- Email: [email protected]
EUS and EUS–FNA diagnosis of suspected pancreatic cystic neoplasms: Is the sum of the parts greater than the CEA?
Pancreatology xxx (2015) 1e8
Contents lists available at ScienceDirect
Pancreatology journal homepage: www.elsevier.com/locate/pan
Original article
EUS and EUSeFNA diagnosis of suspected pancreatic cystic neoplasms: Is the sum of the parts greater than the CEA? K.W. Oppong a, b, *, M.F. Dawwas a, b, R.M. Charnley a, c, V. Wadehra d, K. Elamin a, b, S. White a, c, M. Nayar a, b a
HPB Unit, Freeman Hospital, Newcastle upon Tyne, UK Department of Gastroenterology, Freeman Hospital, Newcastle upon Tyne, UK Department of Surgery, Freeman Hospital, Newcastle upon Tyne, UK d Department of Cellular Pathology, Royal Victoria Infirmary, Newcastle upon Tyne, UK b c
a r t i c l e i n f o
a b s t r a c t
Article history: Available online xxx
Background: Carcinoembryonic antigen (CEA) is suggested as the single most useful EUS/EUSeFNA derived test for the diagnosis of mucinous pancreatic cysts. Study aims: To investigate the yield and diagnostic performance of EUS/EUSeFNA on an intention to diagnose basis and to determine the utility of the recommended CEA and amylase cut-off values. Patients and methods: A retrospective study of a prospectively maintained database of 433 procedures performed in a 10 year period. Diagnostic performance of EUSeFNA was determined in 133 procedures with a definite diagnosis. Results: CEA value was determined in significantly fewer procedures (58.6%) than EUS diagnosis was stated (83.4%; p < 0.0001), cyst fluid appearance recorded (89.4%) or adequate sample for cytology obtained (76.7%; p < 0.005). Median CEA was significantly higher in mucinous cysts than non-mucinous (175 ng/ml vs 3 ng/ml, p < 0.0001) and in malignant cysts compared to benign (8945 ng/ml vs 93 ng/ml, p < 0.001). On an intention-to-diagnose analysis, a CEA cut-off of 110 ng/ml was significantly less accurate (42.8%) than EUS diagnosis (67.7%), cytology (58.6%) or aspirate appearance (66.9%; p < 0.05 for all comparisons). However, the combination of EUS diagnosis, cytology and CEA provided higher sensitivity (91%), specificity (75%) and accuracy (85.7%) than each component test alone (p < 0.05 for all comparisons). Median amylase was significantly higher in benign compared to high-risk mucinous cysts ((11,429 IU/L vs. 113 IU/L; p < 0.05. Conclusion: The combination of EUS, cytology and CEA performed well. Malignant cysts had a higher CEA value than benign cysts. On an intention to diagnose basis a CEA cut-off of 110 ng/ml performed poorly. Copyright © 2015, IAP and EPC. Published by Elsevier India, a division of Reed Elsevier India Pvt. Ltd. All rights reserved.
Keywords: Endoscopic ultrasound Pancreas cysts Tumour markers Fine needle aspiration Cytology Specificity
Introduction The management of pancreatic cysts is a significant challenge in modern pancreatology. These lesions are increasingly identified incidentally because of the use of high quality cross-sectional imaging to investigate disparate abdominal symptoms. In a recent study 2.4% of individuals undergoing abdominal MRI for health screening were found to have a pancreatic cyst [1] A proportion of these cystic lesions will have malignant potential. Mucinous cysts
* Corresponding author. HPB Unit, Freeman Hospital, Newcastle upon Tyne NE7 7DN, UK. E-mail address: Kofi[email protected] (K.W. Oppong).
form a category with variable malignant potential. Correctly identifying which lesions are mucinous; intraductal papillary mucinous neoplasm (IPMN) or mucinous cystic neoplasm (MCN) is an important step in the diagnosis and risk stratification of cystic neoplasms. Endoscopic ultrasound (EUS) and EUS guided fine needle aspiration (FNA) may have a role in the diagnosis of pancreatic cysts. Several large series have found an elevated cyst fluid carcinoembryonic antigen (CEA) to be the single most useful test in predicting mucinous cysts with cut-off values of 192 ng/ml having a sensitivity of 73% [2] and the value of 1093 a sensitivity of 80.8%. The CEA cut-off of 192 ng/ml was recommended in the recent IAP consensus guidance [4]. A very low CEA (cut-off 5 ng/ml) has been
http://dx.doi.org/10.1016/j.pan.2015.08.001 1424-3903/Copyright © 2015, IAP and EPC. Published by Elsevier India, a division of Reed Elsevier India Pvt. Ltd. All rights reserved.
Please cite this article in press as: Oppong KW, et al., EUS and EUSeFNA diagnosis of suspected pancreatic cystic neoplasms: Is the sum of the parts greater than the CEA?, Pancreatology (2015), http://dx.doi.org/10.1016/j.pan.2015.08.001
2
K.W. Oppong et al. / Pancreatology xxx (2015) 1e8
suggested as specific for the diagnosis of serous cyst adenoma (SCA) [5,6]. However the practical utility of cyst fluid CEA may be overstated as sufficient fluid for analysis may not be obtained [7], and diagnostic performance is often reported excluding cysts from which insufficient fluid was aspirated [3] rather than on an intention-to-diagnose basis. Cyst fluid amylase has been suggested as useful in differentiating pseudocysts from non pseudocysts [6] and in differentiating benign from malignant mucinous cysts [8]. The diagnostic utility of cyst fluid cytology in the literature is variable; whilst some series have reported excellent accuracy [9], the largest series [2,3] and a meta-analysis [6] have documented less favourable performance. A visual assessment of the viscosity of aspirated material has also been suggested as useful in categorising a cyst as mucinous [10]. We undertook this study to perform an analysis of EUS/ EUSeFNA on an intention to diagnose basis; assessing the diagnostic performance of recommended CEA cut-offs and comparing the performance of CEA to that of clinical EUS diagnosis, visual assessment of aspirate, and cytology both individually and in combination. We also aimed to assess the utility of cyst fluid amylase in differentiating mucinous from non-mucinous cysts, and pseudocysts from non-pseudocysts. Materials and methods The study population comprised all patients referred to the Hepato-Pancreato-Biliary (HPB) unit at the Freeman Hospital for assessment of suspected pancreatic neoplastic cysts and underwent EUS and FNA over a 10 year period between June 2003 and June 2013. The Freeman Hospital serves as a tertiary referral centre of hepatobiliary and pancreatic surgery for the North East region of England with a population of 3.5 million. All patients were managed by a multidisciplinary team (MDT). There was no specific protocol for the use of EUS. The study period encompassed a time of improving imaging and knowledge of cyst morphology. At any given time point the MDT decision was to use EUS if it was felt that further information to improve clinical decision making would be gained. This study took the form of a review of a prospectively completed database of all EUSeFNA procedures. The database recorded details of cyst size, location, appearance of aspirated fluid, putative diagnosis as recorded in the report as well as CEA, amylase, cytology and histology. All EUS procedures were undertaken after obtaining informed patient consent by two experienced endosonographers (K.W.O and M.K.N). Hitachi EUB-7500 or Preirus US workstations (Hitachi Medical Systems, Wellingborough, UK) and Pentax linear echoendoscopes (Pentax, Slough, UK) were used to carry out a standard EUS assessment of the pancreas. FNA of the cyst was carried out under EUS guidance using 22 or 19 gauge FNA needles (Cook Medical, Limerick, Ireland). The endosonographers were not blind to the results of previous imaging or blood tests. To minimise the risk of infection, intravenous antibiotics were administered during EUSeFNA, and oral ciprofloxacin prescribed for a further 3 days. No more than a single pass was made if possible. A specimen was always sent for cytological analysis. If enough fluid was aspirated (1 ml or greater), a sample was also sent for CEA and amylase assay. Histopathological and cytological assessment was performed by pathologists with a special interest in pancreatic disease. The official report was used as the data source. Cyst classification For the purposes of this study all mucinous cysts (IPMN, MCA) and mucinous cyst adenocarcinoma (MCAC) were considered
