Diagnostic accuracy of fluorine-18-fluorodeoxyglucose positron emission tomography in the evaluation of the primary tumor in patients with cholangiocarcinoma: a meta-analysis.

Objective. To meta-analyze published data about the diagnostic accuracy of fluorine-18-fluorodeoxyglucose (18F-FDG) positron emission tomography (PET) or PET/computed tomography (PET/CT) for primary tumor evaluation in patients with cholangiocarcinoma (CCa). Methods. A comprehensive literature search of studies published through December 31, 2013, was performed. Pooled sensitivity and specificity were calculated on a per patient based analysis. Subgroup analyses considering the device used (PET versus PET/CT) and the localization of the primary tumor (intrahepatic cholangiocarcinoma (IH-CCa), extrahepatic cholangiocarcinoma (EH-CCa), and hilar cholangiocarcinoma (H-CCa)) were carried out. Results. Twenty-three studies including 1232 patients were included in the meta-analysis. Pooled sensitivity and specificity of 18F-FDG-PET or PET/CT were 81% and 82%, respectively. Pooled sensitivity and specificity, respectively, were 80% and 89% for PET, 82% and 75% for PET/CT, 95% and 83% for IH-CCa, 84% and 95% for H-CCa, and 76% and 74% for EH-CCa. Conclusions.   18F-FDG-PET and PET/CT were demonstrated to be accurate diagnostic imaging methods for primary tumor evaluation in patients with CCa. These tools have a better diagnostic accuracy in patients with IH-CCa than in patients with EH-CCa. Further studies are needed to evaluate the accuracy of 18F-FDG-PET or PET/CT in patients with H-CCa.


Introduction
Cholangiocarcinoma (CCa) is a malignant tumor arising from the epithelium of the bile ducts and is usually classified by anatomical and clinical criteria into intrahepatic cholangiocarcinoma (IH-CCa), hilar cholangiocarcinoma (H-CCa), and extrahepatic cholangiocarcinoma (EH-CCa) [1]. CCa has a poor prognosis and surgical resection with appropriate lymph node dissection is advocated as the curative approach in some patients [2]. Consequently, accurate evaluation and staging are critical to provide indication to surgery and to avoid unnecessary surgical interventions [3].
Fluorine-18-fluorodeoxyglucose ( 18 F-FDG) positron emission tomography (PET) and PET/CT have been proposed as noninvasive imaging methods to assess the disease extent in cancer patients [4]. Since 18 F-FDG is a glucose analogue, this radiopharmaceutical may be very useful in detecting malignant lesions which usually present high glucose metabolism. Hybrid PET/CT device allows enhanced detection and characterization of neoplastic lesions, by combining the functional data obtained by PET with morphological data obtained by CT [4].
Several studies have assessed the diagnostic accuracy of 18 F-FDG-PET or PET/CT in the evaluation of primary tumor in patients with CCa, reporting different values of sensitivity and specificity. The purpose of our study is to meta-analyze published data on the diagnostic accuracy of 18 F-FDG-PET or PET/CT in the evaluation of primary tumor in patients with CCa, in order to provide more evidence-based data and to address further studies in this setting.

Search Strategy.
A comprehensive computer literature search of PubMed/MEDLINE and Embase databases was carried out to find relevant published articles concerning the evaluation of primary tumor in patients with CCa. We used a search algorithm based on a combination of terms ("PET" or "positron emission tomography") and ("cholangiocarcinoma" or "cholangiocellular" or "cholangio * " or "biliar" or "biliary" or "bile" or "Klatskin"). Only articles in English language were considered. The search was performed from inception to December 31, 2013. To expand our search, references of the retrieved articles were also screened for additional studies.

Study Selection.
Studies or subsets in studies investigating the role of 18 F-FDG-PET or PET/CT in the evaluation of primary CCa were eligible for inclusion. Case reports, small case series, review articles, letters, editorials, and conference proceedings were excluded. The following inclusion criteria were applied to select studies for this meta-analysis: (1) original studies in which 18 F-FDG-PET or PET/CT were performed in patients with CCa or suspicious CCa; (2) a sample size of at least ten patients with CCa or suspicious CCa; (3) sufficient data to reassess sensitivity and specificity of 18 F-FDG-PET or PET/CT in detecting the primary tumor in patients with CCa; (4) no data overlap.
Three researchers (SA, DAP, and CC) independently reviewed titles and abstracts of the retrieved articles, applying the above-mentioned selection criteria. Articles were rejected if they were clearly ineligible. The same three researchers then independently evaluated the full-text version of the included articles to determine their eligibility for inclusion.

Data
Extraction. Information about basic study (authors, year of publication, and country of origin), study design (prospective or retrospective), patients' characteristics (number of patients with biliary ducts lesions performing 18 F-FDG-PET or PET/CT, mean age, and gender), and technical aspects (injected activity of 18 F-FDG and time between injection and image acquisition) was collected.
Each study was analyzed to retrieve the number of truepositive (TP), true-negative (TN), false-positive (FP), and false-negative (FN) findings of 18 F-FDG-PET or PET/CT in patients with CCa or suspicious CCa, according to the reference standard. Only studies providing such complete information were finally included in the meta-analysis.

