Impact of F-18 Fluorodeoxyglucose PET/CT and PET/MRI on Initial Staging and Changes in Management of Pancreatic Ductal Adenocarcinoma: A Systemic Review and Meta-Analysis

A systemic review and meta-analysis were conducted to investigate the diagnostic ability for staging and impact on management of F-18 fluorodeoxyglucose (FDG) positron emission tomography/computed tomography (PET/CT) and PET/magnetic resonance imaging (MRI) in patients with pancreatic ductal adenocarcinoma. A comprehensive search was performed in four databases to retrieve studies of pancreatic ductal adenocarcinoma patients that have reported the diagnostic ability of FDG PET/CT and PET/MRI for detecting metastasis and the proportion of patients whose management was changed by its results. The sensitivity and specificity for detecting metastasis and the proportion of patients with management changes were pooled using a random-effects model. A total of 10 studies were included. The pooled sensitivity and specificity for detecting lymph node metastasis were 0.55 and 0.94, respectively, while the pooled sensitivity and specificity for detecting distant metastasis were 0.80 and 1.00, respectively. The areas under the summarized receiver operating characteristic curves for detecting lymph node and distant metastasis were 0.88 and 0.92, respectively. The pooled proportion of patients with management changes was 19%. FDG PET/CT and PET/MRI showed high diagnostic accuracy for detecting lymph node and distant metastasis in pancreatic ductal adenocarcinoma patients, and the use of these imaging tools led to management changes in a significant portion of these patients.


Introduction
Pancreatic cancer is notorious for its extremely poor prognosis [1]. A previous study reported the 5-year survival rate of pancreatic cancer to be only 9%, and it is the seventh leading cause of cancer death in both men and women worldwide [1,2]. The only potential curative treatment for pancreatic cancer is radical surgical resection [3]. However, according to several recent reviews, at the time of initial staging work-up, distant metastatic lesions are found in over 50% of pancreatic cancer patients and only 20% of patients have resectable disease [1,3]. In order to choose the most suitable treatment with newly diagnosed pancreatic ductal adenocarcinoma as the patients; (2) FDG PET/CT or PET/MRI as the intervention; (3) no comparator; (4) the diagnostic performance for staging pancreatic cancer, confirmed by histopathological evaluation or imaging studies, and the proportion of patients with management changes as the outcomes; and (5) original articles as the study type. The exclusion criteria were as follows: studies that (1) enrolled any patients with malignant pancreatic diseases other than pancreatic ductal adenocarcinoma, such as neuroendocrine tumors, metastasis, or lymphoma; (2) used PET, not PET/CT, or PET/MRI as the intervention; (3) provided insufficient information for calculating true-positive, false-positive, true-negative, and false-negative rates for estimating diagnostic performance during per-patient analysis or for calculating the proportion of patients with management changes; (4) were publication types other than original articles, including case reports, reviews, letters, editorials, and conference abstracts; (5) were in vitro or animal studies; (6) had study populations that overlapped with other literature; (7) were published in languages other than English or Korean; or (8) were not available in full text.

Data Extraction and Quality Assessment
The following information was extracted from each of the included studies: the first author, publication year, country, study design, number and characteristics (age, sex, and enrollment criteria) of patients, details of the intervention techniques (imaging modality and imaging analytical method), reference standard, and outcomes. The methodological quality of the included studies was assessed using the Quality Assessment of Diagnostic Accuracy Studies (QUADAS)-2 tool [21]. The QUADAS-2 tool consists of 14 questions with 4 key domains-patient selection, index test, reference standard, and flow and timing. Each question was answered with "yes" for a low risk of bias, "no" for a high risk of bias, and "unclear" if insufficient data were reported [21].

Statistical Analysis
The primary outcome of this meta-analysis was the diagnostic accuracies of FDG PET/CT and PET/MRI for detecting lymph node metastasis (N staging) and distant metastasis (M staging) in patients with pancreatic cancer. The secondary outcome was the impact of FDG PET/CT and PET/MRI on patient management, as assessed by calculating the proportion of patients whose therapeutic plan was changed due to the PET/CT or PET/MRI findings.
With the data extracted from the included studies, the pooled sensitivity, specificity, positive likelihood ratio (PLR), negative likelihood ratio (NLR), and diagnostic odds ratio (DOR) in detecting lymph node metastasis and distant metastasis were calculated by using a bivariate meta-analysis method with a random-effects model. Furthermore, summarized receiver operating characteristic (SROC) curves were generated for evaluating the diagnostic abilities of PET/CT and PET/MRI using the calculated area under the curves (AUC). The proportion of patients with management changes was calculated for each study, and the results were meta-analytically pooled using a random-effects model. The I 2 statistic was used to assess the heterogeneity among the included studies. An I 2 value lies from 0% to 100%, with a value of >50% suggesting substantial heterogeneity [22]. In meta-analyses with an I 2 > 50%, subgroup analyses were further performed to investigate the sources of heterogeneity; investigated subgroups were country (Asia vs. non-Asia), study design (prospective vs. retrospective), and analytical method of PET images (qualitative analysis vs. quantitative analysis). Publication bias was assessed using Deek's funnel plot and Egger's test. All statistical analyses were performed using Stata software version 15.0 (StataCorp, College Station, TX, USA).

