- Email: [email protected]
Cyst carcinoembryonic antigen in differentiating pancreatic cysts: A meta-analysis
Digestive and Liver Disease 45 (2013) 920–926
Contents lists available at ScienceDirect
Digestive and Liver Disease journal homepage: www.elsevier.com/locate/dld
Liver, Pancreas and Biliary Tract
Cyst carcinoembryonic antigen in differentiating pancreatic cysts: A meta-analysis Saowanee Ngamruengphong, Michael J. Bartel, Massimo Raimondo ∗ Division of Gastroenterology and Hepatology, Mayo Clinic Florida, Jacksonville, FL, USA
a r t i c l e
i n f o
Article history: Received 5 March 2013 Accepted 2 May 2013 Available online 18 June 2013 Keywords: Endoscopic ultrasonography Fine-needle aspiration Pancreatic cancer
a b s t r a c t Background: Using carcinoembryonic antigen in discriminating between benign and malignant disease remains controversial. Aims: We aim to evaluate the diagnostic accuracy of cyst fluid carcinoembryonic antigen in predicting malignant pancreatic cystic lesions. Methods: We performed a literature search of MEDLINE and EMBASE. We included studies that compared the diagnostic accuracy of carcinoembryonic antigen with histology. Pooled estimates of diagnostic precision were calculated using random-effects models. Results: Eight studies (504 patients) were included. The carcinoembryonic antigen cutoff level for determining a malignant cyst ranged from 109.9 to 6000 ng/mL. Pooled estimates of carcinoembryonic antigen in malignant cysts prediction were poor: pooled sensitivity of 63%, pooled specificity of 63%. The positive likelihood ratio was 1.89 and the negative likelihood ratio was 0.62. The diagnostic odds ratio was 3.84. The area under the summary receiver–operating characteristic curve was 0.70. In subgroup analysis of patients with mucinous cysts (mucinous cystic neoplasm and intraductal papillary mucinous neoplasm; 5 studies, 227 patients), pooled sensitivity was 65%, pooled specificity 66% and diagnostic odds ratio 4.74 respectively. Conclusion: This meta-analysis suggests that the accuracy of carcinoembryonic antigen in differentiating “between benign and malignant” pancreatic cysts was poor. The decision to perform surgical resection for pancreatic cystic lesions should not be based solely on carcinoembryonic antigen level. © 2013 Editrice Gastroenterologica Italiana S.r.l. Published by Elsevier Ltd. All rights reserved.
1. Introduction Cystic lesions of the pancreas have an estimated prevalence of 2.4% in the population and are commonly detected due to widespread use of high quality imaging [1]. Most incidental pancreatic cysts are small and have a benign course that requires conservative treatment. Some may be or have the potential to be malignant. Differentiating benign from malignant pancreatic cysts is clinically important for deciding the optimal management, as decisions regarding surgery may be based on the risk of malignancy in asymptomatic patients. Various modalities including computerized tomography, endoscopic ultrasound (EUS), and magnetic resonance have been advocated to differentiate benign from malignant pancreatic cystic lesions. However, the diagnostic accuracy of these imaging techniques is still not optimal [2,3].
Cyst fluid carcinoembryonic antigen (CEA) has been proposed as a useful marker for mucinous cysts. A cyst fluid CEA level of 192–200 ng/mL had 80% accuracy for differentiating between mucinous and non-mucinous cysts [4,5]. However, the role of cyst fluid CEA in discriminating between benign and malignant disease remains controversial. The aim of this meta-analysis is to determine the diagnostic precision of cyst fluid CEA in discriminating benign from malignant pancreatic cystic neoplasms. 2. Materials and methods This study was performed in compliance with the checklist provided by the Preferred Items for reporting Systematic Reviews and Meta-Analyses (PRISMA; Supplementary Table 1). 2.1. Search strategy
∗ Corresponding author at: Division of Gastroenterology and Hepatology, Mayo Clinic Florida, 4500 San Pablo Road, Jacksonville, FL 32224, USA. Tel.: +1 904 953 6982; fax: +1 904 953 7260. E-mail address: [email protected] (M. Raimondo).
A literature search of the MEDLINE and EMBASE databases was conducted for English-language studies published before October 2012. Systematic search strategies were designed by a consultant health science librarian. The following keywords were used: “pancreas OR pancreatic cystic lesion,” “tumour marker OR
1590-8658/$36.00 © 2013 Editrice Gastroenterologica Italiana S.r.l. Published by Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.dld.2013.05.002
S. Ngamruengphong et al. / Digestive and Liver Disease 45 (2013) 920–926
Carcinoembryonic Antigen OR CEA” and “diagnosis.” A manual search of the cross-references of identified studies and relevant reviews were used for completion of the literature search.
921
as a cause of heterogeneity. The analyses were carried out using the statistical software Meta-DiSc (version 1.4) [9]. 3. Results
2.2. Study eligibility and quality assessment Two reviewers (SN and MJB) independently reviewed each identified trial for fulfilment of predefined inclusion criteria. Studies that compared the diagnostic accuracy of cyst fluid CEA with histological diagnosis were included. Criteria for exclusion were (1) if the study authors did not apply specific cutoff value of cyst fluid CEA to evaluate the sensitivity and specificity in differentiating benign from malignant pancreatic cysts; (2) if there was insufficient information to construct 2 × 2 contingency tables. Articles by the same author or research group were identified for analysis only when different patient populations had been used. In cases of more than one published report on the same population or group of patients, we included only data from the largest number of patients for analysis. We requested additional information and clarification of data from five authors and two responded. All eligible studies identified by the search underwent methodological quality assessment using the 14-item quality assessment of diagnostic accuracy (QUADAS) tool [6]. Disagreement was resolved in a consensus meeting.
2.3. Outcomes of interest Outcomes of interest were the sensitivity, specificity, likelihood ratios, and area under the summary receiver–operating characteristic curves (sROC) of cyst fluid CEA with regards to diagnosis of malignant pancreatic cystic lesions. Malignant cysts were defined as either high grade dysplasia (HGD) or invasive cancer.
