Role of ¹⁸F-FDG PET/CT in Management of High-Grade Salivary Gland Malignancies

Subjects

Thirty-five subjects with high-grade salivary carcinoma were retrieved from the Salivary Gland Malignancy Data Registry at Samsung Medical Center (2002–2005). All pathologic diagnoses were reconfirmed by intradepartment consultations, and the equivocal subjects were discussed with other pathologists to confirm the final pathologic diagnoses. Two subjects were excluded from the pathologic review because the surgical pathology did not meet the criteria of high-grade malignancy of the salivary gland (they were intermediate-grade mucoepidermoid carcinomas). For 14 subjects, high-grade malignancies were prediagnosed on the basis of cytologic analyses, and surgical pathology reconfirmed the high-grade cancer status after surgery. For the remaining 23 subjects, the pathologic diagnoses of high-grade malignancy were confirmed after surgery. Finally, 33 subjects with high-grade salivary carcinoma were included in the analyses, and a total of 37 PET/CT scans with contrast-enhanced CT were obtained (Table 1). Twenty-three imaging studies were performed to evaluate recurrence during the follow-up period, whereas 14 imaging studies were aimed at arriving at the initial diagnosis. Twenty-three PET/CT studies for the evaluation of recurrence were performed on 19 patients. All patients underwent surgical treatment as an initial therapy, and the mean duration between the initial treatment and the PET/CT study ranged from 1 to 180 mo (median, 16 mo). In 6 patients, malignancy was not suggested before the surgery; therefore, PET/CT was performed 1 mo after surgery. Among a total of 37 PET/CT studies, the reference standards were surgical specimens in 25 studies, including 19 patients who underwent neck dissection with or without surgical treatment of the salivary glands. The rest of the studies relied on the clinical follow-up along with the physical examination and other imaging modalities.

The pathologic diagnoses consisted of 5 salivary duct carcinomas, 6 high-grade mucoepidermoid carcinomas, 10 solid-type adenoid cystic carcinomas, 3 squamous cell carcinomas, 3 adenocarcinomas, and 6 carcinoma ex pleomorphic adenoma. The patients' ages ranged from 31 to 83 y (mean, 54.48 y); 22 patients were male and 11 were female. The majority of the primary salivary cancers arose from the parotid gland (66.7%), whereas some were from the submandibular gland (21.2%) and the minor salivary glands (12.1%), including those in the lip and soft palate and 2 sublingual glands. The initial T classifications of the subjects were T1–T2 for 19 patients and T3–T4 for 14 patients. At the initial diagnosis, N0–N1 was the most frequent nodal status of the subjects (81.8%); however, distant metastasis (lung, iliac bone, and paraaortic lymph nodes) was detected in 7 patients (21.2%). During the follow-up, another patient was detected to have distant metastasis on PET/CT.

All patients without distant metastasis were subjected to curative surgical resection with postoperative radiation therapy, whereas 7 patients with distant metastasis underwent palliative chemotherapy. As for the treatment outcome, 18 patients (54.5%) had distant metastasis (lung, liver, bone, abdominal lymph nodes), including 8 subjects that were diagnosed by PET/CT. Unfortunately, PET/CT was not performed on the remaining 10 patients when distant metastasis was detected with the help of conventional studies, so comparative analysis was not available.

The average follow-up period of the study was 24.3 mo (range, 12–47 mo).

Contrast-Enhanced CT Scans

CT scans (LightSpeed Ultra or Ultra 16; GE Healthcare) of the head and neck, focusing on the sites of the salivary gland, were performed using the following parameters: 160 mA·s, 120 keV, a section width of 3.75 mm, and a table feed of 8.75 mm per rotation. For contrast enhancement, 90 mL of an iodinated contrast agent (Iopromide, Ultravist 300; Schering) were injected intravenously at 3 mL/s using an automated injector. The scan delay time was 30 s.

Integrated ¹⁸F-FDG PET/CT Scans

All patients fasted for at least 6 h before PET/CT, which were performed using a Discovery LS PET/CT scanner (GE Healthcare). Whole-body CT was performed by a continuous spiral technique and using 8-slice helical CT with a gantry rotation speed of 0.8 s. The CT scan data were collected using the following parameters: 80 mA·s, 140 keV, a section width of 5 mm, and a table feed of 5 mm per rotation. No intravenous or oral contrast agents were used. After the CT scans and after injecting 370 MBq ¹⁸F-FDG intravenously, an emission scan was performed from the thigh to the head for 5 min per frame, for a total of 45 min. The duration of the uptake phase was 45 min. The attenuation-corrected ¹⁸F-FDG PET images were reconstructed using the CT data with an ordered-subset expectation maximization algorithm (28 subsets, 2 iterations). The images were displayed in a 128 × 128 matrix (pixel size = 4.29 × 4.29 mm, with a slice thickness of 4.25 mm). The separate CT and PET scan data were coregistered accurately using commercial software (eNTEGRA; Elgems). The standardized uptake values (SUVs) were acquired using the attenuation-corrected images, the amount of injected ¹⁸F-FDG, the body weight of each patient, and the cross-calibration factors between ¹⁸F-FDG PET and the dose calibrator.