Study Selection and Characteristics
The process of study selection is depicted in Figure 1. In the systematic literature search, a total of 1457 articles were initially retrieved (Supplementary Table S1). After excluding 384 duplicated articles and 1002 articles based on title and abstract screening, 71 articles were potentially eligible. On full-text assessment for eligibility, 61 studies were excluded; hence, 10 studies comprising 852 patients were finally included in our meta-analysis.
The characteristics of the 10 enrolled studies are shown in Table 1. All included studies were published in English; three studies were prospective [14,15,23], while the remaining seven were retrospective [6,[11][12][13][24][25][26]. The number of enrolled patients in each study ranged from 37 to 261, with two studies showing enrollment of >100 patients [11,15]. Of the 10 included studies, only one was a multi-center study [15], while the remaining were single-center studies. As for imaging methods, one study evaluated the diagnostic performance of PET/MRI [23], while in all other included studies the diagnostic ability of PET/CT was assessed. For the imaging analytical method, six studies used visual assessment for determining lesions with positive FDG uptake [6,14,[23][24][25][26], whereas four studies performed quantitative analysis using the cut-off standardized uptake value (SUV) of 2.5 [11,13], 3.0 [12], or 3.5 [15] for determining malignant involvement. full-text assessment for eligibility, 61 studies were excluded; hence, 10 studies comprising 852 patients were finally included in our meta-analysis. The characteristics of the 10 enrolled studies are shown in Table 1. All included studies were published in English; three studies were prospective [14,15,23], while the remaining seven were retrospective [6,[11][12][13][24][25][26]. The number of enrolled patients in each study ranged from 37 to 261, with two studies showing enrollment of >100 patients [11,15]. Of the 10 included studies, only one was a multi-center study [15], while the remaining were single-center studies. As for imaging methods, one study evaluated the diagnostic performance of PET/MRI [23], while in all other included studies the diagnostic ability of PET/CT was assessed. For the imaging analytical method, six studies used visual assessment for determining lesions with positive FDG uptake [6,14,[23][24][25][26], whereas four studies performed quantitative analysis using the cut-off standardized uptake value (SUV) of 2.5 [11,13], 3.0 [12], or 3.5 [15] for determining malignant involvement.

Quality Assessment
The methodological quality assessment of the included studies is shown in Figure 2 ( Supplementary  Table S2). In the domain of patient selection, six studies were judged to have high or unclear risk of bias, which was mainly due to providing insufficient information regarding consecutive enrollment of patients [6,11,13,[24][25][26]. For the reference standard, three studies were judged to have high or unclear risk of bias because the blinding method was not used in interpreting the references standard results [11,23,26]. For the flow and timing, the major weaknesses were the lack of information regarding the interval between the index tests and reference standard and the lack of a uniform reference test for enrolled patients. Meanwhile, only a single study was judged to be high risk with regard to the concerns about applicability in the domain of patient selection [6]. Furthermore, a single study was considered to have unclear risk with regard to applicability concerns in both the index test [11] and reference standard [23] domains.

Quality Assessment
The methodological quality assessment of the included studies is shown in Figure 2 (Supplementary Table S2). In the domain of patient selection, six studies were judged to have high or unclear risk of bias, which was mainly due to providing insufficient information regarding consecutive enrollment of patients [6,11,13,[24][25][26]. For the reference standard, three studies were judged to have high or unclear risk of bias because the blinding method was not used in interpreting the references standard results [11,23,26]. For the flow and timing, the major weaknesses were the lack of information regarding the interval between the index tests and reference standard and the lack of a uniform reference test for enrolled patients. Meanwhile, only a single study was judged to be high risk with regard to the concerns about applicability in the domain of patient selection [6]. Furthermore, a single study was considered to have unclear risk with regard to applicability concerns in both the index test [11] and reference standard [23] domains.

Management Changes Following PET/CT
Six studies with a total of 650 patients assessed the proportions of patients who underwent management changes following FDG PET/CT staging and compared it to that of conventional staging [6,[11][12][13][14][15]. All six studies performed PET/CT for initial staging of pancreatic cancer. The pooled percentage of patients who underwent management changes following FDG PET/CT was 19% (95% CI: 5-34%, Figure 5). In all six studies, the most common reason of management change was identifying unknown metastatic lesions by PET/CT. Additionally, PET/CT also changed the management plan by detecting secondary primary malignancies and ruling out metastasis of lesions found on other imaging modalities.

