Contrast-enhanced endoscopic ultrasound for differential diagnosis of pancreatic cancer: an updated meta-analysis

Aim We aim to assess the diagnostic value of contrast-enhanced endoscopic ultrasound (CE-EUS) for pancreatic cancer and inflammatory lesions by pooling current evidence. Materials and Methods A systematical search of PubMed, Web of Science and the Cochrane Library was performed from inception to January 2016. Two authors independently screened and extracted detailed data from included studies. A random effect model was adopted to estimate the pooled sensitivity, specificity in order to determine the diagnostic ablitity of CE-EUS. Furthermore, we conducted the meta-regression and subgroup analyses to explore possible heterogeneity. Results Eighteen eligible studies enrolling 1668 patients were finally included in the study. The pooled sensitivity of CE-EUS for distinguishing pancreatic cancers from solid inflammatory masses was 0.93 (95% CI, 0.91–0.94), and the specificity was 0.88 (95% CI, 0.84–0.90). The area under summary receiver operating characteristic curve yielded 0.97. No publication bias was observed by Deeks’ funnel plot in current meta-analysis. Conclusions We provided evidence that CE-EUS is a promising modality for differential diagnosis of pancreatic adenocarcinomas. Further multicenter prospective studies should be carried out to certify its utility.


INTRODUCTION
Endoscopic ultrasonography (EUS) is considered as a valuable diagnostic technology for pancreatic diseases with good spatial resolution [1,2]. Although the sensitivity of EUS is high, its ability to characterize and differentiate solid masses is still limited [3]. However, it is crucial for clinicians to establish or exclude pancreatic malignancy in clinical works. Distinguishing pancreatic cancer from inflammatory lesions remains challenging with conventional EUS. The development of EUS-guided fine needle aspiration (EUS-FNA) makes it possible for characterization of pancreatic lesions with high accuracy [4][5][6]. EUS-FNA is effective in differentiation of pancreatic masses. However, EUS-FNA had its own limitations, including sampling errors and invasive procedure [7,8]. In addition, the sensitivity of EUS-FNA significantly reduced to 54%-73% with the setting of chronic pancreatitis [9,10]. It is still imperative for endoscopists to seek for effective and noninvasive technologies that could differentiate pancreatic cancer accurately.
Contrast-enhanced ultrasonography was used in percutaneously abdominal ultrasound examination since 1995, Contrast-enhanced endoscopic ultrasonography (CE-EUS) is also been performed to determine the pancreatic parenchymal perfusion and microvessels inside lesions of interest with better delineation [11]. The application of contrast agents in EUS had improved characterization ability of pancreatic masses and aided in the differentiation of pancreatic diseases. Several studies had evaluated

Characteristics and quality assessment of included study
Our initial search identified 1475 articles from databases. After applied inclusion and exclusion criteria, 18 eligible studies comprising 1668 participants were included in final analysis. Detailed selection flow was presented in Figure 1, The main characteristics are presented in Table 1. Ten studies were performed in Europe, seven were conducted in Asia and one was in USA. The gold diagnostic standard was based on pathology histology, or follow-up. According to QUADAS-2 criteria, the overall methodological quality of included articles was moderate to high ( Figure 2).

Meta-regression and sensitivity analysis
To explore possible heterogeneity, we performed a meta-regression analysis and results showed that the characteristics of studies were not significantly associated with diagnostic odds ratio ( Table 2). It is illustrated by  (Table 3).

