Coronal nonattenuation-corrected images of a patient with squamous cell carcinoma of the tongue. The primary tumor is clearly identified by highly elevated fluorodeoxyglucose F 18 uptake. There are no metastatic lymph nodes present, as proved later by histological examination.
Nonattenuation-corrected positron emission tomographic (PET) images of a patient with metastatic lymph nodes on the left side (A, tranversal; B, sagittal; and C, coronal). The primary tumor was not found by morphologic procedures or by PET. However, PET demonstrated 1 previously unknown metastatic lymph node contralaterally (C, D, and E) that was not reported as suggestive of metastasis on a magnetic resonance imaging scan.
Kau RJ, Alexiou C, Laubenbacher C, Werner M, Schwaiger M, Arnold W. Lymph Node Detection of Head and Neck Squamous Cell Carcinomas by Positron Emission Tomography With Fluorodeoxyglucose F 18 in a Routine Clinical Setting. Arch Otolaryngol Head Neck Surg. 1999;125(12):1322-1328. doi:10.1001/archotol.125.12.1322
Accurate determination of lymph node involvement is a prerequisite for individualized therapy in patients with squamous cell carcinoma of the head and neck region. In a previous study, we showed that positron emission tomography (PET) with fluorodeoxyglucose F 18 with and without attenuation correction is superior to magnetic resonance imaging for this purpose in a scientific setting.
To evaluate the diagnostic accuracy of a shortened PET protocol (acquisition time, 20 minutes) in a routine clinical setting.
The results of static, nonattenuation-corrected PET performed on patients in 2 bed positions starting 40 minutes after the intravenous injection of 370 MBq of fluorodeoxyglucose F 18 and the results of morphologic procedures (computed tomography and magnetic resonance imaging) were compared prospectively in 70 patients for lymph node staging. Postoperative pathologic findings served as a criterion standard.
An academic medical center.
The diagnostic accuracy of PET for detecting "neck sides" with malignant involvement was superior to morphologic procedures, with a sensitivity and specificity of 87% and 94%, respectively, compared with computed tomographic values of 65% and 47% and magnetic resonance imaging values of 88% and 41%, respectively.
A short PET protocol that is suitable for routine clinical use is superior to morphologic procedures (computed tomography and magnetic resonance imaging) for the detection of lymph node involvement in head and neck squamous cell carcinomas.
ACCURATE pretherapeutic lymph node staging is mandatory for planning surgical strategy (type of neck dissection and 1 or both sides) in patients with resectable head and neck squamous cell carcinomas (SCCs). In patients with nonresectable primary tumors, an individualized radiation field also requires precise definition of metastatic lymph node involvement. Until now, the definition of malignant lymph node involvement was based on morphologic changes as determined by computed tomography (CT) or magnetic resonance imaging (MRI). These procedures mainly lack sensitivity, because of malignant involvement in nonenlarged lymph nodes (false negatives), and specificity, because of reactive, nonmalignant enlargement of lymph nodes (false positives). Metabolic factors such as glycolysis are independent of morphologic changes, and the increase of these factors is measureable in malignant tumors. Thus, previous studies1,2 showed that lymph node staging of head and neck SCCs by positron emission tomography (PET) with fluorodeoxyglucose F 18 is more precise than morphologic procedures such as CT and MRI. To our knowledge, all published studies have been performed in selected patient groups. In this study, we compared the diagnostic accuracy of PET, CT, and MRI for lymph node involvement in nonselected patients in a routine clinical setting.
