Fax: (011) 61-08-9346-4122
Optimizing the multimodal approach to pancreatic cyst fluid diagnosis †
Developing a volume-based triage protocol
Article first published online: 7 SEP 2012
Copyright © 2012 American Cancer Society
Volume 121, Issue 2, pages 86–100, February 2013
How to Cite
Chai, S. M., Herba, K., Kumarasinghe, M. P., de Boer, W. B., Amanuel, B., Grieu-Iacopetta, F., Lim, E. M., Segarajasingam, D., Yusoff, I., Choo, C. and Frost, F. (2013), Optimizing the multimodal approach to pancreatic cyst fluid diagnosis . Cancer Cytopathology, 121: 86–100. doi: 10.1002/cncy.21226
See editorial on pages 57–60, this issue.
- Issue published online: 5 FEB 2013
- Article first published online: 7 SEP 2012
- Manuscript Accepted: 18 JUN 2012
- Manuscript Revised: 10 JUN 2012
- Manuscript Received: 5 APR 2012
- pancreatic cyst;
- carcinoembryonic antigen;
- KRAS mutation;
- cytologic techniques;
- cyst fluid analysis;
- molecular analysis
The objective of this study was to develop a triage algorithm to optimize diagnostic yield from cytology, carcinoembryonic antigen (CEA), and v-Ki-ras2 Kirsten rat sarcoma viral oncogene homolog (KRAS) testing on different components of a single pancreatic cyst fluid specimen. The authors also sought to determine whether cell block supernatant was suitable for CEA and KRAS testing.
Fifty-four pancreatic cysts were triaged according to a volume-dependent protocol to generate fluid (neat and supernatant) and cell block specimens for cytology, comparative CEA, and KRAS testing. Follow-up histology, diagnostic cytology, or a combined clinicopathologic interpretation was recorded as the final diagnosis.
There were 26 mucinous cystic lesions and 28 nonmucinous cystic lesions with volumes ranging from 0.3 mL to 55 mL. Testing different components of the specimens (cell block, neat, and/or supernatant) enabled all laboratory investigations to be performed on 50 of 54 cyst fluids (92.6%). Interpretive concordance was observed in 17 of 17 cases (100%) and in 35 of 40 cases (87.5%) that had multiple components tested for CEA and KRAS mutations, respectively. An elevated CEA level (>192 ng/mL) was the most sensitive test for the detection of a mucinous cystic lesion (62.5%) versus KRAS mutation (56%) and “positive” cytology (61.5%). KRAS mutations were identified in 2 of 25 mucinous cystic lesions (8%) in which cytology and CEA levels were not contributory.
A volume-based protocol using different components of the specimen was able to optimize diagnostic yield in pancreatic cyst fluids. KRAS mutation testing increased diagnostic yield when combined with cytology and CEA analysis. The current results demonstrated that supernatant is comparable to neat fluid and cell block material for CEA and KRAS testing. Cancer (Cancer Cytopathol) 2013. © 2012 American Cancer Society.
Currently, a significant proportion of pancreatic cysts are an incidental finding in patients undergoing high-resolution imaging.1 The increasing number of these lesions being detected coupled with the widespread availability and expertise with endoscopic ultrasound-guided fine-needle aspiration (EUS-FNA) has led to an increase in the number of pancreatic cyst fluid (PCF) specimens submitted to cytopathology laboratories. Separation of non-neoplastic from neoplastic pancreatic cysts, which may require follow-up or surgical management, is central to good patient outcome and effective use of medical resources.
Cytologic evaluation in isolation is hampered by the low quantity and quality of cells in these often scant samples, resulting in reduced sensitivity for the detection of neoplastic cysts, specifically, mucinous cysts such as intraductal papillary mucinous neoplasms (IPMNs) and mucinous cystic neoplasms (MCNs).2-4 The introduction of molecular analysis of PCF specimens to the traditional combination of carcinoembryonic antigen (CEA) and cytologic examination has led to a reappraisal of how these samples are triaged and the comparative roles of these different modalities.5, 6 Some investigators have suggested preferential molecular analysis of PCF in low-volume samples (<1 mL), which are less likely to yield diagnostic information from cytologic examination and CEA analysis.4, 7
The objective of this study was to optimize the multimodal diagnosis of pancreatic cystic lesions (PCLs) in our laboratory by assessing whether v-Ki-ras2 Kirsten rat sarcoma viral oncogene homolog (KRAS) mutation testing increases diagnostic yield and whether it can be performed reliably on either supernatant (SN) and/or cell block (CB) material rather than a neat fluid sample. To do this, we used a volume-dependent protocol for triaging these specimens in the cytopathology laboratory and used CB material for KRAS mutation testing, the resultant SN fluid for both CEA and KRAS mutation testing, and compared this with neat fluid analysis.
MATERIALS AND METHODS
Fifty-four consecutive patients with PCL (from 2010 to 2011) who underwent EUS-FNA and had PCF specimens submitted to the cytopathology laboratory were evaluated. Follow-up information was obtained from clinical review appointments during the study period and through follow-up of PathWest Department of Anatomical Pathology electronic records for histologic diagnosis of pancreatic resection specimens. The study was approved by the institutional review board.