Image Analysis

For the subsite-based analysis, we divided the structures around the salivary gland into subsites according to the American Joint Committee on Cancer Staging (AJCC) Manual criteria (18): the primary tumor size, the extraparenchymal extension and the involvement of skin, the mandible or maxilla bone, the ear canal, the skull base, and the pterygoid plate. If each diagnostic modality predicted the malignant tumors in the affected gland and the pathologic tumor size within an error of 1 cm for the diameter, then this was considered as a true-positive for tumor size. The regional cervical nodes were also subdivided into the cervical lymph node levels (levels I–VI, the periparotid and infraparotid lymph node groups).

For the CT scans, a radiologist who specializes in interpreting head and neck sections determined the extent of the salivary gland lesion and the significance of the cervical lymph nodes, without having any knowledge of the clinical findings. First, the interpretation of the CT scans was focused on the primary lesions of the salivary gland. Salivary lesions with an abnormal enhancement pattern were evaluated to assess the tumor extent, according to the subsites. The regional lymph nodes of the head and neck region were also evaluated according to the accepted criteria (19).

For PET/CT, 1 nuclear medicine physician, who was experienced in interpreting head and neck imaging for >5 y, reviewed the fused PET/CT images without having any knowledge of the clinical findings. The nuclear medicine physician first reviewed the images to determine any abnormal uptake of ¹⁸F-FDG in the salivary gland and neck nodes by using the maximal uptake values with intensity that was higher than that of the surrounding tissues. An abnormal location and asymmetry of the ¹⁸F-FDG uptake was also considered as an important diagnostic criterion for PET/CT. Their interpretation was then revised on the basis of the anatomic information provided by the combined PET/CT images. For example, a nonnodal focal uptake in a blood vessel or muscle that mimicked a lymph node was considered benign. A lymph node with low peripheral attenuation suggesting a fatty hilum on the CT image was considered benign even if it had a high ¹⁸F-FDG uptake. On the contrary, a lymph node containing a lucent portion on the CT image and low ¹⁸F-FDG uptake on the PET image that was suggestive of necrosis was considered malignant even if it had a low ¹⁸F-FDG uptake.

Similar to the interpretation of the CT scans, a nuclear physician determined the status of each subsite from the information of the “stand-alone” PET/CT scans. The interpretation of each subsite was classified as malignant-positive, malignant-negative, or equivocal. During PET/CT, there were 2 equivocal subsites in the evaluation of primary tumors and 9 equivocal sites in the evaluation of distant metastasis. There was no equivocal lesion in the evaluation of the nodes at the neck level. During CT, 4 equivocal lesions were measured when evaluating the distant metastasis. The equivocal subjects were reclassified into malignant-positive or malignant-negative after separately reviewing the images by the same radiologists or the nuclear medicine physician.

Decision Making for Treatment

The Head and Neck Cancer Tumor Board decided the treatment plan for the salivary lesions after a review of the results of the clinical examinations and the CT scans. The team then reviewed the PET/CT images, and any changes in the decision making for treatments were recorded. The final treatment plans were determined after a discussion with each patient, and the available treatment options were outlined.

Statistical Analysis

The reference standards for describing the status of each subsite and the initial tumor stage were derived from the histopathologic results from the biopsies and the surgical excisions in 25 subjects. The clinical and radiologic follow-up for neck nodes and distant sites were also used as a reference in 8 patients.

We determined the sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and diagnostic accuracy of CT and PET/CT for predicting the extent of disease using the subsite-based analysis that was described. Statistical differences between the imaging modalities were analyzed by using the McNemar test, and the 95% confidence levels were determined using Wilson's method. Two-tailed P values < 0.05 were considered to be statistically significant. The clinical decisions for treatment using CT alone and then adding PET/CT were recorded to register the impact of each diagnostic modality on patient care. The additional benefits provided by PET/CT were noted.

Diagnostic Values of Subsite-Based Analysis for High-Grade Salivary Cancer

The maximum SUV of the primary lesions ranged from 1.5 to 23.7, with a mean SUV ± SD of 8.21 ± 3.62 (Table 2). A minimum SUV of 1.5 was noted for the adenoid cystic carcinoma at the submandibluar gland, and a maximum SUV of 23.7 was detected for the mucoepidermoid carcinoma at the submandibular gland.

Within the primary site, there was a total of 67 subsites as revealed by 37 PET/CT and CT scans that were then evaluated (Table 3). Of these, the size of the salivary tumors was analyzed by PET/CT and CT to predict the pathologic tumor size, and this was achieved to within an error of 1 cm for the diameter in 21 subjects. When comparing PET/CT and CT for predicting the pathologic primary tumors, CT had a sensitivity of 64.2%, a specificity of 80.0%, a PPV of 84.3%, a NPV of 57.1%, and a diagnostic accuracy of 70.1%; the corresponding values for the PET/CT scans were 85.7%, 100.0%, 100.0%, 80.6%, and 91.0%, respectively. Thus, the sensitivity, NPV, and overall diagnostic accuracy were each significantly higher for PET/CT than for CT, whereas the differences in the other parameters were close to being statistically significant.