DISCUSSION
It is well known that pancreatic cancer is a lethal disease with dismal prognosis duo to low eraly detection rate. The 5-year survival rate was sharply decreased by advanced stage [30]. Patients were free of symptoms until Oncotarget 66395 www.impactjournals.com/oncotarget they were diagnosed by advanced pancreatic cancer [31]. There is an imperative need to diagnose pancreatic cancer at earlier stages. Regarding accurate differentiation of pancreatic cancers from inflammatory tumor-like lesions still remains a big challenge for clinicians and is crucial for therapeutic decisions [32].
It is well demonstrated by several studies [33][34][35] that EUS is super to other modalities in detection and diagnosis of pancreatic diseases with high sensitivity. However, the ability to characterize solid lesions accurately still remains limited, particularly in the setting of chronic pancreatitis.
CE-EUS, with the intravenously infusion of contrast agents, is a newly developed technology. It can characterize and differentiate pancreatic lesions non-invasively [11]. In general, CE-EUS could be classified as contrast-enhanced Doppler EUS (CD-EUS) and contrast-enhanced harmonic EUS (CH-EUS) according to the method of sonographic assessment [36]. For CD-EUS, intravenous contrast agents would enhance the Doppler signals from vascularity of targeted lesions [37,38]. However, the disadvantage of this technique included the flash and blooming artifacts. Furthermore, the poor ability to depict microvessels with slow flow and parenchymal perfusion also limited its application widely [27,39]. For these limitations, CH-EUS was developed to overcome them. It depicts harmonic signals from contrast agents selectively and filters them from surrounding tissues [40,41]. Thereby, it provides more detailed images of fine vessels with slow flow and parenchymal perfusion in the target lesions [42,43]. This allows pancreatic lesions to be visualized and characterized more accurately. Besides, the second-generation of contrast agents also demonstrated to be relatively safe for patients, even with liver and renal dysfunctions [43][44][45][46].    Low heterogeneity across pooled sensitivity (I 2 < 30%) and High heterogeneity across pooled specificity (I 2 > 50%).   In our meta-analysis, the pooled results supported a great diagnostic value of CE-EUS for characterization and differentiation of pancreatic masses, which were consistent with previous meta-analysis and studies [29]. Compared with previous meta-analysis, the strength of our study is included comprehensive lectures. CE-EUS is extremely useful for patients with negative results of EUS-FNA. Although EUS-FNA still is considered as a gold standard for pancreatic cancer diagnosis, the sensitivity and accuracy is still suboptimal, particularly in the setting of chronic pancreatitis [47]. Several trials reported that CH-EUS could complements EUS-FNA by delineating the outline of the target lesions clearly, thus facilitating EUS-FNA [14,19,48]. Moreover, it could not only improve the sensitivity of EUS-FNA, but also avoid repeated biopsy or surgery. For CH-EUS, a well-known and sensitive diagnostic standard of pancreatic adenocarcinomas is a hypoenhanced image of lesion [19,22]. age olesion [19,22]. However, it is somehow operator-dependent and subjective to the analysis of the enhanced pattern, which might affect the diagnostic accuracy. To avoid this disadvantage, a quantification analysis, time-intensity curve for region of interest, was developed recently [15,18,21,28].
Five of the included studies demonstrated the values of maximum intensity, median intensity, time to peak, intensity reduction rate, the ratio of uptake inside the mass to uptake of the surrounding parenchyma in discrimination of malignances from pancreatitis, solid-pseudopapilliary neoplasm and neuroendocrine tumors [15,18,21,26,28], which makes objective definition of lesion characteristics possible. CH-EUS is also a reproducible method in the evaluation of pancreatic lesions with good interobserver agreement, even for endosonographers with no or limited experience in EUS [16].
The present meta-analysis has some limitations. Significant heterogeneity in specificity and positive likelihood ratio might affect interpretation of the data and conclusions. Serval diagnostic criterion for CE-EUS were adopted in included studies, which might introduce some bias into our conclusion. Furthermore, we cannot exclude the presence of publication bias, although the analysis of the funnel plot indicated that it could not be detected.
In conclusion, CE-EUS, especially for CH-EUS, is a promising tool for differential diagnosis of pancreatic cancer. CE-EUS should be regarded as a promising tool for pancreatic masses characterization, especially when EUS-FNA findings were negative. Further multicenter trials should be carried out to certify its utility.

Search strategy
We searched PubMed, Web of Science and the Cochrane Library from inception to January 2016 for relevant articles comprehensively. Following search terms were adopted: ("contrast-enhanced" OR "contrast medium" OR "echo-enhanced") AND ("pancreatic mass*" OR Oncotarget 66399 www.impactjournals.com/oncotarget "pancreatic cancer" OR "pancreatitis" OR "pancreatitis" OR "pancreatic lesion*" OR "pancreatic adenocarcinoma") AND ("ultrasonograph*" OR "ultrasound" OR "endosonograph*" OR "endosonography" OR "EUS"). We also searched bibliography of articles and reviews to identify additional articles.

Inclusion and exclusion criteria
Articles were considered as eligible if they used CE-EUS for the diagnosis, provision of data for true positive (TP), false positive (FP), false negative and true-negative (TN), the reference standard based on histopathology of samples by EUS-FNA, surgery or a follow-up of at least 6 months. Following studies were excluded: (1) complete data unavailable; (2) overlapping with the selected articles; (3) Case reports, reviews, editorials, comments, abstracts.

Data extraction
Two authors independently extracted following data from each study: authors, year, country, numbers of patient, sex, age, diagnostic standard, contrast agent, contrast mode, gold standard. We adopted the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) for quality assessment [49]. Any discrepancies were resolved by discussions or consensus.

Statistical methods
The pooled estimates of sensitivity, specificity, diagnostic odds ratio were performed by Meta-Disc, version 1.4 (Ramony Cajal Hospital, Madrid, Spain). Heterogeneity across studies was evaluated by the Cochrane Q test and I 2 statistic. We constructed a summary receiver operating characteristic (sROC) curve and calculated the area under sROC (AUC). The metaregression and sensitivity analyses were performed with the following covariates such as numbers of patient (< 60 patients vs. ≥ 60 patients), contrast mode (color/power Doppler vs. harmonic), country (Europe vs. other), analysis of images (quality vs. quantity). Deeks' asymmetry test was used to detect publication bias by Stata version 13.0 (Stata Corporation, College Station, Texas). The two-tailed P value is statistically significant at less than 0.05.

Authors ̕ contributions
Sun LM designed the research and submit the manuscript; He XK and Sun LM collected the data and performed statistical analysis together; He XK wrote the manuscript.

ACKNOWLEDGMENTS AND FUNDING
We express our gratitude to Wen-Jie Dong for her assistant in data management.

CONFLICTS OF INTEREST
The authors declare no conflicts of interests.