2.4. Statistical analysis From the reported data, we extracted sensitivity and specificity of cyst fluid CEA in distinguishing between benign and malignant pancreatic cysts by the use of 2 × 2 contingency tables of correct diagnosis and CEA level above the specified cutoff point. The pooled estimates of sensitivity, specificity, positive likelihood ratios (LR), negative LR, diagnostic odd ratios (DOR) and their 95% confidence intervals (95% CI) were calculated using random-effects model. sROC was performed to assess the interaction between sensitivity and specificity. We used DOR and area under the sROC curve (AUC) to analyze the diagnostic performance of cyst fluid CEA in distinguishing the benign from the malignant group. The DOR was calculated from sensitivity and specificity and was defined as [(true positive rate/false positive rate)]/[(1 − true positive rate/1 − false positive)]. A higher DOR reflected a greater diagnostic precision of cyst fluid CEA in the detection of malignant pancreatic cysts. The AUC represents an overall summary of test performance and a global measure of overall test accuracy. A perfect test will have an AUC close to 1 and poor tests have AUCs close to 0.5 [7,8]. Because the DOR varies across studies with different diagnostic threshold, the sROC curve is asymmetrical. Cochrane Q test and I2 test were used to assess the inter-study heterogeneity. Statistically significant heterogeneity was considered if p < 0.1 and I2 > 50%. Subgroup analyses were performed to assess the diagnostic accuracy of cyst fluid CEA in mucinous cystic neoplasm and intraductal papillary mucinous neoplasm (IPMN). Analysis included only high quality studies (QUADAS score ≥ 11). The type of pancreatic cyst (mucinous only versus mixed type) and study quality (QUADAS score ≥ 11 versus < 11) were included as covariates to examine its possibility
The search strategy identified 844 articles. Of these, 795 articles were excluded after preliminary abstract review and 49 articles underwent detailed full-text evaluation. From these, 8 published articles [10–17] met our inclusion criteria. The number of articles is outlined in Fig. 1 by reason for exclusion at each stage of the eligibility assessment. The reasons for exclusion of each individual article for the 57 articles following detail assessment were summarized in Appendix A. The study by Correa-Gallego et al. [13] was published as a letter to the editor. All other articles were published as full papers. Two studies [13,14] reported the data from Massachusetts General Hospital, Boston, USA. The study by Correa-Gallego et al. [13] evaluated the diagnostic performance of cyst fluid CEA in 72 patients with histologically confirmed IPMNs, whereas Cizginer et al. [14] studied 198 patients with various types of pancreatic cystic lesions, including 66 patients with IPMNs. We included these two studies in the meta-analysis as they were conducted by different authors and had different patient eligibility criteria, although the study population overlapped. However, we performed a post hoc analysis to assess the influence of each study on heterogeneity and summary estimate by including each of these two studies one at a time. There were a total of 504 patients in this analysis (315 with benign and 189 with malignant cysts). The characteristics of the included studies are shown in Table 1. Results from the application of the QUADAS 14-question quality assessment tool revealed that all studies scored ranged from 8 to 12 (Appendix B). Two studies were prospective and 6 were retrospective in design. The cutoff value of cyst fluid CEA used to differentiate benign and malignant cysts ranged from 109.9 to 6000 ng/mL. One study [10] reported the sensitivity and specificity of cystic fluid CEA at two different cutoff levels; 1000 and 6000 ng/mL. Because the DOR of cutoff level of 6000 ng/mL was higher, we used the sensitivity and specificity derived from this cutoff level in the meta-analysis. The pooled sensitivity of cyst fluid CEA level in prediction of malignant pancreatic cysts was 63% (95% CI, 55–69) and pooled specificity was 63% (95% CI, 57–68) (Fig. 2). The positive LR was 1.89 (95% CI, 1.13–3.16) and the negative LR was 0.62 (95% CI, 0.40–0.97). The DOR was 3.84 (95% CI, 1.37–10-74) (Fig. 3). Heterogeneity was observed (I2 = 76%, p < 0.0001). The AUC was 0.70 with an SE of 0.09 (Fig. 4). We performed a post hoc sensitivity analysis to evaluate the impact of including two studies from the overlapping study population (Massachusetts General Hospital, Boston, USA) [13,14]. The pooled sensitivity, specificity, and DOR were not significantly different among the two subgroups and the main analysis. The heterogeneity remained high. In the analysis including the study by Correa-Gallego et al. [13] the pooled sensitivity was 56% (95% CI, 47–65) and pooled specificity was 68% (95% CI, 61–74). The DOR was 4.10 (95% CI, 1.09–15.47). Heterogeneity was observed (I2 = 79%, p < 0.0001). In the analysis including the alternate study by Cizginer et al. [14] instead of the Correa-Gallego et al. study [13], the pooled sensitivity was 64% (95% CI, 56–71) and pooled specificity was 68% (95% CI, 62–74). The DOR was 5.19 (95% CI, 1.87–14.39). Heterogeneity was observed (I2 = 69%, p = 0.003). 3.1. Subgroup analysis and sensitivity analysis A subgroup analysis to evaluate the diagnostic performance of cyst fluid CEA in only patients with mucinous cysts (mucinous
922
Table 1 Characteristics of included studies. Male %
Mean age, range (± SD), years
Size of lesion in mm, range, SD
No of patient with malignant lesion (%): definition of malignant cysts (n)
Surgical pathology (n)
Method of calculating cutoff level of CEA
Cutoff level of CEA (ng/mL)
12
63
19 (30%)
67
19.5 (5–74)
16 (36%): HGD (9), invasive cancer (7)
Literature
≥6000
Retrospective
12
41
14 (34%)
64, range 31–72
NA
10 (24%): HGD (5), invasive cancer (5)
ROC curve
>200
Linder et al. 2006; USA [12] Kucera et al. 2012; USA [17]
Prospective
12
71*
42 (41%)
NA
18–102
14 (20%): MCCa (14)
11
47
32 (68%)
66, range 24–88
24 ± 14 (for Br-IPMN)
21 (44%): HGD (14), invasive cancer (7)
Optimal cutoff from included patients Literature
≥6000
Retrospective
Correa-Gallego et al. 2009; USA [13] Cizginer et al. 2011; USA [14]
Retrospective
9
72
NA
NA
NA
Literature
≥200
Prospective
12
77 (38.8%)
60.6
NA
17 (23%): CIS (12), invasive cancer (5) 55 (35%): CIS, invasive cancer
MCN (6), M-IPMN (9), Br-IPMN (25), mixed-IPMN (4) IPMN (41): M-IPMN (2), Br-IPMN (13), mixed-IPMN (26) MCN (21). MCCa (14), PC (23), SCN (13) IPMN (47): M-IPMN (11), Br-IPMN (24), mixed-IPMN (12) IPMN (72)
ROC curve
>109.9
Pais et al. 2007; USA [15]
Retrospective
10
74***
38 (51%)
Median age 65, range 41–84
Literature
>800
Wu et al. 2007; China [16]
Retrospective
8
85
26 (30%)
Cystadenoma: 50, range 17–85. cystadenocarcinoma: 56, range 27–83.
Malignant group: 60 mm. Benign group: 4.7 mm (p = 0.37) NA
MCN (43), IPMN (66), PC, SCN (8), PNEC (7), others(4) IPMN (74): M-IPMN (22), Br-IPMN (19), mixed-IPMN (33) Cystadenoma (37), cystadenocarcinoma (48)
NA
NA
Study design
Othman et al. 2012; USA [10]
Retrospective
Maire et al. 2008; France [11]
QUADAS scores
198**
21 (28%): invasive cancer (8**** ) 48 (56%): Cystadenocarcinoma (48)
>200
CEA: carcinoembryonic antigen; CIS: carcinoma in situ; HGD: high grade dysplasia; IPMN: intraductal papillary mucinous neoplasm; M-IPMN: main duct IPMN; Br-IPMN: branch duct IPMN; MCCa: mucinous cystadenocarcinoma; MCN: mucinous cystic neoplasm; PC: pseudocyst; PNEC: pancreatic neuroendocrine cyst; QUADAS: quality assessment of diagnostic accuracy; ROC: receiver–operating characteristic curve; SCN: serous cyst; NA: not available. * 35 patients with mucinous cysts. ** 157 patients with CEA analysis. *** 23 patients with CEA analysis. **** 8 patients had fluid CEA tested.