Triage of Specimens
The PCF specimens were received unaltered (neat) by the cytopathology laboratory and triaged according to a volume-dependent research protocol specified in Figure 1A for CEA analysis, cytology, and molecular pathology. Samples ≤1.0 mL were centrifuged, and any resultant deposit was used to manufacture a CB for histologic analysis and KRAS testing. The SN fluid was used for CEA analysis and, volume permitting, comparative KRAS testing. For samples between 1 mL and 5 mL, 0.5 mL of neat fluid was submitted for CEA analysis and, in lower volume samples, such as those <1.5 mL, any remaining fluid was centrifuged to generate SN fluid and a deposit. On the basis of the size of the deposit generated, this material was used to produce 1) Papanicolaou-stained smears and 2) a CB for comparative CEA and KRAS testing. For samples in which the deposit was deemed insufficient for the production of both components, CBs were preferentially manufactured. The SN fluid from these samples was used for comparative CEA analysis and, volume permitting, KRAS testing. For higher volume samples, both SN and neat fluids were routinely submitted for CEA analysis and KRAS testing, as specified in Figure 1A.
Cell Block Preparation
Cellular material was concentrated by centrifugation (at 2000 revolutions per minute for 10 minutes) in an Eppendorf centrifuge 5810 (Eppendorf, Sydney, New South Wales, Australia). After removal of the SN fluid, an equal volume of plasma was added to the remaining cell deposit. This cell suspension was mixed with 0.25 mL of thrombin solution to form a clot, which was then fixed in 10% buffered neutral formalin for subsequent processing in the histopathology laboratory. After this, 4-μm sections of tissue were cut from the paraffin-embedded blocks, transferred onto glass slides, and stained with hematoxylin and eosin.
Carcinoembryonic Antigen Analysis
A minimum of 0.15 mL is required for CEA analysis on the automated ARCHITECT analyzer (Abbott Diagnostics, Sydney, New South Wales, Australia). Despite the low volume required for automated analysis, a minimum of 0.5 mL was submitted whenever possible to allow for sample preparation and retesting if required. All samples were routinely analyzed with and without dilution to ensure there was no analytic interference, and the average of these 2 results was reported. Viscous samples were analyzed in a “1-in-10” dilution on the automated analyzer using a phosphate-buffered saline solution. The minimum reportable level was 1 ng/mL, and interassay coefficients of variation were 11%, 4.4%, and 3.9% at CEA concentrations of 1 ng/mL, 12 ng/mL, and 256 ng/mL, respectively. A value >192 ng/mL was considered supportive of a mucinous cystic lesion (MCL), as previously reported.3
KRAS Mutation Testing
Bidirectional dideoxy sequencing and fluorescent single-strand conformation analysis (F-SSCA) of exon 2, codons 12 and 13 of the KRAS gene were performed to identify the most common mutations.
Fluorescent single-strand conformation analysis
A PCR-based F-SSCA method was used to screen for mutations in exon 2 of the KRAS gene as described previously.8 Briefly, HEX-labeled fluorescent primers (Geneworks, Adelaide, South Australia, Australia) were designed to span the known codon 12/13 mutation hotspot and give rise to a PCR product of 110 base pairs in size. PCR was carried out in a final volume of 15 μL and comprised a mix of 1 times reaction buffer, 0.2 mM deoxynucleotide triphosphate, 3.0 mM MgCl2, 0.5 μM of each primer, 0.4 U Taq Polymerase (Qiagen, Melbourne, Victoria, Australia), and approximately 20 ng of DNA template.
Three microliters of PCR product were heated to 94°C for 5 minutes. One microliter of that mixture was loaded onto a nondenaturing 10% polyacrylamide/2% glycerol gel and run on a Gel Scan 3000 DNA fragment analyzer according to the manufacturer's instructions (Corbett Research, Sydney, New South Wales, Australia). DNA samples that displayed bands additional to the wild-type bands were classified as carrying a mutation. Negative and positive controls comprised of DNA with known KRAS mutations were run with each gel.
Bidirectional dideoxy sequencing
Automated dideoxy DNA sequencing was performed at the Lotterywest Biomedical Facility (Genomics Department of Clinical Immunology, Royal Perth Hospital, Perth, Western Australia, Australia). The forward and reverse primers (Geneworks) were designed to span the known exon 2, codon 12/13 mutation hotspot and give rise to a PCR product of 263 base pairs in size. PCR was carried out in a final volume of 25 μL and comprised a mix of 1 times reaction buffer, 0.2 mM deoxynucleotide triphosphates, 3.0 mM MgCl2, 0.5 μM of each primer, 0.8 U Taq polymerase (Qiagen), and approximately 20 ng of DNA template. The PCR products were purified using the QIAquick PCR purification kit (Qiagen), and 2 ng were used as a template for sequencing using the BigDye Terminator version 3.1 Cycle Sequencing Kit (Applied Biosystems, Melbourne, Victoria, Australia). The 3100 Genetic Analyzer (Applied Biosystems) was used for capillary electrophoresis and sequence analysis. All samples were sequenced in both directions.