Figure 1 presented a subject of initial evaluation. There was only a poorly marginated soft-tissue lesion without definite enhancement in the submandibular area, as seen on the CT scan. But PET/CT showed a high SUV uptake at the same area, suggesting malignancy, and this was proven to be true on the surgical pathology. When imaging studies were performed for the evaluation of recurrence, it was more difficult to differentiate inflammation from recurrence. A patient who had surgical treatment and postoperative radiation therapy presented different results on CT and PET/CT during the follow-up (Fig. 2). PET/CT correctly diagnosed the recurrence in this patient.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1. 

A 50-y-old man with adenocarcinoma in sublingual gland. (A) There was no enhancement in mass at floor of mouth; thus, judging from CT scans, mass is classified as benign lesion. (B) Using PET/CT scans, lesion showed high glucose metabolism (maximum SUV = 10.6) with irregular margins. Interpretation was that the lesion was malignant tumor. (C) Final surgical pathology showed a 5 × 4 cm adenocarcinoma, not otherwise specified.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2. 

A 78-y-old man with salivary duct carcinoma in right parotid gland who underwent CT and PET/CT scans at 24 mo after initial treatment of a wide surgical resection with postoperative radiation therapy. (A) CT scans led to impression of right mastoid inflammation, on basis of finding of an intact bony wall (arrow). (B) In contrast, using PET/CT images, maximum SUV of lesion was 10.8, and tumor invasion into temporal bone was strongly suggested (arrow). Final surgical pathology revealed malignant cells in mastoid bone.

Eighty-six neck node analyses were done to determine the diagnostic values (Table 4). For predicting cervical lymph node metastasis, CT had a sensitivity of 71.8%, a specificity of 94.4%, a PPV of 88.4%, a NPV of 85.0%, and a diagnostic accuracy of 86.0%; the corresponding values for the PET/CT scans were 100.0%, 96.3%, 94.1%, 100.0%, and 97.6%, respectively. Again, the sensitivity, NPV, and overall diagnostic accuracy were significantly higher for PET/CT than for CT for the evaluation of cervical metastasis (Table 4). On the PET images, the maximum SUV of the neck nodes ranged from 1.4 to 27.2, with a mean SUV ± SD of 7.07 ± 5.36.

In the subject with recurrence at multiple cervical lymph nodes, PET/CT detected the metastatic lymph node at the right highest mediastinum, which made a surgical approach deep to that level necessary (Fig. 3). CT failed to provide definitive information for the same lesion.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 3. 

CT and PET/CT scans were performed on 52-y-old-man with primary squamous cell carcinoma in right parotid gland as initial work-up. (A) PET/CT scans demonstrated right lower anterior lymph node metastasis with a maximum SUV of 2.7 (arrow). (B) In contrast, CT scans failed to detect this regional metastasis, on basis of size of nodes and minimal enhancement pattern (arrow). Malignant cells were found in this node, according to final surgical pathology. Another active ¹⁸F-FDG uptake lesion was evident beside left clavicle; it was correctly diagnosed as malignancy on both CT and PET/CT.

Diagnostic Accuracy for Determination of TNM Tumor Stages of High-Grade Salivary Cancer

For predicting the TNM stages, CT had a diagnostic accuracy of 62.1%, as compared with 83.7% for the PET/CT scans; thus, PET/CT was far superior to CT (P = 0.03) (Table 5).

The T and N classifications were determined separately for each diagnostic modality; however, the M classification was determined from the clinical examination, a routine simple chest x-ray, and the CT scans. In the case of PET/CT, all T, N, and M classifications were determined solely by the results of PET/CT.

Impact on Clinical Treatment Planning

The clinical decision making with regard to treatment, as based on CT, was modified by adding the information gained from PET/CT in 16 of the 37 subjects we studied (43.2%) (Table 6). In 5 patients, the extent of surgery indicated for their primary lesions was modified. In 3 of these patients, CT was not able to detect the primary lesions as malignant tumors. Therefore, the information obtained from PET/CT converted a conservative resection to a comprehensive resection of the primary tumor. For 1 patient, PET/CT detected the recurrence in the surgical bed after the initial surgery and radiation therapy, and this allowed salvage surgery to be undertaken. The PET/CT information led to decreasing the extent of surgery in 1 patient; thus, a radical parotidectomy was avoided, including that on the periparotid muscles. PET/CT correctly predicted a final T1 tumor, whereas CT had interpreted the T classification of this tumor as T4.

For the neck nodes, the PET/CT information changed an elective neck dissection regimen in 1 patient to a comprehensive neck dissection. Also for this patient, the peritumor regional lymph node groups (infraparotid and preparotid) were included in the extent of neck dissection, based on the results of the PET/CT scans.

In 8 subjects, PET/CT detected distant metastasis (lung, liver, bone, abdominal lymph nodes); therefore, the treatment was changed from a curative intent to a palliative setting.