Significant Benefit of Multimodal Imaging: PET/CT Compared with PET Alone in Staging and Follow-up of Patients with Ewing Tumors

Patient Population

All patients included in this retrospective study (Table 1) fulfilled the following criteria: presence of histopathologically confirmed Ewing tumor; inclusion in EURO-E.W.I.N.G. 99 (EUROpean Ewing tumour Working Initiative of National Groups--Ewing Tumour Studies 1999), which was approved by the local ethics board; PET/CT examinations between January 2004 and June 2006; and follow-up of at least 6 mo.

All patients or their parents had given their written informed consent for inclusion in the EURO-E.W.I.N.G. 99 study, analysis of their clinical data for research purposes, and PET/CT examinations.

Data Acquisition

All PET/CT studies were performed with a hybrid scanner (Biograph Sensation 16; Siemens Medical Solutions), which consists of a full-ring PET tomograph (lutetium oxyorthosilicate crystals, picoelectronics) and a 16-detector-row CT component. Details concerning PET/CT in pediatric patients have been published elsewhere (18). All patients were studied after fasting for at least 5 h. Blood glucose levels at the time of ¹⁸F-FDG application were less than 120 mg/dL. Body-weight–adapted activity of ¹⁸F-FDG (4 MBq/kg of body weight) was injected intravenously 60 min before the acquisition. Patients were positioned head first and supine in the PET/CT scanner. An identical field of view was chosen from the base of the skull to the feet for the PET scan and the low-dose CT scan. First, a low-dose CT scan with acquisition parameters adjusted for pediatric patients (18) was acquired during mild expiration for attenuation correction and location of anatomic landmarks. If clinically indicated, for example, to detect pulmonary metastases at initial diagnosis or for follow-up studies after systemic therapy, patients underwent additional, diagnostic CT of the chest during full inspiration after the intravenous application of contrast medium and with weight- and size-adjusted acquisition parameters (18). Acquisition of the CT scan(s) was followed by an automatic table movement to the PET position; emission data were acquired for 4 min in each bed position.

Data Reconstruction

Low-dose CT datasets were reconstructed in an overlapping manner at a 3-mm slice thickness with a 1.5-mm reconstruction increment by use of standard soft-tissue reconstruction kernel B30 and standard lung and bone reconstruction kernels B50 and B60; the field of view (<350 mm²) was individually adapted to the body contours. ¹⁸F-FDG PET data were reconstructed iteratively by use of the ordered-subset expectation maximization algorithm with data from the low-dose CT used for attenuation correction.

If a contrast-enhanced CT scan of the chest during full inspiration was obtained, those CT data were reconstructed at a 1-mm slice thickness with an 0.8-mm reconstruction increment by use of standard soft-tissue (B30) and lung (B50) reconstruction kernels and a field of view covering the entire chest cross section.

All reconstructed ¹⁸F-FDG PET and CT data were transferred to a Leonardo workstation (VA 70; Siemens Medical Solutions) for further assessment.

Imaging Analysis

All PET/CT studies included low-dose CT for attenuation correction of PET data; in 91 examinations, chest CT was performed for diagnostic purposes. Fully diagnostic CT was performed only for initial staging, restaging in cases of confirmed or highly likely relapse, and known pulmonary metastases in follow-up. PET and CT data each were interpreted independently by 2 experienced nuclear medicine physicians and 2 experienced radiologists, respectively. The evaluating physicians knew the clinical background but were unaware of the results of other imaging modalities. In a final consensus reading, PET and CT images were analyzed again side by side as well as after software-based image fusion by all 4 reviewers. In each reading, the locations of all abnormalities were recorded, and the lesions were scored with a 5-point scale (28,29): 5, definitely malignant; 4, probably malignant; 3, equivocal; 2, probably benign; and 1, definitely benign or physiologic tracer uptake. When individual patients had multiple organ metastases (e.g., pulmonary metastases), a maximum of 4 lesions per organ were included for analysis to avoid the bias of a few studies with a very large number of lesions.