Quality Assessment. The 2011 Oxford Center for
Evidence-Based Medicine checklist for diagnostic studies was used for quality assessment of the included studies. This checklist has 5 major parts as follows: representative spectrum of the patients, consecutive patient recruitment, ascertainment of the gold standard regardless of the index test results, independent blind comparison between the gold standard and index test results, and enough explanation of the test to permit replication.

Statistical
Analysis. Sensitivity and specificity of 18 F-FDG-PET and PET/CT in the evaluation of primary CCa were obtained from the individual studies, on a per patientbased analysis. We considered as positive a biliary ducts lesion with increased uptake of 18 F-FDG, according to the criteria reported by the different authors. When a positive lesion was histologically confirmed as malignant, this was considered a TP lesion, whereas a histologically confirmed benign lesion was considered as a FP lesion. We considered as negative a lesion with no uptake of 18 F-FDG: when the lesion was histologically confirmed as malignant, this was considered as FN lesion, whereas a histologically confirmed benign lesion was considered as a TN lesion.
Sensitivity was determined according to the following formula: TP/(TP + FN); specificity was determined according to this following formula: TN/(TN + FP). Statistical pooling of the data was performed by means of a random effects model. Pooled data are presented with 95% confidence intervals (95% CI). Heterogeneity between studies was assessed by an 2 index. A summary receiving operator characteristics (ROC) curve was obtained for selected studies and area under the curve (AUC) was calculated to assess the overall accuracy of 18 F-FDG-PET and PET/CT. Subsequently, subgroup analyses were also performed, calculating the pooled sensitivity and specificity of 18 F-FDG-PET and PET/CT in three different groups of primary CCa (IH-CCa, EH-CCa, and H-CCa) and in two groups based on the different device used (PET or PET/CT).
For publication bias evaluation, funnel plots, Egger's regression intercept, and Duval and Tweedie's method were used [5].
Statistical analyses were performed using Meta-DiSc statistical software version 1.4.
The reference standard used to validate the 18 F-FDG-PET or PET/CT findings in the included studies was quite different ( Table 4). The results of the quality assessment of the studies included in this systematic review, according to the 2011 Oxford Center for Evidence-Based Medicine checklist for diagnostic studies, are shown in Table 4. Sensitivity and specificity values of 18 F-FDG-PET or PET/CT on a per patient-based analysis ranged from 59 to 100% and from 63 to 100%, with pooled estimates of 81% (95% CI: 78-83%) and 82% (95% CI: 75-87%), respectively. The area under the summary ROC curve was 0.89. The included studies showed statistical heterogeneity in their estimate of sensitivity ( 2 : 63.7%).
To reduce the heterogeneity, subgroup analyses considering the different device used (PET or PET/CT) were performed ( Figure 4). In studies in which 18 F-FDG-PET was used, values of sensitivity (thirteen eligible studies) and specificity (seven eligible studies) on a per patient-based analysis ranged from 60 to 95% and from 67 to 95%, respectively, with pooled estimates of 80% (95% CI: 76-83%) and 89% (95% CI: 80-95%), respectively. Statistical heterogeneity was found only in their estimate of sensitivity ( 2 : 63%). The area under the ROC curve was 0.92.
In studies in which hybrid 18 F-FDG-PET/CT was used, values of sensitivity (ten eligible studies) and specificity (six eligible studies) on a per patient-based analysis ranged from 59 to 100% and from 63 to 100%, respectively, with pooled estimates of 82% (95% CI: 78-85%) and 75% (95% CI: 65-84%), respectively. Statistical heterogeneity was found only in their estimate of sensitivity ( 2 : 67%). The area under the ROC curve was 0.81.
Finally, subgroup analyses considering different anatomic sites of CCa (IH-CCa, EH-CCa, and H-CCa) were carried out ( Figure 3). In patients with IH-CCa, values of sensitivity (nine eligible studies) and specificity (five eligible studies) on a per patient-based analysis ranged from 91 to 100% and from 80 to 100%, respectively, with pooled estimates of 95% (95% CI: 91-98%) and 83% (95% CI: 64-94%), respectively. No statistical heterogeneity was found, among the included studies, in both the estimate of sensitivity and the estimate of specificity ( 2 : 0%). The area under the ROC curve was 0.95.