The cyst location and FNA route (transduodenal or transgastric) were noted for each patient to facilitate the interpretation of contaminant versus lesional epithelium. The presence of sheets or fragments of epithelial cells with nonmucinous cytoplasm, a recognizable brush border, and interspersed goblet cells was indicative of contaminant duodenal epithelium.9 Contaminant gastric epithelium was characterized by flat sheets of epithelial cells with uniform basal nuclei, apical mucin caps, and the absence of a brush border.9 The quantity and quality (thin or thick “colloid-like”) of the mucin and the presence or absence of inflammatory cells and macrophages were used to determine whether this was lesional or contaminant in nature.9, 10
MCLs, serous cystadenomas, pancreatic neuroendocrine tumors (PENs), and pseudocysts (PCs) were diagnosed according to conventional cytologic criteria, as specified in Table 1 and illustrated in Figure 2.9 No attempt was made to separate IPMNs and MCNs because of the lack of reproducible cytologic criteria for distinguishing between the 2 entities.2, 9-11
|Cyst Type||Clinical Presentation||EUS Findings||Cytology||CEA, ng/mL|
|Pseudocyst||Any age, M=W; preceding episode of acute pancreatitis or trauma||Distributed throughout pancreas; unilocular and thick walled with internal debris||Macrophages, some containing hemosiderin or hematoidin pigment, inflammatory cells, proteinaceous cyst background||<192|
|SCA||Aged >65 y, W>M||Distributed throughout pancreas. Microcystic: innumerable, 1-5 mm cysts with central fibrous nidus and calcification; larger cysts observed peripherally; oligocystic: less common, lobulated, unilocular cyst||Small, flat sheets or clusters of uniform, cuboidal cells with round nuclei, even chromatin, and small amounts of pale-to-clear cytoplasm; hemosiderin-laden macrophages and lymphocytes in a clean background||<192 (Often <5 ng/mL)|
|MCL (MCN and IPMN) without malignancy||MCN: Ages 40-50 y, W>M; IPMN: Aged >65 y, M>W||MCN: Usually located body/tail; multiloculated, thick-walled, egg-shell rim calcification with no pancreatic duct communication; IPMN: Main duct (mucin extrusion from papilla, diffuse dilation of the main pancreatic duct, papillary projections into the main pancreatic duct) or branch duct (most commonly located in head, uncinate, and neck; “cluster of grapes” appearance, communication with main pancreatic duct)||Extracellular mucin, mucinous epithelium with varying degrees of atypia; branching aggregates, small clusters, and occasional papillary structures; no necrosis in the background.||≥192|
|MCL (MCN and IPMN) with malignancy||MCN: Patients with associated invasive carcinoma, on average, are 5-10 y older than those with noninvasive disease, W>M; IPMN: Patients with associated invasive carcinoma, on average, are 3-5 y older than those with noninvasive disease, M>W||Findings similar to above in addition to more extensive thickening of the wall, presence of a mural nodule, or adjacent solid mass||Extracellular mucin, mucinous epithelium with varying degrees of cytological and architectural atypia; increase in nuclear-to-cytoplasmic ratio, nuclear irregularity, chromatin abnormalities, and nucleolar prominence; necrosis and inflammation in the background||≥192|
|PEN with cystic degeneration||Any age, W>M||Well demarcated, hypoechoic mass with cystic change||Dispersed and aggregated cells with round-to-oval nuclei, “salt-and-pepper” chromatin, and delicate to finely granular cytoplasm||<192|
Endoscopic Ultrasound Assessment
EUS criteria for distinguishing MCLs from non-MCLs and IPMNs from MCNs have been described previously and are listed in Table 1.3, 4 Briefly, MCLs were more likely to have macrocystic septations or adjacent solid areas in contrast to non-MCLs, which often were unilocular, honeycombed, or thick walled. Intraductal papillary mucinous neoplasms were more likely to be located in the head of the pancreas and to exhibit dilatation of the main or side pancreatic duct, in contrast to MCNs, which were more likely to be located in the pancreatic body/tail, to have peripheral calcifications, and to have no communication with the ductal system. Gastrointestinal tract contamination during EUS-FNA was minimized by aspiration of gastric and duodenal contents before passage of the needle down the endoscope, withdrawal of the needle stylet only after cyst wall puncture, and expelling any trapped contents from the needle before aspiration.
Excisional histopathology or cytology that was diagnostic of 1) an MCL, 2) a specific non-MCL, or 3) a malignancy with EUS evidence of a cyst was considered definitive for classification. For cases in which these criteria were not fulfilled, a combined clinicopathologic diagnosis was generated from clinical details, imaging findings (including EUS), cytology, and CEA results. The clinicopathologic diagnosis was generated without knowledge of the KRAS mutation status. Criteria used for classification of the various cysts are presented in Table 1 with correlative illustrations in Figure 2.
Sensitivity, specificity, positive predictive value, and negative predictive value were used to assess the dichotomous classification models. Differences in categorical data were assessed by using the chi-square test and the Fisher exact test as necessary. Continuous variables were analyzed using the Mann-Whitney test. Statistical significance was determined at P < .05. All statistical analyses were conducted in the statistical software package R (version 2.11.1; R Foundation for Statistical Computing, Vienna, Austria).
