Abstract
The accurate detection of lung carcinoma and the determination of its stage remain significant clinical problems. 18F-FDG PET has been shown to improve detection and staging of lung cancer and to prevent unnecessary invasive procedures. Positron imaging with dual-head gamma cameras may not be as sensitive as PET, but recent studies have shown good results with these cameras. Methods: In the present study, we investigated 100 patients, 76 of whom were male and 24 female (mean age ± SD, 60.7 ± 9.4 y), with suspected non-small cell lung cancer. 18F-FDG scanning was performed using a dual-head coincidence camera 1 h after the intravenous injection of 185 MBq of 18F-FDG. For 46 patients, attenuation correction was also performed. Two independent observers unaware of clinical status analyzed all imaging studies. TNM classification was assigned after surgical staging. Results: In 44 patients with clinically suspected bronchogenic carcinoma, no evidence of malignancy was found. However, in 56 patients a pulmonary neoplasm was demonstrated. At interobserver analysis, a κ value of 0.94 (P < 0.0001) was found for detection of the primary tumor and a κ value of 0.63 (P < 0.0001) was found for mediastinal staging. A sensitivity of 96%, a specificity of 93%, and an accuracy of 95% were found for detection of pulmonary neoplasm. Assessment of lymph node involvement showed a sensitivity of 50%, a specificity of 92%, and an accuracy of 77%. The sensitivity of CT in assessing lymph node involvement was 36%, the specificity was 86%, and the accuracy was 67%. Attenuation correction provided more anatomic information, but no differences were seen between attenuation-corrected and non-attenuation-corrected images for detecting lesions or lymph node involvement. Conclusion: The present study confirms earlier data showing that 18F-FDG scans obtained with dual-head coincidence cameras are useful in the detection of non-small cell lung cancer and less suitable for staging of lymph node involvement, with accuracy comparable to that of CT.
- non-small cell lung cancer
- 18F-FDG
- dual-head coincidence camera
- interobserver variability
- mediastinal staging
Lung cancer is among the most frequent and most lethal of cancers in both men and women. Lung cancer makes up 22% of all cancers in men and 8% of all cancers in women. Five-year survival for non-small cell lung cancer (NSCLC) depends on the stage. For stages I, II, III, and IV, the 5-y survival is 42%, 22%, 5%, and 3%, respectively (1). Many lung malignancies present as single pulmonary nodules (SPNs) and are potentially cured by resection. Unfortunately, up to one third of patients with SPNs who undergo surgery are found to have benign lesions at thoracotomy (2). In both detection and staging of lung cancer, conventional techniques lack sensitivity and specificity for adequate clinical work-up (3). 18F-FDG PET allows the visualization of metabolic activity in tissue. Several studies have already shown that 18F-FDG PET is the most accurate noninvasive imaging test for the diagnosis of pulmonary nodules. Moreover, recent studies showed that PET better assesses the stage of NSCLC and provides more accurate prognostic stratification than does conventional staging (4–8). Although PET has the advantage of high spatial resolution and allows quantification, its limited availability may prevent routine clinical application. In contrast, coincidence imaging with dual-head gamma cameras may allow routine applications, because of more widespread availability. Previous studies performed using 18F-FDG PET and dual-head gamma cameras showed excellent results for staging and detection in head and neck cancer and encouraging results for mediastinal staging in lung cancer (9–11). In the present study, we investigated the sensitivity, specificity, accuracy, positive predictive value, and negative predictive value of 18F-FDG PET in detection and nodal staging of NSCLC using a dual-head coincidence camera. The 18F-FDG PET images were analyzed by 2 independent observers.
MATERIALS AND METHODS
Patients
A total of 100 consecutive nondiabetic patients (76 male and 24 female; mean age ± SD, 60.7 ± 9.4 y) suspected of having NSCLC were studied retrospectively. Suspicion of malignancy was based on clinical course and radiologic examination (planar chest radiography, CT, or MRI). A clinical diagnosis of lung cancer was histologically confirmed at surgical staging or CT-guided biopsy. If no histologic diagnosis was available, the pulmonary nodule was considered benign if during a year of follow-up no signs of progression or metastases were seen on radiologic examination.
Imaging Protocol
For all 100 patients, 18F-FDG PET was performed after an overnight fast. Blood glucose levels were measured for all patients using a routine clinical test. A blood glucose level higher than 10 mmol/L was used as an exclusion criterion. 18F-FDG PET was performed, using a SPECT scanner with PET capability (Vertex-MCD; ADAC Laboratories, Milpitas, CA), 1 h after the intravenous administration of 185 MBq of 18F-FDG. The spatial resolution of this scanner is 5 mm. The acquisition (peak to peak and Compton to peak; full-field detector mask) involved a 180° rotation of each detector with 32 stops at 45 s per stop and a 128 × 128 matrix. A transmission scan was obtained with an external 137Cs-point source, to measure photon attenuation and to correct subsequent emission scans. PET images were generated using iterative reconstruction by an ordered-subsets expectation maximization algorithm with 2 iterations and 8 order subsets, followed by Wiener filtering.