Reference Methods

The site of each lesion was regarded as malignant when the site met at least one of the following criteria: the site was confirmed by histopathology, the site was confirmed by other imaging modalities (MRI, diagnostic CT, or bone scintigraphy), the site exhibited progressive disease in the follow-up period, or the site was accompanied by multiple PET- and CT-positive sites that met any of the first 3 criteria.

On follow-up PET scans, an increase in ¹⁸F-FDG uptake in a lesion was considered to represent the progression of tumor activity. On follow-up CT scans, an increase in the size of a whole lesion or of the extraosseous part of a lesion or a progressive permeative lytic lesion with a periosteal reaction was considered to be a sign of malignancy.

Statistical Analysis and Data Interpretation

Statistical analysis was performed on a lesion basis as well as on an examination basis. For examination-based analysis, lesion-based data were aggregated by examination, assigning the worst evaluation among all lesions within an examination, respectively (i.e., the highest single score of all identified lesions within an examination was assigned to the score of the whole examination). The results of PET and PET/CT were compared by use of a 5 × 5 contingency table. Differences in diagnostic performance were assessed with a marginal homogeneity test.

Sensitivity, specificity, accuracy, positive predictive value (PPV), and negative predictive value (NPV) were calculated in 3 different ways. In the first method, equivocal cases were not considered (scores 1 and 2: benign; scores 4 and 5: malignant). This method represents common clinical practice. Usually, equivocal foci were reevaluated with other modalities or observed with very close follow-up examinations. In the second method, equivocal foci were included and considered to be benign foci (scores 1–3: benign; scores 4 and 5: malignant). In the third method, equivocal foci were included and considered to be malignant foci (scores 1 and 2: benign; scores 3–5: malignant).

The ordinal data obtained with the 5-point scale were analyzed to create receiver operating characteristic (ROC) curves (30). To compare the areas under the curve (AUC) for PET and PET/CT, a nonparametric approach was applied (31). In conventional ROC analysis, the localization of detected lesions is ignored. Consequently, lesions are classified as true positive, even if the correct localization of a lesion has failed. In an alternative approach, the true-positive classification of a lesion requires not only its detection but also its correct localization (L-ROC analysis) (32). Therefore, L-ROC and ROC analyses were performed for the PET datasets. In addition, ROC analyses of different lesion subgroups, such as lesions located in lung tissue, soft tissue, and bones, were carried out.

P values of ≤0.01 were designated as statistically significant.

Additional Foci Detected by CT

To increase the sensitivity of imaging modalities, it is necessary to reduce the number of false-negative results. Several lesions are seen on low-dose CT but not on PET (e.g., lung lesions) and vice versa (e.g., osseous lesions) (33). A total of 124 new foci detected by CT (20%) were not seen by PET alone; 65% of the new foci detected by CT were located in the lungs. These findings are in concordance with those of earlier PET studies in patients with Ewing tumor, which showed a low sensitivity of PET for the detection of pulmonary metastases (24,25,27). Although none of the new foci detected by CT were of physiologic origin, 55% were scored as probably or definitely benign. A large portion, however, were scored as equivocal (36%), and only 19% were scored as probably or definitely malignant. However, among the 124 new lesions detected by CT, 23 additional true-positive foci were not detected by PET alone (an additional 12% true-positive malignant foci among the total of 193 malignant lesions). These results are relevant for staging and further therapeutic management.

Mislocalization

Mislocalization is one of the major problems of interpretation of images obtained by PET alone. Typical examples are intrapulmonary metastases falsely localized by PET interpretation in adjacent structures, such as the pleura, ribs, or mediastinum (34). The additional anatomic information provided by the CT component allows mislocalization to be avoided. Especially in pediatric patients, anatomic correlation is very helpful because of the smaller anatomy and the different organ proportions of pediatric patients relative to adult patients.