S. Ngamruengphong et al. / Digestive and Liver Disease 45 (2013) 920–926
Number of patients
Study author; country
S. Ngamruengphong et al. / Digestive and Liver Disease 45 (2013) 920–926
923
Fig. 1. Articles selection diagram. CEA: carcinoembryonic antigen, ERCP: endoscopic retrograde cholangio-pancreatography.
cystic neoplasm and IPMN) was performed. Five studies involving 227 patients reported sensitivity and specificity of CEA in prediction of malignancy among mucinous cystic lesions [10,11,13,15,17], the pooled sensitivity of cyst fluid CEA level was 65% (95% CI, 57–73) and pooled specificity was 66% (95% CI, 59–72). The DOR was 4.74 (95% CI, 1.46–15.37). The AUC was 0.69 with an SE of 0.12. In subgroup analysis including only 5 studies with higher quality (QUADAS ≥ 11) [10–12,14,17], the pooled sensitivity was 69% (95% CI, 59–77); pooled specificity was 63% (95% CI, 59–70), and DOR was 5.15 (95% CI, 1.38–19.28). There was still significant heterogeneity among studies. Sensitivity analysis was performed to assess the possible sources of heterogeneity across the studies by weight meta-regression analysis of DOR with the following covariates as predictor variables: type of pancreatic cyst (mucinous only versus mixed type) and study quality (QUADAS score ≥ 11 versus < 11). Both type of pancreatic cyst and study quality did not have statistically significant effect on the overall DOR. 4. Discussion Cyst fluid CEA has proven to be a helpful marker in differentiating between mucinous and non-mucinous pancreatic cysts [4,14,17,18]. Whether elevated cyst fluid CEA is predictive of malignant cyst remains controversial. This information is crucial because mucinous lesions with no worrisome features (e.g. presence of symptoms, solid component within cyst, mural nodule), particularly in high surgical risk patients, do not require immediate surgical resection and may be monitored closely for signs of
malignant transformation [5]. In this review, pooled accuracy estimates of cyst fluid CEA in pancreatic cystic lesions and subgroup analysis including only mucinous cysts fell short of achieving the optimal sensitivity and specificity values. Therefore the usefulness in differentiating benign from malignant pancreatic cysts was limited. Our findings suggested that clinical decision making, particularly to perform surgery, should not base solely on a degree of cyst fluid CEA elevation. Our findings support the current guidelines which do not recommend the use of cyst fluid CEA to diagnose malignant pancreatic cysts [5,19]. Although, initial studies have suggested that high cyst fluid CEA is a predictor of malignant pancreatic cysts [11,12], several subsequent studies have reported that the level of cyst fluid CEA concentrations between benign and malignant cystic lesions do overlap [10,13,15,17,18,20]. In fact, Kucera et al. [17] reported the relationship of cyst fluid CEA concentration and the degree of dysplasia; the mean cyst fluid CEA concentration increased as histologic grade progressed from low grade to high grade dysplasia. However, the cyst fluid CEA declined markedly once invasive cancer had developed. The underlying pathophysiological mechanisms of this change are unknown. The role of CEA in the follow-up of patients with pancreatic cysts is even debated. Although, the recently revised guidelines recommend follow-up CEA level in patients with non-resected IPMN [5], the available evidence does not support this statement. Several studies on long-term follow-up patients with pancreatic cysts, including IPMN, who were managed conservatively found no correlation between cyst fluid CEA level and cyst growth [10,20]. The conclusion of our study strengthens the previous findings to
924
S. Ngamruengphong et al. / Digestive and Liver Disease 45 (2013) 920–926
Fig. 2. Forest plot of sensitivity and specificity estimates for cyst fluid carcinoembryonic antigen in predicting malignant pancreatic cysts for all studies.
limit the use of cyst fluid CEA in differentiating mucinous from non-mucinous pancreatic cyst, and forgo its use for prediction of malignant pancreatic cysts or monitoring non-surgical patients. Other than cyst tumour marker, several techniques have been used to improve detection of malignant pancreatic cysts. Certain cyst features in imaging studies were highly associated with malignancy. In a Meta analyses, the pooled odds ratios for malignancy of cyst size > 3 cm, the presence of mural nodules and dilatation of the main pancreatic duct was 62.4 (95%CI 30–126), 9.3 (95%CI 5.3–16.1), and 7.27 (95%CI 3.0–17.4), respectively [21]. Thus, these high risk characteristics should be carefully considered in the differential diagnosis between benign and malignant pancreatic cysts, mainly in the setting of IPMN. Recently, using a cytological
threshold of high-grade atypical epithelial cells as a predictor of malignancy in mucinous cyst has been shown to improved sensitivity from 29% to 72% with specificity of 85% [22]. Moreover, the identification of high-grade atypical epithelial cells in aspirated cystic fluid from a presumed benign small branch duct IPMN because of the absence of high risk radiological features improved the detection of malignancy by 50% [23]. Studies of molecular analysis of the cyst fluid indicated that the detection of KRAS and GNAS mutations can be helpful in predicting malignancy in a pancreatic cystic lesion [24,25]. However, the utility of these tests have not yet been established in routine clinical practice. As with any meta-analysis, our study has limitations. There was significant heterogeneity among the studies. Although we
Fig. 3. Forest plot of diagnostic odds ratio of cyst fluid carcinoembryonic antigen in predicting malignant pancreatic cysts for all studies. OR: odds ratio.
S. Ngamruengphong et al. / Digestive and Liver Disease 45 (2013) 920–926
925
Fig. 4. The summary receiver–operating characteristic curve for cyst fluid carcinoembryonic antigen in predicting malignant pancreatic cysts. The area under curve represents the overall diagnostic accuracy. Each circle represents a study in the meta-analysis. Sample size is indicated by the size of the circle. sROC: summary receiver–operating characteristic, AUC: area under curve, SE: standard error.