Management Changes Following PET/CT
Six studies with a total of 650 patients assessed the proportions of patients who underwent management changes following FDG PET/CT staging and compared it to that of conventional staging [6,[11][12][13][14][15]. All six studies performed PET/CT for initial staging of pancreatic cancer. The pooled percentage of patients who underwent management changes following FDG PET/CT was 19% (95% CI: 5-34%, Figure 5). In all six studies, the most common reason of management change was identifying unknown metastatic lesions by PET/CT. Additionally, PET/CT also changed the management plan by detecting secondary primary malignancies and ruling out metastasis of lesions found on other imaging modalities.
Among the six studies, two prospective studies have evaluated the cost-effectiveness of PET/CT in staging of pancreatic cancer [14,15]. In a study by Heinrich et al. [14], PET/CT was cost saving by avoiding patients from unnecessary surgery, showing cost-savings of USD 1066 per patient. In another study by Ghaneh et al. [15], PET/CT was both less costly and more effective when compared to contrast-enhanced CT. A subgroup of pancreatic cancer patients with resection showed the most cost-effective results, showing cost-savings of GBP 1542 per patient. management changes following FDG PET/CT staging and compared it to that of conventional staging [6,[11][12][13][14][15]. All six studies performed PET/CT for initial staging of pancreatic cancer. The pooled percentage of patients who underwent management changes following FDG PET/CT was 19% (95% CI: 5-34%, Figure 5). In all six studies, the most common reason of management change was identifying unknown metastatic lesions by PET/CT. Additionally, PET/CT also changed the management plan by detecting secondary primary malignancies and ruling out metastasis of lesions found on other imaging modalities.

Subgroup Analysis
Subgroup analyses was performed based on the country, study design, and the analytical method of each study (Table 2). For the diagnostic ability of FDG PET/CT and PET/MRI for lymph node metastasis, because there was only one study performed in a non-Asian country and only one with a prospective study design, we performed a subgroup analysis only based on the analytical method. On this subgroup analysis, studies that used qualitative analytical methods showed a higher sensitivity for detecting both lymph node metastasis (0.64 vs. 0.40, respectively) and distant metastasis (0.84 vs. 0.64, respectively) than those that used quantitative analytical methods (p < 0.05 for both). Meanwhile, all analyzed factors failed to explain the heterogeneity in the specificity for detecting lymph node and distant metastasis. In subgroup analysis of the proportion of subjects who underwent management changes, studies in non-Asian countries, with a prospective study design, and with qualitative analytical methods revealed higher proportions of patients with management changes following PET/CT (28.2 vs. 9.8%, respectively, for country; 39.9 vs. 10.2%, respectively, for study design; 25.5 vs. 14.8%, respectively, for analytical method) than those in Asian countries, with a retrospective design, and with quantitative analytical methods (p < 0.05 for all).