One hundred eleven consecutive patients (95 men and 16 women; age range, 29-78 years), in whom head and neck SCCs were diagnosed by a biopsy, were included in this study. Exclusion criteria for PET were prior therapy or possible pregnancy. All patients underwent morphologic imaging (CT or MRI) and PET with fluorodeoxyglucose F 18 within 2 weeks before surgery. In addition, all patients underwent panendoscopy before PET. Following these investigations, the therapeutic regimen (surgery alone, surgery and radiotherapy, combined radiotherapy and chemotherapy, or primary radiotherapy and type of neck dissection) was decided by an interdisciplinary conference with experienced ears, nose, and throat surgeons, oncologists, and radiation therapists (W.A., R.J.K., and C.A.). The results of PET scans were known to the members of this conference. According to this conference, 70 of the 111 patients underwent surgery with or without postsurgical radiotherapy. Informed written consent was obtained from all patients included in this study.
Fluorodeoxyglucose F 18 was produced by a method modified from the synthesis used by Hamacher et al.3 All patients, except 5 individuals with known diabetes, fasted for at least 4 hours. Before PET, the plasma glucose level of the patients was measured with a standard clinical test. For PET imaging, a scanner (model ECAT EXACT or ECAT 951/R; Siemens/CTI, Munich, Germany) was used, which encompasses a 16.2- or a 10.8-cm axial field of view and yields 31 or 47 image planes per bed position, respectively, with a slice separation of 3.38 mm.4 The scanning procedure was started at 40 minutes after the intravenous injection of 300 to 370 MBq of fluorodeoxyglucose F 18 at the base of the skull. After an acquisition time of 10 minutes per bed position, the table was moved to the next position, proceeding toward the feet of the patient. Thirty-four patients studied via the ECAT EXACT scanner (with the larger axial field of view) and 27 of 36 patients studied via the ECAT 951/R scanner needed just 2 bed positions, resulting in a total scanning time of 20 minutes. For a small subset of patients (n=9), there was a suggestion of malignant lung involvement, and we used 3 bed positions via the ECAT 951/R scanner to cover the head and neck region and the whole pulmonary region. These investigations did not show any pulmonary abnormalities. The PET images were generated using filtered back projection and corrected for decay. Suspected malignant involvement was evaluated visually as focally increased fluorodeoxyglucose F 18 uptake by comparing images with anatomical maps and normal fluorodeoxyglucose F 18 distribution in the head and neck region, as described by Jabour et al.5 Interpretation was done by PET-experienced nuclear physicians (C.L. and M.S.) blinded to patients' history and findings of previous imaging results.
All patients underwent contrast-enhanced imaging by CT, MRI, or both, which was done on an outpatient basis using a wide variety of scanners and readers. The hard copies of these examinations and the written reports of the radiologists who did not have any knowledge of the background of the patients and the PET results were routinely used in our department for preoperative evaluation. These conventional images were of the same quality as the real-time views. Lymph nodes were classified as positive for malignant neoplams if their largest diameter was more than 1.5 cm or if indirect signs of malignant involvement, like such as inhomogeneity, were found.6,7
The localization of dissected tissue was documented by the surgeons (W.A., R.J.K., and C.A.) to allow correlation between histological findings and preoperative imaging results. After the specimens were embedded in paraffin, routine staining was performed. Tumor and lymph node status were coded according to the TNM (tumor-node-metastasis) classification.8 The results of preoperative examinations were compared with the results of histopathological classification, which served as the criterion standard.
Values are given as the mean ± SD. Data were expressed by sensitivity and specificity and positive and negative predictive values. Noncontinuous variables were compared with the χ2 test, depending on sample sizes. Statistical significance was defined as P ≤ .05.
Based on the results of imaging modalities and clinical examination, 41 (36.9%) of the 111 patients were diagnosed as having an inoperable disease and were referred for radiation therapy alone or combined radiation therapy and chemotherapy. In these patients, no further histopathological data were available. Of the 111 examined patients, 70 (63.1%) had surgically curable lesions and underwent surgery within 2 weeks after undergoing the PET scan. In these patients, all data concerning histopathological classification were available. In these 70 patients, there were 17 T1, 25 T2, 12 T3, and 16 T4 tumors (Table 1). Tumor volume ranged from 0.3 to 238.0 mL, representing globes with diameters of 0.8 to 8.6 cm. Of the 70 patients, 34 (48.6%) underwent ipsilateral radical neck dissection only, and 36 (51.4%) underwent ipsilateral radical neck dissection and contralateral functional neck dissection. The decision of whether contralateral functional neck dissection was necessary was made by the ears, nose, and throat surgeons (W.A., R.J.K., and C.A.) after evaluating the results of all preoperative imaging modalities.
