Parotid cancer: Impact of changes from the 1997 to the 2002 American Joint Committee on Cancer classification on outcome prediction




The TNM classification [American Joint Committee on Cancer (AJCC)] of salivary gland cancer was revised again in 2002. In the present study, the outcome prediction of the new TNM system was compared with the old 1997 TNM system in 202 patients with primary parotid cancer.


All patients treated from 1986 to 2006 were reclassified in both TNM systems. Disease-free survival (DFS) rates were calculated. The prognostic validity of both the TNM system and other factors were analyzed univariately (log-rank test) and multivariately (Cox regression).


AJCC TNM stage changes from 1997 to 2002 altered the disease staging in 35% of the patients. Concerning DFS, the new TNM 2002 staging allowed significantly better separation of stage III, but not of stage I versus stage II. The TNM 2002 staging was the most powerful predictor for DFS according to multivariate analysis. The 1997 system showed no independent significance. The subclassification of the new stage IV was not satisfactory; no clear distinction of IVA versus III, and IVA versus IVB was possible.


The TNM 2002 staging is more valid than the 1997 system, but a significant problem was observed in separating stage I from stage II, and within the stage IV subgroups. Cancer 2008. © 2008 American Cancer Society.

The TNM classification [American Joint Committee on Cancer (AJCC)] for salivary gland carcinoma was revised in 2002.1 The previous revision had been performed in 1997. In an analysis of 134 patients with parotid cancer, it was stated that the 1997 system had a more reliable prognostic value than the 1987 system.2 Nevertheless, only a minority of patients could be classified as stage III with the 1997 system. No significant differences between the survival curves of stage I versus II and III versus IV were observed.3 The studies also disclosed that it is difficult to prove the validity of the TNM classification for parotid malignancies, principally because major salivary gland cancer (including also submandibular and sublingual gland cancer) is a rare disease, accounting for only approximately 1% to 3% of all head and neck cancers.4 In the 2002 TNM system, no changes were made to the N and M classifications. But to improve the internal consistency of T staging across all sites, the description for T3 was revised. Now all tumors >4 cm are considered T3, but the maximal size is no longer limited. In accordance with other tumor sites, T4 is divided in resectable (T4a) and unresectable (T4b) tumors (Table 1). The former stage I is split into stages I and II, and in the 2002 system stage III is enlarged by inclusion of T3N0M0 and T3N1M0. In accordance with other tumor localizations, stage IV is now separated into stage IVA, IVB, and IVC (Table 2). To date, it is unclear whether these changes have improved the validity of the TNM classification for parotid cancer.

Table 1. 1997 and 2002 TNM Classification for Major Salivary Gland Tumors
19972002 Changes
T1Tumor ≤2 cm without extraparenchymal extension
T2Tumor >2-4 cm without extraparenchymal extension
T3Tumor having extraparenchymal extension without 7th nerve involvement and/or >4-6 cmT3Tumor having extraparenchymal extension without 7th nerve involvement and/or >4 cm
T4Tumor invades base of skull, 7th nerve, and/or >6 cmT4aTumor invades skin, mandible, ear canal, and/or facial nerve
 T4bTumor invades skull base and/or pterygoid plates and/or carotid artery
N1Metastasis in a single lymph node, ≤3 cm
N2Metastasis in a single lymph node, 3-6 cm; Multiple ipsilateral, bilateral or contralateral lymph nodes ≤6 cm
N3Metastasis in a lymph node >6 cm
M1Distant metastasis
Table 2. 1997 and 2002 TNM Stage Grouping for Major Salivary Gland Tumors
19972002 Changes
Stage IT1N0M0 T1N0M0
Stage IIT3N0M0 T2N0M0
Stage IIIT1N1M0 T3N0M0
Stage IVT4N0M0IVAT4aN0M0
Any T N2M0T2N2M0
Any T N3M0T3N2M0
Any T any N M1T4aN2M0
Any T N3M0
  IVCAny T any N M1

In this retrospective study, we evaluated whether the 6th edition of the TNM classification (2002) provided more reliable prognostic information compared with the 5th edition (1997) in patients with parotid cancer, based on data from 202 patients treated in a single center.