attempted to address heterogeneity by pre-specified subgroup analysis and meta-regression, significant heterogeneity still exists. The heterogeneity that we were unable to account for could be due to minor variations in patient populations, methods of sampling, techniques used to assay the samples and proportion of patient with malignant disease. Also, small sample sizes in the included studies could potentially be subject to selection bias. It should be noted that a single published study by Kucera et al. [17] reported that the degree of CEA elevation is different between HGD and invasive cancer. If this finding is true, then defining a “malignant cyst” as HGD and invasive cancer in most of the studies might be inappropriate. This could be subjected to spectrum bias i.e. a phenomenon that tests sensitivity and specificity may vary depending on the severity of disease in the population in which the test has been applied [26]. Because none of the included studies reported the diagnostic accuracy results from patients with HGD separately from the group with invasive cancer, we were unable to perform the analysis to validate this observation. In addition, studies with negative results were likely to be unpublished and this could lead to publication bias. Statistical tests for funnel plot asymmetry, such as the Egger and Begg tests that are often used in meta-analyses of randomized clinical trials, are not effective to detect potential publication bias in diagnostic accuracy [27]. Even though publication bias does exist, the overall diagnostic precision would have been lower than our reported results. To the best of our knowledge, the validation of the assay methods used to measure the CEA level in the cyst fluid matrix has never been described in the literature. The composition of the cystic fluid is different in cysts of variable nature and therefore the method of validation of the fluid cyst matrix can be different from the serum and complex to be applied for a general laboratory. Thus, the cutoff values for CEA in the cystic fluid can be of limited value and difficult to compare in different studies. However, the overall low DOR still suggests that the clinical utility of cyst fluid CEA should not be used in this setting. Because of very limited availability of information, we did not include studies that evaluated the diagnostic performance of cyst fluid CEA
combined with other tumour marker, such as CA 72.4, in this analysis. In conclusion, we found that the level of cyst fluid CEA does not differentiate benign from malignant pancreatic cysts and by itself should not be used for guiding surgical decision making. The clinical value of cyst fluid CEA should be limited only to distinguish mucinous from non-mucinous cystic lesions. Large, multicentric, well-designed trials are needed to further characterize the role of cyst fluid tumour marker and molecular analysis in the evaluation of pancreatic cysts. Financial disclosures No financial support was received for this study. Conflict of interest The authors declare no conflict of interest. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.dld.2013.05.002. References [1] de Jong K, Nio CY, Hermans JJ, et al. High prevalence of pancreatic cysts detected by screening magnetic resonance imaging examinations. Clinical Gastroenterology and Hepatology 2010;8:806–11. [2] Cellier C, Cuillerier E, Palazzo L, et al. Intraductal papillary and mucinous tumors of the pancreas: accuracy of preoperative computed tomography, endoscopic retrograde pancreatography and endoscopic ultrasonography, and long-term outcome in a large surgical series. Gastrointestinal Endoscopy 1998;47:42–9. [3] Baba T, Yamaguchi T, Ishihara T, et al. Distinguishing benign from malignant intraductal papillary mucinous tumors of the pancreas by imaging techniques. Pancreas 2004;29:212–7. [4] Brugge WR, Lewandrowski K, Lee-Lewandrowski E, et al. Diagnosis of pancreatic cystic neoplasms: a report of the cooperative pancreatic cyst study. Gastroenterology 2004;126:1330–6.
926
S. Ngamruengphong et al. / Digestive and Liver Disease 45 (2013) 920–926
[5] Tanaka M, Fernandez-del Castillo C, Adsay V, et al. International consensus guidelines 2012 for the management of IPMN and MCN of the pancreas. Pancreatology 2012;12:183–97. [6] Whiting P, Rutjes AW, Reitsma JB, et al. The development of QUADAS: a tool for the quality assessment of studies of diagnostic accuracy included in systematic reviews. BMC Medical Research Methodology 2003;3:25. [7] Walter SD. Properties of the summary receiver operating characteristic (SROC) curve for diagnostic test data. Statistics in Medicine 2002;21:1237–56. [8] Rosman AS, Korsten MA. Application of summary receiver operating characteristics (sROC) analysis to diagnostic clinical testing. Advances in Medical Sciences 2007;52:76–82. [9] Zamora J, Abraira V, Muriel A, et al. Meta-DiSc: a software for meta-analysis of test accuracy data. BMC Medical Research Methodology 2006;6:31. [10] Othman MO, Patel M, Dabizzi E, et al. Carcino embryonic antigen and longterm follow-up of mucinous pancreatic cysts including intraductal papillary mucinous neoplasm. Digestive and Liver Disease 2012;44:844–8. [11] Maire F, Voitot H, Aubert A, et al. Intraductal papillary mucinous neoplasms of the pancreas: performance of pancreatic fluid analysis for positive diagnosis and the prediction of malignancy. American Journal of Gastroenterology 2008;103:2871–7. [12] Linder JD, Geenen JE, Catalano MF. Cyst fluid analysis obtained by EUS-guided FNA in the evaluation of discrete cystic neoplasms of the pancreas: a prospective single-center experience. Gastrointestinal Endoscopy 2006;64:697–702. [13] Correa-Gallego C, Warshaw AL, Fernandez-del Castillo C. Fluid CEA in IPMNs: a useful test or the flip of a coin? American Journal of Gastroenterology 2009;104:796–7. [14] Cizginer S, Turner B, Bilge AR, et al. Cyst fluid carcinoembryonic antigen is an accurate diagnostic marker of pancreatic mucinous cysts. Pancreas 2011;40:1024–8. [15] Pais SA, Attasaranya S, Leblanc JK, et al. Role of endoscopic ultrasound in the diagnosis of intraductal papillary mucinous neoplasms: correlation with surgical histopathology. Clinical Gastroenterology and Hepatology 2007;5:489–95. [16] Wu H, Yan LN, Cheng NS, et al. Role of cystic fluid in diagnosis of the pancreatic cystadenoma and cystadenocarcinoma. Hepato-Gastroenterology 2007;54:1915–8.
[17] Kucera S, Centeno BA, Springett G, et al. Cyst fluid carcinoembryonic antigen level is not predictive of invasive cancer in patients with intraductal papillary mucinous neoplasm of the pancreas. Journal of Oncology Practice 2012;13:409–13. [18] Park WG, Mascarenhas R, Palaez-Luna M, et al. Diagnostic performance of cyst fluid carcinoembryonic antigen and amylase in histologically confirmed pancreatic cysts. Pancreas 2011;40:42–5. [19] Khalid A, Brugge W. ACG practice guidelines for the diagnosis and management of neoplastic pancreatic cysts. American Journal of Gastroenterology 2007;102:2339–49. [20] Nagula S, Kennedy T, Schattner MA, et al. Evaluation of cyst fluid CEA analysis in the diagnosis of mucinous cysts of the pancreas. Journal of Gastrointestinal Surgery 2010;14:1997–2003. [21] Anand N, Sampath K, Wu BU. Cyst features and risk of malignancy in intraductal papillary mucinous neoplasms of the pancreas: a meta-analysis. Clinical Gastroenterology and Hepatology 2013. [22] Pitman MB, Genevay M, Yaeger K, et al. High-grade atypical epithelial cells in pancreatic mucinous cysts are a more accurate predictor of malignancy than “positive” cytology. Cancer Cytopathology 2010;118:434–40. [23] Genevay M, Mino-Kenudson M, Yaeger K, et al. Cytology adds value to imaging studies for risk assessment of malignancy in pancreatic mucinous cysts. Annals of Surgery 2011;254:977–83. [24] Wu J, Matthaei H, Maitra A, et al. Recurrent GNAS mutations define an unexpected pathway for pancreatic cyst development. Science Translational Medicine 2011;3:92ra66. [25] Khalid A, Zahid M, Finkelstein SD, et al. Pancreatic cyst fluid DNA analysis in evaluating pancreatic cysts: a report of the PANDA study. Gastrointestinal Endoscopy 2009;69:1095–102. [26] Goehring C, Perrier A, Morabia A. Spectrum bias: a quantitative and graphical analysis of the variability of medical diagnostic test performance. Statistics in Medicine 2004;23:125–35. [27] Deeks JJ, Macaskill P, Irwig L. The performance of tests of publication bias and other sample size effects in systematic reviews of diagnostic test accuracy was assessed. Journal of Clinical Epidemiology 2005;58:882–93.