All PET images were evaluated qualitatively by 2 independent observers experienced in describing PET scans and unaware of clinical status and radiologic data. In cases of disagreement, a consensus was obtained. Uptake in the lung lesions and mediastinum was scored with reference to liver uptake: 0 = no uptake, 1 = uptake equal to that in the liver, and 2 = uptake more than that in the liver. Uptake of 1 or more was considered to indicate malignancy. Because of the limited resolution with 18F-FDG, the same lymph node categories as described in the study of Pieterman et al. were used to compare the results of CT, 18F-FDG imaging, and histopathologic examination (Fig. 1) (6).
Surgical Lymph Node Staging
Cervical mediastinoscopy was performed within 3 wk of imaging studies, with the patient under general anesthesia. The left and right upper paratracheal nodes (nodes 2L and 2R, according to the classification of Naruke, the paratracheal node (node 3), the left and right lower paratracheal nodes (nodes 4L and 4R), and the subcarinal node (node 7) were routinely sampled. At the left parasternal mediastinoscopy, the subaortic node (node 5) and the para-aortic node (node 6) were sampled. In patients whose N2 nodes were not accessible by mediastinoscopy, these nodes were sampled using video-assisted thoracic surgery. Thoracotomy, if indicated, was usually performed within 2 wk of mediastinoscopy. The precise anatomic location of the removed lymph nodes was recorded using the lymph node map and the definitions of the American Thoracic Society (12,13).
CT
Chest CT was performed on a spiral CT scanner (Philips Medical Systems, Best, The Netherlands) using a 5-mm slice thickness and a recovery index of 2 mm. A power injector was used to achieve a 2 mL/s bolus injection of contrast medium during 45 s. The scan delay was 25 s. Each CT scan was evaluated by an experienced radiologist. Mediastinal lymph nodes were considered to be involved if they exceeded 10 mm in short-axis diameter.
Statistical Analysis
Sensitivity, specificity, accuracy, positive predictive value, and negative predictive value were calculated for 18F-FDG PET and CT, using the mediastinoscopy or thoracotomy as the gold standard.
Agreement statistics (κ) were analyzed to compare the 2 observers. A value of 1 indicates perfect agreement. A value of 0 indicates chance agreement. The McNemar test was used to calculate any significant difference in the sensitivity and specificity of CT and 18F-FDG for staging the mediastinum.
RESULTS
18F-FDG PET was performed without attenuation correction for the first 54 patients and with attenuation correction for the following 46. The clinical data of all patients are summarized in Table 1. Fifty-six patients had NSCLC (26 patients with squamous cell carcinoma, 18 with adenocarcinoma, 9 with large cell neuroendocrine carcinoma, and 3 with large cell undifferentiated carcinoma), and for all patients the diagnosis was confirmed histologically. The mean tumor size was 3.3 cm (±1.8 cm), with a range of 0.5–9 cm; size was determined at pathologic examination. Two tumors of 0.5–1 cm were seen, 7 tumors of 1–2 cm, 9 of 2–3 cm, and 21 larger than 3 cm. In 17 patients, the tumor size could not be established because they could not undergo surgery.
In 35 patients with a benign lesion, the tumor size could be established on radiologic examination; mean size was 2.3 cm (±1.4), with 2 lesions of 0.5–1 cm, 11 of 1–2 cm, 15 of 2–3 cm, and 7 larger than 3 cm. In the remaining 9 patients, no tumor size could be established reliably.
The sensitivity and specificity of 18F-FDG PET for the detection of lung cancer were 96% and 93%, respectively (Table 2). False-negative lesions were seen in 2 patients: one with a 1.1-cm lesion with little uptake and another with a 1.7-cm lesion with diffuse uptake in the lung due to diffuse metastases. False-positive lesions were seen in 3 patients: in one because of infection, in another because of tuberculosis, and in a third because of sarcoidosis.
In the staging of mediastinal lymph node involvement, 18F-FDG PET showed a sensitivity of 50% and a specificity of 92% (Table 2). CT had a sensitivity of 36% and specificity of 86% for the detection of mediastinal lymph node involvement (Table 2). CT and 18F-FDG PET did not significantly differ in sensitivity and specificity for staging the mediastinum (P = 0.2 and P = 0.4, McNemar test).
The scoring of the 2 observers was nearly the same for detection of the primary tumor (κ = 0.94 [P < 0.0001], indicating almost perfect agreement). In the group of patients with lung cancer, the scoring of the 2 observers agreed well for detection of mediastinal lymph node involvement (κ = 0.63 [P < 0.0001]).