Discussion
In patients with pancreatic cancer, contrast-enhanced CT is the preferred primary imaging modality for the initial evaluation [27]. Additionally, MRI and endoscopic ultrasound have been commonly used to delineate primary tumors, to evaluate blood vessel involvement, and to detect metastatic lesions [27,28]. Although most pancreatic cancer lesions showed increased FDG uptake, the potential benefits of FDG PET/CT in staging pancreatic cancer remains contentious [9,27,29]. In previous studies, staging FDG PET/CT showed a high diagnostic accuracy for detecting metastatic lesions of pancreatic cancer [13,14,25,26], and it has been recommended in patients with localized disease for detecting metastatic lesions in guidelines from the Japan Pancreas Society, the United Kingdom National Institute for Health and Care Excellence, and the National Comprehensive Cancer Network [30][31][32]. By contrast, because of the relatively small proportion of patients in whom additional metastatic lesions were found only by FDG PET/CT, other studies have suggested that PET/CT has a limited role in the staging work-up of pancreatic cancer [11,12]. In guidelines from the American Society of Clinical Oncology and the European Society for Medical Oncology, the use of FDG PET/CT is not routinely recommended for the management of pancreatic cancer patients [28,33]. In the present meta-analysis, FDG PET/CT and PET/MRI showed only moderate sensitivity for detecting lymph node metastasis. Meanwhile, PET/CT and PET/MRI demonstrated high specificity for detecting lymph node metastasis and high sensitivity and specificity for detecting distant metastasis, showing high AUC of the SROC values for both lymph node and distant metastasis. Furthermore, the findings from FDG PET/CT and PET/MRI led to management changes in 19% of pancreatic cancer patients mainly by identifying unknown metastatic lesions, and PET/CT was cost-effective for pancreatic cancer staging with cost-savings of more than USD 1000 per patient. The results of our study suggest that FDG PET/CT and PET/MRI could be a valuable diagnostic imaging modality for initial staging in patients with pancreatic cancer and could have a significant impact on determining therapeutic plans for these patients. Considering that, in the included studies, FDG PET/CT and PET/MRI both had a high diagnostic accuracy for distant metastasis, and most patients who underwent management changes following FDG PET/CT or PET/MRI were upstaged to stage IV [6,[12][13][14][15]26], FDG PET/CT and PET/MRI should be routinely recommended in patients with localized pancreatic cancer who plan to receive curative treatment.
In previous meta-analyses of studies of pancreatic cancer patients, the sensitivity and specificity of FDG PET and PET/CT were 32-67% and 75-81%, respectively, for identifying lymph node metastasis and 57-67% and 96-100%, respectively, for distant metastasis [4,34]. Since those meta-analyses mainly included studies that utilized a PET scanner or studies that enrolled patients with malignant pancreatic diseases other than pancreatic ductal adenocarcinoma [4,16,34], it is difficult to make a direct comparison of our results with the results of these previous meta-analyses. However, our results demonstrated higher specificity for detecting lymph node metastasis and higher sensitivity for detecting distant metastasis with FDG PET/CT or PET/MRI than the results of those previous meta-analyses [4,34]. Given that anatomical information gleaned from CT images from PET/CT scans can increase the specificity for lymph node staging and the sensitivity for detecting metastatic lesions [5,18,25,35], these results may suggest an incremental increase in the diagnostic value of PET/CT and PET/MRI compared with PET alone.
Our meta-analyses found substantial heterogeneity across the included studies. Notably, in subgroup analyses performed to investigate the source of this heterogeneity, the method of analyzing PET images was found to be a significant factor for heterogeneity in all meta-analyses. Qualitative analysis is subjective and known to be dependent on the clinical experience of the reader, which might lead to significant inter-reader discrepancies [36][37][38]. Hence, quantitative analysis using SUV has been widely used for objective assessment of malignant diseases and has shown comparable results to those of qualitative analysis [36][37][38]. However, the results of our subgroup analysis revealed that studies that used qualitative analytical methods showed significantly higher sensitivity for lymph node and distant metastasis and a higher proportion of patients with management changes than studies using quantitative methods. This might be due to the different cut-off values used in the individual studies that utilized quantitative analysis, as there is no determined cut-off SUV for defining metastatic lesions [36]. Further studies would be needed to clarify the effects of analytical methods on PET/CT and PET/MRI results. Another potential source of the heterogeneity between studies for the proportion of subjects who underwent management changes was study design. In the subgroup analysis, the prospective studies had significantly higher proportions of patients with management changes than the retrospective ones. Given that retrospective studies use the data that might be collected for another objective and are often assumed to have more bias than prospective ones [39], this result might imply the underestimation of the effect of FDG PET/CT on therapeutic planning in the retrospective studies. Additionally, variations in the country where the studies were performed was another potential cause of heterogeneity seen between studies for the proportion of patients who underwent management changes.
The present analysis had several limitations that need to be addressed. First, the number of studies included in this meta-analysis was relatively small, and most of the included studies were retrospectively performed (7 out of 10 studies) or single-center studies (9 out of 10 studies). Further studies are necessary which include a large number of prospective and multi-center studies. Second, the definition of a positive lesion on PET/CT, as well as reference standards, varied among the included studies which could affect the accuracy of the results. Third, due to insufficient information regarding the stages of those patients who underwent management changes following PET/CT, the pooled proportion of patients with management changes could not be stratified by initial stage. However, given this limitation, our results could support the clinical use of PET/CT in pancreatic cancer irrespective of the initial stage. Fourth, the definition of changes in patient management following PET/CT staging might differ among the enrolled studies. Furthermore, although the diagnostic performance of PET/MRI was similar to that of PET/CT, because only a single study evaluated the diagnostic performance of PET/MRI [23], we could not perform subgroup analysis to compare diagnostic ability between PET/MRI and PET/CT. Lastly, there was substantial heterogeneity across the studies, and we were unable to find the potential source for the heterogeneity in specificity. Hence, the general application of our pooled results might be limited to specific clinical conditions.

Conclusions
In the present meta-analysis, FDG PET/CT and PET/MRI showed high specificity for detecting lymph node metastasis and high sensitivity and specificity for identifying distant metastasis in patients with pancreatic ductal adenocarcinoma. Furthermore, FDG PET/CT and PET/MRI had a significant impact on the management of pancreatic ductal adenocarcinoma, with our results showing a pooled proportion of 19% of patients who underwent management changes following imaging. Based on our results, FDG PET/CT and PET/MRI would be considered to be one of the routine imaging examinations used in the staging work-up of pancreatic cancer. However, because of the relatively small number of included studies with substantial heterogeneity among them, further prospective studies with larger populations are needed to further elucidate our results.

Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.