Pancreatic cystic neoplasms (PCNs) are increasingly found in clinical practice, due to an aging population and the routine use of high-quality abdominal imaging.1 The importance of PCNs is related to malignant potential, high frequency, and significant morbidity and mortality of surgical treatment. Therefore, there is an urgent need to find noninvasive and reliable markers of malignant and high-risk premalignant PCNs.
In clinical practice, after clinical and imaging findings of a potentially significant lesion, including mucinous premalignant or malignant cysts, endoscopic ultrasound with fine-needle aspiration (EUS-FNA) for cystic fluid analysis for carcinoembryonic antigen (CEA) and cytology became standard in decision making. Carcinoembryonic antigen is the most accurate for diagnosing mucinous cysts, which are premalignant lesions, whereas cytology is highly specific for malignancy diagnosis.2 Treatment options, including surgery, follow-up, or no additional evaluation, rely on imaging and pancreatic cystic fluid (PCF) analysis, but a significant part remains indeterminate, with approximately one-third of preoperative diagnosis being incorrect.3,4
In this clinical context, pancreatic cyst fluid analysis for molecular markers has shown that KRAS mutations may be specific for mucinous cysts5–7 and that simultaneous KRAS/GNAS mutations are specific of intraductal papillary mucinous neoplasms (IPMNs).8,9 Currently, next-generation sequencing (NGS), a very sensitive technique for detection of genetic mutations, can be considered in indeterminate PCNs or if it modifies patient management.10 Numerous studies have shown that DNA molecular analysis of aspirates obtained by EUS-FNA provides a better characterization of PCNs compared with current methods used in clinics.11–19 However, these studies have generally included a limited number of patients, and results are not consistent among studies. Currently, the integration of molecular analysis in routine clinical practice is still a matter of debate.
We therefore performed a systematic review and meta-analysis of all previous studies with KRAS mutational analysis performed by NGS in PCF obtained preoperatively by EUS-FNA. All samples with a surgical pathology as reference standard for diagnosis were evaluated. Our aim was to investigate the accuracy of KRAS mutational analysis for diagnosis of mucinous and significant (mucinous and malignant) PCNs and compare it with routine standard diagnosis, with CEA and cytology.
MATERIALS AND METHODS
Search Strategy and Eligibility Criteria
The systematic review and meta-analysis reported here were conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines,20 and the protocol was registered at PROSPERO (CRD42018097268). A comprehensive search of databases, including MEDLINE, Scopus, Web of Science, and SCIELO, for the past 18 years (January 1, 2000, to March 31, 2018) and restricted to human studies was performed. No language restrictions were applied. The following search terms were used: “pancreas,” “cyst,” “molecular,” and “analysis.” Additional search of related articles and hand search of references of all selected studies were performed, adding additional publications.
Published studies were included in the meta-analysis if they (1) analyzed mutational analysis of KRAS using highly sensitive techniques, such as NGS; (2) analyzed a cohort of patients with pancreatic cysts and symptomatic or incidental findings; (3) cysts were evaluated by EUS-FNA with PCF analysis; and (4) all patients had a definitive diagnosis with a surgically resected specimen.
Exclusion criteria were as follows: (1) studies on molecular markers other than KRAS mutation; (2) studies involving solid pancreatic lesions; (3) studies performed in PCF not obtained by EUS-FNA; (4) studies with cytology and clinical information as standard criterion of diagnosis without a surgical pathology specimen as reference standard; and (5) reviews, case reports, letters to editor, exploratory studies, and articles published only in abstract form.
Two reviewers (S.F. and A.L.) independently judged study eligibility, and disagreements were resolved by consensus.
Histological Criteria and Tests Under Investigation
Based on the World Health Organization tumor classification, PCN diagnosis were reviewed and classified into 1 of 3 groups: (1) malignant cysts (adenocarcinoma or high-grade dysplasia in IPMNs and mucinous cystic neoplasms, secondary cystic adenocarcinomas, and cystic pancreatic neuroendocrine tumors); (2) premalignant mucinous cysts (IPMNs and mucinous cystic neoplasms with low- or intermediate-grade dysplasia); and (3) benign cysts (serous cystadenomas, pseudocysts, and other benign cysts).
