The patient characteristics and scan results for malignant lesions are shown in Table 1, and those for benign lesions are shown in Table 2. In total, 23 lesions were assessed (16 based on histologic verification and seven based on clinical or radiologic follow-up) in 21 patients (13 girls and eight boys; age range, 1–17 years). The average age was 11.8 ± 1.4 years for patients with malignant lesions and 11.9 ± 0.9 years for patients with benign lesions (p > 0.05).

There was considerable overlap in the early SUV between benign and malignant lesions. All 13 malignant lesions (100%) had an early SUV of 2.5 or higher (Table 1 and Fig. 1), whereas only three of 10 (33.3%) benign lesions had an early SUV of less than 2.5 (Table 2 and Fig. 2). For patients with malignant disease, the SUV that was measured between the first and second imaging time points increased such that the mean early and delayed SUVs were 7.3 ± 1.2 and 10.9 ± 2.7, respectively, although the difference between the mean SUVs was not statistically significant (p > 0.05; n = 13). In patients with benign lesions, the mean early and delayed SUVs remained stable between the two imaging time points (4.5 ± 0.8 and 4.2 ± 1.0, respectively; p > 0.05; n = 10). When comparing the SUVs between malignant and benign groups, the mean delayed SUV was observed to be significantly greater for the malignant tumor group compared with the benign group (p < 0.05), whereas no significant difference in the early SUV was observed between the two groups.

The average calculated retention index for the malignant group was significantly greater than that for patients with benign disease (37.1% ± 10.8% vs −9.9% ± 7.1%; p < 0.01) (Fig. 3). By using a retention index of 10% or higher between the early and delayed scans as a criterion for malignancy, 10 of 13 neoplastic lesions were identified correctly (Table 1). Similarly, eight of 10 nonmalignant lesions were diagnosed correctly as being benign (Table 2). Three malignancies were incorrectly classified as being benign, whereas two lesions meeting criteria for the diagnosis of malignancy were later confirmed to represent infection or inflammation. Therefore, the sensitivity, specificity, and accuracy for this method were 76.9%, 80.0%, and 78.2%, respectively. In comparison, the sensitivity, specificity, and accuracy for the conventional single–time-point PET approach when using an early or delayed SUV of 2.5 or higher as a cutoff alone for distinguishing between malignant and benign conditions were 100%, 30.0%, and 69.5%, and 100%, 40%, and 73.9%, respectively.

Combined FDG PET/CT is being used more frequently in the evaluation of pediatric patients with cancer, and numerous studies have shown that FDG PET/CT has very high sensitivity with regard to staging, therapy monitoring, and detection of relapse [2]. Indeed, FDG is the most commonly available PET radiopharmaceutical, and its biologic uptake is substantially increased in most types of pediatric tumors because of an increase in glucose metabolism and glucose transporters [2]. Unfortunately, FDG is not tumor specific, and increased accumulation may be seen in a variety of benign entities, including infections, drug toxicity, granulocyte colony-stimulating factor therapy, radiation therapy, postoperative or postbiopsy changes, fracture, degenerative changes, or injection leakage [1]. FDG PET image interpretation is further complicated in pediatric patients because of the normal physiologic uptake of the radiotracer by the thymus gland, brown adipose tissue, skeletal growth centers, and hematopoietic bone marrow. Recognition of these findings and correlation with clinical and other radiologic findings (e.g., PET/CT fusion images) is therefore paramount for the accurate interpretation of PET data in children [17]. Alternatively, numerous studies in adult patients have found that the difference in the time course of FDG uptake can be used to improve the ability of PET to distinguish benign lesions from some types of malignancies in adults [3–10]. Although the role of dual–time-point FDG PET/CT in pediatric patients is less well established, the results of the current study show significant advantages of early and delayed FDG PET/CT in differentiating malignant and benign diseases in pediatric patients.

An SUV threshold of 2.5 or higher has been proposed as the optimal threshold for separating malignant from benign lesions [2]. However, a considerable overlap in the early or delayed SUV_max was observed among most benign and malignant lesions. The degree of overlap became less evident with delayed imaging, likely because the uptake of FDG continues to increase in malignant tissues for several hours after FDG injection, whereas benign lesions show a decrease or remain stable over time. Nevertheless, numerous studies have similarly found a significant degree of overlap between SUVs of malignant and benign lesions, which can cause considerable difficulty in correctly interpreting FDG PET data. Aoki and colleagues [18], for example, examined 52 children and adults with benign (n = 33) or malignant (n = 19) bone tumors, including osteosarcoma and chondrosarcoma. Although both malignancies had high SUVs (> 3.0), there was considerable overlap in the SUVs of other nonmalignant pathologic entities, such as fibrous dysplasias, sarcoidosis, and Langerhans cell histocytosis, and a cutoff level for differentiating malignant from benign bone disease could not be applied. Dimitrakopoulou-Strauss et al. [19] similarly reported considerable overlap in the SUVs of malignant bone and soft-tissue tumors (e.g., osteosarcoma and Ewing sarcoma; mean SUV, 3.7; range, 0.4–12.3) and those of benign pathologic abnormalities (e.g., osteomyelitis, fibroma, granuloma, and others; mean SUV, 1.1; range, 0.4–3.5). A more recent study by Kodama et al. [20] similarly reported SUVs of less than 3.0 in five of 16 patients with neuroblastoma or clear cell sarcoma, whereas some patients presenting with benign processes such as lymphadenitis had moderate SUVs equal to 3.2. Consequently, an SUV at a single time point alone (whether an early or delayed image) may not be reliable enough on its own to characterize malignant lesions of different organ systems.