Figure 1 shows coronal nonattenuation-corrected images of a patient with an SCC of the tongue. The primary tumor is clearly identified by highly elevated fluorodeoxyglucose F 18 uptake. There are no metastatic lymph nodes present, as proved later by histological examination.
Of 111 primary tumors, 102 (91.9%) were clearly visualized by PET. The following primary tumors could not be detected.
One tumor (diameter, 1.4 cm) was situated close to the submandibular gland, which by itself demonstrated a high fluorodeoxyglucose F 18 uptake. By reevaluating the images during a second reading, but with knowledge of the results of the other imaging modalities, this tumor could be clearly identified.
A smaller tumor (diameter, 0.8 cm) was located in the nasopharynx and masked by diffuse high fluorodeoxyglucose F 18 uptake due to an inflammation of the maxillary sinus.
In 5 patients, the histopathological findings of the primary tumor revealed a microinvasive, superficial SCC classified as a carcinoma in situ.
In 2 patients, metastatic involvement of cervical lymph nodes was proved by biopsy and PET, but the primary tumor was not found in the head and neck region by morphologic procedures (CT or MRI) or panendoscopy. A PET investigation examining the whole body of the respective patients also could not find a primary tumor. Therefore, we determined that these patients had carcinoma of unknown primary site syndrome.
Figure 2 shows nonattenuation-corrected PET images of a patient with metastatic lymph nodes on the left side (A, transversal; B, sagittal; and C, coronal). The primary tumor was not found by morphologic procedures or by PET. However, PET demonstrated 1 previously unknown metastatic lymph node contralaterally (C, D, and E) that was not reported as suggestive of malignant lymph node involvement on an MRI scan.
Preoperative lymph node status was assessed by morphologic procedures (CT or MRI) and PET with fluorodeoxyglucose F 18. There were 106 neck dissections in 70 patients (bilateral neck dissections, n=36; and unilateral neck dissections, n=34). In all 106 "neck sides," histopathological data were available, and all patients underwent PET before surgery, 58 of which were examined by CT and 63 by MRI. In 15 neck sides that were examined by CT and MRI, both methods agreed for all neck sides.
Of 1281 histologically examined lymph nodes, 98 (7.6%) were proved to show metastases. In the 70 patients with 106 neck dissections, 40 neck sides had malignant lymph node involvement, which resulted in a prevalence of 37.7%. The prevalences in the patient groups examined with CT and MRI were in the same range (Table 2).
Based on involved neck sides, we calculated the sensitivity, specificity, positive predictive value, and negative predictive value of the respective investigations (Table 3). Five false-negative neck sides were found by PET; 3 were also misclassified by morphologic imaging procedures. Positron emission tomographic scans also produced 4 false-positive results (Table 3). Two of these 4 patients had unspecific reactive changes probably due to previous biopsies of these lymph nodes shortly before the PET scan, and 2 had only a histologically proved sinus histiocytosis.
Computed tomographic scans produced 27 false results, with 13 true-positive and 18 true-negative identifications of side involvement, which resulted in a sensitivity of 65% and a specificity of 47% (PET vs CT, P<.001, χ2 test). The sensitivity (88%) of MRI scans was approximately in the same range as PET, but the specificity (40%) was significantly lower (PET vs MRI, P<.001, χ2 test) due to 22 false-positive results.