Overall, there were 54 samples collected from 52 patients (27 women and 25 men). On follow-up histopathology, diagnostic cytology, or clinicopathologic diagnosis, there were 26 MCLs (48.1%), 16 PCs (29.6%), 9 serous cystadenomas (16.7%), 2 ductal retention cysts (3.7%), and 1 cystic PEN (1.9%). The patients ranged in age from 8 years to 85 years (mean age, 62 years). Patients with MCLs were older than those with non-MCLs (mean age, 66 years and 58 years, respectively; P = .1273). There was no significant difference between the number of men and women in the MCL and non-MCL groups.
Pancreatic Cyst Fluid Volumes
Specimen volumes ranged from 0.3 mL to 55 mL (Table 2). Although the mean volume of fluid for non-MCLs (9.0 mL) was greater than that for MCLs (3.2 mL), this difference was not statistically significant (P = .1856). When PCs were isolated and compared as a group with MCLs, the difference in mean volume (12.9 mL vs 3.2 mL) became statistically significant (P = .0106). Samples ≤1 mL accounted for 9 of 26 MCLs (34.6%) and 7 of 28 non-MCLs (25%), and 5 of the latter cysts were serous in nature.
All 3 laboratory investigations (cytology, CEA, and KRAS mutation testing) were completed in 50 of 54 PCF specimens (92.6%). Of the 4 patient who had an incomplete set of investigations, 2 samples were unsuitable for biochemistry, and 2 samples did not generate an adequate quantity of DNA for KRAS mutation testing.
Carcinoembryonic Antigen Analysis
Cyst fluid CEA was obtained in 52 of 54 patients (96.3%). In 27 patients, there was a reading generated from both SN and neat components. The higher of the CEA-SN and CEA neat (CEA-N) values designated as CEA final (CEA-F) was used for further data analysis. CEA values ranged from <1 ng/mL to 359,570 ng/mL. The median CEA-F in patients with MCLs was significantly higher than that in patients with non-MCLs (290 ng/mL vs 30 ng/mL, respectively; P = .0001). This relation was maintained for both neat and SN fluids (Table 2). Fifteen of 24 patients with MCLs (62.5%) had CEA values >192 ng/mL, and 25 of 28 patients (89.3%) with non-MCLs had CEA levels ≤192 ng/mL, resulting in an overall sensitivity of 62.5% and specificity of 89.3% for the detection of MCL. It is noteworthy that 2 of 9 patients with MCL who had CEA levels ≤192 ng/mL had readings of 171 ng/mL and 180 ng/mL, whereas the remaining 7 patients had CEA levels that ranged from 1 ng/mL to 70 ng/mL. A CEA level >192 ng/mL was reported in 1 PC and in 1 ductal retention cyst. The first of these 2 patients was a man aged 42 years who had a history of recurrent pancreatitis and a pancreatic head cyst that had the cytologic features of a PC. The patient had recurrent bouts of pancreatitis after the initial EUS-FNA; and, upon surveillance, the cyst reduced in size. The ductal retention cyst was an incidental finding in a woman aged 40 years and was characterized on EUS examination on 2 separate occasions as a unilocular uncinate cyst. Cytology demonstrated small numbers of degenerate epithelial cells with cyst debris. Subsequent resection and histopathology revealed a unilocular cyst lined by cuboidal to flat, nonmucinous epithelial cells without papillae, ovarian-type stroma, or dysplasia. Table 3 presents the sensitivity, specificity, and positive and negative predictive values of the various laboratory investigations.
|Age: Mean, y||66||58||.1273|
|Sample volume: Mean, mL||3.2||9.0||.1856|
|Median CEA, ng/mL|
|Percentage (95% CI)|
|Variable||CEA-F >192 ng/mL||Positive for KRAS Mutation||Cytologya|
|Sensitivity||62.5 (40.6-81.2)||56.0 (34.9-75.6)||61.5 (40.7-79.1)|
|Specificity||89.3 (71.8-97.7)||100.0 (87.2-100)||100.0 (85-100)|
|PPV||83.3 (58.6-96.4)||100.00 (76.8-100)||100.0 (75.9-100)|
|NPV||73.5 (55.6-87.1)||69.23 (54.1-84.6)||73.7 (56.6-86)|
Six of 9 patients with (66.7%) serous cystadenoma had a CEA <1 ng/mL, whereas the other 3 had readings of 3 ng/mL, 4 ng/mL, and 6 ng/mL. Only 2 other patients with non-MCL (1 PC and 1 cystic PEN) had values within this range (both <1 ng/mL).
Of the 27 patients who had SN and neat fluid available to compare, 17 produced different results from the 2 components (Table 4). In 13 patients, CEA-N yielded a higher value; and, in 4 patients, CEA-SN had a higher value. None of these differences in any individual patient affected the final interpretation of the result.
|Case No.||Supernatant||Neat||Difference (%)|
KRAS Mutation Testing
Mutation analysis for KRAS was possible in 52 of 54 patients (96.3%); and 2 patients had insufficient DNA for assessment (1 MCL and 1 serous cystadenoma). Fourteen patients who had KRAS mutations in codon 12 were identified, and all occurred in MCLs (14 of 25 lesions; 56%). There were no false-positive KRAS results. KRAS mutations were associated significantly with MCLs (P < .0001). In 12 of 52 patients (23.1%), testing was performed on the CB only; in 15 of 52 patients (28.8%), 2 components were tested (5 neat + CB, 7 SN + CB, 3 SN + neat); and, in 25 of 52 patients (48.1%), all 3 components were tested. Thirty-five of the 40 patients who had multiple components tested had concordant results. In 5 patients, differing results were obtained from the various components, as specified in Table 5. A representative DNA sequence analysis from fluid and CB samples for 1 discordant case is provided in Figure 3.
