus depth from the ulcer base. Perineural invasion was defined as tumor invasion of the perineural sheath and/or epineurium (Fig. 2). To qualify as an intravascular or intralymphatic tumor embolus, the tumor cells had to be either associated with

FIG. 2. Perineural and early intraneural invasion (arrow) (H & E, X3 15).

No. 6

PROGNOSTIC FACTORS IN TONGUE AND MOUTHCA

h ~ w etn a/.

1197

FIG.3. Vascular invasion. Notice the points of attachment of the tumor to the vascular wall (arrows) and early fibrin deposition (asterisk) (H & E, X315).

fibrin deposits (thrombus) or attached to the wall of the vessel (Figs. 3 and 4). In the case of metastatic disease, tumor within the lymph nodes were confirmed, and the presence or absence of extracapsular spread outside of the lymph node was recorded. Classes of tumor thickness were created using the data obtained from the ocular micrometer measurements. Tu-

FIG.4. Lymphatic invasion (H & E, X315).

mors were grouped into the following thickness categories: Class A, 0 to I3 mm; Class B, > 3 mm I 7 mm; and Class C, > 7 mm. The impact of tumor stage, tumor thickness, perineural invasion, intralymphatic and intravascular emboli, cervical metastases, and extracapsular nodal spread on survival was assessed by univariate analysis using the Kaplan-

CANCERSeptember 1.5 1989

1198 TABLE1. Tumor Staging*

both for those variables assessed with survival and those assessed for regional recurrence.

Tumor size

Nodal status

T1

T2

T3

T4

NO N1 N2 N3

29

5 4

0

2 1 0 1

73

0 0

37 5 1 0

Total

30

43

10

4

87

1

1

Results 11

2 1

* All tumors were MO. Meier product limit method. The generalized Wilcoxon test was then used to determine statistical significance. The relationship of tumor location, tumor stage, tumor thickness, perineural invasion, and intralymphatic and intravascular tumor emboli to the occurrence of metastatic disease in clinically NO necks was assessed. This included patients found to have metastatic lymph nodes by elective neck dissection at the time of initial surgery, and those who were observed and later developed metastatic disease in the neck. Chi-square analysis was performed to assess statistical significance.The relationship of tumor thickness and tumor stage to perineural invasion and intralymphatic and intravascular tumor emboli was also assessed and tested for significance with chi-square analysis. Multivariate analysis was done using stepwise logistic regression

Total patients 231 Dead of disease 2

.-... :.........

L-,

...... ..Total patients 35/

!

cn w 2.40

L

4

___ - -

Vol. 64

Dead of disease 13 -Total patients 101 Dead of disease 4

= T,

........

= T,

-----

= T, P VALUE = ,0009

SURVIVAL IN MONTHS FIG. 5. Life tables. Tumor size: death from tumor.

One hundred forty-seven patients were initially evaluated. Of these, 87 patients met all necessary criteria for inclusion in the study. This included 39 tumors of the floor of mouth and 48 tumors of the anterior tongue. The patients ranged in age from 50 to 70 years. There were 30 women and 57 men.

Size of Primary Tumor There were 30 TI tumors, 43 T2 tumors, ten T3 tumors, and four T4 tumors (Table 1). No patients had distant metastases at diagnosis. All were followed greater than 2 years after surgery. In patients who were initially staged NO, regional metastases were discovered by elective neck dissections or developed during the follow-up period in eight of 29 (28%) TlNO tumors, 20 of 37 (54%)T2NO tumors, four of five (80%)T3NO tumors, and zero of 23 T4NO tumors. The difference was statistically significant with P = 0.02. Nine patients with T1 tumors died, seven from other causes and two from disease (adjusted survival, 2 1/23; 9 1%). Of the two patients who died from disease, one died of local recurrence and, on review of the pathology slides, was found to have a positive margin. The other patient died of regional disease and had a 6-mm thick tumor. Twenty-one patients with T2 tumors died, 13 from disease and eight from other causes (adjusted survival, 22/35; 63%). Of the 13 who died from disease, one died of local disease, seven died of regional disease, four died with local, regional, and distant disease, and one died of regional and distant disease. Four patients with T3 tumors died, all from disease (survival, six often; 60%).A local recurrence was noted in one patient; regional disease developed in four patients. All patients with T4 tumors died of other causes. A significant decrease in survival with increasing tumor stage was seen both for death from all causes ( P = 0.01) and death due to disease ( P = 0.0009) (Fig. 5). In patients dead of disease, there was a significant difference between T1 and T2 tumors ( P = 0.01). There was not a statistical difference in survival between T2 and T3 tumors (P > 0.05). There were too few T4 tumors for statistical analysis.

