| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Head and Neck Imaging |
, MD,
Johannes A. Langendijk, MD, PhD,
Barry Schenk, MD,
Menno Van Wattingen, MD,
Dirk L. Knol, PhD,
C. René Leemans, MD, PhD and
Jonas A. Castelijns, BSc, MD, PhD
1 From the Departments of Radiology (R.L., B.S., M.V.W., J.A.C.), Radiation Oncology (J.A.L.), Clinical Epidemiology and Biostatistics (D.L.K.), and Otolaryngology/Head-Neck Surgery (C.R.L.), VU University Medical Center, De Boelelaan 1117, 1081 HV Amsterdam, the Netherlands. Received June 25, 2003; revision requested September 3; final revision received November 25; accepted January 5, 2004. Address correspondence to R.L. (e-mail: redina.ljumanovic@vumc.nl).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: Pretreatment MR images acquired in 84 patients with supraglottic carcinoma treated with curative radiation therapy were reviewed for tumor involvement of laryngeal sites including glottis, subglottis, pre-epiglottic space, laryngeal cartilages, and hypopharynx, and for extralaryngeal extension. The volume of each tumor was estimated, and mean tumor volume was calculated for the group of tumors in each T staging category.
RESULTS: Results of univariate analysis showed MR imaging–determined primary tumor volume (P = .03), involvement of pre-epiglottic space (P = .008), abnormal signal intensity in thyroid cartilage (P = .04), and extralaryngeal extension beyond thyroid and/or cricoid cartilage (P = .02) to be significant predictors of local control rate. Results of multivariate analysis with the Cox regression model confirmed statistical significance for invasion of pre-epiglottic space (P = .004) and for abnormal signal intensities in thyroid cartilage adjacent to the anterior commissure (P = .04) and in cricoid cartilage (P = .01). Five-year local control rates were calculated from the regression coefficients of three independent MR imaging prognostic factors, and three prognostic groups were identified on the basis of these control rates. The 5-year local control rate in the high-risk group was 35%, significantly lower than the rates in the intermediate- and low-risk groups (60% and 89%, respectively; P = .002).
CONCLUSION: MR imaging–determined pre-epiglottic space involvement and abnormal signal intensities in the thyroid cartilage adjacent to the anterior commissure and/or the cricoid cartilage are strong predictors of local outcome in supraglottic carcinoma treated with definitive radiation therapy.
© RSNA, 2004
Index terms: Cartilage, MR, 27.12141 • Larynx, CT, 27.1211 • Larynx, MR, 27.12141 • Larynx, neoplasms, 27.373 • Larynx, therapeutic radiology
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
Any cancer staging system, including the widely used TNM classification system advocated by the American Joint Committee on Cancer and the Union Internationale Contre le Cancer, has shortcomings (3,4). Computed tomography (CT) and magnetic resonance (MR) imaging are useful for supplementing clinical staging methods. The depiction of tumor extension in the laryngeal cartilages and intra- and extralaryngeal tissues and of neck lymph node metastases at CT and/or MR imaging provides valuable additional staging information. Although both CT and MR imaging enable the quantification of tumor extent and calculation of tumor volume, these characteristics are not routinely assessed. Because clinical staging information may not correlate directly with the results achieved with radiation therapy, it is important to develop an improved method for determining patient prognosis.