The index test was molecular analysis with KRAS mutation, because it is the most frequent mutation. The comparators were as follows: (1) CEA (cutoff value >192 ng/mL) for diagnosis of mucinous cystic lesions and (2) cytology that was considered positive if samples were read as atypical, suspicious, positive, or malignant. Cytology was considered negative if samples were read as indeterminate, acellular, or negative for malignancy. It should be noted that a diagnosis of atypia in a cytological evaluation does not warrant a malignancy diagnosis requiring surgery.
The primary outcome was to assess the diagnostic accuracy of KRAS mutation in PCF for diagnosis of malignant and significant PCNs. The secondary outcome was to compare the accuracy of KRAS mutation with current standard of diagnosis, with PCF analysis for CEA and cytology, in malignant and significant PCNs.
Data Extraction and Quality Assessment
Selected articles' data were extracted independently by 2 reviewers (S.F. and A.L.), who were blinded to publication details, onto a predefined worksheet. Disagreements were discussed and reviewed by a third reviewer (L.P.).
Data extraction included the name of first author, publication year, study design (prospective, cross-sectional, retrospective), sample size (all patients included in the study), number of patients referred for surgery (surgical cohort), number of malignant lesions, distribution of cyst types (malignant, premalignant, benign), number of patients with a CEA of greater than 192 ng/mL, a positive cytology, and KRAS mutation detection.
Methodological quality of primary studies included was assessed by 2 authors (S.F. and A.L.) using the modified QUADAS-2 (Quality Assessment Tool for Diagnostic Accuracy Studies version 2) tool,21 which evaluates the quality of articles for systematic reviews of diagnostic accuracy studies in 4 domains, including patient selection, index test, reference standard, and flow and timing, for risk of bias and applicability concerns.
Statistical Analysis and Data Synthesis
Our reference standard was surgical specimen that classified PCNs into 3 groups: malignant, premalignant, and benign cysts. This resulted in a 2 × 3 table: positive or negative test result in each of the 3 groups, for each of the 3 tests, KRAS (index test), cytology, and CEA (comparator tests).
To calculate test accuracy and to reflect the categories that are used in clinical practice and guide management, we constructed 2 × 2 tables, to evaluate the ability of the index test and comparator tests to discriminate malignant from nonmalignant (all cysts except those proven to be malignant) and significant (proven malignant and premalignant cysts) from nonsignificant cysts (proven benign cysts).
The data of the 2 × 2 tables were used to calculate sensitivity and specificity for each study. We present individual study results graphically by plotting the estimates of sensitivity and specificity (and their 95% confidence intervals [CIs]) in both forest plots and on the summary receiver operating characteristic (sROC) curve plots. The area under the curve (AUC) is equal to the probability that if a pair of relevant and nonrelevant cysts is selected at random, the relevant cyst will have a higher or positive test result than the nonrelevant cyst. Pooled estimates of the sensitivity and specificity were obtained by DerSimonian-Laird method (random-effects model) to incorporate variation among studies, when data are heterogeneous.
Heterogeneity was investigated in the first instance through visual examination of forest plots of sensitivities and specificities and through visual examination of the ROC plot of the raw data. Last, we used the χ2 test to evaluate if the differences across the studies were greater than expected by chance alone. A low P value suggested presence of heterogeneity. In addition, we used the statistic I2 of Higgins that allowed us to quantify the amount of heterogeneity.22,23 The scale of I2 has a range of 0% to 100% and values of 25%, 50%, and 75% are considered low, moderate, and high heterogeneity, respectively.
To analyze the publication bias in meta-analyses of sensitivity and specificity, we used Deeks' test. This test, developed for diagnostic test accuracy (DTA), is the least biased and is recommended in the Cochrane Handbook for Systematic Reviews of Diagnostic Test Accuracy.24,25
We used Meta-DiSc (version 1.4; Meta-analysis of Diagnostic and Screening Tests)26 for assessment of diagnostic yield of the studies and SPSS Statistics (version 23; IBM Corp, Armonk, NY) for Deeks' test.
Search Results and Characteristics of the Studies Included
Our search found 496 study titles and abstracts. Figure 1 describes the selection process of the articles included in this study.
After abstract screening and full-text review, 16 studies met the inclusion criteria and were considered suitable for qualitative and quantitative analyses. The design was retrospective in 13 and prospective in 3 articles, with 3 studies published from years 2005 to 2009, 8 from years 2010 to 2014, and 5 from years 2015 to 2018.