Contents lists available at ScienceDirect
Digestive and Liver Disease journal homepage: www.elsevier.com/locate/dld
Liver, Pancreas and Biliary Tract
Cyst carcinoembryonic antigen in differentiating pancreatic cysts: A meta-analysis Saowanee Ngamruengphong, Michael J. Bartel, Massimo Raimondo ∗ Division of Gastroenterology and Hepatology, Mayo Clinic Florida, Jacksonville, FL, USA
a r t i c l e
i n f o
Article history: Received 5 March 2013 Accepted 2 May 2013 Available online 18 June 2013 Keywords: Endoscopic ultrasonography Fine-needle aspiration Pancreatic cancer
a b s t r a c t Background: Using carcinoembryonic antigen in discriminating between benign and malignant disease remains controversial. Aims: We aim to evaluate the diagnostic accuracy of cyst fluid carcinoembryonic antigen in predicting malignant pancreatic cystic lesions. Methods: We performed a literature search of MEDLINE and EMBASE. We included studies that compared the diagnostic accuracy of carcinoembryonic antigen with histology. Pooled estimates of diagnostic precision were calculated using random-effects models. Results: Eight studies (504 patients) were included. The carcinoembryonic antigen cutoff level for determining a malignant cyst ranged from 109.9 to 6000 ng/mL. Pooled estimates of carcinoembryonic antigen in malignant cysts prediction were poor: pooled sensitivity of 63%, pooled specificity of 63%. The positive likelihood ratio was 1.89 and the negative likelihood ratio was 0.62. The diagnostic odds ratio was 3.84. The area under the summary receiver–operating characteristic curve was 0.70. In subgroup analysis of patients with mucinous cysts (mucinous cystic neoplasm and intraductal papillary mucinous neoplasm; 5 studies, 227 patients), pooled sensitivity was 65%, pooled specificity 66% and diagnostic odds ratio 4.74 respectively. Conclusion: This meta-analysis suggests that the accuracy of carcinoembryonic antigen in differentiating “between benign and malignant” pancreatic cysts was poor. The decision to perform surgical resection for pancreatic cystic lesions should not be based solely on carcinoembryonic antigen level. © 2013 Editrice Gastroenterologica Italiana S.r.l. Published by Elsevier Ltd. All rights reserved.
1. Introduction Cystic lesions of the pancreas have an estimated prevalence of 2.4% in the population and are commonly detected due to widespread use of high quality imaging [1]. Most incidental pancreatic cysts are small and have a benign course that requires conservative treatment. Some may be or have the potential to be malignant. Differentiating benign from malignant pancreatic cysts is clinically important for deciding the optimal management, as decisions regarding surgery may be based on the risk of malignancy in asymptomatic patients. Various modalities including computerized tomography, endoscopic ultrasound (EUS), and magnetic resonance have been advocated to differentiate benign from malignant pancreatic cystic lesions. However, the diagnostic accuracy of these imaging techniques is still not optimal [2,3].
Cyst fluid carcinoembryonic antigen (CEA) has been proposed as a useful marker for mucinous cysts. A cyst fluid CEA level of 192–200 ng/mL had 80% accuracy for differentiating between mucinous and non-mucinous cysts [4,5]. However, the role of cyst fluid CEA in discriminating between benign and malignant disease remains controversial. The aim of this meta-analysis is to determine the diagnostic precision of cyst fluid CEA in discriminating benign from malignant pancreatic cystic neoplasms. 2. Materials and methods This study was performed in compliance with the checklist provided by the Preferred Items for reporting Systematic Reviews and Meta-Analyses (PRISMA; Supplementary Table 1). 2.1. Search strategy
∗ Corresponding author at: Division of Gastroenterology and Hepatology, Mayo Clinic Florida, 4500 San Pablo Road, Jacksonville, FL 32224, USA. Tel.: +1 904 953 6982; fax: +1 904 953 7260. E-mail address: [email protected] (M. Raimondo).
A literature search of the MEDLINE and EMBASE databases was conducted for English-language studies published before October 2012. Systematic search strategies were designed by a consultant health science librarian. The following keywords were used: “pancreas OR pancreatic cystic lesion,” “tumour marker OR
1590-8658/$36.00 © 2013 Editrice Gastroenterologica Italiana S.r.l. Published by Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.dld.2013.05.002
S. Ngamruengphong et al. / Digestive and Liver Disease 45 (2013) 920–926
Carcinoembryonic Antigen OR CEA” and “diagnosis.” A manual search of the cross-references of identified studies and relevant reviews were used for completion of the literature search.
921
as a cause of heterogeneity. The analyses were carried out using the statistical software Meta-DiSc (version 1.4) [9]. 3. Results
2.2. Study eligibility and quality assessment Two reviewers (SN and MJB) independently reviewed each identified trial for fulfilment of predefined inclusion criteria. Studies that compared the diagnostic accuracy of cyst fluid CEA with histological diagnosis were included. Criteria for exclusion were (1) if the study authors did not apply specific cutoff value of cyst fluid CEA to evaluate the sensitivity and specificity in differentiating benign from malignant pancreatic cysts; (2) if there was insufficient information to construct 2 × 2 contingency tables. Articles by the same author or research group were identified for analysis only when different patient populations had been used. In cases of more than one published report on the same population or group of patients, we included only data from the largest number of patients for analysis. We requested additional information and clarification of data from five authors and two responded. All eligible studies identified by the search underwent methodological quality assessment using the 14-item quality assessment of diagnostic accuracy (QUADAS) tool [6]. Disagreement was resolved in a consensus meeting.
2.3. Outcomes of interest Outcomes of interest were the sensitivity, specificity, likelihood ratios, and area under the summary receiver–operating characteristic curves (sROC) of cyst fluid CEA with regards to diagnosis of malignant pancreatic cystic lesions. Malignant cysts were defined as either high grade dysplasia (HGD) or invasive cancer.