Discussion
To the best of our knowledge, this meta-analysis is the first to evaluate the diagnostic accuracy of 18 F-FDG-PET or PET/CT in the evaluation of primary tumor in patients with CCa [26]. Several studies have used 18 F-FDG-PET or PET/CT in this setting reporting different values of sensitivity and specificity. However, many of these studies have limited power, analyzing only relatively small numbers of patients. In order to derive more robust estimates of the diagnostic accuracy of 18 F-FDG-PET or PET/CT in this setting we pooled published studies. or "biliar, " or "biliary, " or "bile, " or "the klatskin") " , " or cholangio * Figure 1: Plot of the literature search.
A systematic review process was adopted in ascertaining studies, thereby avoiding selection bias.
Pooled results of our meta-analysis indicate that 18 F-FDG-PET or PET/CT have a good sensitivity (81%) and specificity (82%) in the evaluation of primary tumor in patients with CCa. Furthermore, the value of the AUC (0.89) demonstrates that 18 F-FDG-PET or PET/CT are accurate diagnostic methods in this setting. Considering patients with all anatomical localizations of primary CCa, independently of the device used (PET or PET/CT), significant heterogeneity between the studies in their estimate of sensitivity was found ( 2 : 63.7%). In order to reduce possible source of heterogeneity, subgroup analyses considering different device used (PET or PET/CT) and patients with different anatomical localizations (IH-, H-, and EH-CCa) were performed.
These subgroup analyses provide differences in the diagnostic accuracy data for various anatomical localizations. 18 F-FDG-PET and PET/CT seem to be more sensitive and specific in the evaluation of primary tumor in patients with IH-CCA than in patients with H-CCA and EH-CCA.
In particular 18 F-FDG-PET and PET/CT have a moderate diagnostic accuracy in evaluating primary EH-CCa (sensitivity of 76% and specificity of 74%). In this setting, sensitivity and specificity of 18 F-FDG-PET and PET/CT may be affected by FN (due to the confounding anatomical localization of extrahepatic bile ducts) and FP (due to inflammation of extrahepatic bile ducts). Larger use of hybrid PET/CT and, consequently, further studies about the role of PET/CT in evaluation of primary tumour in patients with EH-CCA may improve these results.
Conversely, the diagnostic accuracy of 18 F-FDG-PET and PET/CT in primary IH-CCA (sensitivity of 95% and specificity of 83%) seems to be better than in the other anatomical localizations of primary CCa. Possible explanations are the easier individuation of illness in the liver parenchyma and the small number of FP cases (intrahepatic noncancerous disease positive with 18 F-FDG-PET). Further studies are needed to evaluate if different histological types of IH-CCA (nodular or mass-forming type, infiltrating type, and intraluminal type) could cause different diagnostic accuracy of 18 F-FDG-PET and PET/CT in this setting.
Finally, the diagnostic accuracy of 18 F-FDG-PET and PET/CT in evaluating primary H-CCa is good (sensitivity of 84% and specificity of 95%). Nevertheless, we cannot exclude that the low number of the included studies in this subgroup analysis may have influenced the results. FP findings (due to the presence of 18 F-FDG-avid lymph nodes in the hepatic hilum) and FN results (due to the difficult anatomical localization of the hepatic hilum) should be considered. More studies are needed to further evaluate sensitivity and specificity of 18 F-FDG-PET and PET/CT in primary H-CCa.
However, performing these subgroup analyses has been useful in demonstrating that the anatomical localization of primary tumor (IH-CCa, EH-CCa, or H-CCA) is a source of heterogeneity among the studies. In fact, no significant heterogeneity was found in the subgroup analyses performed, except in the calculation of pooled sensitivity of 18 F-FDG-PET or PET/CT in primary EH-CCA.
Pooled sensitivity is similar in the subgroup analyses regarding different device used (80% for PET and 82% for PET/CT, resp.). Nevertheless, heterogeneity was found in these groups, in particular for the calculation of pooled sensitivity, suggesting that, beyond the device used, other   factors (such as the anatomical localization of the primary CCa) seem to be a stronger source of heterogeneity. PET alone seems to be more specific than PET/CT (89% and 75%, resp.). A possible explanation of these surprising findings could be the higher number of patients with primary EH-CCa included in the studies which performed PET/CT compared to those which performed PET only.
Finally, regarding the diagnostic workup of patients with CCa, 18 F-FDG-PET and PET/CT may have little diagnostic advantage over traditional imaging modalities in detecting the primary CCA [3]. 18 F-FDG-PET and PET/CT can be complementary to CT and MR in the diagnosing and staging of CCA [20]. Since 18 F-FDG-PET imaging is a wholebody scanning technique, it allows detection of unsuspected     Li et al. [19] 2008 41    metastatic lymph nodes or distant spread that may lead to major changes in the surgical management of patients with biliary tract cancer [25]. Nevertheless, the diagnostic performance of 18 FDG-PET or PET/CT in detecting metastatic lymph nodes or distant spread was not object of our analysis. This study has several limitations. Different anatomical classifications of CCa were used by several studies. For example, it is likely that some H-CCa were classified as EH-CCa by some studies. Other possible limitations of our metaanalysis could be the heterogeneity between the included studies (nevertheless subgroup analyses were performed to reduce the heterogeneity) and the possible publication bias. We assessed publication bias in our meta-analysis using qualitative and quantitative methods (Egger's regression and Duval and Tweedie's method). Funnel plots showed the importance of possible publication bias in particular for the estimation of pooled sensitivity ( Figure 2).
Overall, 18 F-FDG-PET and PET/CT were demonstrated to be accurate noninvasive tools in the evaluation of primary tumors in patients with CCa. Furthermore, more studies in patients with H-CCa and cost-effectiveness analyses of the role of 18 F-FDG-PET or PET/CT in this setting are needed.