together, cystic ductal adenocarcinoma were included in this category. SCA, pseudocyst, solid pseudopapillary neoplasm (SPN) and neuroendocrine tumours (NET) were classified as nonmucinous. In terms of immediate malignant risk: MCAC, malignant IPMN, malignant cystic NET, cystic ductal adenocarcinoma and cysts exhibiting high grade dysplasia were classified as high risk. EUS For the purposes of this study the diagnosis stated in the official EUS report was used as the EUS diagnosis. The EUS report documented the appearance of the cyst as well as the appearance of the aspirated fluid/material and whether it appeared mucoid/viscid or thin/serous. The EUS diagnosis recorded in the report was arrived at based on cyst appearance and informed by the nature of the aspirate. Cysts with a classical microcystic appearance and a nonmucinous aspirate were classified as SCA. Cysts with a mucoid aspirate were classified as IPMN if a communication with the pancreatic duct (PD) was identified and or the PD was dilated. A diagnosis of MCA was considered in females with a noncommunicating mucinous cyst. Unilocular or minimally septated cysts were classified as macrocystic serous cyst adenoma if no communication was demonstrated with the PD and the aspirated fluid was thin and clear or straw coloured. Thick-walled unilocular cysts containing debris and a thin aspirate were classified as pseudocysts (in individuals with a prior history compatible with pancreatitis). If no diagnosis was stated in the report none was derived. If a differential diagnosis was recorded rather than a single diagnosis this was only included if both diagnosis allowed categorisation as either mucinous or non-mucinous. Cytology The official cytology report was used as the data source, samples were classified as non-diagnostic (acellular and non-mucinous) or diagnostic (mucinous or non-mucinous epithelium, extracellular mucin or inflammatory cells). Cysts were categorised as mucinous if mucinous epithelium, or extracellular mucin without gastrointestinal contaminant cells was documented and non-mucinous if diagnostic cytology documented inflammatory cells only or nonmucinous epithelium. Data collection and analysis All the data was collected prospectively and entered into a database. The study group comprised patients with a final definitive tissue diagnosis. This included diagnostic malignant cytology, resection histology or biopsy histology. Sensitivity, specificity, and accuracy were determined for visual assessment of aspirate, cytology; diagnosis recorded on the EUS report; as well as for the CEA cut-offs of 192 ng/ml, 110 ng/ml and 30 ng/ml reported from previous series [2,3,8] to differentiate between mucinous and non-mucinous cysts. Receiver operating characteristic (ROC) curves were plotted for CEA (mucinous versus non mucinous) and for amylase (pseudocyst versus non pseudocyst); cut off values were selected to maximise the proportion of correct cyst classification. The performance of combinations of tests was also assessed. Using combination testing a cyst was deemed as mucinous (or malignant) if any of the constituent tests classified it as such. All analyses were performed using the entire dataset of procedures with established tissue diagnosis (henceforth referred to as “intention-to-diagnose analysis”) as well as the subset with a recorded EUS diagnosis recorded; adequate cyst specimen cytology or sufficient aspirate for CEA assay (henceforth referred to as “complete case analysis”). For the purposes of the intention-todiagnose analysis procedures with missing data were considered either false negative (in the case of a mucinous/malignant cyst) or
Please cite this article in press as: Oppong KW, et al., EUS and EUSeFNA diagnosis of suspected pancreatic cystic neoplasms: Is the sum of the parts greater than the CEA?, Pancreatology (2015), http://dx.doi.org/10.1016/j.pan.2015.08.001
K.W. Oppong et al. / Pancreatology xxx (2015) 1e8
false positive (in the case of a non-mucinous cyst). Statistical comparison of quantitative data was made using the ManneWhitney U test. Comparison of gender proportion was performed using Chi Squared test. Comparison of diagnostic sensitivity specificity and accuracy was performed using Cochran's Q test. If the Cochrane Q test was positive (p < 0.05) a test for pairwise comparison of variables according to Sheskin was automatically performed. Test performance was calculated on an intention to diagnose basis as well as per patient tested. Statistical analysis was performed using MedCalc Statistical Software version 13.1.0 (MedCalc Software bvba, Ostend, Belgium; http://www.medcalc. org; 2014). Ethics All procedures were done as a part of standard patient care as directed by the HPB MDT and not to a research protocol. Data collection was performed as part of our ongoing quality monitoring. In accordance with U.K. National Research Ethics Service guidelines formal ethical review was not required. Results During the study period 433 EUSeFNA procedures were performed on 388 individuals; 37 individuals had 2 or more procedures (Fig. 1). Adequate samples for cytology were obtained in 262 cases (60.5%). Cyst aspirate appearance was recorded in 392 procedures (90.5%). A CEA value was determined for 248 procedures (57.3%) and an amylase value in 254 (58.7%). One hundred and thirty three procedures (119 individuals) had a final definitive tissue diagnosis and formed the study group. Diagnosis was based on surgical resection in 102, malignant cytology in 8, biopsy histology in 6, and a cytological diagnosis of NET in 3 (Fig. 1 details flow of patients) The average age was 61.4 years (range 19e84), and the median cyst size was 4.0 cm. There were 79 mucinous cysts (31 benign IPMN, 19 MCA, 8 MCAC, 7 adenocarcinoma, 6 undifferentiated mucinous malignancy, 4 IPMN with high grade dysplasia, 3 malignant IPMN and 1
3
undifferentiated malignancy). There were 40 non-mucinous cysts (12 SCA, 11 pseudocysts, 11 NET, 2 gastrointestinal stromal tumours (GIST), 1 cystic lymphangioma, 1 phaeochromocytoma, 1 SPN and 1 Schwannoma). Considering all individuals with a definitive diagnosis there was a female preponderance 82/119 (68.9%, p < 0.001). CEA (58.6%) and amylase (57.9%) were determined in significantly fewer procedures than EUS diagnosis (91.7%), fluid appearance (89.5%) or adequate cytology (76.7%) (p < 0.0001).
Fluid analysis Carcinoembryonic antigen CEA values were determined in 78 (58.6%) of the study group comprising 50 mucinous and 28 non-mucinous cysts. Fig. 1 is a box plot of CEA values by cyst type. Table 1 summarises the differences between cyst categories. Median CEA was significantly higher in mucinous (175 ng/ml) than non-mucinous (3 ng/ml) p < 0.0001 and in high risk mucinous cysts (8945 ng/ml) compared to benign mucinous cysts (93 ng/ml) p < 0.001. An ROC analysis was performed classifying all mucinous cysts together. The area under the curve (AUC) was 0.89. The cut off level that produced the highest accuracy was 7 ng/ml. At this level the sensitivity was 94%, specificity 75% and accuracy 87.1% (Table 2). The diagnostic performance of 3 CEA cut-off (complete case and intention to diagnose) of 3 large series 192 ng/ml, 110 ng/ml and 30 ng/ml are shown in Table 2. With a Cochran's Q test we found a significant difference among the sensitivity and accuracy of cutoffs. The internally derived cut-off of 7 ng/ml provided superior discrimination of mucinous from non-mucinous cysts than the cutoffs of 192 ng/ml and 110 ng/ml. The discriminative performance of CEA cut-offs of 30; 110; and 192 ng/ml was similar. In terms of diagnosing SCA a very low CEA has been suggested as showing good diagnostic performance [5,11]. In our series 21 cases had a value below 5 ng/ml (Fig. 1). Of these 6 were SCA, 1 SCA had a value above 5 ng/ml. All NET had a low CEA value with 7/8 below 5 ng/ml. The diagnostic performance of a cut-off of 5 ng/ml in diagnosing SCA was sensitivity 85.7% specificity 78.8%.
Fig. 1. Flow chart of patients in the study and adequacy assessment.
Please cite this article in press as: Oppong KW, et al., EUS and EUSeFNA diagnosis of suspected pancreatic cystic neoplasms: Is the sum of the parts greater than the CEA?, Pancreatology (2015), http://dx.doi.org/10.1016/j.pan.2015.08.001
4
K.W. Oppong et al. / Pancreatology xxx (2015) 1e8
Table 1 Comparison of median levels of cyst fluid CEA & amylase using different categories of cyst. Cyst category
CEA, ng/ml
Non-mucinous cysts Mucinous cysts (including PDAC) Benign mucinous cysts High risk mucinous cysts (HGD & malignant) Benign cysts High risk cyst (HGD & malignant) Non-pseudocysts Pseudocysts
Amylase, IU/L
n
Median (IQR)
P value
n
Median (IQR)
P value
28 50 33 17 61 17 69 9
3 175 93 6432 29 6432 82 54
0.0001
28 49 32 17 60 17 68 9
161 2624 11429 113 3103 113 207 32800
ns
(1e12.5) (35e2265) (28e686) (246.7e29510.7) (3.0e141.2) (246.7e29510.7) (3.75e774.2) (15.25e113.5)
0.0004 0.0004 ns
(50.5e21703) (62.5e21730) (655e33341) (32e1235.7) (62e26495) (32e1235.7) (50.5e16245.5) (22525.5e46330)
0.0031 0.0175 0.0011
Table 2 Complete case and intention-to-diagnose analyses of the diagnostic performance for EUS diagnosis; aspirate appearance; aspirate cytology; and aspirate CEA for discriminating mucinous from non-mucinous cysts.