In the current study, which used an SUV threshold of 2.5 for separating malignant from benign lesions, single–time-point PET showed good sensitivity (100%), with all malignant lesions having an SUV of 2.5 or higher at early or delayed imaging. However, many benign lesions were also observed to have increased FDG uptake, resulting in a high false-positive rate and relatively poor specificity of 30–40%. In comparison with single–time-point PET, dual–time-point FDG PET/CT has the advantage of calculating a retention index, which, when applied using a threshold of 10% or higher, was shown to increase the specificity to 80%, whereas the sensitivity decreased moderately to 77%. This finding is comparable to the study by Tian et al. [11], who similarly compared single–time-point PET versus dual–time-point FDG PET/ CT and applied an SUV threshold of 2.5 and a retention index of 10% for distinguishing benign from malignant bone lesions in adult patients. They similarly found that the specificity increased from 44% to 90.6%, whereas the sensitivity decreased from 96% to 76% when the results of delayed imaging were considered [11]. Thus, these data suggest that malignancy should be considered high on the differential diagnosis when a negative lesion on early imaging (i.e., with an SUV < 2.5) subsequently shows a retention index of 10% or higher on delayed imaging, whereas a nonmalignant process might be considered if the subsequent image reveals a decrease in the SUV or becomes negative on delayed imaging. Overall, dual–time-point FDG PET/ CT may provide more information than early imaging alone in distinguishing benign from malignant lesions. Moreover, dual–time-point FDG PET/CT can be realistically and easily incorporated into the daily work routine because the delayed images, in our experience, only required a short amount of camera time to acquire (i.e., approximately 10–15 minutes).

Brown adipose tissue (i.e., brown fat) is another potential source of false-positive findings on FDG PET, especially in children, in whom FDG uptake in brown fat is more common than in adults [1, 21]. Indeed, Alkhawaldeh and Alavi [22] recently found that there is a progressive increase in FDG uptake with most hyper-metabolic brown fat areas on dual–time-point FDG PET/CT that mimic malignant lesions. Although the uptake of FDG in metabolically active brown fat is typically uniform and symmetric, brown fat can have wide variability in distribution and degree of FDG uptake. These factors therefore need to be taken into account when trying to differentiate brown fat and malignancy, and careful attention to anatomic location on CT is needed to avoid misinterpretation. We did not attempt to discriminate between brown fat and malignancy using dual– time-point FDG PET/CT; however, the use of combined FDG PET/CT fusion imaging, along with strategies for preventing brown fat uptake, including room temperature control and the use of warm blankets [21, 23, 24], has enabled our group to improve the anatomic localization of brown fat activity and to exclude the presence of underlying soft-tissue abnormality.

Although dual–time-point FDG PET/CT was more specific than single–time-point PET for the detection of malignancy, several benign lesions also exhibited increased SUVs over time. These cases included one patient with an oral cavity lesion on FDG PET and with a history of acute lymphoblastic leukemia but who had received dental work several days before the PET scan (patient 9 in Table 2), and another patient (patient 8 in Table 2) who was being evaluated for malignancy but who was subsequently diagnosed with pulmonary histoplasmosis due to infection with Histoplasma capsulatum. Likewise, not all malignant lesions revealed increasing SUVs over time. Indeed, two large malignant lesions (rhabdomyosarcoma and anaplastic large cell lymphoma; patients 2 and 10 in Table 1) with large photopenic areas on FDG PET, possibly representing areas of necrosis within the tumor, had a decrease in SUV over time, suggesting that caution should be exercised when interpreting dual–time-point FDG PET/CT images in lesions with nonsymmetric or nonhomogenous uptake of the radiotracer. These cases illustrate that dual–time-point FDG PET/CT may still have shortcomings in its ability to discriminate between tumor and infection or inflammation and should be considered as complementary to standard investigations and clinical workup in the evaluation of malignancy in children.

In conclusion, dual–time-point FDG PET/ CT is a simple and noninvasive diagnostic tool that has improved accuracy and provides more information than single–time-point imaging in distinguishing malignant from nonmalignant lesions in pediatric patients, and it is feasible enough to incorporate into the daily work routine. The inclusion of a larger number of cases, including those with different pathologic subtypes, is recommended for future studies. These studies are currently planned by our group.