Classes of Tumor Thickness Distribution of tumor thickness by tumor stage is seen in Table 2. There were 25 patients in Class A, 33 patients in Class B, and 29 patients in Class C. In those patients

PROGNOSTIC FACTORS IN TONGUE AND MOUTHCA

No. 6

TABLE2. Thickness Class by Tumor Size

Brown et d.

*

1199

TABLE3. Prognostic Factors for Development of Regional Disease in NO Necks

Tumor size

Class

Depth (mm)

TI

T2

T3

T4 Variable

A

B C

0 to 5 3 >3 to 5 7 >7

16 12 2

7 20 16

2 0 8

0 1 3

25 33 29

30

43

10

4

87

clinically staged NO, nine of 24 patients (38%)in Class A developed regional disease as did 12 of 29 (41%) in Class B and 1 I of 20 (55%) in Class C . The presence of regional disease was not statistically different among classes of tumor thickness (P = 0.48) (Table 3). In Class A, eight patients died, one from recurrent local disease and seven (39%)from other causes. Adjusted survival with a minimum of 2 years follow-up was 94%(17/ 18 patients). In Class B, eight patients died of tumor, whereas seven (27%)died of other causes. Two-year determinate survival was 69%(1 8/26 patients). Of those patients who died from tumor, one had a local recurrence, seven patients had regional recurrences, and two patients had distant metastases. In Class C, ten patients died of disease, whereas five (21%) died of other causes. The 2year determinate survival was 58% (14/24 patients). Of the ten patients who died of disease, there were four local recurrences and nine regional (one contralateral) recurrences. Increases in tumor thickness were significantly correlated with decreased survival for death from all causes (P = 0.004) as well as death from disease ( P = 0.001) (Fig. 6). Although Class A survival was statistically improved compared to Class B (P = 0.0006), the difference between Class B and Class C only approached statistical significance (P= 0.053).

No. with regional disease

Total patients

8 20 4 0 9 12 11 12

29 37 5 2 24 29 20 17

38 41 55 70

0.01

7

8

88

0.006

T1 T2 T3 T4 Thickness class A Thickness class B Thickness class C Perineural invasion Intralymphatic tumor emboli

28 54 80

P value 0.02 0.48

Regional disease: those patients with positive nodes after elective neck dissection and those patients who were observed and then developed regional recurrence.

eight died of disease for an adjusted survival of 82%(37/ 45 patients). Perineural invasion had a statistically significant decrease in survival in both death from all causes (P = 0.04) and death due to disease (P= 0.0003, Fig. 7).

Intralymphatic Tumor Emboli Eleven of 87 patients ( 13%)had intralymphatic tumor emboli. Of those patients with NO necks, seven of eight (88%)with intralymphatic emboli developed regional dis-

Total patients 181 Dead of Disease 1

_. .

.Total patients 26/ Dead of disease 8

:.