As previously mentioned, 36 patients underwent a contralateral functional neck dissection, as decided based on surgical the ears, nose, and throat surgeons (W.A., R.J.K., and C.A.) after evaluating preoperative imaging modalities and performing a clinical investigation or based on surgical experience. Positron emission tomographic scanning was done in all of these cases, CT scanning in 21, and MRI scanning in 21 (6 patients underwent a CT scan and an MRI scan). Postoperative histomorphologic examination results revealed a carcinoma in 4 contralateral neck sides. Positron emission tomographic scans showed 29 true-negative, 3 false-positive, and all of the 4 true-positive neck sides. This resulted in a sensitivity of 100% and a specificity of 90.6%. Magnetic resonance imaging scans revealed 7 true-negative, 11 false-positive, 2 true-positive, and 1 false-negative neck sides, resulting in a sensitivity of 66.7% and a specificity of 38.9%. Computed tomographic scans showed 11 true-negative, 8 false-positive, 2 false-negative, and no true-positive neck sides. This revealed a sensitivity of 0% and a specificity of 57.9%.
Postoperative histopathologic examination results of our patient group revealed N0 or N1 lymph node staging in 66 neck sides (42 patients). Positron emission tomographic scanning was done in all of these cases, CT scanning in 41, and MRI scanning in 35 (10 neck sides were examined by a CT scan and an MRI scan). Positron emission tomographic scans showed 55 true-negative, 3 false-positive, 1 false-negative, and 7 true-positive neck sides. This resulted in a sensitivity of 87.5% and a specificity of 94.8%. Magnetic resonance imaging scans revealed 15 true-negative, 17 false-positive, 1 false-negative, and 2 true-positive neck sides, resulting in a sensitivity of 66.7% and a specificity of 46.9%. Computed tomographic scans showed 17 true-negative, 17 false-positive, 1 false-negative, and 6 true-positive neck sides. This revealed a sensitivity of 85.7% and a specificity of 50.0%.
Head and neck carcinomas constitute approximately 5% of all malignant neoplasms worldwide, and their frequency is increasing.9 Squamous cell carcinomas represent most malignant tumors.10 Lymph node involvement is the most important prognostic factor affecting the survival of patients with head and neck cancer.11,12 The success of surgical treatment depends on the complete excision of all tumor tissue, and precise preoperative TNM staging is mandatory and is based on physical, morphologic, and endoscopic examination.10 Imaging procedures (CT and MRI) are used for the detection and localization of the primary tumor and regional lymph node involvement and for the determination of their relation to adjacent anatomical structures.13 Differentiation between reactive enlargement of lymph nodes and tumor-infiltrated nodes may be difficult based on radiological criteria.14 Fluorodeoxyglucose F 18 is a marker of tumor viability, based on the increased glycolysis that is associated with malignant neoplasms compared with normal tissues. Head and neck carcinomas have high glycolytic activity and increased fluorodeoxyglucose F 18 uptake.5
In this prospective study, we demonstrated in a nonselected patient group that a short PET protocol, which is suitable for routine clinical use, is superior to morphologic procedures (CT or MRI) for lymph node staging of head and neck SCCs. In our PET investigation, we revealed a higher sensitivity and specificity compared with CT and MRI (Table 3). The results of the present PET investigation were within the range of other studies concerning lymph node staging in head and neck carcinomas (Table 4) and were confirmed in previous studies.1,2 The clinical value of PET among our patients lacking obvious clinical nodal disease (eliminating N2a, N2b, N2c, and N3) showed the same percentage for sensitivity (87.5%) and specificity (94.8%) compared with the whole patient group. The sensitivity of CT increased to 85.7%, and the specificity was in the same range (50.0%), whereas the sensitivity of the MRI investigation decreased to 66.7% (specificity, 46.9%).
Compared with the values reported in the literature,17,18 the specificity and sensitiviy for CT and MRI are quite low in our study. A reason for this is that those studies were performed in a highly selected patient group in a scientific setting with a high prevalence of malignant lymph nodes. Our lower values for specificity and sensitivity for CT and MRI have been found in a nonselected patient group in a routine clinical setting. According to Bias theorem, sensitivity and specificity will decrease with decreasing prevalence.