Pretreatment CT findings in the primary tumor, including tumor volume, have been described as potential effective predictors of local control in a variety of laryngeal tumors treated with radiation therapy alone (5–13). It has been suggested that tumor volume is an important factor that helps determine the outcome of primary radiation therapy in laryngeal carcinoma and that it may be a better predictor of local failure than the standard TNM classification (6–9,11). Mancuso et al (11) stated that the pretreatment CT measurement of tumor volume permits stratification of patients with supraglottic cancer treated with radiation therapy alone into groups in which local control is more likely and less likely. The visualization of tumor extension (ie, involvement of submucosal spaces and soft tissues) in supraglottic carcinoma, however, is difficult with CT (14). MR imaging enables a soft-tissue contrast resolution much higher than that achievable with CT, and, consequently, MR imaging may provide much better delineation of tumor tissue. Furthermore, MR imaging has a higher sensitivity than does CT for depiction of pathologic changes in cartilage. Reports about the prognostic value of MR imaging characteristics, however, are fewer than those about the value of CT findings (15,16). We have reported the results of preliminary evaluations of the prognostic importance of various MR imaging characteristics for the success of radiation therapy in laryngeal cancer patients, but these results were not stratified according to subsites in the larynx (17–20). In these studies, an increased risk of posttherapeutic tumor recurrence was found for patients in whom the tumor had invaded the laryngeal cartilages. A pretreatment MR finding of large tumor volume, particularly in the presence of abnormal signal intensity in cartilage, appeared to indicate an adverse prognosis with regard to tumor recurrence. There have been indications that the MR imaging depiction of abnormal signal intensity in cartilage in patients with a large tumor mass (>5 cm3) may be associated with a substantially worse prognosis (18,20). To the best of our knowledge, however, no study previously has been performed to correlate MR imaging findings with local outcome in supraglottic cancer.
The purpose of our study was to retrospectively assess the prognostic meaning of pretreatment MR imaging–related tumor characteristics for local outcome in supraglottic squamous cell carcinoma treated with definitive radiation therapy.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
From a group of patients (n = 332) with supraglottic squamous cell carcinoma for whom data were accrued between December 1984 and January 2001, we selected 84 patients who met the study selection criteria. Patient age ranged from 41 to 88 years (median, 64 years). There were 59 male patients (70%; mean age, 61 years; age range, 38–78 years) and 25 female patients (30%; mean age, 65 years; age range, 40–87 years). The pretreatment clinical characteristics of disease in the 84 patients are listed in Table 1. The stage of disease in all patients was appraised clinically and radiologically according to the recommendations of the Union Internationale Contre le Cancer (21). Hemoglobin levels were obtained before initiation of radiation therapy in 83 of 84 patients. Sixty-four (77%) of the 83 patients had a normal pretreatment hemoglobin concentration (female patients, 7.5–10.0 mmol/L; male patients, 8.7–11.0 mmol/L), and 19 (23%) had a hemoglobin concentration lower than the normal range. None of these patients received adjuvant chemotherapy.
|
Radiation Therapy and Follow-up
The mean interval between the MR imaging examination and the start of radiation therapy was 38 days. Patients with T1 and T2 supraglottic lesions (40 patients) were irradiated to a total dose of 58–72 Gy (mean for patients with T1 lesions, 66 Gy; mean for patients with T2 lesions, 68 Gy) in increments of 2.0–2.5 Gy. Patients with T3 and T4 supraglottic lesions (44 patients) received a total radiation dose of 60–76 Gy (mean, 68 Gy) in increments of 2.0–2.5 Gy.
Patients were followed up at regular intervals by the otolaryngologist or head and neck surgeon and by the radiation oncologist every 2 months in the first 2 years after radiation therapy, every 3 months in the 3rd year, and every 4–6 months thereafter. The minimum follow-up period was 2 years. The mean follow-up period (with follow-up ending either at local treatment failure or at last patient contact) was 3.7 years (range, 0.4–15.0 years).
Evaluation of MR Imaging Characteristics
All pretreatment MR images were reviewed and evaluated by a single radiologist with 11 years of experience in head and neck imaging (J.A.C.). MR images were assessed for ten characteristics, including primary tumor volume, presence of glottic and subglottic extension, involvement of pre-epiglottic space, abnormal signal intensity in or destruction of cartilages adjacent to tumor tissue (ie, abnormal signal intensity in cartilage at the anterior commissure, in thyroid cartilage, and/or in cricoid cartilage; and extralaryngeal extension beyond cartilage at the anterior commissure, thyroid cartilage, and/or cricoid cartilage), and hypopharyngeal extension. Tumor extension was depicted on T1-weighted images as an area with intermediate signal intensity that contrasted markedly with the high signal intensity of fat and that was somewhat lower in intensity than the signal of muscle (22). Abnormal signal intensity in cartilage was determined with the combined use of T1- and T2-weighted images at the same levels. On T1-weighted images, the tumor appeared as an area with intermediate signal intensity, in marked contrast with the high-signal-intensity bone marrow of ossified cartilage. On T2-weighted images, tumor extension into cartilage was depicted as an area with increased signal intensity that contrasted markedly with the signal intensity of nonossified cartilage (15,23). Abnormal signal intensity in thyroid cartilage at the anterior commissure was measured as a separate characteristic because of the frequency of thyroid cartilage involvement in this area (18). Extralaryngeal extension through cartilage adjacent to the anterior commissure, which was considered to indicate tumor invasion beyond the cartilaginous framework and into contiguous soft tissues, also was measured separately. Involvement of the hypopharynx was defined as tumor invasion in the lateral wall of the piriform sinus and beyond the cricoid cartilage.