These 16 studies included a total of 3429 patients, of which 731 (21%) underwent surgical resection and had a surgical pathology specimen available as reference standard and were included for analysis. Patients of studies in which data were available for the overall series (aggregating results of surgical and clinical surveillance cohorts) but were not discretely available for the surgical cohort were excluded from analysis.
The characteristics of the studies, surgical pathology diagnoses, and cystic fluid analysis details are in Table 1.
Results of methodological quality of the primary studies included are presented in Figure 2, which was sketched with templates available at www.quadas.org. All studies included in this review showed a “low-risk” classification, as the index test (KRAS mutation analysis) and the reference standard (surgical pathology specimen) were reliable and mentioned in all studies. However, a “high risk” of selection bias was demonstrated in a patient and in flow and timing because only a small proportion of the patients evaluated in all studies, except 1, were included in the analysis. In fact, many patients were excluded in all studies as the inclusion criterion requiring surgical pathology as diagnostic reference was not met. Applicability concerns regarding patient selection were also significant in all studies, because the subgroup of PCNs referred for surgery is more often malignant than for patients with pancreatic cysts on clinical surveillance, which would also be targeted with this review.
Fourteen articles were included in the meta-analysis for diagnostic accuracy of KRAS mutation. For each of the 2 definitions of relevant cyst, forest plots of sensitivity and specificity with heterogeneity denoted are shown in Figure 3.
The definitions of malignant and significant cysts resulted in a different range of specificity and sensitivity of the studies included. In the first case, both sensitivity and specificity varied from 0% to 100% and in the second case sensitivity varied from 12% to 100% and specificity from 50% to 100%. In the first subgroup, the wide range of sensitivity was largely due to chance variation because of small numbers of patients with the target condition (proven malignant cysts) in the different studies (median, 6; range, 1–31). For instance, if there was only 1 patient with a proven malignant cyst in a study, and this patient had a positive test, the sensitivity would be 100%, but if he/she had a negative test result, the sensitivity would be 0%. Small numbers of patients with nonmalignant cysts in some studies (median, 10; range, 1–111) also led to a wide range of specificity.
For each of the 2 subgroups, there occurred a moderate heterogeneity in sensitivity (I2 = 46.8% vs I2 = 65.0%) and specificity (I2 = 52.5% vs I2 = 34.3%), and therefore random-effects models were used. In malignant cysts, the pooled sensitivity was 0.43 (95% CI, 0.34–0.53), and the pooled specificity was 0.62 (95% CI, 0.56–0.68). In significant cysts, the sensitivity was 0.46 (95% CI, 0.42–0.51) with a specificity of 0.97 (95% CI, 0.92–0.99).
Figure 4 displays the sROC curves of KRAS analysis, showing the sensitivity of the individual articles mapped on the vertical scale, 1-specificity on the horizontal scale, and summary (sensitivity, 1-specificity) point marked, as well as the sROC curve and the confidence region for the summary (sensitivity, 1-specificity) points. The area under the sROC curve ± SE was 0.5551 ± 0.0659 in malignant cysts and 0.5290 ± 0.1424 in significant cysts. The results of the studies had greater variation in malignant cysts, as shown by the wide confidence region.
The median prevalence of malignant cysts and significant cysts was 29.6% and 82.5%, respectively (range, 9.1%–85.7% and 50%–100%, respectively). This prevalence was based on the proportion of proven malignant and proven significant cysts in the studies.
Twelve articles were included in the meta-analysis for diagnostic accuracy of cytology. Figure 5 shows the forest plots of sensitivity and specificity for the 2 defined subgroups of cysts. The forest plots for cytology show variable sensitivities within the articles, from 0 to 1, which can be due to the small numbers of patients with the target condition in some studies.
In the malignant and significant cysts groups, respectively, there were 6 and 2 studies, respectively, that recorded sensitivity of cytology as scoring greater than or equal to 0.5.
For each of the 2 subgroups, there existed heterogeneity in sensitivity (I2 = 69.2% vs I2 = 60.4%) and specificity (I2 = 64.9% vs I2 = 27.7%), and therefore random-effects models were used. In malignant cysts, the pooled sensitivity was 0.37 (95% CI, 0.27–0.48), and the pooled specificity was 0.96 (95% CI, 0.93–0.98). In significant cysts, the sensitivity was 0.19 (95% CI, 0.13–0.25) with a specificity of 0.94 (95% CI, 0.86–0.98).