We reviewed the file records of all patients who presented with previously untreated primary parotid carcinoma between 1985 and 2006 at the Department of Otorhinolaryngology, Head and Neck Surgery at the University of Cologne, Germany. Patients with lymphoma, ipsilateral skin cancer, secondary invasion of the parotid gland, and parotid metastasis of other tumors were not included in the study. We identified 202 cases admitted during the selected time frame. Demographic and follow-up clinical information was obtained from the review of the patients' medical histories and pathology reports. Standard treatment in patients without distant metastasis consisted of parotidectomy and neck dissection. Radical parotidectomy was performed in cases of facial nerve infiltration. In patients with clinical lymph node metastases after the staging procedure (cN+), a radical or radical-modified neck dissection was performed. Most patients without clinical lymph node metastases (cN0) received a selective or radical-modified neck dissection. Normally, patients with stage III or stage IV tumors and patients with adenoid cystic carcinoma received postoperative radiotherapy, which was administered with daily doses of 1.8-2.0 Gy, 5 days per week for a total dose of 60-65 Gy.

All staging was based on the clinical, radiographic, and pathologic descriptions. A review of the clinical descriptions was performed to control the T staging concerning the potential extraparenchymal extension of the primary tumor. A review of the pathology reports concerning the exact tumor size was performed to adjust the T staging if necessary. Tumor stage of each patient was recoded as described in the 5th (1997) and 6th (2002) editions of the AJCC Cancer Staging Manual (Tables 1 and 2). If we refer to the 1997 TNM system, we will use the term “old” stage system, and in reference to the 2002 TNM system, we will use the term “new” stage system. The histopathologic subtype of the tumor and its classification as low-grade or high-grade were determined by the World Health Organization classification from 1991.

Follow-up data were based on the last documented visit. The period from date of histologic diagnosis to last follow-up assessment or death was used to calculate survival and local control rates. Statistical analysis of the data was performed by using SPSS software for medical statistics, version 15.0.1. The disease-free survival (DFS) curves were calculated by using the Kaplan-Meier method. Comparisons between survival curves were made by using the 2-sided log-rank test. P values <.05 were considered to be statistically significant. Prognostic factors associated with an impact (P < .2) on DFS were included in the multivariate Cox regression analysis. Odds ratios were obtained associated with 95% confidence intervals.


Patients and Treatment

A total of 202 patients met the study eligibility criteria and were included for analysis. The study population comprised 107 (53%) women and 95 (47%) men, with a median age of 60.7 years (range, 7 years-88 years). A lateral parotidectomy was performed in 26 (13%) patients, total parotidectomy in 101 (50%) patients, radical parotidectomy in 73 (37%) patients, and only a biopsy in 2 (1%) patients. A selective neck dissection was performed in 83 (41% patients) and a radical/radical-modified neck dissection in 84 (42%) patients. Thirty-three (16%) patients did not receive a neck dissection. Postoperative radiotherapy was performed in 83 (47%) patients. Histologic characteristics of the cohort are given in Table 3. The surgical margins were negative in 151 (75%) patients and positive in 51 (25%) patients. Thirty-five (17%) patients showed extracapsular lymph node spread in the neck.

Table 3. Histopathologic Types of Parotid Cancer
Histopathologic TypeNo. of Patients (%)
Adenocarcinoma35 (17)
Adenoid cystic carcinoma29 (14)
Mucoepidermoid carcinoma25 (12)
Acinic cell carcinoma23 (11)
Squamous cell carcinoma17 (8)
Undifferentiated carcinoma16 (8)
Carcinoma ex pleomorphic adenoma14 (7)
Salivary duct carcinoma9 (5)
Epithelial-myoepithelial carcinoma8 (4)
Myoepithelial carcinoma6 (3)
Lymphoepithelial carcinoma3 (2)
Miscellaneous17 (8)
High-grade tumors120 (59)
Low-grade tumors76 (38)
Grade unknown6 (3)

The follow-up time of the study population ranged from 0.23 months to 241 months (median: 24.17 months). The median follow-up of living patients was 24.40 months. During follow-up, 65 (32%) patients experienced recurrences, and 39 (19%) patients died.

TNM 1997 and 2002 Classifications

The distribution of the entire patient population into different T classifications in relation to N classifications is given in Table 4. The T classification changed in 32 (16%) patients from the 1997 staging to the 2002 staging: These patients were down-staged from T4 to T3. From the remaining 55 patients with stage T4 according to the 1997 classification, most (ie, 50) patients were classified with the new 2002 system as T4a and only 5 patients as T4b. A comparison of TNM stage grouping using the 1997 classification in relation to the new classification from 2002 is shown in Table 5. The stage classification changed in 71 (35%) patients from the 1997 to the 2002 classification; 28 patients were upstaged from old stage I to new stage II, and 30 patients from old stage II to new stage III. Thirteen patients were downstaged from old stage IV to new stage III. From the remaining 96 patients with stage IV according to the 1997 classification, 77 patients were classified as stage IVA, 6 patients as stage IVB, and 13 patients as stage IVC.