2.4. Statistical analysis From the reported data, we extracted sensitivity and specificity of cyst fluid CEA in distinguishing between benign and malignant pancreatic cysts by the use of 2 × 2 contingency tables of correct diagnosis and CEA level above the specified cutoff point. The pooled estimates of sensitivity, specificity, positive likelihood ratios (LR), negative LR, diagnostic odd ratios (DOR) and their 95% confidence intervals (95% CI) were calculated using random-effects model. sROC was performed to assess the interaction between sensitivity and specificity. We used DOR and area under the sROC curve (AUC) to analyze the diagnostic performance of cyst fluid CEA in distinguishing the benign from the malignant group. The DOR was calculated from sensitivity and specificity and was defined as [(true positive rate/false positive rate)]/[(1 − true positive rate/1 − false positive)]. A higher DOR reflected a greater diagnostic precision of cyst fluid CEA in the detection of malignant pancreatic cysts. The AUC represents an overall summary of test performance and a global measure of overall test accuracy. A perfect test will have an AUC close to 1 and poor tests have AUCs close to 0.5 [7,8]. Because the DOR varies across studies with different diagnostic threshold, the sROC curve is asymmetrical. Cochrane Q test and I2 test were used to assess the inter-study heterogeneity. Statistically significant heterogeneity was considered if p < 0.1 and I2 > 50%. Subgroup analyses were performed to assess the diagnostic accuracy of cyst fluid CEA in mucinous cystic neoplasm and intraductal papillary mucinous neoplasm (IPMN). Analysis included only high quality studies (QUADAS score ≥ 11). The type of pancreatic cyst (mucinous only versus mixed type) and study quality (QUADAS score ≥ 11 versus < 11) were included as covariates to examine its possibility
The search strategy identified 844 articles. Of these, 795 articles were excluded after preliminary abstract review and 49 articles underwent detailed full-text evaluation. From these, 8 published articles [10–17] met our inclusion criteria. The number of articles is outlined in Fig. 1 by reason for exclusion at each stage of the eligibility assessment. The reasons for exclusion of each individual article for the 57 articles following detail assessment were summarized in Appendix A. The study by Correa-Gallego et al. [13] was published as a letter to the editor. All other articles were published as full papers. Two studies [13,14] reported the data from Massachusetts General Hospital, Boston, USA. The study by Correa-Gallego et al. [13] evaluated the diagnostic performance of cyst fluid CEA in 72 patients with histologically confirmed IPMNs, whereas Cizginer et al. [14] studied 198 patients with various types of pancreatic cystic lesions, including 66 patients with IPMNs. We included these two studies in the meta-analysis as they were conducted by different authors and had different patient eligibility criteria, although the study population overlapped. However, we performed a post hoc analysis to assess the influence of each study on heterogeneity and summary estimate by including each of these two studies one at a time. There were a total of 504 patients in this analysis (315 with benign and 189 with malignant cysts). The characteristics of the included studies are shown in Table 1. Results from the application of the QUADAS 14-question quality assessment tool revealed that all studies scored ranged from 8 to 12 (Appendix B). Two studies were prospective and 6 were retrospective in design. The cutoff value of cyst fluid CEA used to differentiate benign and malignant cysts ranged from 109.9 to 6000 ng/mL. One study [10] reported the sensitivity and specificity of cystic fluid CEA at two different cutoff levels; 1000 and 6000 ng/mL. Because the DOR of cutoff level of 6000 ng/mL was higher, we used the sensitivity and specificity derived from this cutoff level in the meta-analysis. The pooled sensitivity of cyst fluid CEA level in prediction of malignant pancreatic cysts was 63% (95% CI, 55–69) and pooled specificity was 63% (95% CI, 57–68) (Fig. 2). The positive LR was 1.89 (95% CI, 1.13–3.16) and the negative LR was 0.62 (95% CI, 0.40–0.97). The DOR was 3.84 (95% CI, 1.37–10-74) (Fig. 3). Heterogeneity was observed (I2 = 76%, p < 0.0001). The AUC was 0.70 with an SE of 0.09 (Fig. 4). We performed a post hoc sensitivity analysis to evaluate the impact of including two studies from the overlapping study population (Massachusetts General Hospital, Boston, USA) [13,14]. The pooled sensitivity, specificity, and DOR were not significantly different among the two subgroups and the main analysis. The heterogeneity remained high. In the analysis including the study by Correa-Gallego et al. [13] the pooled sensitivity was 56% (95% CI, 47–65) and pooled specificity was 68% (95% CI, 61–74). The DOR was 4.10 (95% CI, 1.09–15.47). Heterogeneity was observed (I2 = 79%, p < 0.0001). In the analysis including the alternate study by Cizginer et al. [14] instead of the Correa-Gallego et al. study [13], the pooled sensitivity was 64% (95% CI, 56–71) and pooled specificity was 68% (95% CI, 62–74). The DOR was 5.19 (95% CI, 1.87–14.39). Heterogeneity was observed (I2 = 69%, p = 0.003). 3.1. Subgroup analysis and sensitivity analysis A subgroup analysis to evaluate the diagnostic performance of cyst fluid CEA in only patients with mucinous cysts (mucinous
922
Table 1 Characteristics of included studies. Male %
Mean age, range (± SD), years
Size of lesion in mm, range, SD
No of patient with malignant lesion (%): definition of malignant cysts (n)
Surgical pathology (n)
Method of calculating cutoff level of CEA
Cutoff level of CEA (ng/mL)
12
63
19 (30%)
67
19.5 (5–74)
16 (36%): HGD (9), invasive cancer (7)
Literature
≥6000
Retrospective
12
41
14 (34%)
64, range 31–72
NA
10 (24%): HGD (5), invasive cancer (5)
ROC curve
>200
Linder et al. 2006; USA [12] Kucera et al. 2012; USA [17]
Prospective
12
71*
42 (41%)
NA
18–102
14 (20%): MCCa (14)
11
47
32 (68%)
66, range 24–88
24 ± 14 (for Br-IPMN)
21 (44%): HGD (14), invasive cancer (7)
Optimal cutoff from included patients Literature
≥6000
Retrospective
Correa-Gallego et al. 2009; USA [13] Cizginer et al. 2011; USA [14]
Retrospective
9
72
NA
NA
NA
Literature
≥200
Prospective
12
77 (38.8%)
60.6
NA
17 (23%): CIS (12), invasive cancer (5) 55 (35%): CIS, invasive cancer
MCN (6), M-IPMN (9), Br-IPMN (25), mixed-IPMN (4) IPMN (41): M-IPMN (2), Br-IPMN (13), mixed-IPMN (26) MCN (21). MCCa (14), PC (23), SCN (13) IPMN (47): M-IPMN (11), Br-IPMN (24), mixed-IPMN (12) IPMN (72)
ROC curve
>109.9
Pais et al. 2007; USA [15]
Retrospective
10
74***
38 (51%)
Median age 65, range 41–84
Literature
>800
Wu et al. 2007; China [16]
Retrospective
8
85
26 (30%)
Cystadenoma: 50, range 17–85. cystadenocarcinoma: 56, range 27–83.
Malignant group: 60 mm. Benign group: 4.7 mm (p = 0.37) NA
MCN (43), IPMN (66), PC, SCN (8), PNEC (7), others(4) IPMN (74): M-IPMN (22), Br-IPMN (19), mixed-IPMN (33) Cystadenoma (37), cystadenocarcinoma (48)
NA
NA
Study design
Othman et al. 2012; USA [10]
Retrospective
Maire et al. 2008; France [11]
QUADAS scores
198**
21 (28%): invasive cancer (8**** ) 48 (56%): Cystadenocarcinoma (48)
>200
CEA: carcinoembryonic antigen; CIS: carcinoma in situ; HGD: high grade dysplasia; IPMN: intraductal papillary mucinous neoplasm; M-IPMN: main duct IPMN; Br-IPMN: branch duct IPMN; MCCa: mucinous cystadenocarcinoma; MCN: mucinous cystic neoplasm; PC: pseudocyst; PNEC: pancreatic neuroendocrine cyst; QUADAS: quality assessment of diagnostic accuracy; ROC: receiver–operating characteristic curve; SCN: serous cyst; NA: not available. * 35 patients with mucinous cysts. ** 157 patients with CEA analysis. *** 23 patients with CEA analysis. **** 8 patients had fluid CEA tested.