L- 1

Perineural Invasion Twenty-six of 87 patients (30%) demonstrated perineural invasion. In patients initially staged NO, 12 of 17 patients (7 I %) with perineural invasion developed regional disease, whereas 20 of 56 patients (36%)without perineural invasion developed regional disease. This difference was statistically significant ( P = 0.01) (Table 3). Of the 26 patients with perineural invasion, I I died from recurrent disease and three died from other causes. Adjusted 2-year survival was 52% (1 2/23 patients). Of those patients who died from tumor, two died from local disease, seven died from regional disease (one patient with recurrence in the contralateral neck), and two died from locoregional disease. Of those 61 patients without perineural invasion, 16 (26%)died of other causes whereas

Percent

I

(I)

L

I-

z

- - .Total patients 241 Dead of Disease 10

W

gW A 0 1 a

.........

----

Thickness Class A = 0 to =3mm Thickness Class B = 3.1 to = 7mm Thickness Class C = 7mm P VALUE = ,001

O.OL0

'

i

2b. 4b.SURVIVAL ' sb. ab. Id0 ' 4 0 ' ,do' IN MONTHS I

I

I

FIG.6. Life tables. Classes of tumor thickness:death from tumor.

1200

CANCERSeptember 15 1989 1

-- r ,

1

1

1

1

1

1

1

1

1

1

1

1

1

1.0 7 b.

-

r, --I 1?

Multivariate analysis with the covariates:Class of tumor thickness, tumor stage, perineural invasion, intralymphatic tumor emboli, initial nodal status, and extracapsular spread for death from tumor was performed. Only the class of tumor thickness was statistically significant ( P = 0.002). The results were unchanged when looking at only the covariates: class of tumor thickness, tumor stage, and perineural invasion.

b7

L--

1

I I

Total patients 23/ Dead of disease 11

---- INVASION PRESENCE OF PERINEURAL

-

ABSCENCE OF PERINEURAL INVASION P VALUE = .0003

0.ot 0.0

I

I

20.

I

I

I

I

1

1

I

"

I

60. 80. 100 120 SURVIVAL IN MONTHS

40.

' 140I 1

FIG.7. Life tables. Perineural invasion: death from tumor.

ease whereas 25 of 65 (38%)without intralymphatic tumor emboli developed regional disease. The difference was highly significant ( P = 0.006) (Table 3). Three of 11 patients died from tumor whereas two ( 18%)died from other causes. Two-year determinate survival was 67% (six of nine patients) in patients with intralymphatic tumor emboli as compared to 73% (43/59) in patients without tumor emboli. This difference is not statistically significant.

Intravasciilar Tumor Emboli Only three patients had intravascular tumor emboli and no patients died during the study. Therefore, this variable was not used in any statistical analysis.

Lymph Nodes With Extracapsular Spread Only eight patients had lymph nodes with extracapsular spread identified after a neck dissection. Of those patients, TABLE4. Elective Neck Dissection Versus Observation

Group 1 Group 2

two died from disease and two died of other causes. Both patients who died from disease had regional recurrence. The small numbers precluded statistical analysis.

Multivariate Analysis of Survival

Total patients 451 Dead of disease 8

I I

Vol. 64

Total patients

DOD patients

Percent

21 7 28

9 2 I1

43 29

P value = 0.89 Group 1: patients with initial elective neck dissection; Group 2: patients necks observed initially; DOD: dead of disease.

Multivariate Analysis of Regional Recurrence Analysis of the covariates: Class of tumor thickness, perineural invasion, tumor size, and intralymphatictumor emboli for the development of regional disease was performed. Analysis of all patients, including those with clinically positive nodes on initial exam, revealed that perineural invasion ( P = 0.001),intralymphatictumor emboli ( P = 0.006),and tumor stage ( P = 0.0 12) were significantly associated with regional disease. The classes of tumor thickness did not show a statistical correlation with regional recurrence. Multivariate analysis was also performed using the covariates above to address patients with clinically negative necks only. In that analysis, intralymphatic tumor emboli ( P = 0.006), perineural invasion ( P = 0.009), and tumor stage ( P = 0.017) were still significant variables for the development of regional recurrence. Elective Neck Versus Observation The survival of those patients who had elective neck dissections with occult positive nodes (Group I) versus those patients who were observed, developed regional recurrence, and subsequently received treatment (Group 11) was compared. Unfortunately, the patient population is too small to allow meaningful statistical analysis. Table 4 shows the numbers obtained in Group I; nine of 21 patients (43%)died from disease. In Group 11, two of seven patients died from disease (29%).This suggests decreased survival in patients who were observed initially until disease recurred in the neck. Finally, tumor thickness and tumor size were correlated with perineural invasion and intralymphatic tumor emboli. The class of tumor thickness showed a significant correlation with perineural invasion ( P = 0.00 l), although there was no significant correlation with intralymphatic tumor emboli (Table 5). Tumor size showed no significant correlation with intralymphatic tumor emboli or perineural invasion.