Table 4. TN Distribution: AJCC 1997 Versus AJCC 2002*
  • *

    Changes between AJCC 1997 and 2002 are in bold.

AJCC 1997      
AJCC 2002      
Table 5. TNM Stage Correlation: AJCC 1997 Versus AJCC 2002*
AJCC 1997AJCC 2002
  • *

    Changes between AJCC 1997 and 2002 are in bold.


Disease-free Survival

In this case series, the 5-year and 10-year DFS rates were 86% and 71%, respectively. The 5-year and 10-year DFS curves of patients according to the 1997 and 2002 classification are shown for the various T classifications in Figure 1 and for the various disease stages in Figure 2. In both staging systems, there was no clear separation between T1 and T2 patients. The separation between T2 and T3 patients did not change in the 2002 staging system. But in contrast to the 1997 system, the 2002 system allowed a significant separation between T3 and T4. The survival curves were only significantly worse in T4 in comparison to T3 in the 2002 system. There was no significant difference between the T4a and T4b patients in the 2002 system. Concerning the stage grouping, stage III was not different from stage I in the 1997 system, but the DFS was significantly better in old stage I than old stage III. There was no significant difference between stage III and stage IV in the 1997 system. In contrast, the new 2002 system did not allow a separation between new stage I and stage II patients. New stage III was now better defined, with significantly worse survival curves than new stage II. The separation between new stage III and new stage IV improved, but was still not significantly different. There was no clear separation within the new groups stage IVA, IVB, and IVC.

Figure 1.

(A) Disease-free survival rates of T classification subgroups according to the 1997 TNM staging are shown: T1 versus T2, P = .719; T2 versus T3, P = .045; T3 versus T4, P = .179. (B) Disease-free survival rates of T classification subgroups according to the 2002 TNM staging are shown: T1 versus T2, P = .719; T2 versus T3, P = .044; T3 versus T4, P = .002; T3 versus T4a, P = .003; T4a versus T4b, P = .847.

Figure 2.

(A) Disease-free survival rates of disease stage subgroups according to the 1997 TNM staging are shown. Stage I versus stage II, P = .034; stage II versus stage III, P = .505; stage III versus stage IV, P = .209. (B) Disease-free survival rates of disease stage subgroups according to the 2002 TNM staging are shown. Stage I versus stage II, P = .707; stage II versus stage III, P = .034; stage III versus stage IV, P = .165; stage III versus stage IVA, P = .562; stage IVA versus stage IVB, P = .236; stage IVB versus stage IVC, P = .171.

Other factors negatively influencing DFS on univariate analysis, in order of significance, were lymph node extracapsular spread (P < .001), positive surgical margins (P < .001), postoperative radiotherapy (P = .001), high-grade histology (P = .019), and male sex (P = .043). The parameters age >60.7 years (P = .116) and neck dissection (P = .337) did not reach statistical significance. The results of the multivariate analysis are presented in Table 6. In using the 1997 tumor staging system in the analysis, the only independent prognostic factor was extracapsular lymph node tumor spread. The 1997 tumor staging itself was not a significant independent prognostic factor for DFS. In contrast, in using the 2002 tumor staging in the analysis, the 2002 system and no other factor was a powerful prognostic factor for DFS.

Table 6. Cox Regression Analysis of the Prognostic Factors for DFS
FactorβSEPOdds Ratio95% Confidence Interval
  • β indicates β estimator; SE, standard error of the β estimator.

  • *

    Reference category.

  • Significant (P < .05) values.

  • Postoperative radiotherapy.

1997 Tumor stage.282
 1997 Stage IV*1
 1997 Stage I−1.0410.571.0680.3530.115-1.081
 1997 Stage II−0.0020.517.9980.9980.363-2.748
 1997 Stage III0.2791.069.7941.3220.163-10.755
Age, y
Tumor grade
 Low grade*1
 High grade0.3520.350.2711.4210.716-2.823
Surgical margins
Extracapsular spread
2002 Tumor stage<.001
 2002 Stage IVC*1
 2002 Stage I−3.2001.124.0040.0410.005-0.369
 Stage II−2.7320.758<.0010.0650.015-0.287
 Stage III−1.7410.557.0020.1750.059-0.523
 Stage IVA−2.0670.478<.0010.1270.050-0.323
 Stage IVB−0.3350.324.2500.4260.100-1.820
Age, y
Tumor grade
 Low grade*1
 High grade0.2480.348.4761.2810.648
Surgical margins
Extracapsular spread