As documented in Table 3, we found 5 false-negative neck sides by PET investigation. In 2 patients, the postoperative histopathological findings revealed a single small lymph node metastasis of less than 0.5 cm in only 1 ipsilateral regional lymph node per case (N1). One patient displayed even a micrometastasis (TNM supplement 1993)19 of less than 0.2 cm, ie, pN1(micrometastasis). However, the technical resolution of PET with fluorodeoxyglucose F 18 (full width at half maximum) is 0.5 cm. The malignant involvement itself within these 2- to 3-cm-large lymph nodes covered just a small part (<20%). These "false-negative lymph nodes" could therefore not be detected because of the small size of malignant involvement. In the remaining 2 patients, there were 2 lymph nodes on the respective neck side that histologically revealed an extensive necrosis. Because of this necrosis, a reduced glucose metabolism in these lymph nodes is assumed and, consequently, a reduced uptake of fluorodeoxyglucose F 18, which resulted in a negative PET imaging result. In all of these 5 patients, the primary tumor could be clearly visualized by PET. For the conventional imaging procedures (CT and MRI), there was a suggestion of lymph node involvement in 2 of these patients. Postoperative histomorphologic examination results of our patient group showed a malignant lymph node involvement in 4 contralateral neck sides, which were all detected by PET. This resulted in a sensitivity of 100%. Since PET allows the localization of the malignant involvement of lymph nodes, resulting in a high sensitivity and specificity (as shown in this study), it may be useful to select preoperatively between regional and comprehensive neck dissections, according to the PET findings. This should, however, be confirmed in a multicenter study.
Positron emission tomographic scans revealed 4 false-positive neck sides. Two of these 4 patients had unspecific reactive changes probably due to biopsies of these lymph nodes shortly before PET, and 2 had a histologically proved sinus histiocytosis.
According to the investigations by Myers et al,20 PET revealed an accuracy of 100% in the evaluation of the N0 neck in 8 patients with SCC of the oral cavity. In our study, we had 30 patients with an SCC of the oral cavity, and we performed 44 neck dissections in these patients. Positron emission tomographic scans overstaged (false positive) 2 neck sides, and this resulted in an accuracy of 95.5%.
Rege et al21 have shown PET to be a useful diagnostic modality for examining patients with a carcinoma of unknown primary site syndrome. They studied 60 patients with biopsy-proved cancers of the head and neck. In 4 patients with an unknown primary tumor site, PET localized the primary tumor in 2, whereas MRI did not localize a tumor for any of these patients. We were able detect the primary tumor with PET in 1 of 3 patients with a suggestion of carcinoma of unknown primary site syndrome.
A further diagnostic use for PET with fluorodeoxyglucose F 18 is the identification of local recurrence in patients with cancer. The diagnostic question of recurrent cancer or soft tissue changes after irradiation is often difficult to determine with imaging techniques such as CT and MRI. In a group of 13 patients, Paulus et al22 were able to correctly identify all recurrences using PET with fluorodeoxyglucose F 18.
This study underscores that PET with fluorodeoxyglucose F 18, using a short protocol of just 20 minutes of scanning time, is suitable for clinical routine use and produces the same sensitivity and specificity as more sophisticated PET protocols. The assessment of nodal status of head and neck cancer by PET is excellent and superior to morphologic variables derived from MRI or CT.
Accepted for publication May 20, 1999.
This study was supported by the Margarete Ammon Foundation, Munich, Germany.
We thank the staff of the positron emission tomographic suite for their technical support in performing the positron emission tomographic studies, and the cyclotron and radiochemistry staff for preparing fluorodeoxyglucose F 18.
Reprints: Reinhardt J. Kau, MD, Hals-Nasen-Ohren-Klinik und Poliklinik der Technischen Universität München, Klinikum rechts der Isar, Ismaningerstrasse 22, D-81675 München, Germany.