|Case No.||CPD||Age, y||Sex||Cyst Location||EUS||Cytology||Histology/ Clinical FU||Volume, mL||CEA-F, ng/mL||Components Tested for KRAS||KRAS Status||Mutation Type||Concordance Between Tested Components|
|1||MCL (MCN)||79||W||Tail||MCN||MCL||MCN||Not tested||Not tested||N, CB||MUT||Gly12Asp||Discordant (N−/CB+)|
|2||MCL (IPMN)||79||M||Head||IPMN||MCL-Mal||Stable on FU EUS||2.5||50||CB||MUT||Gly12Val||NA|
|3||MCL (Mal)||70||W||Tail||MCN-Mal||MCL-Mal||Metastatic disease on FU PET scan||4||359,570||CB||MUT||Gly12Val||NA|
|4||PC||58||M||Tail||PC||INT||Re-drained on FU EUS||55||160||N, SN||WT||NA||Concordant|
|6||PC||64||M||Body/tail||PC||PC||Reduction in size on MRI||8||100||N, SN, CB||WT||NA||Concordant|
|7||MCL (Mal)||66||W||Head||MCL-Mal||MCL-Mal||Metastatic disease on FU CT||3.5||Not tested||N, SN, CB||WT||NA||Concordant|
|8||PC||43||M||Head||PC||PC||Recurrent pancreatitis 1yr after initial EUS||Not tested||360||SN, CB||WT||NA||Concordant|
|9||MCL (IPMN)||52||W||Body||IPMN||ND||Stable on FU EUS||1||10||N, CB||MUT||Gly12Arg||Concordant|
|10||MCL (IPMN)||61||M||Tail||IPMN||ND||Stable on FU EUS||5||171||N, SN||MUT||Gly12Arg||Concordant|
|11||MCL (IPMN)||72||W||Head||IPMN||INT||Stable on FU EUS and CT||10||34||N, SN, CB||WT||NA||Concordant (fluids-/insufficient DNA on CB)|
|12||PC||74||M||Head||PC||PC||Not available||10||120||N, SN, CB||WT||NA||Concordant|
|13||MCL (IPMN)||74||M||Tail||IPMN||INT||Stable on FU EUS||5||81,300||N, SN, CB||WT||NA||Concordant|
|14||PC||58||M||Tail||PC||INT||Increase in size on FU CT||50||110||N, SN, CB||WT||NA||Concordant|
|15||MCL (MCN)||67||M||Body||MCN||INT||MCN||5||18,000||N, SN, CB||MUT||Gly12Cys||Discordant (N−/SN−/CB+)|
|17||PC||8||M||Body||PC||PC||Cyst resolved on FU CT||25||<1||N, SN, CB||WT||NA||Concordant|
|18||PC||73||M||Body||PC||PC||Deceased secondary to complications of pancreatitis||5||47||N, SN, CB||WT||NA||Concordant|
|19||MCL||68||M||Tail||MCL||MCL||Persistent high-risk features on FU EUS||1||2200||CB||MUT||Gly12Cys||NA|
|20||SCA||62||M||Uncinate||SCA||ND||Stable on FU EUS and CT||2||<1||N, CB||WT||NA||Concordant|
|21||MCL (Mal)||53||M||Tail||MCN-Mal||MCL-Mal||Metastatic disease on FU PET||0.4||4000||CB||MUT||Gly12Asp||NA|
|22||MCL (Mal)||82||M||Head||IPMN-Mal||ND||Not available||1.5||9||SN, CB||MUT||Gly12Val (SN), Gly12Asp (CB)||Concordant|
|23||MCL (Mal)||65||W||Tail||MCL-Mal||ND||Not available||2||200||SN, CB||MUT||Gly12Asp||Discordant (SN−/CB+)|
|24||SCA||74||W||Head||SCA||SCA||Stable on FU MRI||0.5||<1||CB||WT||0||NA|
|25||MCL (IPMN)||79||W||Head and tail||IPMN||MCL||Not available||0.5||5700||CB||Insufficient DNA||NA||NA|
|26||MCL (IPMN)||59||W||Body||IPMN||INT||IPMN||0.3||360||SN, CB||MUT||Gly12Val||Discordant (SN+/CB−)|
|27||PC||64||M||Uncinate||PC||PC||Not available||25||33||N, SN, CB||WT||NA||Concordant|
|28||MCL (Mal)||51||W||Tail||MCL-Mal||MCL-Mal||Metastatic disease at presentation||3||23,990||N, SN, CB||MUT||Gly12Val||Concordant|
|29||PC||66||W||Tail||PC||PC||Cyst resolved on FU CT||1||125||N, CB||WT||NA||Concordant|
|30||MCL (IPMN)||68||W||Head||IPMN||MCL||Stable on FU MRI||5||200||N, SN, CB||WT||NA||Concordant|
|31||PC||30||M||Head||PC||INT||Not available||3||15||N, SN, CB||WT||NA||Concordant|
|32||MCL (Mal)||77||M||Head, body, and tail||IPMN-Mal||MCL||Metastatic disease on FU imaging||1||620||CB||WT||NA||NA|
|33||MCL (IPMN)||75||W||Neck||IPMN||ND||Not available||0.3||340||CB||WT||NA||NA|
|34||MCL (IPMN)||68||M||Neck||IPMN||MCL||Stable on FU MRI||2||180||SN, CB||WT||NA||Concordant|
|36||PC||79||W||Tail||PC||PC||Severe complications post pancreatitis||4||65||N, SN, CB||WT||NA||Concordant|
|37||MCL (MCN)||40||W||Body/tail||MCN||MCL||MCN||6||34,900||N, SN, CB||WT||NA||Concordant|