MR images obtained before 1994 were digitized with a film scanner and made available for review on a large-screen monitor by using custom-written dedicated computer software. From 1994 to 2001, digital MR images were obtained directly. For the evaluation of tumor volume, the primary lesion was outlined manually on T1-weighted MR images (22). The volume of the tumor was calculated in cubic centimeters by multiplying the value of the tumor area in each section in which the tumor was present by the sum of the section thickness and intersection gap and then summing the resultant values.
Statistical Analyses
Statistical analyses were performed by using software (SPSS, version 11.0; SPSS, Chicago, Ill). Local tumor control was estimated from the 1st day of radiation therapy. In the univariate analysis, curves for local control rate were estimated with the Kaplan-Meier method (actuarial life-table analysis) and compared by using the log-rank test. A multivariate analysis was performed with the Cox proportional hazards model to identify radiologic covariates, with correction for potential clinical confounders that were significantly associated with local tumor control (backward elimination). The following variables were entered into the statistical model: patient sex (male vs female), age (0–64 vs >64 years), vocal cord mobility (normal vs impaired vs fixed), hemoglobin concentration (normal vs low), histopathologic grade (well vs moderately vs poorly differentiated), primary tumor volume (0–3 vs >3–10 vs >10 cm3), glottic extension (absent vs present), subglottic extension (absent vs present), extension in pre-epiglottic space (absent vs present), abnormal signal intensity in cartilage at the anterior commissure (absent vs present), abnormal signal intensity in thyroid cartilage (absent vs present), abnormal signal intensity in cricoid cartilage (absent vs present), extralaryngeal extension at the anterior commissure (absent vs present), extralaryngeal extension beyond thyroid and/or cricoid cartilage (absent vs present), and hypopharyngeal extension (absent vs present). In the statistical analyses, extralaryngeal extension beyond thyroid and cricoid cartilages was characterized as either absent or present in at least one of these structures. P of less than .05 was considered to indicate a statistically significant difference.
Prognostic Model
The MR imaging characteristics that were identified as independent prognostic factors for local control in the multivariate analysis were pooled in a prognostic model to assess a subset of patients with a possibly very poor outcome with regard to local control. We calculated the risk score for each risk factor present for each patient by multiplying the regression coefficient, obtained from the assessment with the multivariate model, by 5 and then rounding to the nearest integer. On the basis of the total risk score, which was calculated by summing the risk scores of the various risk factors, three prognostic groups were defined. In addition, the local tumor control rate for these three groups was estimated by using the Cox proportional hazards model with correction for confounding factors.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
|
|
|
|
|
|
|
|
|
|
|
To identify a subset of patients at high risk for local recurrence, we developed a prognostic model by using the regression coefficient of the three MR imaging–determined characteristics identified as independent prognostic factors for local control (Table 6). The estimated curves for local tumor control in the presence of individual characteristics are shown in Figures 5–7. The total risk score calculated by summing the risk scores of these three factors for each patient ranged from 0 to 18.5 points (median, 7.5 points). Then the patients were grouped according to risk score into three categories: low risk (risk score, 0–5 points; 28 patients), intermediate risk (risk score, 6–11 points; 40 patients), and high risk (risk score, 
12 points; 16 patients) (Fig 8). Four (14%) of the patients in the low-risk group, 15 (37%) in the intermediate-risk group, and nine (56%) in the high-risk group developed a local recurrence. The local control rate in the high-risk group was 35% after 5 years, significantly lower than the 5-year rates for the intermediate-risk and low-risk groups (60% and 89%, respectively; P = .002) (Fig 9).