The results were plotted as a symmetrical sROC curve (Fig. 4). The area under the sROC curve ± SE was 0.7788 ± 0.1309 in malignant and 0.4805 ± 0.1542 in significant cysts.
The median prevalence of malignant and significant cysts was 29.4% and 86.4%, respectively (range, 9.1%–85.7% and 50%–100%, respectively).
CEA Greater Than 192 ng/mL
Eight articles were included in the meta-analysis for diagnostic accuracy of CEA. Because only 4 articles (with few patients) allowed the evaluation of accuracy of CEA greater than 192 ng/mL for diagnosis of malignant cysts, we restricted the analysis to significant cysts. Figure 6 shows the forest plots of sensitivity and specificity. The forest plots for cytology showed a sensitivity from 0.5 to 0.82 and a specificity from 0.5 to 1.
There existed homogeneity in sensitivity (I2 = 21.2%) and specificity (I2 = 0%). The pooled sensitivity was 0.58 (95% CI, 0.52–0.65), and the pooled specificity was 0.90 (95% CI, 0.76–0.97).
The area under the sROC curve ± SE was 0.6903 ± 0.1228.
The median prevalence of significant cysts was 89.7% (range, 81.6%–100%).
Regression analyses of funnel plots were not statistically significant (P > 0.05), suggesting that publication bias was not a major determinant.
In this systematic review and meta-analysis, we performed a comparative analysis of the current standard tests in PCF obtained by EUS-FNA (CEA and cytology) and molecular analysis (KRAS mutation) in PCNs. The comparative analysis included all studies, evaluating the 3 tests separately.
Our meta-analysis is the largest published and included 731 patients, all with molecular analysis performed by NGS preoperatively, and all patients with a surgical pathology specimen as reference standard for diagnosis. We analyzed these 3 markers, for diagnosis of significant as compared with benign cysts and for diagnosis of malignant versus nonmalignant cysts, because relevant clinical decisions apply to these categories.
The comparative analysis of KRAS, cytology, and CEA for cyst diagnosis showed that cytology alone had the highest accuracy (AUC = 0.7788) for the diagnosis of malignant cysts, and CEA, the highest accuracy (AUC = 0.6903) for the diagnosis of significant cysts. KRAS mutational analysis had the worst performance for both groups of lesions with AUC = 0.551 for malignant and AUC = 0.46 for significant cysts. The specificity of KRAS for diagnosis of significant cysts was high (97%), which makes it useful to diagnose these lesions, but because of low sensitivity (46%), KRAS should not be used to exclude the diagnosis, as false-negative (FN) results are common. Similar results for KRAS were previously published by Guo et al,36 who analyzed several molecular tests for improving differential diagnosis of PCNs.
As DNA testing continues to evolve, questions remain about its accuracy, how it influences patient management, and in what order it should be performed to better support clinical decisions. Previous studies6 have shown that DNA testing combined with clinical features increased correct PCNs diagnosis compared with either one. With the multiple recent advances in biomarkers, particularly DNA-based mutations, molecular genetics will probably prove to be useful in management of PCNs.37 In a previous meta-analysis, cytology in preoperative diagnosis of PCNs has shown low sensitivity for diagnosis,38 recommending additional tests to improve diagnosis. Another published meta-analysis evaluating diagnostic accuracy of EUS-FNA with CEA and cytology in differentiating mucinous cysts has demonstrated to be accurate to confirm the diagnosis but performs poorly in excluding it.39 The role of KRAS as individual screening test has also been analyzed before40 with poor accuracy and added benefit coming from a combined approach with cytology. Finally, a recently published meta-analysis supporting KRAS, GNAS, and RNF43 mutations as diagnostic markers of IPMNs41 used different methods for mutation detection, different tumor materials, and clinicopathologic data as reference standard for diagnosis, which may limit its clinical application in pancreatic cystic lesions, in which mutational analysis is performed solely in cystic fluid.
In our study, the pooled sensitivities of KRAS, cytology, and CEA, besides being limited, also varied considerably. On the other hand, specificity was uniformly high for the tests analyzed, particularly for KRAS and CEA for diagnosis of significant cysts and cytology for both malignant and significant cysts.
By estimating the pooled sensitivities, we sought to determine which of the tests had a better performance.