Sensitivity - Complete case - Intention-to-diagnose Specificity - Complete case - Intention-to-diagnose Accuracy - Complete case - Intention-to-diagnose AUC - Complete case
EUS diagnosis (n ¼ 122)
Cytology (n ¼ 102)
Aspirate appearance (n ¼ 119)
CEA >7 ng/ml (n ¼ 78)
CEA >30 ng/ml (n ¼ 78)
CEA >110 ng/ml (n ¼ 78)
CEA >192 ng/ml (n ¼ 78)
83.9% (68/81) 76.4% (68/89)
72.2% (52/72) 58.4% (52/89)
75.9% (60/79) 67.4% (60/89)
94% (47/50)m 52.8% (47/89)
39/50 (78%) 39/89 (43.8%)
62.0% (31/50)m 31/89 (34.8%)
48% (24/50)m 26.9% (24/89)
73.3% (22/30) 50% (22/44)
86.7% (26/30) 59.1% (26/44)
72.5% (29/40) 65.9% (29/44)
75% (21/28) 47.7% (21/44)
78.6% (22/28) 50% (22/44)
92.9% (26/28) 59.1% (26/44)
96.4% (27/28) 27/44 (61.3%)
81.1% (90/111) 67.7% (90/133)*^
76.5% (78/102) 58.6% (78/133)*
74.8% (89/119) 66.9% (89/133)*^
87.1% (68/78)m 51.1% (68/133)
78.2% (61/78) 45.8% (61/133)^
73.1% (57/78)m 42.8% (57/133)^
65.4% (51/78)m 51/133 (38.3%)*
0.79
0.79
0.74
0.89
0.78
0.77
0.72
m
Complete case (CEA cutoffs): Cutoff of 7 ng/ml significantly more sensitive and accurate than 192 and 110 ng/ml p < 0.05 (n ¼ 78). Complete case (all tests): no significant difference in accuracy of matched cases (n ¼ 45). Intention to diagnose: CEA cut-off of 192 ng/ml significantly less accurate than EUS diagnosis, cytology or aspirate appearance p < 0.05*. CEA cut-offs of 110 and 30 ng/ml were significantly less accurate than EUS diagnosis and aspirate appearance p < 0.05^.
Amylase Amylase values were determined in 77/133 (57.9%) procedures, comprising 43 mucinous and 34 non mucinous (including 9 pseudocysts) cysts. Fig. 2 is a box plot of cyst fluid amylase by cyst type. Table 1 summarises the differences between cyst categories. The median amylase value in pseudocysts (32,800 IU/L) was significantly higher than in non pseudocysts 207 IU/L p < 0.01. An ROC curve was constructed, the calculated optimal cut-off for differentiating between pseudocysts and non-pseudocysts was 3977 IU/L, providing a sensitivity of 100%, and specificity 66%. The area under the ROC curve was 0.84. Median amylase was also significantly higher in benign mucinous cysts (11,429 IU/L) than high risk mucinous cysts (113 IU/L) and benign cysts (3101 IU/L) than high risk cysts (113 IU/L) both p < 0.05. The currently recommended cutoff of 250 IU/L for identifying pseudocysts had a sensitivity of 9/9 (100%), specificity of 35/68 (51.6%) and accuracy of 44/77 (57.1%) (see Fig. 3). EUS An EUS diagnosis or differential was recorded in 122/133 (91.7%). Excluding reports with inconsistent differential diagnosis (i.e. categorisation as both mucinous and non-mucinous), there were 111 (83.4%) procedures where the cyst was unambiguously categorised as mucinous or non-mucinous. Of the 111 cysts, 81 were mucinous and 30 non-mucinous. Diagnostic performance on a complete case and intention to diagnose analysis is detailed in Table 2. Cytology Adequate samples for cytology were obtained in 102/133 (76.7%) of procedures. Diagnostic performance is detailed in Table 2.
Aspirate visual appearance Aspirate appearance was recorded in 119/133 (89.5%) of procedures. Diagnostic performance is detailed in Table 2. Comparisons On a complete case (45 matched cases) basis using the Cochran's Q test there was no significant difference in accuracy among EUS diagnosis, aspirate appearance, cytology or the CEA cut-off values of 30, 110 and 192 ng/ml (Table 2). However on an intention to diagnose basis we found a significant difference. A CEA cut-off of 192 ng/ml was significantly less accurate than EUS diagnosis, aspirate appearance and cytology; p < 0.05 whilst cut-offs of 30 and 110 ng/ml were less accurate than EUS diagnosis and aspirate appearance (Table 2). Combination testing The CEA cut-off 110 ng/ml was used. The combinations resulted in improved diagnostic performance (Table 3). Using the Cochran's Q test we found a significant difference among the accuracy of the individual tests and test combinations (detailed in Table 3). All combinations tested were significantly more accurate (p < 0.05) than CEA cut-offs of 7, 110 or 192 ng/ml. In addition the diagnostic accuracy of EUS diagnosis þ cytology þ aspirate appearance þ CEA > 110 ng/ml (85.7%) and of EUS diagnosis þ cytology þ CEA > 110 ng/ml (85.7%) were significantly higher than the individual tests by themselves (p < 0.05): EUS diagnosis (67.7%), cytology (58.6%), and aspirate appearance (66.9%). EUS diagnosis þ cytology þ aspirate appearance þ CEA > 110 ng/ml (85.7%) and EUS diagnosis þ cytology þ CEA > 110 ng/ml (85.7%) were significantly more accurate than CEA > 110 ng/ml þ cytology (p < 0.05).
Please cite this article in press as: Oppong KW, et al., EUS and EUSeFNA diagnosis of suspected pancreatic cystic neoplasms: Is the sum of the parts greater than the CEA?, Pancreatology (2015), http://dx.doi.org/10.1016/j.pan.2015.08.001
K.W. Oppong et al. / Pancreatology xxx (2015) 1e8
5
Fig. 2. Box plot of cyst fluid CEA by cyst type. Y axis is on a logarithmic scale (nanograms per millilitre). There are 4 lines representing the cut-off levels of 192, 110, 30 and 7 ng/ml. The high risk category includes malignant cysts and those with foci of high grade dysplasia (IPMN, MCAC and PDAC). The Other category is made up of two GIST, one cystic lymphangioma and one phaeochromocytoma. All categories are mutually exclusive.
Discussion The role of EUS/EUSeFNA in the management of suspected neoplastic pancreatic cysts remains to be fully defined. There is
however a consensus that the decision on its use is best made in a multidisciplinary team setting [12,13] For the majority of patients who do not have symptoms or imaging findings mandating [4,14] surgery, an accurate diagnosis (particularly the discrimination of
Fig. 3. Box plot of cyst fluid amylase by cyst type. Y axis is on a logarithmic scale (international units per millilitre). There is a line representing the cut-off of 250 IU/L The high risk category includes malignant cysts and those with foci of high grade dysplasia (IPMN, MCAC and PDAC). The Other category is made up of two GIST, one cystic lymphangioma and one phaeochromocytoma. All categories are mutually exclusive.
Please cite this article in press as: Oppong KW, et al., EUS and EUSeFNA diagnosis of suspected pancreatic cystic neoplasms: Is the sum of the parts greater than the CEA?, Pancreatology (2015), http://dx.doi.org/10.1016/j.pan.2015.08.001
6
K.W. Oppong et al. / Pancreatology xxx (2015) 1e8
Table 3 Intention-to-diagnose analysis of the diagnostic performance of multiple test combinations for discriminating mucinous from non-mucinous cysts (utilising a CEA cut-off of 110 ng/ml).
Sensitivity Specificity Accuracy
Cytology þ aspirate appearance
CEA > 110 ng/ml þ Cytology
EUS diagnosis þ Cytology
Cytology þ CEA > 110 ng/ml þ aspirate appearance
EUS diagnosis þ cytology þ aspirate appearance þ CEA > 110 ng/ml
EUS diagnosis & cytology þ CEA > 110 ng/ml
84.2% (75/89) 72.7% (32/44) 80.4% (107/133)
77.5% (69/89) 75% (33/44) 76.7%*m (102/133)
86.5% (77/89) 68.2% (30/44) 80.4% (107/133)
89.9% (80/89) 72.7% (32/44) 84.2% (112/133)
93.2% (83/89) 70.5% (31/44) 85.7%* (114/133)
91.0% (81/89) 75% (33/44) 85.7%m (114/133)
*p < 0.05. p < 0.05.