|
|
|
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
Tumor Volume and Deep Tissue Infiltration
In our study, the P value for correlation between MR imaging–determined tumor volume and local control did not reach statistical significance. Similar results were found by others using CT as the imaging method (6,10). Those authors found no statistically significant difference for CT-determined tumor volume as an independent predictor of the local control rate in multivariate analysis. The studies by Gilbert et al (9), Mancuso et al (11), and Mendenhall et al (24) showed that tumor volume as measured on CT images could be used to stratify patients into groups of those with tumors very likely or much less likely to be controlled at the primary site with definitive radiation therapy. These previous findings do not corroborate our results, because we used more sensitive imaging characteristics, such as an abnormal signal intensity in cartilage, which has been shown to be a predictor of poor local outcome in laryngeal cancers (17–19). Although MR imaging has higher soft-tissue resolution than CT, CT may enable more accurate assessment of tumor volume than does MR imaging with an intersection gap. However, in mostly large supraglottic cancers, the intersection gap may not play a major role in tumor measurements. Furthermore, the correlation between primary tumor volume and probability of local control is not absolute; it may be influenced by pathophysiologic factors such as blood flow, nutrient supply, and cellular metabolic microenvironment, in which significant variations may occur among locations within the same tumor and between tumors (25).
Giron et al (26) have concluded that MR imaging is the method of choice today for staging of laryngeal malignancies. MR imaging enables very accurate visualization of deep tissue such as intrinsic laryngeal musculature, which is important for the recognition of subtle tumor extension in supraglottic laryngeal carcinoma. Univariate and multivariate modeling–based data from our patient population with supraglottic cancer revealed a strong relationship between the degree of invasion of the pre-epiglottic space and local tumor control probability. Fletcher and Hamberger (27) reported the results of a study in 173 patients treated with primary radiation therapy for supraglottic carcinoma and stated that the pre-epiglottic space is poorly vascularized. They suggested that the anoxic compartment of tumors penetrating this space must be substantial, which would make the tumors comparatively resistant to irradiation. This might explain the relatively greater resistance to radiation therapy and the higher cure rates with surgery in such tumors, observed by Dursun et al (28). The pre-epiglottic space is composed predominantly of fat, which has low attenuation on CT images and high signal intensity on MR images. Both techniques are reported to be equal in the evaluation of pre-epiglottic involvement (29). As reviewed by Hermans et al (6), pretreatment CT findings of involvement of the pre-epiglottic space showed significant correlation with local outcome. In our study, invasion of the pre-epiglottic space was the strongest independent predictor of local control of supraglottic cancer.
Cartilage Involvement on MR Images
Katsounakis et al (30) wrote that MR imaging is superior to CT for staging of tumors, especially those confined to the supraglottis, because of enhanced detection of cartilaginous involvement, and may therefore lead to over- or underestimation in TNM classification. MR has been reported to be highly sensitive in depicting cartilage invasion and to be more accurate than CT (15,16). In a preliminary study, invasion of cartilage depicted on CT images was not found to be a significant predictor of local recurrence (6). In our study, abnormal MR imaging signal intensity patterns in thyroid and cricoid cartilages were found to be significant factors influencing the local outcome of supraglottic lesions.
In our patients, the cartilage most often involved was thyroid cartilage (but not that adjacent to the anterior commissure), and involvement of this cartilage was associated with poorer local control (univariate analysis for predictive value, P = .04). As reported by Castelijns et al (20), an abnormal MR signal intensity pattern in cartilage, combined with large tumor volume, was associated with a significantly worse prognosis (Fisher exact test, P < .05). Multivariate analysis demonstrated that abnormal signal intensity in cricoid cartilage is a strong independent prognostic factor associated with tumor recurrence. However, an abnormal MR signal intensity pattern in laryngeal cartilage does not automatically imply a need for laryngectomy, especially in tumors with smaller volume. It is incorrect to postulate that radiation therapy cannot cure a substantial number of lesions in which cartilage is involved; minimal cartilage involvement in patients with low-stage tumors does not indicate a bad prognosis (17,18).