For a group of 100 patients with a pancreatic cyst and a prevalence of malignant cysts of 30%, the presence of a KRAS mutation would diagnose 13 (true-positive [TP]) and miss 17 (FN), and 27 (false-positive [FP]) would be unnecessarily operated. For a prevalence of significant cysts of 86%, 40 would be correctly diagnosed (TP), 46 would be missed by KRAS (FN), and none would be unnecessarily referred for surgery/surveillance (FP).
With respect to cytology, a positive result for the diagnosis of malignant cysts in a group of 100 patients with a prevalence of 30% of malignant cysts would diagnose 11 (TP) and would miss 19 (FN) patients, and 3 (FP) would be unnecessarily referred for surgery. For significant cysts, with a prevalence of 86%, a positive cytology would diagnose 16 (TP) PCNs and would miss 70 (FN) PCNs, and none (FP) would be unnecessarily referred for surgery/surveillance.
If 100 patients with PCNs evaluated with a CEA of greater than 192 ng/mL in PCF and a prevalence of significant cysts of 86%, 52 (TP) would be diagnosed by CEA, and 36 (FN) would be missed by the test, with 1 (FP) that would be unnecessarily referred for surgery/follow-up.
Although both KRAS and CEA are useful for mucinous cyst diagnosis that were classified in this meta-analysis as significant, based on our results we can conclude that CEA would miss fewer PCNs (lower FN rate) with only 1 FP. Concerning malignancy diagnosis, cytology is the best diagnostic test, because although KRAS mutation can diagnose more malignant cysts (13 vs 11), it would have significantly higher numbers of FP diagnosis (27 vs 1). We can conclude that KRAS mutation is not better than CEA for significant cyst diagnosis and that cytology is the most accurate test for malignancy diagnosis.
However, we should remember that in routine clinical practice a major pitfall for PCNs diagnosis is the frequently scant volume of PCF obtained, precluding routine PCF testing. As mutation analysis requires less volume of PCF, it may be an alternative test in these circumstances. This major advantage of molecular analysis was not possible to evaluate because the volume of cystic fluid obtained was not available in most studies analyzed.
Additionally, combining KRAS mutation with conventional testing increased the sensitivity of PCN diagnosis without compromising specificity. We extracted data from the studies analyzed in this meta-analysis to evaluate the added value of KRAS in conjunction with cytology and CEA, but the available data were limited to 4 studies27,28,32,34 (Table 1), making the analysis inconclusive.
The strengths of our work are the use of strict exclusion criteria, with all analyzed patients with an analyzed surgical pathology as the reference standard and avoiding bias related to methodological limitations of the studies evaluated. We chose to include only patients with a surgical pathology as the reference standard because histopathology is the criterion standard for diagnosis of neoplasia. This is an important strength of our systematic review and provides a more realistic and accurate estimate for the index and comparative tests evaluated. In previous studies of accuracy of cytology including both surgical pathology and clinical follow-up39 as reference standard, pooled sensitivities were 12% higher than in studies with exclusive surgical pathology40 as reference standard in the diagnosis of mucinous cysts, with overestimation of test accuracy.
Limitations of this study include incomplete reporting in DTA in primary studies, with no separate information for distinction of malignant and mucinous cysts in 2 studies5,31 and in another 2 studies for distinction of benign and premalignant mucinous cysts.29,33 These 4 studies were included in the group of 7 studies with more patients analyzed in the meta-analysis. Another limitation is the time elapsed between the index tests and the reference standard. The final diagnosis could have been made at different time intervals from the tests. If the time between index tests and reference standard is too long, the true diseased status of the patient may have changed by the time the reference standard was assessed. Finally, the low number of malignant cysts per study (0–13), except for 4 studies,5,7,9,30 may contribute to part of the heterogeneity in the sensitivity observed.
With the increasing diagnosis of asymptomatic PCNs, some with malignant potential, there is a growing need to find accurate biomarkers of malignancy in these lesions, to reduce surgeries on benign cysts and still diagnose and resect early malignant lesions with favorable prognosis. DNA molecular markers, particularly KRAS mutation, which is an early event in pancreatic carcinogenesis, have the potential to fulfill this need, but clinicians should be aware of their current limitations in diagnostic performance and type of lesions identified.
Certainly, the significant costs, logistic difficulties in collecting and preserving material for future molecular analysis in busy general hospitals, and the technical complexity of the test make its generalized use difficult in clinical practice. Moreover, large multicenter validation studies are still missing.