The 46 images of patients for whom attenuation correction was performed were not scored differently from the non-attenuation-corrected images. Attenuation correction provided more information only about anatomic borders, which are less well defined in images without attenuation correction. In judging mediastinal lymph node involvement, both observers preferred the images without attenuation correction (Fig. 2).
DISCUSSION
The results we describe here for the diagnosis and nodal staging of NSCLC confirm earlier data showing that 18F-FDG PET with coincidence imaging is accurate for differentiation between benign and malignant pulmonary nodules (14–17). A sensitivity of 97% and a specificity of 93% were found. There were 3 false-positive studies, which were due to inflammatory diseases. Because the study was retrospective, the patient selection may have been biased. Patients with known extrapulmonary tumors were not excluded; probably because of the sample size, we did not see lung metastases in our group. Patients with small cell lung cancer are often diagnosed in an earlier stage of the diagnostic tract and therefore do not undergo further staging and are treated with chemotherapy. In the whole population, we saw 2 patients with small cell lung cancer (not mentioned in our article, both true-positive).
Recent studies have shown that the introduction of 18F-FDG imaging with a dedicated PET camera is cost effective in the management of solitary pulmonary nodules (18,19). Therefore, considering the lower costs of a dual-head PET camera and the results of our study, 18F-FDG PET with a dual-head gamma camera may be a useful imaging technique in the analysis of SPNs when dedicated PET is absent. Several studies have shown that a full-ring 18F-FDG PET system is the most accurate noninvasive method for preoperative mediastinal staging of NSCLC, with a sensitivity of 91% and specificity of 86% (6). Moreover, different studies have shown that inclusion of 18F-FDG PET in the work-up to surgery is cost effective (20,21). Because of the limited availability of full-ring PET scanners, dual-head coincidence SPECT cameras have been used for 18F-FDG imaging in lung cancer and for staging.
To be as clinically useful as dedicated PET systems, dual-head gamma cameras with 18F-FDG PET need to provide accurate information about the mediastinal stage. A few studies have described the sensitivity and specificity of the use of a dual-head coincidence gamma camera in staging of the mediastinum. A study by Tatsumi et al. reported a sensitivity of 78% and a specificity of 93% in determining the stage of mediastinal lymph nodes (15). In another study, by Shreve et al., a sensitivity of 58% was reported for detection of metastatic disease in mediastinal lymph nodes (22). In an earlier study from our department, by Stokkel et al., a sensitivity of 90% and a specificity of 97% were found for the assessment of lymph node involvement in patients with NSCLC (23). In the present study, we could not reproduce our earlier results. However, the sample size was small, and the confidence intervals of our earlier data were consequently wide and overlapped those of the present study. Considering the lack of diagnostic accuracy for dual-head gamma cameras with PET capability in the staging of mediastinal lymph nodes, dedicated PET cameras should be preferred.
The spatial resolution of dual-head gamma cameras with PET capability is about 6–8 mm transversely and 6–10 mm axially for a line source in water (9). However, these dual-head systems are less sensitive for detection of annihilation radiation, and the counting rate is limited because of the thin sodium iodine detectors (9). These limitations may result in poor image quality and low contrast resolution. Moreover, gamma camera-based PET systems use only 2 detectors; therefore, the camera has to rotate for 30–40 min around the patient, increasing the possibility of movement artifacts. All these factors may cause the lack of sensitivity for small lesions in the mediastinum.
Measured attenuation correction using a transmission scan improves image quality and, with the absence of the artifacts present on non-attenuation-corrected images, better delineates anatomic structures. In dedicated PET cameras, attenuation correction did not improve the diagnostic accuracy of 18F-FDG (24,25). In contrast, 2 recent studies found better image quality and minimization of the limits of dual-head PET for the detection of small lesions when attenuation correction was used for oncologic patients (26,27). In our study, attenuation correction indeed gave more information about anatomic location, but the analysis of the nonattenuated data and the attenuated images did not differ. However, in our study we did not separately describe the images with and without attenuation; therefore, our results concerning the effect of attenuation correction should be interpreted cautiously.
CONCLUSION
The present study showed excellent sensitivity and specificity for 18F-FDG PET with a dual-head gamma camera in the analysis of SPNs, thus indicating that these systems do have a role in the diagnostic procedure for SPNs. For routine clinical use in the work-up of patients with NSCLC, 18F-FDG PET with a dual-head gamma camera proved less suitable, having an accuracy comparable to that of CT.
Acknowledgments
The authors gratefully acknowledge the fruitful comments by Sally Collyer.
Footnotes
Received May 2, 2002; revision accepted Sep. 20, 2002.
For correspondence or reprints contact: Henk Stevens, MD, PhD, Department of Nuclear Medicine, HP E02.222, University Medical Center Utrecht, Heidelberglaan 100, 3508 GA Utrecht, The Netherlands.
E-mail: H.Stevens{at}azu.nl