m
mucinous from non-mucinous cyst) is important as it facilitates risk stratification and decisions regarding the need for surveillance, surveillance interval and duration. However this remains problematic; in one series from a very experienced, high-volume centre, preoperative diagnosis was incorrect in one-third of incidentally discovered cysts and 5% of resected cysts were non-neoplastic [12]. Another series documented a pre-operative accurate diagnosis rate of 60.9% [15]. The high resolution images obtained with EUS initially raised the possibility that neoplastic cysts could be accurately differentiated on morphology alone [16,17] however it is now recognised that considerable overlap exists in cyst morphology [18]. Whilst mural nodules and associated masses are suggestive of malignancy [18e21], other EUS morphological features are not specific. EUS morphology has a reported sensitivity of between 35 and 90% [2,3,9,19,22e26] with the largest single centre series reporting an accuracy of 48% [3]. In addition, interobserver agreement of EUS assessment of pancreatic cysts is not good. A recent study found moderate agreement between expert endosonographers [27] whilst an earlier study found little more than chance interobserver agreement among experienced endosonographers for the diagnosis of neoplastic versus non-neoplastic cysts [2]. EUS morphology alone, therefore, cannot be relied on to accurately differentiate mucinous from non-mucinous cysts. There are a number of factors in a non-blinded study such as this (and the majority of such prior studies) that are a potential bias and might lead to an overestimation of EUS performance these include; knowledge of prior imaging and knowledge of patient demographics (age and gender specifically) however this reflects everyday clinical practice. In the present study a clear mucinous/non-mucinous diagnosis was recorded on the EUS report in the majority of procedures 111/ 133 (83.4%). This diagnosis was based on a combination of cyst morphology and aspirate appearance. On an intention-to-diagnose basis, the sensitivity of the clinical EUS diagnosis was modest (76.4%), with poor specificity (50%) and modest accuracy (67.7%), but these performance characteristics compare favourably to those reported previously [2,9,19,22e24] for EUS morphology alone. Cyst FNA in addition to enabling cytology and fluid analysis (if sufficient volume is aspirated) allows a visual assessment of cyst content which can be a useful adjunct with a clear mucoid/viscid aspirate suggesting a mucinous cyst. However, a thin non-mucoid aspirate does not exclude a mucinous neoplasm [18,28]. Previous investigators have suggested that viscosity may be a useful marker for differentiating mucinous and non-mucinous lesions [29,30]. Measuring viscosity using commercially available tests is not straightforward as a significant volume of aspirate is needed [31]. Debris, blood, and tissue in pancreatic-cyst fluid can also interfere with the measurement of fluid viscosity [31]. Visual assessment of aspirate viscosity utilising the “string sign” has been suggested as a measure of cyst fluid mucin content [10]. In the present study, aspirate appearance was recorded in a high proportion of cases (89.4%) and accuracy on an intention-to-diagnose basis was 67%.
CEA is the most widely investigated and used tumour marker for cyst classification. A number of different cut-off values have been reported in the literature including 192 ng/ml [2], 109 ng/ml [3], 800 ng/ml [6], 50 ng/ml [32] and 30 ng/ml [8]. Interpretation of the literature for use in everyday practice is further complicated by the fact that some studies have assessed diagnostic performance in differentiating mucinous (MCA, MCAC and IPMN) from nonmucinous [2,3,8,9] whilst others have excluded IPMN [6,30]. No CEA assay is validated for use with aspirated cyst fluid, in addition a recent study found that reproducibility of CEA analysis was variable [33]. In the present study, in which IPMN lesions were not excluded and cystic PDAC were included with mucinous cysts, we found a cut-off of 7 ng/ml to provide the largest area under the ROC curve with an accuracy of 87.1%. This is one of the lowest optimal CEA cutoffs reported, further increasing the heterogeneity of results in the literature, and reflects the fact that a significant number of mucinous cysts, primarily IPMNs, had a low CEA value. The discriminative ability of our internally-derived cut-off was similar to that of 110 ng/ml, derived from the largest single centre series [3], but superior to that of 192 ng/ml. In our study sufficient fluid for CEA analysis was obtained in 58.6% of cases, compared to 49% in a recent Dutch study [4], and 78e79% in the studies by Cizginer et al. [3] and Khashab et al. [25] the performance of CEA on an intention-to-diagnose basis in our cohort was therefore less impressive with a cut-off of 7 ng/ml having an accuracy of 55.9% and a cut-off of 110 ng/ml 47.3%. In the present series, we also documented significantly higher CEA levels in high-risk mucinous cysts compared to benign ones. Previous studies of this issue have reported inconsistent findings [3,8,30,34,35]. The variable rate of obtaining fluid for analysis, lack of clarity in the literature as to whether higher values are seen in high risk cysts and the possible impact of differing assays are all factors that limit the utility of a CEA cut-off used in isolation. The diagnostic performance of cytology for cysts is recognised to be lower than for solid pancreatic lesion [36]. The modest performance of cytology in the diagnosis of mucinous cysts is primarily due to the relatively high proportion of acellular aspirates [3]. A number of studies have reported high rate of acellular aspirates, with two recent series documenting rates of 53e69% [7,37]. Our inadequate aspirate rate of 23.3% was similar to that reported in another recent large series [25]. A recent meta-analysis of 18 studies reported a pooled sensitivity for cytology of 54% and specificity of 93% [38]. This compared with an intention-todiagnose sensitivity of 58.4% and specificity of 59.1% in the present study. We found a significant difference in median amylase levels between pseudocysts and other cysts (32,800 IU/L vs. 207; p < 0.01). Utilising the recommended [6,39] amylase cut-off of 250 IU/L gave a sensitivity of 100%, specificity of 51.5% and an accuracy of 57.1% in line with the findings of previous studies [6,8]. High amylase levels were recorded not only in IPMN as would be expected but also in MCN, supporting the notion of microscopic communication between the pancreatic ductal system and some MCN [8]. We found
Please cite this article in press as: Oppong KW, et al., EUS and EUSeFNA diagnosis of suspected pancreatic cystic neoplasms: Is the sum of the parts greater than the CEA?, Pancreatology (2015), http://dx.doi.org/10.1016/j.pan.2015.08.001
K.W. Oppong et al. / Pancreatology xxx (2015) 1e8
amylase levels to be significantly higher in benign (11,429 IU/L) compared to high-risk mucinous cysts (113 IU/L). Similar results were reported from the Mayo clinic [8]. Intention-to-diagnose analysis reveals the true utility of a diagnostic test. On this basis a CEA cut-off of 110 ng/ml was significantly less accurate (42.8%) than EUS diagnosis (67.7%), cytology (58.6%) or aspirate appearance (66.9%) However, combination testing improved diagnostic performance, yielding very similar results to those reported by the largest single centre series to date [3]. The combination of EUS diagnosis, cytology and CEA resulted in a sensitivity, specificity and accuracy of 91%, 75%, 85.7% and this was more accurate than EUS, cytology, CEA 110 ng/ml or aspirate appearance individually. Potential limitations to this study include the fact that it is a retrospective single tertiary centre study. In addition, as with all such studies, there is a selection bias because the study group is restricted to those with a definitive histological diagnosis. Strengths of the study are that analysis was performed on an intention-to-diagnose basis and the selection process for EUS/ EUSeFNA was by the multi-disciplinary team. Adjuncts to the EUS/ EUSeFNA procedure including molecular analysis of fluid and needle-based confocal laser endomicroscopy (nCLE have been studied. A recent study found no difference in accuracy between CEA/cytology and molecular analysis [40], initial results with nCLE [41,42] are promising, but data is relatively sparse and the rate of pancreatitis appears higher than with EUSeFNA. Further studies are required to delineate its role. In conclusion, in our series a CEA cut-off of 110 ng/ml used in isolation performed poorly in part due to the significant proportion of cases in which sufficient fluid for analysis was not obtained. This in addition to the heterogeneity of cut-off values would suggest caution in using CEA in isolation. However combination with the other results from the EUS/EUSeFNA procedure resulted in good diagnostic performance on an intention-to-treat basis. References [1] de Jong K, Nio CY, Hermans JJ, Dijkgraaf MG, Gouma DJ, van Eijck CHJ, et al. High prevalence of pancreatic cysts detected by screening magnetic resonance imaging examinations. Clin Gastroenterol Hepatol 2010;8:806e11. [2] Brugge WR, Lewandrowski K, Lee-Lewandrowski E, Centeno BA, Szydlo T, Regan S, et al. Diagnosis of pancreatic cystic neoplasms: a report of the cooperative pancreatic cyst study. Gastroenterology 2004;126:1330e6. [3] Cizginer S, Turner BG, Turner B, Bilge AR, Karaca C, Pitman MB, et al. Cyst fluid carcinoembryonic antigen is an accurate diagnostic marker of pancreatic mucinous cysts. Pancreas 2011;40(7):1024e8. ndez-Del Castillo C, Adsay V, Chari S, Falconi M, Jang J-Y, et al. [4] Tanaka M, Ferna International consensus guidelines 2012 for the management of IPMN and MCN of the pancreas. Pancreatology 2012;12:183e97. vy P, Ruszniewski P, Bernades P. Diagnostic [5] Hammel P, Voitot H, Vilgrain V, Le value of CA 72-4 and carcinoembryonic antigen determination in the fluid of pancreatic cystic lesions. Eur J Gastroenterol Hepatol 1998;10:345e8. [6] Waaij LA, Van Der Dullemen HM, Van Porte RJ. Cyst fluid analysis in the differential diagnosis of pancreatic cystic lesions: a pooled analysis. Gastrointest Endosc 2005;62:383e9. [7] de Jong K, Poley J-WJ, van Hooft JE, Visser M, Bruno MJ, Fockens P. Endoscopic ultrasound-guided fine-needle aspiration of pancreatic cystic lesions provides inadequate material for cytology and laboratory analysis: initial results from a prospective study. Endoscopy 2011;43:585e90. [8] Park WG-U, Mascarenhas R, Palaez-Luna M, Smyrk TC, O’Kane D, Clain JE, et al. Diagnostic performance of cyst fluid carcinoembryonic antigen and amylase in histologically confirmed pancreatic cysts. Pancreas 2011;40:42e5. [9] Frossard JL, Amouyal P, Amouyal G, Palazzo L, Amaris J, Soldan M, et al. Performance of endosonography-guided fine needle aspiration and biopsy in the diagnosis of pancreatic cystic lesions. Am J Gastroenterol 2003;98:1516e24. [10] Leung KK, Ross WA, Evans D, Fleming J, Lin E, Tamm EP, et al. Pancreatic cystic neoplasm: the role of cyst morphology, cyst fluid analysis, and expectant management. Ann Surg Oncol 2009;16:2818e24. [11] Hammel P, Levy P, Voitot H, Levy M, Vilgrain V, Zins M. Preoperative cyst fluid analysis is useful for the differential diagnosis of cystic lesions of the pancreas. Gastroenterology 1995;108:1230e5. ndez-del Castillo C. Pancreatic cystic neoplasms: management [12] Farrell JJ, Ferna and unanswered questions. Gastroenterology 2013;144:1303e15.