Additionally, there is a need for more trials to confirm their clinical relevance in patient outcomes, such as early cancer diagnosis, number of surgeries of benign lesions avoided, and prognostic value in numerous cysts that require periodic surveillance.
Moreover, for successful massive implementation of molecular markers in pancreatic cyst clinics, a validation of KRAS mutation as a complementary test to patients with an unavailable CEA level and a nondiagnostic cytology will be insufficient. Its development as a universal, highly accurate, first-line test with clinical impact in cyst diagnosis and patient management will be required. Next-generation sequencing reliably allows analysis of multiple gene panels both in PCF and peripheral blood and offers an attractive option to increase the accuracy of molecular analysis in diagnosis and risk stratification of these lesions.42
Finally, with current evidence, KRAS can only be recommended as a second-line test in the case that CEA and cytology of PCF are nondiagnostic. It would be useful to determine the additional value of the KRAS in combination with the other tests and to evaluate the adequate order of the tests, in order to maximize the diagnoses of malignant and/or significant cysts.
The intended use and clinical role of KRAS mutational analysis in the present should be limited to patients with an undefined CEA level and a nondiagnostic cytology, serving only as a complementary diagnostic test due to its limited accuracy. KRAS has lower diagnostic accuracy than CEA and cytology and should not replace standard EUS-FNA analysis. Clinicians should be aware of a significant rate of FP results of KRAS mutation if the diagnosis of a malignant cyst is under consideration.
1. Stark A, Donahue TR, Reber HA, et al. Pancreatic cyst
disease: a review. JAMA
2. Cizginer S, Turner BG, Bilge AR, et al. Cyst fluid carcinoembryonic antigen is an accurate diagnostic marker of pancreatic mucinous cysts. Pancreas
3. Correa-Gallego C, Ferrone CR, Thayer SP, et al. Incidental pancreatic cysts: do we really know what we are watching? Pancreatology
4. Cho CS, Russ AJ, Loeffler AG, et al. Pre-operative classification of pancreatic cystic neoplasms: the clinical significance of diagnostic inaccuracy. Ann Surg Oncol
5. Kadayifci A, Al-Haddad M, Atar M, et al. The value of KRAS
mutation testing with CEA
for the diagnosis of pancreatic mucinous cysts. Endosc Int Open
6. Springer S, Wang Y, Dal Molin M, et al. A combination of molecular markers and clinical features improve the classification of pancreatic cysts. Gastroenterology
7. Singhi AD, Mcgrath K, Brand RE, et al. Preoperative next-generation sequencing of pancreatic cyst
fluid is highly accurate in cyst classification and detection of advanced neoplasia. Gut
8. Wu J, Matthaei H, Maitra A, et al. Recurrent GNAS
mutations define an unexpected pathway for pancreatic cyst
development. Sci Transl Med
9. Singhi AD, Nikiforova MN, Fasanella KE, et al. Preoperative GNAS
testing in the diagnosis of pancreatic mucinous cysts. Clin Cancer Res
10. European Study Group on Cystic Tumours of the Pancreas. European evidence-based guidelines on pancreatic cystic neoplasms. Gut
11. Khalid A, McGrath KM, Zahid M, et al. The role of pancreatic cyst
fluid molecular analysis in predicting cyst pathology. Clin Gastroenterol Hepatol
12. Schoedel KE, Finkelstein SD, Ohori NP. K-ras and microsatellite marker analysis of fine-needle aspirates from intraductal papillary mucinous neoplasms of the pancreas. Diagn Cytopathol
13. Sreenarasimhaiah J, Lara LF, Jazrawi SF, et al. A comparative analysis of pancreas cyst fluid CEA
and histology with DNA mutational analysis in the detection of mucin producing or malignant cysts. JOP
14. Shen J, Brugge WR, Dimaio CJ, et al. Molecular analysis of pancreatic cyst
15. Sawhney MS, Devarajan S, O'Farrel P, et al. Comparison of carcinoembryonic antigen and molecular analysis in pancreatic cyst
fluid. Gastrointest Endosc