|38||RC||39||W||Uncinate||MCN||INT||RC||4.5||1500||N, SN, CB||WT||NA||Concordant|
|39||MCL (IPMN)||75||M||Head||IPMN||MCL||Stable on FU CT||1||1||SN, CB||WT||NA||Concordant|
|40||RC||39||W||Uncinate||MCN||ND||RC||3||1940||N, SN, CB||WT||NA||Concordant|
|41||PEN||68||W||Tail||MCN||PEN||Concurrent chemotherapy for metastatic ovarian carcinoma||8||<1||N, SN, CB||WT||NA||Concordant|
|42||MCL (IPMN)||58||M||Tail||PC||MCL||Stable on FU EUS and imaging||5.5||70||N, SN, CB||MUT||Gly12Asp||Concordant (N+/SN+/insufficient DNA on CB)|
|43||PC||61||W||Body||PC||PC||Cyst resolved on FU imaging. Second PC detected||2||60||N, SN, CB||WT||NA||Concordant|
|44||MCL (Mal)||67||W||Head/body||IPMN-Mal||MCL-Mal||Locally advanced on presentation||13||8900||N, SN, CB||MUT||Gly12Asp||Discordant (N+/SN+/CB−)|
|45||SCA||76||W||Tail||SCA||INT||Stable on FU CT||1||6||CB||WT||NA||NA|
|46||SCA||53||W||Uncinate||SCA||INT||Not available||1||3||SN, CB||Insufficient DNA||NA||NA|
|47||MCL (IPMN)||61||W||Head||IPMN||MCL||Cyst smaller but with thickened wall on FU EUS||0.4||240||CB||WT||NA||NA|
|49||PC||38||M||Body||PC||PC||Stable on FU MRI||15||40||N, SN||WT||NA||Concordant|
|50||PC||42||W||Body||PC||PC||Reduction in size on FU CT||5||110||N, SN, CB||WT||NA||Concordant|
|51||SCA||65||W||Tail||SCA||SCA||SCA||2||<1||N, SN, CB||WT||NA||Concordant|
|52||MCL (MCN)||59||W||Tail||MCN||ND||Not available||2||25||SN, CB||WT||NA||Concordant|
|53||PC||27||W||Tail||PC||PC||Stable on FU MRI||3||18||N, SN, CB||WT||NA||Concordant|
|54||SCA||76||W||Head||SCA||SCA||Not available||6||<1||N, SN, CB||WT||NA||Concordant|
KRAS Mutation-Negative Mucinous Cystic Lesions
Eleven of 25 patients with MCL (44%) were negative for KRAS mutations. Of those 11 lesions, 4 were identified by both cytology and elevated CEA, 3 were identified by cytology alone, and 2 were identified by elevated CEA alone. Two cysts with EUS features of branch duct-type IPMN (BD-IPMN) were negative on cytology and CEA analysis. Follow-up computed tomography and EUS examinations 12 months after the initial EUS-FNA revealed consistent features of BD-IPMN in both cysts.
KRAS Mutation-Positive Mucinous Cystic Lesions
Fourteen of 25 patients with MCL (56%) were positive for KRAS mutations. Five of those patients had concurrent positive cytology and elevated CEA, 3 had concurrent positive cytology, and 3 had concurrent elevated CEA. The remaining 3 patients were not identified as mucinous on cytology or CEA analysis. Two of those patients had features of BD-IPMN on repeat EUS examinations at 6 monthly intervals and the third patient had adenocarcinoma with mucinous features diagnosed on subsequent EUS-FNA of the mixed cystic solid lesion.
KRAS Testing in Low-Volume Samples
Two of the 16 PCF samples that were ≤1 mL in volume had “nondiagnostic” KRAS tests (1 non-MCL and 1 MCL). Four of 8 patients (50%) with low-volume MCL who had adequate DNA for testing did not have KRAS mutations but were identified by cytology and elevated CEA in 1 patient and by either cytology or elevated CEA in 3 patients. Of the 4 patients with low-volume MCL who had mutations identified, 3 tumors also were identified as mucinous with concurrent cytology and/or elevated CEA, and 1 tumor was classified according to consistent EUS features of a BD-IPMN on 2 separate occasions 12 months apart.
Microscopic evaluation was diagnostic of a specific non-MCL or MCL in 32 of 54 patients (59.3%), inconclusive in 12 of 54 patients (22.2%), and nondiagnostic in 10 of 54 patients (18.5%). Cytology identified 16 of 26 MCLs, demonstrating a sensitivity of 61.5%. When there was adequate material for assessment (44 of 54 patients), cytologic evaluation had excellent specificity (100%) for the diagnosis of MCL.