7
[13] Robinson SM, Scott J, Oppong KW, White SA. What to do for the incidental pancreatic cystic lesion? Surg Oncol 2014;23(3):117e25. €ppel G, Werner J, McKay C, et al. Eu[14] Del Chiaro M, Verbeke C, Salvia R, Klo ropean experts consensus statement on cystic tumours of the pancreas. Dig Liver Dis 2013;45:703e11. [15] Del Chiaro M, Segersv€ ard R, Pozzi Mucelli R, Rangelova E, Kartalis N, Ansorge C, et al. Comparison of preoperative conference-based diagnosis with histology of cystic tumors of the pancreas. Ann Surg Oncol 2014;21:1539e44. [16] Gress F, Gottlieb K, Cummings O, Sherman S, Lehman G. Endoscopic ultrasound characteristics of mucinous cystic neoplasms of the pancreas. Am J Gastroenterol 2000;95:961e5. [17] Koito K, Namieno T, Nagakawa T, Shyonai T, Hirokawa N, Morita K. Solitary cystic tumor of the pancreas: EUS-pathologic correlation. Gastrointest Endosc 1997;45:268e76. [18] Pitman MB, Lewandrowski K, Shen J, Sahani D, Brugge W, Fernandez-del Castillo C. Pancreatic cysts: preoperative diagnosis and clinical management. Cancer Cytopathol 2010;118:1e13. [19] Ahmad NA, Kochman ML, Lewis JD, Ginsberg GG. Can EUS alone differentiate between malignant and benign cystic lesions of the pancreas? Am J Gastroenterol 2001;96:3295e300. [20] Song MH, Lee SK, Kim M-H, Lee HJ, Kim K-P, Kim HJ. EUS in the evaluation of pancreatic cystic lesions. Gastrointest Endosc 2003;57:891e6. [21] Sahani DV, Lin DJ, Venkatesan AM, Sainani N, Mino-Kenudson M, Brugge WR. Multidisciplinary approach to diagnosis and management of intraductal papillary mucinous neoplasms of the pancreas. Clin Gastroenterol Hepatol 2009;7:259e69. [22] Ahmad NA, Kochman ML, Brensinger C, Brugge WR, Faigel DO, Gress FG. Interobserver agreement among endosonographers for the diagnosis of neoplastic versus non-neoplastic pancreatic cystic lesions. Gastrointest Endosc 2003;58:59e64. [23] Hernandez LV, Mishra G, Forsmark C, Draganov PV, Petersen JM, Hochwald SN. Role of endoscopic ultrasound (EUS) and EUS-guided fine needle aspiration in the diagnosis and treatment of cystic lesions of the pancreas. Pancreas 2002;25:222e8. [24] Sedlack R, Affi A, Vazquez-Sequeiros E, Norton ID, Clain JE, Wiersema MJ. Utility of EUS in the evaluation of cystic pancreatic lesions. Gastrointest Endosc 2002;56:543e7. [25] Khashab M a, Kim K, Lennon AM, Shin EJ, Tignor AS, Amateau SK, et al. Should we do EUS/FNA on patients with pancreatic cysts? The incremental diagnostic yield of EUS over CT/MRI for prediction of cystic neoplasms. Pancreas 2013;00:1e5. [26] de Jong K, van Hooft JE, Nio CY, Gouma DJ, Dijkgraaf MGW, Bruno MJ, et al. Accuracy of preoperative workup in a prospective series of surgically resected cystic pancreatic lesions. Scand J Gastroenterol 2012;8e9:1056e63. [27] de Jong K, Verlaan T, Dijkgraaf MG, Poley JW, van Dullemen H, Bruno MJ, et al. Interobserver agreement for endosonography in the diagnosis of pancreatic cysts. Endoscopy 2011;43:579e84. [28] Al-Haddad M, Schmidt MC, Sandrasegaran K, Dewitt J. Diagnosis and treatment of cystic pancreatic tumors. Clin Gastroenterol Hepatol 2011;9:635e48. [29] Lewandrowski KB, Warshaw AL, Compton CC, Pins MR, Southern JF. Variability in cyst fluid carcinoembryonic antigen level, fluid viscosity, amylase content, and cytologic findings among multiple loculi of a pancreatic mucinous cystic neoplasm. Am J Clin Pathol 1993;100:425e7. [30] 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. Gastrointest Endosc 2006;64:697e702. [31] Brugge W. The mistletoe and cyst-fluid analysis: a sticky insight. Gastrointest Endosc 2006;64:703e4. [32] Oh H-C, Kang H, Brugge WR. Cyst fluid amylase and CEA levels in the differential diagnosis of pancreatic cysts: a single-center experience with histologically proven cysts. Dig Dis Sci 2014;59(12):3111e6. [33] Boot CS, Mahon BS, Bramhall SR, Clark PM. Validity of carcinoembryonic antigen and carbohydrate antigen 19-9 measurements in pancreatic cyst fluid with a serum-based immunoassay. Clin Chem 2010;56:1351e2. [34] Kucera S, Centeno BA, Springett G, Malafa MP, Chen YA, Weber J, et al. Cyst fluid carcinoembryonic antigen level is not predictive of invasive cancer in patients with intraductal papillary mucinous neoplasm of the pancreas. JOP 2012;13:409e13. [35] Buscaglia JM, Giday SA, Kantsevoy SV, Jagannath SB, Magno P, Wolfgang CL, et al. Patient- and cyst-related factors for improved prediction of malignancy within cystic lesions of the pancreas. Pancreatology 2009;9:631e8. [36] Wadehra V. The challenge of pancreatic endoscopic ultrasound-guided fine needle aspiration cytology. Cytopathology 2013;24:143e9. [37] Lim LG, Lakhtakia S, Ang TL, Vu CKF, Dy F, Chong VH, et al. Factors determining diagnostic yield of endoscopic ultrasound guided fine-needle aspiration for pancreatic cystic lesions: a multicentre Asian study. Dig Dis Sci 2013;58: 1751e7. [38] Thornton GD, McPhail MJW, Nayagam S, Hewitt MJ, Vlavianos P, Monahan KJ Endoscopic ultrasound guided fine needle aspiration for the diagnosis of pancreatic cystic neoplasms: a meta-analysis. Pancreatology 13:48e7. [39] Khalid A, Brugge W. ACG practice guidelines for the diagnosis and management of neoplastic pancreatic cysts. Am J Gastroenterol 2007;102:2339e49. [40] Al-Haddad M, DeWitt J, Sherman S, Schmidt CM, LeBlanc JK, McHenry L, et al. Performance characteristics of molecular (DNA) analysis for the diagnosis of mucinous pancreatic cysts. Gastrointest Endosc 2014;79:79e87.
Please cite this article in press as: Oppong KW, et al., EUS and EUSeFNA diagnosis of suspected pancreatic cystic neoplasms: Is the sum of the parts greater than the CEA?, Pancreatology (2015), http://dx.doi.org/10.1016/j.pan.2015.08.001
8
K.W. Oppong et al. / Pancreatology xxx (2015) 1e8
[41] Konda VJ, Meining A, Jamil LH, Giovannini M, Hwang JH, Wallace MB, et al. A pilot study of in vivo identification of pancreatic cystic neoplasms with needle-based confocal laser endomicroscopy under endosonographic guidance. Endoscopy 2013;45:1006e13.
[42] Lemaistre A, Pujol B, Caillol F, Lucidarme D, Morellon-mialhe B, Fumex F, et al. A novel approach to the diagnosis of pancreatic serous cystadenoma: needlebased confocal laser endomicroscopy. Endoscopy 2015;47(1):26e32.