No. 6

PROGNOSTIC FACTORSIN TONGUEAND MOUTHCA

Discussion This study evaluated multiple factors that have been considered relevant in predicting survival and recurrence of squamous cell carcinoma of the head and neck. In univariate analysis, tumor thickness, tumor size (T stage), and perineural invasion all showed a statistically significant correlation with survival. Tumor thickness, however, was not correlated with the development of regional recurrence. Similar correlations of tumor thickness with survival have been found by others. Spiro et al.' found a significant difference in survival comparing tumors < 2 mm to those > 2 mm thick. Multivariate analysis comparing tumor thickness to tumor size revealed that tumor thickness was a more important predictor of prognosis. They were also able to correlate thickness with regional recurrence. Similarly, Mohit-Tabataba et aL2showed a significant increase in cervical lymph node metastases with lesions < 1.5 mm thick as compared to those > 1.5 mm. Urist et ~ l .studied ~, buccal mucosa carcinomas and found a significant decrease in survival when comparing tumors < 6 mm in tumor thickness to those > 6 mm. One difference between this study and those cited above is the larger values for tumor thickness used by us. Only seven patients in our study had tumors measuring 2 mm or less whereas Spiro et al.' and Mohit-Tabataba et af.* had 43 of 105 patients and 57 of 84 patients with T1 and T2 tumors, respectively, measuring less than 2 mm. In those seven patients, there were no regional metastases and no deaths from disease, which is in agreement with the other studies. Close et aL6 also looked at tumor thickness along with vascular invasion and perineural invasion to determine their predictive value for lymph node metastasis. In their study, only intravascular invasion was found to correlate with regional recurrence. Close et aL6 suggest that their patient population had larger tumor sizes and, therefore, were more likely to have thicker tumors. The same holds true for our study. An important consideration, however, is the difficulties in measuring tumor thickness. With malignant melanoma, as defined by Bre~ lo w,~ strict criteria for measuring tumors are defined with the measurement taken from the deepest point of invasion to the top of the granular cell layer of the overlying epidermis. Measurements are taken at right angles to the skin surface. Breslow' noted two main problems in measurement: poor sampling of the lesion and variation in apparent thickness due to changes in the angle of sectioning. In measuring mucosal tumors in a retrospective study, there often is not a mucosal surface in the slide to use as a reference point. Furthermore, there is often a significant amount of keratin and debris

*

Brown et d.

1201

TABLE 5. Relationship of Thickness Class and Penneural Invasion Perineural invasion Thickness class A