In this series, 4 of 26 MCLs (15.4%) and 3 of 27 non-MCLs (11.1%) were resected. Histopathology confirmed that the 4 MCLs were 2 MCNs with low-grade dysplasia and 2 IPMNs with low-grade and intermediate-grade dysplasia. Of the remaining 22 patients with unresected MCLs, eight had an associated solid mass on initial EUS examination. Apart from 1 patient who had a diagnosis established on subsequent FNA of the solid component rather than the cystic component, MCL-associated, invasive carcinoma was diagnosed on cyst fluid cytology in the remaining 7 patients. These patients were not suitable for surgical resection because of locally advanced or metastatic disease. Another 3 patients who were classified as having MCL choose not to undergo surgical resection because of personal reasons (1 patient) or because of the presence of severe comorbid conditions (2 cases). Of the unresected MCLs, all but 4 were classified using a minimum of 2 diagnostic modalities. The follow-up imaging in these 4 patients was concordant with the original EUS assessment. Two serous cystadenomas and 1 ductal retention cyst represented the 3 non-MCL that were resected. Follow-up clinical and/or imaging information for the patients is provided in Table 5.
No false-positive laboratory or EUS results were identified among the patients with MCL who underwent resection. Apart from an elevated CEA level in the patient who had a ductal retention cyst, no other false-positive laboratory results were reported among the patients with resected non-MCLs. Integrated test results from the different components that were generated using the volume-dependent approach correctly identified 24 of 26 MCLs (92.3%).
The laboratory investigation of PCF by cytology and biochemistry aims primarily to separate non-neoplastic from neoplastic cysts that require either regular surveillance imaging or surgical management. Accurate diagnosis of neoplastic mucinous cysts can be challenging due to factors such as low cellularity and frequent absence of mucin. More recently, molecular analysis of PCF has been introduced, and studies assessing the commercially available PathFinder TG test (RedPath Integrated Pathology, Pittsburgh, Pa) generally indicate that it is a useful adjunctive test that can increase diagnostic yield from these often low-volume samples.4-6, 10, 12 The added value of molecular testing, however, may be small when critically compared with the combination of cytology and CEA. In the current overall series of 54 patients, the presence of KRAS mutation was able to support an interpretation of MCL in 3.7% of patients, a figure rate to the 4.8% of patients reported by Toll et al in their cohort of pancreatic cysts ≤3 cm in size.7 Within the MCL subgroup, KRAS mutation testing was able to provide supportive evidence for a mucinous lesion in 2 of 25 cases (8%) in which cytology and CEA analyses were not contributory. Along similar lines, Shen et al demonstrated that molecular testing identified only 1 extra mucinous cyst and 1 benign nonmucinous cyst in their cohort of 35 patients compared with the combination of CEA and cytology.4 Both false-negative and false-positive molecular results also have been reported compared with combined clinicopathologic criteria and excisional histology.4, 13 In the context of these findings, it appears that molecular analysis is best used and interpreted in conjunction with concurrent clinical, imaging, cytology, and biochemical information for each patient.4, 13 Therefore, we sought to increase the diagnostic yield of PCF by using a volume-dependent protocol to split the fluid into several components (smears, CB, neat, and SN fluid) for triage, thereby allowing multimodal assessment of a single specimen. To the best of our knowledge, ours is the only study to date that has assessed the utility of generating different components from 1 PCF specimen for concurrent CEA analysis and KRAS mutation testing.
Neoplastic mucinous cysts (IPMNs and MCNs), similar to solid pancreatic cancers, carry KRAS mutations in up to 80% of patients.14-18 Testing for KRAS mutations alone can increase the diagnostic yield of EUS-FNA specimens from solid pancreatic cancers.19, 20 Results from the Pancreatic Cyst DNA (PANDA) study that also were replicated in our laboratory indicate that KRAS mutations have high specificity (96%-100%) but only moderate sensitivity (44%-45%) for the detection of MCL, indicating that, similar to the more complex PathFinder TG analysis, its optimal use is in conjunction with other ancillary tests.6 KRAS mutation testing has been optimized in many larger centers in the context of adjuvant therapy for colorectal and lung cancers and, thus, is widely available and probably is a more cost-effective ancillary study for most pathology laboratories to introduce in contrast to PathFinder TG analysis for PCF specimens.5, 6
To maximize the available material for multiple ancillary tests, we used SN, neat, and CB samples for KRAS testing and were able to demonstrate 87.5% concordance between these various components. Of the 5 discordant cases, 3 had KRAS mutations in the CB that were not detected in the concurrent fluid sample(s), and 2 had KRAS mutations in the fluid sample(s) that were not detected in the concurrent CB. In the context of triaging material for multiple ancillary studies, these findings indicate that, in the vast majority of cases, either sample can provide adequate DNA for molecular analysis; however, if sufficient fluid is available for testing, then all sample types should be analyzed to increase the sensitivity of mutation testing in difficult cases.