Please cite this article in press as: Oppong KW, et al., EUS and EUSeFNA diagnosis of suspected pancreatic cystic neoplasms: Is the sum of the parts greater than the CEA?, Pancreatology (2015), http://dx.doi.org/10.1016/j.pan.2015.08.001
Contents lists available at ScienceDirect
Pancreatology journal homepage: www.elsevier.com/locate/pan
Original article
EUS and EUSeFNA diagnosis of suspected pancreatic cystic neoplasms: Is the sum of the parts greater than the CEA? K.W. Oppong a, b, *, M.F. Dawwas a, b, R.M. Charnley a, c, V. Wadehra d, K. Elamin a, b, S. White a, c, M. Nayar a, b a
HPB Unit, Freeman Hospital, Newcastle upon Tyne, UK Department of Gastroenterology, Freeman Hospital, Newcastle upon Tyne, UK Department of Surgery, Freeman Hospital, Newcastle upon Tyne, UK d Department of Cellular Pathology, Royal Victoria Infirmary, Newcastle upon Tyne, UK b c
a r t i c l e i n f o
a b s t r a c t
Article history: Available online xxx
Background: Carcinoembryonic antigen (CEA) is suggested as the single most useful EUS/EUSeFNA derived test for the diagnosis of mucinous pancreatic cysts. Study aims: To investigate the yield and diagnostic performance of EUS/EUSeFNA on an intention to diagnose basis and to determine the utility of the recommended CEA and amylase cut-off values. Patients and methods: A retrospective study of a prospectively maintained database of 433 procedures performed in a 10 year period. Diagnostic performance of EUSeFNA was determined in 133 procedures with a definite diagnosis. Results: CEA value was determined in significantly fewer procedures (58.6%) than EUS diagnosis was stated (83.4%; p < 0.0001), cyst fluid appearance recorded (89.4%) or adequate sample for cytology obtained (76.7%; p < 0.005). Median CEA was significantly higher in mucinous cysts than non-mucinous (175 ng/ml vs 3 ng/ml, p < 0.0001) and in malignant cysts compared to benign (8945 ng/ml vs 93 ng/ml, p < 0.001). On an intention-to-diagnose analysis, a CEA cut-off of 110 ng/ml was significantly less accurate (42.8%) than EUS diagnosis (67.7%), cytology (58.6%) or aspirate appearance (66.9%; p < 0.05 for all comparisons). However, the combination of EUS diagnosis, cytology and CEA provided higher sensitivity (91%), specificity (75%) and accuracy (85.7%) than each component test alone (p < 0.05 for all comparisons). Median amylase was significantly higher in benign compared to high-risk mucinous cysts ((11,429 IU/L vs. 113 IU/L; p < 0.05. Conclusion: The combination of EUS, cytology and CEA performed well. Malignant cysts had a higher CEA value than benign cysts. On an intention to diagnose basis a CEA cut-off of 110 ng/ml performed poorly. Copyright © 2015, IAP and EPC. Published by Elsevier India, a division of Reed Elsevier India Pvt. Ltd. All rights reserved.
Keywords: Endoscopic ultrasound Pancreas cysts Tumour markers Fine needle aspiration Cytology Specificity
Introduction The management of pancreatic cysts is a significant challenge in modern pancreatology. These lesions are increasingly identified incidentally because of the use of high quality cross-sectional imaging to investigate disparate abdominal symptoms. In a recent study 2.4% of individuals undergoing abdominal MRI for health screening were found to have a pancreatic cyst [1] A proportion of these cystic lesions will have malignant potential. Mucinous cysts
* Corresponding author. HPB Unit, Freeman Hospital, Newcastle upon Tyne NE7 7DN, UK. E-mail address: Kofi[email protected] (K.W. Oppong).
form a category with variable malignant potential. Correctly identifying which lesions are mucinous; intraductal papillary mucinous neoplasm (IPMN) or mucinous cystic neoplasm (MCN) is an important step in the diagnosis and risk stratification of cystic neoplasms. Endoscopic ultrasound (EUS) and EUS guided fine needle aspiration (FNA) may have a role in the diagnosis of pancreatic cysts. Several large series have found an elevated cyst fluid carcinoembryonic antigen (CEA) to be the single most useful test in predicting mucinous cysts with cut-off values of 192 ng/ml having a sensitivity of 73% [2] and the value of 1093 a sensitivity of 80.8%. The CEA cut-off of 192 ng/ml was recommended in the recent IAP consensus guidance [4]. A very low CEA (cut-off 5 ng/ml) has been
http://dx.doi.org/10.1016/j.pan.2015.08.001 1424-3903/Copyright © 2015, IAP and EPC. Published by Elsevier India, a division of Reed Elsevier India Pvt. Ltd. All rights reserved.
Please cite this article in press as: Oppong KW, et al., EUS and EUSeFNA diagnosis of suspected pancreatic cystic neoplasms: Is the sum of the parts greater than the CEA?, Pancreatology (2015), http://dx.doi.org/10.1016/j.pan.2015.08.001
2
K.W. Oppong et al. / Pancreatology xxx (2015) 1e8
suggested as specific for the diagnosis of serous cyst adenoma (SCA) [5,6]. However the practical utility of cyst fluid CEA may be overstated as sufficient fluid for analysis may not be obtained [7], and diagnostic performance is often reported excluding cysts from which insufficient fluid was aspirated [3] rather than on an intention-to-diagnose basis. Cyst fluid amylase has been suggested as useful in differentiating pseudocysts from non pseudocysts [6] and in differentiating benign from malignant mucinous cysts [8]. The diagnostic utility of cyst fluid cytology in the literature is variable; whilst some series have reported excellent accuracy [9], the largest series [2,3] and a meta-analysis [6] have documented less favourable performance. A visual assessment of the viscosity of aspirated material has also been suggested as useful in categorising a cyst as mucinous [10]. We undertook this study to perform an analysis of EUS/ EUSeFNA on an intention to diagnose basis; assessing the diagnostic performance of recommended CEA cut-offs and comparing the performance of CEA to that of clinical EUS diagnosis, visual assessment of aspirate, and cytology both individually and in combination. We also aimed to assess the utility of cyst fluid amylase in differentiating mucinous from non-mucinous cysts, and pseudocysts from non-pseudocysts. Materials and methods The study population comprised all patients referred to the Hepato-Pancreato-Biliary (HPB) unit at the Freeman Hospital for assessment of suspected pancreatic neoplastic cysts and underwent EUS and FNA over a 10 year period between June 2003 and June 2013. The Freeman Hospital serves as a tertiary referral centre of hepatobiliary and pancreatic surgery for the North East region of England with a population of 3.5 million. All patients were managed by a multidisciplinary team (MDT). There was no specific protocol for the use of EUS. The study period encompassed a time of improving imaging and knowledge of cyst morphology. At any given time point the MDT decision was to use EUS if it was felt that further information to improve clinical decision making would be gained. This study took the form of a review of a prospectively completed database of all EUSeFNA procedures. The database recorded details of cyst size, location, appearance of aspirated fluid, putative diagnosis as recorded in the report as well as CEA, amylase, cytology and histology. All EUS procedures were undertaken after obtaining informed patient consent by two experienced endosonographers (K.W.O and M.K.N). Hitachi EUB-7500 or Preirus US workstations (Hitachi Medical Systems, Wellingborough, UK) and Pentax linear echoendoscopes (Pentax, Slough, UK) were used to carry out a standard EUS assessment of the pancreas. FNA of the cyst was carried out under EUS guidance using 22 or 19 gauge FNA needles (Cook Medical, Limerick, Ireland). The endosonographers were not blind to the results of previous imaging or blood tests. To minimise the risk of infection, intravenous antibiotics were administered during EUSeFNA, and oral ciprofloxacin prescribed for a further 3 days. No more than a single pass was made if possible. A specimen was always sent for cytological analysis. If enough fluid was aspirated (1 ml or greater), a sample was also sent for CEA and amylase assay. Histopathological and cytological assessment was performed by pathologists with a special interest in pancreatic disease. The official report was used as the data source. Cyst classification For the purposes of this study all mucinous cysts (IPMN, MCA) and mucinous cyst adenocarcinoma (MCAC) were considered