B C

Yes

No

Percent

Total

2 4 20

23 29 9

8 12 67

25 33 29

26

61

87

on the surface, which as noted earlier, were not included in our measurements. For some patients, we had three different measurements depending on whether one chose the mucosal surface, the tumor surface, or the ulcer base as a starting point. In this study, we always chose the largest measurement, which might also account for our larger thickness measurements. Another problem was tangential cuts. In some of our cases, recuts of the paraffin blocks were done when a confident estimate of perpendicular measurement could not be done. Finally, and perhaps most important, were errors identified in sectioning of the tumor. There were some tumors in which only the periphery was sampled. not evaluating the greatest depth in the center (Fig. 1, diagram 3). Therefore, we could not be sure the deepest portion of the tumor was being measured. It thus becomes apparent why different institutions may have different results based on tumor thickness. Only with a prospective study which identifies precise guidelines for processing the tumor and measuring thickness can comparisons among institutions be made. It was expected that perineural invasion would correlate with both survival and regional recurrence. This association has been confirmed by Again, it is important to have a precise definition of perineural invasion. Larson et aL8 clearly showed that once tumor involves the perineural space, it can spread in both a longitudinal and radial fashion through the planes of least resistance. In the presence of perineural spread, the likelihood of tumor extending beyond margins of resection is increased. Thus, tumor can continue to grow leading to local, regional, and distant recurrence. It is also important to note the strong relationship between perineural invasion and tumor thickness suggesting that perineural invasion is an important factor responsible for decreased survival with increasing thickness. Identifying perineural invasion is a tedious task involving a careful review of all tumor slides. The presence of perineural invasion should make one to strongly consider adding elective neck dissection and/or radiation therapy to the resection of the primary tumor. The presence of intralymphatic tumor emboli correlated with the development of regional disease, although

1202

CANCERSeptember 15 1989

there was no correlation with survival. Regardless, our study supports elective treatment of the neck when intralymphatic tumor emboli are found. Although Close el aL6 were able to find a significant number of patients with intravascular tumor emboli, they did not require tumor attachment to the vessel well or thrombus formation. Given our criteria, there were too few patients for any conclusions to be drawn. One question our study attempted to answer was the management of the NO neck. Our data shows a strong positive correlation between tumor size, perineural invasion, intralymphatic tumor emboli, and the development of regional recurrence. Furthermore, if we only looked at tumors that were < 2 mm as Spiro et al.' did, tumor thickness correlates as well. Although our numbers were small and not statistically significant, there is a suggestion that patients undergoing elective neck dissections who are found to have occult nodes have an increased survival as compared to those patients who are observed and treated when recurrence develops. This finding is suggested by earlier studies in our institution as welL9 Therefore, in the presence of tumor size > 2 cm, tumor thickness > 2 mm, perineural invasion or intralymphatic

Vol. 64

tumor emboli, one should strongly consider elective treatment of the neck. REFERENCES I . Spiro HR, Huvos AG, Wong GY et a/. Predictive value of tumor thickness in squamous carcinoma confined to the tongue and floor of the mouth. Am J Surg 1986; 152:345-350. 2. Mohit-Tabatabai MA, Sobel MJ. Rush BF et a/. Relation ofthickness of floor of mouth Stage 1 and I I cancers to regional metastasis. A m JSitrg 1986; 152:351-353. 3. Urist MM, OBrien CJ, Soong SJ ef a/. Squamous cell carcinoma of the buccal mucosa: Analysis of prognostic factors. Am J Sitrg 1987; 154:411-414. 4. So0 KC, Carter RL, OBrien CJ et d.Prognostic implications of perineural spread in squamous carcinomas of the head and neck. Laryngoscope 1986; 96:1145-I 148. 5. Goepfert H , Dichtel WJ, Medina TE ef al. Perineural invasion in squamous cell carcinoma of the head and neck. Am J Surg 1984; 148: 542-547. 6. Close LG, Burns DK, Reisch J et a/. Microvascular invasion in cancer of the oral cavity and oropharynx. Arch Otolaryngol Head Neck Szrrg 1987; 113:1191-1195. 7. Breslow A. Prognostic factors in the treatment of cutaneous melanoma. J Cutan Pathol 1979; 6908-212. 8 . Larson DL, Rodin AE, Roberts DK ef al. Perineural lymphatics: Myth or fact. A m J Surg 1966; 112488-492. 9. Cunningham MJ, Johnson JT, Myers EN ef a/. Cervical lymph node metastasis after local excision of early squamous cell carcinoma of the oral cavity. Am JSurg 1986; 152:361-366.