This study analyzed the validity of the 2002 TNM classification in a representative sample of patients with parotid cancer. Although parotid cancer represents the majority of major salivary gland cancer, our results may not be representative for submandibular or sublingual gland cancer. Even if our study is limited by its retrospective nature, we present a large case series from a single tertiary referral center with homogeneous treatment design and consistent follow-up protocol. Therefore, it would be useful for such a study to be duplicated in datasets of other institutions to confirm the results. Taking into account the rarity of the disease, our database included a considerable number of patients with long follow-up and is distributed homogeneously on all 4 T classifications. The 10-year DFS rate of 71% is somewhat more favorable than the results of other large series of patients with parotid or major salivary gland carcinoma. The overall survival rate was 80%. Details on overall survival are not shown in the article, but the results of the comparison between the 1997 and the 2002 TNM systems regarding overall survival were identical (unpublished data). In the literature, 10-year DFS rates range from 35% to 61%, and the 10-year overall survival rates from 39% to 74%.5-10

The present study confirms findings of recent studies that tumor size and its extraparenchymal extension, facial nerve paralysis, cervical lymph node metastasis and its extracapsular spread, the histopathologic subtype and its grading, and surgical margins are the most important risk factors.7, 10, 11 The 2002 TNM classification considers all these factors aside from the histopathologic factors: subtype, grading, surgical margins, and extracapsular lymph node spread. This may be a reason why the 2002 TNM classification still does not give a significant separation of the survival curves for each stage in parotid cancer. The group of parotid cancers is very inhomogeneous, including more than 25 subtypes. Although the histologic classification system, and especially the grading, is not as simple as for thyroid cancer, it may be an improvement to include more histopathologic features (as was already done for thyroid cancer) in the next revision of the TNM classification. We tried to pool several histological subtypes beyond the low/high-grade system to improve the separation of the survival curves, but in our sample we could not detect any group with better outcome (unpublished data). Several calculation models have been published that include tumor grading, surgical margins, or perineural growth as outcome predictors for parotid cancer.7, 10, 12

The first goal in modifying the major salivary gland cancer staging system was to obtain a clearer definition of stage III. For this, the definition of T3 was widened by removing the upper 6-cm limit. Second, T3N0M0 was up-staged and T3N1M0 was down-staged to stage III. In our series, this alteration in the 2002 system led to the most pronounced improvement in the new TNM classification. Stage III is now better defined and separated from stage II and stage IV for DFS prediction.

Because of the extension of stage III, it was obligatory for the TNM system to reclassify the early disease stages I and II also. The 1997 stage I had to be split into stage I and stage II. Now, the 2002 system stage I consists only of T1N0M0 and stage II only of T2N0M0. This might be helpful for other head and neck sites, but it was not for parotid cancer. In the present series, no significant difference in the survival curves for DFS were observed for new stage I and new stage II patients. We performed univariate and multivariate analysis of the data of the stage I and II patients alone (data not shown). We could not find any significant differences in these 2 groups concerning other prognostic factors. Hence, we unfortunately cannot provide a factor that would improve prognosis accuracy for these 2 groups in a future TNM classification. A recent study found that T1N0M0 and T2N0M0 patients have the best survival rates.3 Therefore, these 2 groups are pooled together in outcome studies for parotid cancer.10

The second goal in modifying the system appears to have been to include the factor “resectability” into the 2002 classification. It has been shown that skull base infiltration dramatically decreases the survival rate of T4 patients with major salivary gland cancer.3 Therefore, there was a theoretical rationale to subdivide T4 into T4a and T4b, and to subdivide stage IV into the 3 subclasses A-C. But there is no common agreement on an accurate definition of unresectability for parotid cancer. This factor largely depends on the experience of the surgeon and the frequency of surgery for rare parotid cancer in the particular institution. In this regard, it might be explainable why most of the T4b tumors in our series underwent surgery as primary treatment. Furthermore, the classification as T4a versus T4b largely depends on the radiologic tools used for tumor staging. Consequent computed tomography and magnetic resonance imaging would probably reveal more T4b tumors. Nevertheless, parotid tumor invasion in the skull base, pterygoid plates, or carotid artery are rare events in a rare tumor. Therefore, in the current series only a small number of patients (5 T4b patients and 6 stage IVB patients) fell into these subgroups. All of these factors together might explain why the Kaplan-Meier curves did not show a distinct separation for T4a versus T4b patients. It might also explain why the separation between stages IVA, IVB, and IVC did not reach a significant level.

In the current study, we retrospectively staged 202 patients with parotid cancer who were treated in a single center, and found that the 2002 modifications made the TNM classification more useful than the 1997 version. In particular, T3 and stage III are better represented. But the overall stage grouping with the 2002 system was still less than satisfactory; separation in the early stages, stage I and stage II, and of the most advanced stages IVA, IVB, and IVC, should be a focus in future revisions.