16. Khalid A, Zahid M, Finkelstein SD, et al. Pancreatic cyst
fluid DNA analysis in evaluating pancreatic cysts: a report of the PANDA study. Gastrointest Endosc
17. Talar-Wojnarowska R, Pazurek M, Durko L, et al. A comparative analysis of K-ras mutation and carcinoembryonic antigen in pancreatic cyst
18. Chai SM, Herba K, Kumarasinghe MP, et al. Optimizing the multimodal approach to pancreatic cyst
fluid diagnosis: developing a volume-based triage protocol. Cancer Cytopathol
19. Mertz H. K-ras mutations correlate with atypical cytology
and elevated CEA
levels in pancreatic cystic neoplasms. Dig Dis Sci
20. Moher D, Shamseer L, Clarke M, et al. Preferred reporting items for systematic review and meta-analysis protocols (PRISMA-P) 2015 statement. Syst Rev
21. Whiting PF, Rutjes AW, Westwood ME, et al. QUADAS 2—a revised tool for the quality assessment of diagnostic accuracy studies. Ann Intern Med
22. Higgins JP, Thompson SG, Deeks JJ, et al. Measuring inconsistency in meta-analyses. BMJ
23. Higgins JP, Thompson SG. Quantifying heterogeneity in a meta-analysis. Stat Med
25. 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. J Clin Epidemiol
26. Zamora J, Abraira V, Muriel A, et al. Meta-DiSc: a software for meta-analysis of test accuracy data. BMC Med Res Methodol
27. Toll AD, Kowalski T, Loren D, et al. The added value of molecular testing in small pancreatic cysts. JOP
28. Panarelli NC, Sela R, Schreiner AM, et al. Commercial molecular panels are of limited utility in the classification of pancreatic cystic lesions. Am J Surg Pathol
29. Rockacy MJ, Zahid M, McGrath KM, et al. Association between KRAS
mutation, detected in pancreatic cyst
fluid, and long-term outcomes of patients. Clin Gastroenterol Hepatol
30. Nikiforova MN, Khalid A, Fasanella KE, et al. Integration of KRAS
testing in the diagnosis of pancreatic cystic lesions: a clinical experience of 618 pancreatic cysts. Mod Pathol
31. Al-Haddad M, DeWitt J, Sherman S, et al. Performance characteristics of molecular (DNA) analysis for the diagnosis of mucinous pancreatic cysts. Gastrointest Endosc
32. Kung JS, Lopez OA, McCoy EM, et al. Fluid genetic analyses predict the biological behavior of pancreatic cysts: three-year experience. JOP
33. Winner M, Sethi A, Poneros JM, et al. The role of molecular analysis in the diagnosis and surveillance of pancreatic cystic neoplasms. JOP
34. Jones M, Zheng Z, Wang J, et al. Impact of next-generation sequencing on the clinical diagnosis of pancreatic cysts. Gastrointest Endosc
35. Singhi AD, Zeh HJ, Brand RE, et al. American Gastroenterological Association guidelines are inaccurate in detecting pancreatic cysts with advanced neoplasia: a clinicopathologic study of 225 patients with supporting molecular data. Gastrointest Endosc
36. Guo X, Zhan X, Li Z. Molecular analyses of aspirated cystic fluid for the differential diagnosis of cystic lesions of the pancreas: a systematic review and meta-analysis. Gastroenterol Res Pract
37. Kulzer M, Singhi AD, Furlan A, et al. Current concepts in molecular genetics and management guidelines for pancreatic cystic neoplasms: an essential update for radiologists. Abdom Radiol (NY)
38. Thosani N, Thosani S, Qiao W. Role of EUS-FNA
in the diagnosis of mucinous pancreatic cystic lesions: a systematic review and meta-analysis. Dig Dis Sci
39. Thornton GD, McPhail MJ, Nayagam S, et al. Endoscopic ultrasound guided fine needle aspiration for the diagnosis of pancreatic cystic neoplasms: a meta-analysis. Pancreatology
40. Gillis A, Cipollone I, Cousins G, et al. Does EUS-FNA
molecular analysis carry additional value when compared to cytology
in the diagnosis of pancreatic cystic neoplasm? A systematic review. HPB (Oxford)
41. Lee JH, Kim Y, Choi JW, et al. KRAS
, and RNF43
mutations in intraductal papillary mucinous neoplasm of the pancreas: a meta- analysis. Springerplus
42. de Biase D, Visani M, Acquaviva G, et al. The role of next-generation sequencing in the cytologic diagnosis of pancreatic lesions. Arch Pathol Lab Med