S. Ngamruengphong et al. / Digestive and Liver Disease 45 (2013) 920–926
Number of patients
Study author; country
S. Ngamruengphong et al. / Digestive and Liver Disease 45 (2013) 920–926
923
Fig. 1. Articles selection diagram. CEA: carcinoembryonic antigen, ERCP: endoscopic retrograde cholangio-pancreatography.
cystic neoplasm and IPMN) was performed. Five studies involving 227 patients reported sensitivity and specificity of CEA in prediction of malignancy among mucinous cystic lesions [10,11,13,15,17], the pooled sensitivity of cyst fluid CEA level was 65% (95% CI, 57–73) and pooled specificity was 66% (95% CI, 59–72). The DOR was 4.74 (95% CI, 1.46–15.37). The AUC was 0.69 with an SE of 0.12. In subgroup analysis including only 5 studies with higher quality (QUADAS ≥ 11) [10–12,14,17], the pooled sensitivity was 69% (95% CI, 59–77); pooled specificity was 63% (95% CI, 59–70), and DOR was 5.15 (95% CI, 1.38–19.28). There was still significant heterogeneity among studies. Sensitivity analysis was performed to assess the possible sources of heterogeneity across the studies by weight meta-regression analysis of DOR with the following covariates as predictor variables: type of pancreatic cyst (mucinous only versus mixed type) and study quality (QUADAS score ≥ 11 versus < 11). Both type of pancreatic cyst and study quality did not have statistically significant effect on the overall DOR. 4. Discussion Cyst fluid CEA has proven to be a helpful marker in differentiating between mucinous and non-mucinous pancreatic cysts [4,14,17,18]. Whether elevated cyst fluid CEA is predictive of malignant cyst remains controversial. This information is crucial because mucinous lesions with no worrisome features (e.g. presence of symptoms, solid component within cyst, mural nodule), particularly in high surgical risk patients, do not require immediate surgical resection and may be monitored closely for signs of
malignant transformation [5]. In this review, pooled accuracy estimates of cyst fluid CEA in pancreatic cystic lesions and subgroup analysis including only mucinous cysts fell short of achieving the optimal sensitivity and specificity values. Therefore the usefulness in differentiating benign from malignant pancreatic cysts was limited. Our findings suggested that clinical decision making, particularly to perform surgery, should not base solely on a degree of cyst fluid CEA elevation. Our findings support the current guidelines which do not recommend the use of cyst fluid CEA to diagnose malignant pancreatic cysts [5,19]. Although, initial studies have suggested that high cyst fluid CEA is a predictor of malignant pancreatic cysts [11,12], several subsequent studies have reported that the level of cyst fluid CEA concentrations between benign and malignant cystic lesions do overlap [10,13,15,17,18,20]. In fact, Kucera et al. [17] reported the relationship of cyst fluid CEA concentration and the degree of dysplasia; the mean cyst fluid CEA concentration increased as histologic grade progressed from low grade to high grade dysplasia. However, the cyst fluid CEA declined markedly once invasive cancer had developed. The underlying pathophysiological mechanisms of this change are unknown. The role of CEA in the follow-up of patients with pancreatic cysts is even debated. Although, the recently revised guidelines recommend follow-up CEA level in patients with non-resected IPMN [5], the available evidence does not support this statement. Several studies on long-term follow-up patients with pancreatic cysts, including IPMN, who were managed conservatively found no correlation between cyst fluid CEA level and cyst growth [10,20]. The conclusion of our study strengthens the previous findings to
924
S. Ngamruengphong et al. / Digestive and Liver Disease 45 (2013) 920–926
Fig. 2. Forest plot of sensitivity and specificity estimates for cyst fluid carcinoembryonic antigen in predicting malignant pancreatic cysts for all studies.
limit the use of cyst fluid CEA in differentiating mucinous from non-mucinous pancreatic cyst, and forgo its use for prediction of malignant pancreatic cysts or monitoring non-surgical patients. Other than cyst tumour marker, several techniques have been used to improve detection of malignant pancreatic cysts. Certain cyst features in imaging studies were highly associated with malignancy. In a Meta analyses, the pooled odds ratios for malignancy of cyst size > 3 cm, the presence of mural nodules and dilatation of the main pancreatic duct was 62.4 (95%CI 30–126), 9.3 (95%CI 5.3–16.1), and 7.27 (95%CI 3.0–17.4), respectively [21]. Thus, these high risk characteristics should be carefully considered in the differential diagnosis between benign and malignant pancreatic cysts, mainly in the setting of IPMN. Recently, using a cytological
threshold of high-grade atypical epithelial cells as a predictor of malignancy in mucinous cyst has been shown to improved sensitivity from 29% to 72% with specificity of 85% [22]. Moreover, the identification of high-grade atypical epithelial cells in aspirated cystic fluid from a presumed benign small branch duct IPMN because of the absence of high risk radiological features improved the detection of malignancy by 50% [23]. Studies of molecular analysis of the cyst fluid indicated that the detection of KRAS and GNAS mutations can be helpful in predicting malignancy in a pancreatic cystic lesion [24,25]. However, the utility of these tests have not yet been established in routine clinical practice. As with any meta-analysis, our study has limitations. There was significant heterogeneity among the studies. Although we
Fig. 3. Forest plot of diagnostic odds ratio of cyst fluid carcinoembryonic antigen in predicting malignant pancreatic cysts for all studies. OR: odds ratio.
S. Ngamruengphong et al. / Digestive and Liver Disease 45 (2013) 920–926
925
Fig. 4. The summary receiver–operating characteristic curve for cyst fluid carcinoembryonic antigen in predicting malignant pancreatic cysts. The area under curve represents the overall diagnostic accuracy. Each circle represents a study in the meta-analysis. Sample size is indicated by the size of the circle. sROC: summary receiver–operating characteristic, AUC: area under curve, SE: standard error.