To quantify the diagnostic performance of PET, including correct localization, L-ROC curves were generated. Simultaneously detecting and properly localizing lesions reduced the AUCs of the L-ROC curves to 0.66 in the lesion-based analysis and to 0.78 when additional foci detected by CT were excluded. The corresponding values for PET/CT were 0.92 and 0.92 and underlined the outstanding diagnostic performance of the hybrid technique.

In 8% of all foci detected by PET, the localization had to be changed after the CT information was considered. The majority of mislocalized foci (90%) were actually located in the skeleton, soft tissues, or physiologic foci. The question that arises is whether such changes in localization have an impact on patient staging and therapeutic management. In lesions with changes in localization, the precise anatomic information provided by CT helps to reduce the frequency of equivocal lesions (PET, 14%; PET/CT, 4%) and increases the frequency of probably benign or physiologic foci (PET, 6%; PET/CT, 27%). The frequency of malignant or probably malignant foci does not change when anatomic information is included. Furthermore, localization of metastases is a major prognostic factor in Ewing tumor. Therefore, therapeutic management for patients with metastases depends on the affected organs (lungs, pleura, bone, bone marrow, and other locations) (5–7). In particular, the prognosis in patients with osseous lesions is very poor (5,7), and it is essential that these patients receive intensive therapy. The identification of all metastatic sites and the correct localization of lesions are important for staging and for making therapeutic management decisions, because metastasectomy or external radiation of all tumor foci provides a curative approach.

Organ-Specific Analyses

ROC analyses of different organ-specific subgroups showed that PET/CT was significantly more accurate than PET in all evaluated locations and that pulmonary foci displayed the greatest increase in AUCs (PET, 0.66; PET/CT, 0.89; P < 0.0001). These findings are in accordance with those of former studies showing that the sensitivity of PET alone is low and inferior to that of spiral CT in detecting pulmonary metastases and may be attributable to respiration movements during emission acquisition (23,24). The PET/CT technique offers both the high sensitivity of PET in the detection of osseous and soft-tissue lesions (23,25) and the high sensitivity of CT in the detection of lung metastases within a single examination.

A comparison of low-dose CT and full diagnostic CT of the chest with regard to the detection of pulmonary metastases was not addressed in the present study. The question as to whether the sensitivity of low-dose CT is high enough to exclude pulmonary metastases remains and should be evaluated in further studies.

Radiation Exposure

Although PET plays an important role in the diagnostic process for patients with Ewing tumor, radiation exposure during an examination often is of great concern (18,19). For pediatric patients, special weight-adapted protocols are used, with an effective doses of about 5 mSv in small children and in adolescents (35,36). The effective dose of low-dose CT scans is reduced to as little as 1/26 that of diagnostic CT scans in adults, which is about 0.7 mSv for whole-body low-dose CT (37). However, the results of the present study and our clinical experience show that the improved diagnostic results obtained with hybrid PET/CT with low-dose CT are more valuable for making therapeutic management decisions and outweigh the risk that additional radiation exposure may harm the patient.

Limitations

Only a small number of PET and CT findings could be compared with histopathologic findings, possibly biasing the specificity of the results. This lack of a histologic gold standard is a common problem in imaging studies and in clinical situations. Therefore, the status “positive for malignancy” was based on a clinically accepted concept (11,13,14). A falsely high sensitivity of both PET and PET/CT could be attributable to the fact that metastases that were not revealed by any imaging modality were not considered. To minimize potential biases, a follow-up of at least 6 mo was adopted to exclude metastatic disease.

Furthermore, other routine staging methods, that is, MRI of the primary tumor site and bone scintigraphy, were considered. The aim of the present study was to evaluate whether PET and PET/CT were able to depict viable tumor tissue and to provide correct information for staging and restaging. A systematic comparison with MRI, especially whole-body MRI, has not yet been performed.