MR images of 15 patients showed abnormal signal intensity in thyroid cartilage adjacent to the anterior commissure, and this abnormality turned out to be an independent prognostic factor in local outcome of supraglottic carcinoma. Maheshwar and Gaffney (31) showed that anterior commissure involvement was a predictor of poor response to radiation therapy. None of the cases in their study included CT images obtained in patients with glottic carcinoma, however, and we believe that treatment failure may have resulted from underestimation of the tumor stage.
No conclusions can be drawn concerning extralaryngeal extension of the tumor beyond these cartilages, because of the limited number of patients with extralaryngeal abnormalities. The prognosis for patients with extralaryngeal extension depicted at diagnostic imaging may be expected to be even worse. Nevertheless, this characteristic did not adversely affect the likelihood of local control in our patients, probably because such abnormalities are treated primarily with laryngectomy at our institution.
There are, however, some limitations to our study. First, MR examinations were performed by using different techniques (0.6-T vs 1.0-T MR imaging systems). In our opinion, this did not influence contrast between tumor tissue and surrounding tissue and therefore did not interfere with adequate diagnosis of the extent of tumor tissue. Second, the review of MR images by a single experienced observer did not enable the collection of information regarding interobserver variance, which should be investigated in the future.
The TNM classification system was developed, at least in part, for prognostic purposes. Regarding tumor extension in cartilages, the most recent revision in TNM classification guidelines (21) states only that tumors with "thyroid cartilage erosion" and extension "through cartilage" would be classified as T3 and T4, respectively. On the basis of our findings, these phrases may be refined in terms more meaningful for MR imaging, to "abnormal signal in (all) cartilages" and "extralaryngeal extension beyond cartilages," respectively.
In conclusion, patients with MR imaging–determined pre-epiglottic space invasion and abnormal signal intensity in thyroid cartilage at the anterior commissure and/or in cricoid cartilage are at high risk for local failure after radiation therapy. Tumors in such patients are probably better treated initially with surgery, with or without postoperative radiation therapy. We believe that MR imaging findings are therefore essential in the work-up for T staging of supraglottic cancer.
|
FOOTNOTES |
|---|
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  S. Bisdas, Z. Rumboldt, J. Wagenblast, M. Baghi, T.S. Koh, M. Hambek, T.J. Vogl, and M.G. Mack Response and Progression-Free Survival in Oropharynx Squamous Cell Carcinoma Assessed by Pretreatment Perfusion CT: Comparison with Tumor Volume Measurements AJNR Am. J. Neuroradiol., April 1, 2009; 30(4): 793 - 799. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Becker, P. Zbaren, J. W. Casselman, R. Kohler, P. Dulguerov, and C. D. Becker Neoplastic Invasion of Laryngeal Cartilage: Reassessment of Criteria for Diagnosis at MR Imaging Radiology, November 1, 2008; 249(2): 551 - 559. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. Ljumanovic, P.J.W. Pouwels, J.A. Langendijk, D.L. Knol, P. van der Valk, C.R. Leemans, and J.A. Castelijns Has the Degree of Contrast Enhancement with MR Imaging in Laryngeal Carcinoma Added Value to Anatomic Parameters Regarding Prediction of Response to Radiation Therapy? AJNR Am. J. Neuroradiol., September 1, 2007; 28(8): 1540 - 1546. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D. G. Pfister, S. A. Laurie, G. S. Weinstein, W. M. Mendenhall, D. J. Adelstein, K. K. Ang, G. L. Clayman, S. G. Fisher, A. A. Forastiere, L. B. Harrison, et al. American Society of Clinical Oncology Clinical Practice Guideline for the Use of Larynx-Preservation Strategies in the Treatment of Laryngeal Cancer J. Clin. Oncol., August 1, 2006; 24(22): 3693 - 3704. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