Similarly, in the use of SN and neat fluids for CEA quantification, none of 27 patients who had both fluids analyzed had a difference in the interpretation of the final results. An elevated CEA was strongly supportive of an MCL, with overall higher sensitivity (68%) than KRAS mutation (45%) in the PANDA study.6 Similarly, other studies have demonstrated that CEA has superior sensitivity for the detection of MCL (66%-82%) in contrast to molecular analysis by PathFinder TG (33%-81%).12, 13, 21 This supports our preference for CEA analysis in all patients. However, CEA levels do not correspond to malignancy in mucinous cysts, an issue that is addressed by more comprehensive molecular tests like PathFinder TG.3, 6, 21, 22 For instances in which such analysis is not available, cytologic evaluation reportedly is highly accurate for the diagnosis of a malignant cyst with close to 100% specificity reported in some studies.3, 6, 21, 23, 24 An area that is problematic for cytology is the identification of preinvasive mucinous cysts with intermediate to high-grade dysplasia, because these lesions do not usually have a solid component accessible to FNA and may shed few cells into the cyst fluid.6 The observation and interpretation of “atypical epithelial cells” presenting either singly or in small, tight clusters may assist in the identification of these cysts, which are at increased risk of invasive malignancy.24
To maximize the usefulness of CEA analysis in PCF, it is important for each laboratory to optimize the cutoff level used for the separation of MCLs and non-MCLs. The range of CEA values observed in our MCL falling below the 192 ng/mL cutoff suggests that the sensitivity of the analysis may be improved by lowering the previously reported level. In support of this, Cizginer and coinvestigators recently reported an optimal CEA cutoff of 109.9 ng/mL for the diagnosis of a mucinous cyst.21
Some reports advocate the preferential use of molecular testing in samples <1 mL in volume, the rationale being that cytology and CEA analysis are unlikely to be meaningful or possible in these samples.4, 7 Those investigators and those from the PANDA study state that a major advantage that molecular testing has over cytology and CEA is the requirement for a lower volume, ie, 0.2 mL, of fluid. In our setting, a minimum of 0.15 mL is required for CEA analysis. The use of different laboratory techniques and automated systems is likely to account for this difference. On the basis of the finding that 7 of 8 low-volume (≤1 mL) mucinous cysts in our study were detected by cytology and/or elevated CEA, in contrast to only 4 of 8 by KRAS mutation analysis alone, we suggest that material should be submitted preferentially for CEA and cytology but, optimally, for all 3 investigations, including KRAS testing.
Using the method of centrifuging samples to generate at least an SN sample for CEA and a CB for cytologic evaluation and KRAS testing meant that a multimodal approach was possible in the majority of cases, including low-volume samples. The ability to view paraffin sections of the CB before submission for molecular testing also assisted in the interpretation of negative KRAS results, especially in the context of strong clinical and/or laboratory evidence of a MCL. Ancillary histochemical (eg, periodic acid-Schiff with diastase digestion) staining and immunohistochemical tests also may be requested on the CB if required. The ability to perform multiple tests in the majority of our specimens meant that 92.3% of mucinous cysts were identified with at least 1 diagnostic modality.
It is noteworthy that we observed a very low CEA level in serous cystadenomas (all samples had readings <6 ng/mL), a feature that was shared only by 2 other non-MCLs (1 cystic PEN and 1 PC). This finding was reported previously and, again, may be a reproducible feature of these cysts, which often generate low volumes and are negative for KRAS mutations.25
Like many other studies of PCF diagnosis, the weakness of our current study rests in the definitive diagnosis of MCL.4 The majority of patients with these cysts do not undergo resection, especially in a prospective series like ours, in which large numbers of PCs are included. In a small number of patients with MCL, EUS assessment was the only diagnostic investigation before KRAS mutation testing. Although EUS examination in isolation is suboptimal (56% sensitivity and 45% specificity), its sensitivity is comparable to that of any individual laboratory investigation.3 In our series, it is noteworthy that EUS was correct in identifying 3 of 4 resected MCLs.
Deficiencies in the preoperative diagnosis of PCL serve to highlight the importance of optimizing a multimodal approach by effective triage of material for ancillary studies. We have demonstrated that testing for KRAS mutations alone has added value above the traditional combination of cytology and CEA in a proportion (2 of 25 samples; 8%) of cases and advocate that samples should be provided for molecular analysis whenever an MCL is part of the differential diagnosis. We have demonstrated that it is feasible to generate multiple components from the 1 specimen to facilitate multimodal ancillary testing and that these samples provide reliable CEA and KRAS mutation results, even in low-volume specimens.
We thank the PathWest cytopathology technical staff for their expertise in the preparation of the pancreatic cyst fluid specimens and Ms. Charley Budgeon and Mr. Martin Firth from the Center for Applied Statistics at the University of Western Australia for their assistance with the statistical analysis.
This work was funded internally by PathWest Laboratory Medicine.
CONFLICT OF INTEREST DISCLOSURES
The authors made no disclosures.
- 7The added value of molecular testing in small pancreatic cysts. J Pancreas. 2010; 9: 582-586., , , .
- 9Endoscopic ultrasound-guided fine needle aspiration cytology of the pancreas: a morphological and multimodal approach to the diagnosis of solid and cystic mass lesions. Cytopathology. 2007; 18: 331-347., .
- 13A comparative analysis of pancreatic cyst fluid CEA and histology with DNA mutational analysis in the detection of mucin producing or malignant cysts. J Pancreas. 2009; 10: 163-168., , , , .