together, cystic ductal adenocarcinoma were included in this category. SCA, pseudocyst, solid pseudopapillary neoplasm (SPN) and neuroendocrine tumours (NET) were classified as nonmucinous. In terms of immediate malignant risk: MCAC, malignant IPMN, malignant cystic NET, cystic ductal adenocarcinoma and cysts exhibiting high grade dysplasia were classified as high risk. EUS For the purposes of this study the diagnosis stated in the official EUS report was used as the EUS diagnosis. The EUS report documented the appearance of the cyst as well as the appearance of the aspirated fluid/material and whether it appeared mucoid/viscid or thin/serous. The EUS diagnosis recorded in the report was arrived at based on cyst appearance and informed by the nature of the aspirate. Cysts with a classical microcystic appearance and a nonmucinous aspirate were classified as SCA. Cysts with a mucoid aspirate were classified as IPMN if a communication with the pancreatic duct (PD) was identified and or the PD was dilated. A diagnosis of MCA was considered in females with a noncommunicating mucinous cyst. Unilocular or minimally septated cysts were classified as macrocystic serous cyst adenoma if no communication was demonstrated with the PD and the aspirated fluid was thin and clear or straw coloured. Thick-walled unilocular cysts containing debris and a thin aspirate were classified as pseudocysts (in individuals with a prior history compatible with pancreatitis). If no diagnosis was stated in the report none was derived. If a differential diagnosis was recorded rather than a single diagnosis this was only included if both diagnosis allowed categorisation as either mucinous or non-mucinous. Cytology The official cytology report was used as the data source, samples were classified as non-diagnostic (acellular and non-mucinous) or diagnostic (mucinous or non-mucinous epithelium, extracellular mucin or inflammatory cells). Cysts were categorised as mucinous if mucinous epithelium, or extracellular mucin without gastrointestinal contaminant cells was documented and non-mucinous if diagnostic cytology documented inflammatory cells only or nonmucinous epithelium. Data collection and analysis All the data was collected prospectively and entered into a database. The study group comprised patients with a final definitive tissue diagnosis. This included diagnostic malignant cytology, resection histology or biopsy histology. Sensitivity, specificity, and accuracy were determined for visual assessment of aspirate, cytology; diagnosis recorded on the EUS report; as well as for the CEA cut-offs of 192 ng/ml, 110 ng/ml and 30 ng/ml reported from previous series [2,3,8] to differentiate between mucinous and non-mucinous cysts. Receiver operating characteristic (ROC) curves were plotted for CEA (mucinous versus non mucinous) and for amylase (pseudocyst versus non pseudocyst); cut off values were selected to maximise the proportion of correct cyst classification. The performance of combinations of tests was also assessed. Using combination testing a cyst was deemed as mucinous (or malignant) if any of the constituent tests classified it as such. All analyses were performed using the entire dataset of procedures with established tissue diagnosis (henceforth referred to as “intention-to-diagnose analysis”) as well as the subset with a recorded EUS diagnosis recorded; adequate cyst specimen cytology or sufficient aspirate for CEA assay (henceforth referred to as “complete case analysis”). For the purposes of the intention-todiagnose analysis procedures with missing data were considered either false negative (in the case of a mucinous/malignant cyst) or
Please cite this article in press as: Oppong KW, et al., EUS and EUSeFNA diagnosis of suspected pancreatic cystic neoplasms: Is the sum of the parts greater than the CEA?, Pancreatology (2015), http://dx.doi.org/10.1016/j.pan.2015.08.001
K.W. Oppong et al. / Pancreatology xxx (2015) 1e8
false positive (in the case of a non-mucinous cyst). Statistical comparison of quantitative data was made using the ManneWhitney U test. Comparison of gender proportion was performed using Chi Squared test. Comparison of diagnostic sensitivity specificity and accuracy was performed using Cochran's Q test. If the Cochrane Q test was positive (p < 0.05) a test for pairwise comparison of variables according to Sheskin was automatically performed. Test performance was calculated on an intention to diagnose basis as well as per patient tested. Statistical analysis was performed using MedCalc Statistical Software version 13.1.0 (MedCalc Software bvba, Ostend, Belgium; http://www.medcalc. org; 2014). Ethics All procedures were done as a part of standard patient care as directed by the HPB MDT and not to a research protocol. Data collection was performed as part of our ongoing quality monitoring. In accordance with U.K. National Research Ethics Service guidelines formal ethical review was not required. Results During the study period 433 EUSeFNA procedures were performed on 388 individuals; 37 individuals had 2 or more procedures (Fig. 1). Adequate samples for cytology were obtained in 262 cases (60.5%). Cyst aspirate appearance was recorded in 392 procedures (90.5%). A CEA value was determined for 248 procedures (57.3%) and an amylase value in 254 (58.7%). One hundred and thirty three procedures (119 individuals) had a final definitive tissue diagnosis and formed the study group. Diagnosis was based on surgical resection in 102, malignant cytology in 8, biopsy histology in 6, and a cytological diagnosis of NET in 3 (Fig. 1 details flow of patients) The average age was 61.4 years (range 19e84), and the median cyst size was 4.0 cm. There were 79 mucinous cysts (31 benign IPMN, 19 MCA, 8 MCAC, 7 adenocarcinoma, 6 undifferentiated mucinous malignancy, 4 IPMN with high grade dysplasia, 3 malignant IPMN and 1
3
undifferentiated malignancy). There were 40 non-mucinous cysts (12 SCA, 11 pseudocysts, 11 NET, 2 gastrointestinal stromal tumours (GIST), 1 cystic lymphangioma, 1 phaeochromocytoma, 1 SPN and 1 Schwannoma). Considering all individuals with a definitive diagnosis there was a female preponderance 82/119 (68.9%, p < 0.001). CEA (58.6%) and amylase (57.9%) were determined in significantly fewer procedures than EUS diagnosis (91.7%), fluid appearance (89.5%) or adequate cytology (76.7%) (p < 0.0001).
Fluid analysis Carcinoembryonic antigen CEA values were determined in 78 (58.6%) of the study group comprising 50 mucinous and 28 non-mucinous cysts. Fig. 1 is a box plot of CEA values by cyst type. Table 1 summarises the differences between cyst categories. Median CEA was significantly higher in mucinous (175 ng/ml) than non-mucinous (3 ng/ml) p < 0.0001 and in high risk mucinous cysts (8945 ng/ml) compared to benign mucinous cysts (93 ng/ml) p < 0.001. An ROC analysis was performed classifying all mucinous cysts together. The area under the curve (AUC) was 0.89. The cut off level that produced the highest accuracy was 7 ng/ml. At this level the sensitivity was 94%, specificity 75% and accuracy 87.1% (Table 2). The diagnostic performance of 3 CEA cut-off (complete case and intention to diagnose) of 3 large series 192 ng/ml, 110 ng/ml and 30 ng/ml are shown in Table 2. With a Cochran's Q test we found a significant difference among the sensitivity and accuracy of cutoffs. The internally derived cut-off of 7 ng/ml provided superior discrimination of mucinous from non-mucinous cysts than the cutoffs of 192 ng/ml and 110 ng/ml. The discriminative performance of CEA cut-offs of 30; 110; and 192 ng/ml was similar. In terms of diagnosing SCA a very low CEA has been suggested as showing good diagnostic performance [5,11]. In our series 21 cases had a value below 5 ng/ml (Fig. 1). Of these 6 were SCA, 1 SCA had a value above 5 ng/ml. All NET had a low CEA value with 7/8 below 5 ng/ml. The diagnostic performance of a cut-off of 5 ng/ml in diagnosing SCA was sensitivity 85.7% specificity 78.8%.
Fig. 1. Flow chart of patients in the study and adequacy assessment.
Please cite this article in press as: Oppong KW, et al., EUS and EUSeFNA diagnosis of suspected pancreatic cystic neoplasms: Is the sum of the parts greater than the CEA?, Pancreatology (2015), http://dx.doi.org/10.1016/j.pan.2015.08.001
4
K.W. Oppong et al. / Pancreatology xxx (2015) 1e8
Table 1 Comparison of median levels of cyst fluid CEA & amylase using different categories of cyst. Cyst category
CEA, ng/ml
Non-mucinous cysts Mucinous cysts (including PDAC) Benign mucinous cysts High risk mucinous cysts (HGD & malignant) Benign cysts High risk cyst (HGD & malignant) Non-pseudocysts Pseudocysts
Amylase, IU/L
n
Median (IQR)
P value
n
Median (IQR)
P value
28 50 33 17 61 17 69 9
3 175 93 6432 29 6432 82 54
0.0001
28 49 32 17 60 17 68 9
161 2624 11429 113 3103 113 207 32800
ns
(1e12.5) (35e2265) (28e686) (246.7e29510.7) (3.0e141.2) (246.7e29510.7) (3.75e774.2) (15.25e113.5)
0.0004 0.0004 ns
(50.5e21703) (62.5e21730) (655e33341) (32e1235.7) (62e26495) (32e1235.7) (50.5e16245.5) (22525.5e46330)
0.0031 0.0175 0.0011
Table 2 Complete case and intention-to-diagnose analyses of the diagnostic performance for EUS diagnosis; aspirate appearance; aspirate cytology; and aspirate CEA for discriminating mucinous from non-mucinous cysts.