attempted to address heterogeneity by pre-specified subgroup analysis and meta-regression, significant heterogeneity still exists. The heterogeneity that we were unable to account for could be due to minor variations in patient populations, methods of sampling, techniques used to assay the samples and proportion of patient with malignant disease. Also, small sample sizes in the included studies could potentially be subject to selection bias. It should be noted that a single published study by Kucera et al. [17] reported that the degree of CEA elevation is different between HGD and invasive cancer. If this finding is true, then defining a “malignant cyst” as HGD and invasive cancer in most of the studies might be inappropriate. This could be subjected to spectrum bias i.e. a phenomenon that tests sensitivity and specificity may vary depending on the severity of disease in the population in which the test has been applied [26]. Because none of the included studies reported the diagnostic accuracy results from patients with HGD separately from the group with invasive cancer, we were unable to perform the analysis to validate this observation. In addition, studies with negative results were likely to be unpublished and this could lead to publication bias. Statistical tests for funnel plot asymmetry, such as the Egger and Begg tests that are often used in meta-analyses of randomized clinical trials, are not effective to detect potential publication bias in diagnostic accuracy [27]. Even though publication bias does exist, the overall diagnostic precision would have been lower than our reported results. To the best of our knowledge, the validation of the assay methods used to measure the CEA level in the cyst fluid matrix has never been described in the literature. The composition of the cystic fluid is different in cysts of variable nature and therefore the method of validation of the fluid cyst matrix can be different from the serum and complex to be applied for a general laboratory. Thus, the cutoff values for CEA in the cystic fluid can be of limited value and difficult to compare in different studies. However, the overall low DOR still suggests that the clinical utility of cyst fluid CEA should not be used in this setting. Because of very limited availability of information, we did not include studies that evaluated the diagnostic performance of cyst fluid CEA
combined with other tumour marker, such as CA 72.4, in this analysis. In conclusion, we found that the level of cyst fluid CEA does not differentiate benign from malignant pancreatic cysts and by itself should not be used for guiding surgical decision making. The clinical value of cyst fluid CEA should be limited only to distinguish mucinous from non-mucinous cystic lesions. Large, multicentric, well-designed trials are needed to further characterize the role of cyst fluid tumour marker and molecular analysis in the evaluation of pancreatic cysts. Financial disclosures No financial support was received for this study. Conflict of interest The authors declare no conflict of interest. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.dld.2013.05.002. References [1] de Jong K, Nio CY, Hermans JJ, et al. High prevalence of pancreatic cysts detected by screening magnetic resonance imaging examinations. Clinical Gastroenterology and Hepatology 2010;8:806–11. [2] Cellier C, Cuillerier E, Palazzo L, et al. Intraductal papillary and mucinous tumors of the pancreas: accuracy of preoperative computed tomography, endoscopic retrograde pancreatography and endoscopic ultrasonography, and long-term outcome in a large surgical series. Gastrointestinal Endoscopy 1998;47:42–9. [3] Baba T, Yamaguchi T, Ishihara T, et al. Distinguishing benign from malignant intraductal papillary mucinous tumors of the pancreas by imaging techniques. Pancreas 2004;29:212–7. [4] Brugge WR, Lewandrowski K, Lee-Lewandrowski E, et al. Diagnosis of pancreatic cystic neoplasms: a report of the cooperative pancreatic cyst study. Gastroenterology 2004;126:1330–6.
926
S. Ngamruengphong et al. / Digestive and Liver Disease 45 (2013) 920–926
[5] Tanaka M, Fernandez-del Castillo C, Adsay V, et al. International consensus guidelines 2012 for the management of IPMN and MCN of the pancreas. Pancreatology 2012;12:183–97. [6] Whiting P, Rutjes AW, Reitsma JB, et al. The development of QUADAS: a tool for the quality assessment of studies of diagnostic accuracy included in systematic reviews. BMC Medical Research Methodology 2003;3:25. [7] Walter SD. Properties of the summary receiver operating characteristic (SROC) curve for diagnostic test data. Statistics in Medicine 2002;21:1237–56. [8] Rosman AS, Korsten MA. Application of summary receiver operating characteristics (sROC) analysis to diagnostic clinical testing. Advances in Medical Sciences 2007;52:76–82. [9] Zamora J, Abraira V, Muriel A, et al. Meta-DiSc: a software for meta-analysis of test accuracy data. BMC Medical Research Methodology 2006;6:31. [10] Othman MO, Patel M, Dabizzi E, et al. Carcino embryonic antigen and longterm follow-up of mucinous pancreatic cysts including intraductal papillary mucinous neoplasm. Digestive and Liver Disease 2012;44:844–8. [11] Maire F, Voitot H, Aubert A, et al. Intraductal papillary mucinous neoplasms of the pancreas: performance of pancreatic fluid analysis for positive diagnosis and the prediction of malignancy. American Journal of Gastroenterology 2008;103:2871–7. [12] Linder JD, Geenen JE, Catalano MF. Cyst fluid analysis obtained by EUS-guided FNA in the evaluation of discrete cystic neoplasms of the pancreas: a prospective single-center experience. Gastrointestinal Endoscopy 2006;64:697–702. [13] Correa-Gallego C, Warshaw AL, Fernandez-del Castillo C. Fluid CEA in IPMNs: a useful test or the flip of a coin? American Journal of Gastroenterology 2009;104:796–7. [14] Cizginer S, Turner B, Bilge AR, et al. Cyst fluid carcinoembryonic antigen is an accurate diagnostic marker of pancreatic mucinous cysts. Pancreas 2011;40:1024–8. [15] Pais SA, Attasaranya S, Leblanc JK, et al. Role of endoscopic ultrasound in the diagnosis of intraductal papillary mucinous neoplasms: correlation with surgical histopathology. Clinical Gastroenterology and Hepatology 2007;5:489–95. [16] Wu H, Yan LN, Cheng NS, et al. Role of cystic fluid in diagnosis of the pancreatic cystadenoma and cystadenocarcinoma. Hepato-Gastroenterology 2007;54:1915–8.
[17] Kucera S, Centeno BA, Springett G, et al. Cyst fluid carcinoembryonic antigen level is not predictive of invasive cancer in patients with intraductal papillary mucinous neoplasm of the pancreas. Journal of Oncology Practice 2012;13:409–13. [18] Park WG, Mascarenhas R, Palaez-Luna M, et al. Diagnostic performance of cyst fluid carcinoembryonic antigen and amylase in histologically confirmed pancreatic cysts. Pancreas 2011;40:42–5. [19] Khalid A, Brugge W. ACG practice guidelines for the diagnosis and management of neoplastic pancreatic cysts. American Journal of Gastroenterology 2007;102:2339–49. [20] Nagula S, Kennedy T, Schattner MA, et al. Evaluation of cyst fluid CEA analysis in the diagnosis of mucinous cysts of the pancreas. Journal of Gastrointestinal Surgery 2010;14:1997–2003. [21] Anand N, Sampath K, Wu BU. Cyst features and risk of malignancy in intraductal papillary mucinous neoplasms of the pancreas: a meta-analysis. Clinical Gastroenterology and Hepatology 2013. [22] Pitman MB, Genevay M, Yaeger K, et al. High-grade atypical epithelial cells in pancreatic mucinous cysts are a more accurate predictor of malignancy than “positive” cytology. Cancer Cytopathology 2010;118:434–40. [23] Genevay M, Mino-Kenudson M, Yaeger K, et al. Cytology adds value to imaging studies for risk assessment of malignancy in pancreatic mucinous cysts. Annals of Surgery 2011;254:977–83. [24] Wu J, Matthaei H, Maitra A, et al. Recurrent GNAS mutations define an unexpected pathway for pancreatic cyst development. Science Translational Medicine 2011;3:92ra66. [25] Khalid A, Zahid M, Finkelstein SD, et al. Pancreatic cyst fluid DNA analysis in evaluating pancreatic cysts: a report of the PANDA study. Gastrointestinal Endoscopy 2009;69:1095–102. [26] Goehring C, Perrier A, Morabia A. Spectrum bias: a quantitative and graphical analysis of the variability of medical diagnostic test performance. Statistics in Medicine 2004;23:125–35. [27] Deeks JJ, Macaskill P, Irwig L. The performance of tests of publication bias and other sample size effects in systematic reviews of diagnostic test accuracy was assessed. Journal of Clinical Epidemiology 2005;58:882–93.