A flow diagram describing patient enrollment and the study protocol is shown in Figure 1.
This prospective study was conducted in patients aged >18 years with histopathologically proven cancers who underwent [F18]FDG whole body PET/CT scan at the Division of Nuclear Medicine, Department of Radiology, Faculty of Medicine Siriraj Hospital, Mahidol University, Bangkok, Thailand during the 1 December 2014 to 30 November 2017 study period. Patients with severe renal dysfunction (glomerular filtration rate [GFR] <20 ml/minute; n=3), contraindications for contrast material (n=2), and marked hyperglycemia (fasting blood sugar [FBS] >200 mg/dl; n=2) were excluded.
All PET/CT studies were performed at least 4 weeks after biopsy or surgery, 6 weeks after chemotherapy, 12 weeks after radiotherapy, and 2 weeks after granulocyte-colony stimulating factor (GCSF) treatment to prevent false-positive results due to inflammatory changes. Clinical, histopathological, radiological follow-up, and other correlative investigation were used as reference standards to evaluate the diagnostic accuracy of PET/CT.
All data were acquired using an integrated PET/CT system (Discovery®; GE Healthcare, Milwaukee, WI, USA) that integrates a 16-detector row CT scanner with a lutetium oxyorthosilicate (LSO)-based PET scanner. Patients fasted at least 6 hours prior to [F-18]FDG administration, and blood glucose level was checked using a glucose meter (Stat Strip®; Nova Biomedical, Waltham, MA, USA) just before [F-18]FDG administration. Tracer injection was performed only in patients whose blood glucose level was less than 150 mg/dL. Whole-body emission images were obtained 60 minutes after intravenous administration of [F-18]FDG (0.14-0.2 mCi/kg), with an average injected dose of 381.19±76.4 MBq (10.3±2.0 mCi).
In all patients, 2 ml/kg of non-ionic contrast material using either Iopamidol (Iopamiro®; Bracco Imaging, Milano, Italy) or Iohexol (Omnipaque®; GE Healthcare, Milwaukee, WI, USA) or Ioversol (Optiray®; Guerbet, Cedex, France) or Iodixanol (Visipaque®, GE Healthcare, Milwaukee, WI, USA) was intravenously injected. Whole body CT (30-300 mAs using Automatic Exposure Control (AEC) and smart mA, 14.0 noise index, 120 kVp, and helical thickness of 2.5 mm collimation) was performed immediately before and after contrast injection. Contrast-enhanced CT scan was performed after contrast injection using the appropriate delay time (40 seconds for lesion at head/neck/chest, and 70 seconds for lesion at abdomen). PET data was acquired in 3-D mode, 3 minutes per bed position, and reconstruction was performed using an ordered subset expectation maximization algorithm (3D Iterative [VUE Point]).
The reconstructed, attenuation-corrected images of all PET/CT datasets were reviewed by one board-certified nuclear medicine physician (14-years’ experience) and one diagnostic radiologist (11-years’ experience), both of whom were blinded to patient clinical information, using an AW Workstation (GE Healthcare, Milwaukee, WI, USA). Images without attenuation correction were available for evaluation in cases with suspicious artifacts. For non-contrast PET/CT, a lesion was defined as either a focus of increased [F-18]FDG uptake compared with background, or as morphologic change with features that increase suspicion of the presence of a tumor. For contrast-enhanced PET/CT, abnormal enhancement was added to the criteria adopted for non-contrast PET/CT (e.g., enhancement >15 HU in pulmonary nodule sized >8 mm).(10-13) Lymph node metastasis was considered in lymph nodes with FDG avidity and when their shortest axial diameter was >11 mm in the jugulodigastric region and >10 mm in the cervical/abdominal/pelvic region (>5 mm in rectal cancer), or if irregular border or central necrosis was evident, or if there was a cluster of three or more lymph nodes of borderline size.(14-16) Distant metastasis was defined as a focally increased [F-18]FDG activity compared with background, with associated soft tissue mass outside of the primary lesion or bony destruction. Equivocal lesion detected by PET or CT that failed to satisfy any of the aforementioned diagnostic criteria was designated as an indeterminate lesion.
All whole body PET/CT studies were assessed using an 8-point scale, as follows; 0 = no abnormality detected; 1 = focal FDG uptake without CT abnormality; 2 = focal FDG uptake with CT abnormality, favoring benign; 3 = focal FDG uptake with CT abnormality, favoring malignant; 4 = CT abnormality without FDG avidity, favoring benign; 5 = CT abnormality without FDG avidity, favoring malignant; 6 = focal FDG uptake with CT abnormality, indeterminate; and, 7 = CT abnormality without FDG avidity, indeterminate. These lesions were then classified as indeterminate (scores 1, 6, or 7); definite benign lesion (scores 2 or 4); or, definite malignant lesion (scores 3 or 5). Both observers reviewed PET/NCCT first, followed by PET/CECT and PET/NCCT-CECT, with at least a 3-week interval between each of the 3 sets to prevent recall bias.
Study outcome was assessed by comparing total number, characterization scores (0-7), and diagnostic confidence (determinate vs. indeterminate) for all lesions detected by the three PET/CT techniques. Both patient-based and lesion-based data were collected and analyzed. To compare diagnostic performance, we compared the interpretation obtained from PET/CT studies with the corresponding results from intra-operative findings, pathologic study, or change in imaging findings during a minimum follow-up period of 6 months. Among patients with no intraoperative findings, lesions that decreased in size or that remain unchanged without receiving any further treatment were considered benign, while progressive lesions were considered to be malignant.
Statistical analysis and sample size calculation
All data were analyzed using the statistical software package PASW Statistics for Windows, Version 18.0 (SPSS, Inc., Chicago, IL, USA) and MedCal version 18.2.1 (MedCalc Software bvba, Ostend, Belgium; http://www.medcalc.org; 2018). Categorical variables were analyzed using chi-square test, and continuous variables were tested using one-way analysis of variance (ANOVA) with post hoc analysis. McNemar’s test was used to determine the statistical significance of differences in lesion detection accuracy by PET/NCCT, PET/CECT, and PET/NCCT-CECT. Kappa statistic was used to analyze agreement of interpretation between PET/CT techniques. For all tests, a p-value less than 0.05 was considered statistically significant. Receiver operating characteristic (ROC) curve analysis was performed to assess diagnostic performance relative to the results obtained from intra-operative finding, pathological study, or change in imaging findings from each technique. The sensitivity, specificity, accuracy, likelihood ratio, positive predictive value (PPV), and negative predictive value (NPV) of all techniques were determined. Net reclassification improvement (NRI) of PET/CECT as compared to PET/NCCT was also calculated, with malignancy from final diagnosis being considered an event.
The sample size for this study was calculated using data from a study by Cantwell, et al.(2) Using a detection rate of liver lesion by non-contrast enhanced PET/CT in patients with colorectal cancer of 70%, a significance level of 0.05, and a power of 80%, the calculated sample size was 33. To cover the three most common types of cancer for which PET/CT study is requested and to compensate for a 10% loss of data for any reason, the calculated minimum sample size was 110 patients.