Sensitivity - Complete case - Intention-to-diagnose Specificity - Complete case - Intention-to-diagnose Accuracy - Complete case - Intention-to-diagnose AUC - Complete case
EUS diagnosis (n ¼ 122)
Cytology (n ¼ 102)
Aspirate appearance (n ¼ 119)
CEA >7 ng/ml (n ¼ 78)
CEA >30 ng/ml (n ¼ 78)
CEA >110 ng/ml (n ¼ 78)
CEA >192 ng/ml (n ¼ 78)
83.9% (68/81) 76.4% (68/89)
72.2% (52/72) 58.4% (52/89)
75.9% (60/79) 67.4% (60/89)
94% (47/50)m 52.8% (47/89)
39/50 (78%) 39/89 (43.8%)
62.0% (31/50)m 31/89 (34.8%)
48% (24/50)m 26.9% (24/89)
73.3% (22/30) 50% (22/44)
86.7% (26/30) 59.1% (26/44)
72.5% (29/40) 65.9% (29/44)
75% (21/28) 47.7% (21/44)
78.6% (22/28) 50% (22/44)
92.9% (26/28) 59.1% (26/44)
96.4% (27/28) 27/44 (61.3%)
81.1% (90/111) 67.7% (90/133)*^
76.5% (78/102) 58.6% (78/133)*
74.8% (89/119) 66.9% (89/133)*^
87.1% (68/78)m 51.1% (68/133)
78.2% (61/78) 45.8% (61/133)^
73.1% (57/78)m 42.8% (57/133)^
65.4% (51/78)m 51/133 (38.3%)*
0.79
0.79
0.74
0.89
0.78
0.77
0.72
m
Complete case (CEA cutoffs): Cutoff of 7 ng/ml significantly more sensitive and accurate than 192 and 110 ng/ml p < 0.05 (n ¼ 78). Complete case (all tests): no significant difference in accuracy of matched cases (n ¼ 45). Intention to diagnose: CEA cut-off of 192 ng/ml significantly less accurate than EUS diagnosis, cytology or aspirate appearance p < 0.05*. CEA cut-offs of 110 and 30 ng/ml were significantly less accurate than EUS diagnosis and aspirate appearance p < 0.05^.
Amylase Amylase values were determined in 77/133 (57.9%) procedures, comprising 43 mucinous and 34 non mucinous (including 9 pseudocysts) cysts. Fig. 2 is a box plot of cyst fluid amylase by cyst type. Table 1 summarises the differences between cyst categories. The median amylase value in pseudocysts (32,800 IU/L) was significantly higher than in non pseudocysts 207 IU/L p < 0.01. An ROC curve was constructed, the calculated optimal cut-off for differentiating between pseudocysts and non-pseudocysts was 3977 IU/L, providing a sensitivity of 100%, and specificity 66%. The area under the ROC curve was 0.84. Median amylase was also significantly higher in benign mucinous cysts (11,429 IU/L) than high risk mucinous cysts (113 IU/L) and benign cysts (3101 IU/L) than high risk cysts (113 IU/L) both p < 0.05. The currently recommended cutoff of 250 IU/L for identifying pseudocysts had a sensitivity of 9/9 (100%), specificity of 35/68 (51.6%) and accuracy of 44/77 (57.1%) (see Fig. 3). EUS An EUS diagnosis or differential was recorded in 122/133 (91.7%). Excluding reports with inconsistent differential diagnosis (i.e. categorisation as both mucinous and non-mucinous), there were 111 (83.4%) procedures where the cyst was unambiguously categorised as mucinous or non-mucinous. Of the 111 cysts, 81 were mucinous and 30 non-mucinous. Diagnostic performance on a complete case and intention to diagnose analysis is detailed in Table 2. Cytology Adequate samples for cytology were obtained in 102/133 (76.7%) of procedures. Diagnostic performance is detailed in Table 2.
Aspirate visual appearance Aspirate appearance was recorded in 119/133 (89.5%) of procedures. Diagnostic performance is detailed in Table 2. Comparisons On a complete case (45 matched cases) basis using the Cochran's Q test there was no significant difference in accuracy among EUS diagnosis, aspirate appearance, cytology or the CEA cut-off values of 30, 110 and 192 ng/ml (Table 2). However on an intention to diagnose basis we found a significant difference. A CEA cut-off of 192 ng/ml was significantly less accurate than EUS diagnosis, aspirate appearance and cytology; p < 0.05 whilst cut-offs of 30 and 110 ng/ml were less accurate than EUS diagnosis and aspirate appearance (Table 2). Combination testing The CEA cut-off 110 ng/ml was used. The combinations resulted in improved diagnostic performance (Table 3). Using the Cochran's Q test we found a significant difference among the accuracy of the individual tests and test combinations (detailed in Table 3). All combinations tested were significantly more accurate (p < 0.05) than CEA cut-offs of 7, 110 or 192 ng/ml. In addition the diagnostic accuracy of EUS diagnosis þ cytology þ aspirate appearance þ CEA > 110 ng/ml (85.7%) and of EUS diagnosis þ cytology þ CEA > 110 ng/ml (85.7%) were significantly higher than the individual tests by themselves (p < 0.05): EUS diagnosis (67.7%), cytology (58.6%), and aspirate appearance (66.9%). EUS diagnosis þ cytology þ aspirate appearance þ CEA > 110 ng/ml (85.7%) and EUS diagnosis þ cytology þ CEA > 110 ng/ml (85.7%) were significantly more accurate than CEA > 110 ng/ml þ cytology (p < 0.05).
Please cite this article in press as: Oppong KW, et al., EUS and EUSeFNA diagnosis of suspected pancreatic cystic neoplasms: Is the sum of the parts greater than the CEA?, Pancreatology (2015), http://dx.doi.org/10.1016/j.pan.2015.08.001
K.W. Oppong et al. / Pancreatology xxx (2015) 1e8
5
Fig. 2. Box plot of cyst fluid CEA by cyst type. Y axis is on a logarithmic scale (nanograms per millilitre). There are 4 lines representing the cut-off levels of 192, 110, 30 and 7 ng/ml. The high risk category includes malignant cysts and those with foci of high grade dysplasia (IPMN, MCAC and PDAC). The Other category is made up of two GIST, one cystic lymphangioma and one phaeochromocytoma. All categories are mutually exclusive.
Discussion The role of EUS/EUSeFNA in the management of suspected neoplastic pancreatic cysts remains to be fully defined. There is
however a consensus that the decision on its use is best made in a multidisciplinary team setting [12,13] For the majority of patients who do not have symptoms or imaging findings mandating [4,14] surgery, an accurate diagnosis (particularly the discrimination of
Fig. 3. Box plot of cyst fluid amylase by cyst type. Y axis is on a logarithmic scale (international units per millilitre). There is a line representing the cut-off of 250 IU/L The high risk category includes malignant cysts and those with foci of high grade dysplasia (IPMN, MCAC and PDAC). The Other category is made up of two GIST, one cystic lymphangioma and one phaeochromocytoma. All categories are mutually exclusive.
Please cite this article in press as: Oppong KW, et al., EUS and EUSeFNA diagnosis of suspected pancreatic cystic neoplasms: Is the sum of the parts greater than the CEA?, Pancreatology (2015), http://dx.doi.org/10.1016/j.pan.2015.08.001
6
K.W. Oppong et al. / Pancreatology xxx (2015) 1e8
Table 3 Intention-to-diagnose analysis of the diagnostic performance of multiple test combinations for discriminating mucinous from non-mucinous cysts (utilising a CEA cut-off of 110 ng/ml).
Sensitivity Specificity Accuracy
Cytology þ aspirate appearance
CEA > 110 ng/ml þ Cytology
EUS diagnosis þ Cytology
Cytology þ CEA > 110 ng/ml þ aspirate appearance
EUS diagnosis þ cytology þ aspirate appearance þ CEA > 110 ng/ml
EUS diagnosis & cytology þ CEA > 110 ng/ml
84.2% (75/89) 72.7% (32/44) 80.4% (107/133)
77.5% (69/89) 75% (33/44) 76.7%*m (102/133)
86.5% (77/89) 68.2% (30/44) 80.4% (107/133)
89.9% (80/89) 72.7% (32/44) 84.2% (112/133)
93.2% (83/89) 70.5% (31/44) 85.7%* (114/133)
91.0% (81/89) 75% (33/44) 85.7%m (114/133)
*p < 0.05. p < 0.05.
m
mucinous from non-mucinous cyst) is important as it facilitates risk stratification and decisions regarding the need for surveillance, surveillance interval and duration. However this remains problematic; in one series from a very experienced, high-volume centre, preoperative diagnosis was incorrect in one-third of incidentally discovered cysts and 5% of resected cysts were non-neoplastic [12]. Another series documented a pre-operative accurate diagnosis rate of 60.9% [15]. The high resolution images obtained with EUS initially raised the possibility that neoplastic cysts could be accurately differentiated on morphology alone [16,17] however it is now recognised that considerable overlap exists in cyst morphology [18]. Whilst mural nodules and associated masses are suggestive of malignancy [18e21], other EUS morphological features are not specific. EUS morphology has a reported sensitivity of between 35 and 90% [2,3,9,19,22e26] with the largest single centre series reporting an accuracy of 48% [3]. In addition, interobserver agreement of EUS assessment of pancreatic cysts is not good. A recent study found moderate agreement between expert endosonographers [27] whilst an earlier study found little more than chance interobserver agreement among experienced endosonographers for the diagnosis of neoplastic versus non-neoplastic cysts [2]. EUS morphology alone, therefore, cannot be relied on to accurately differentiate mucinous from non-mucinous cysts. There are a number of factors in a non-blinded study such as this (and the majority of such prior studies) that are a potential bias and might lead to an overestimation of EUS performance these include; knowledge of prior imaging and knowledge of patient demographics (age and gender specifically) however this reflects everyday clinical practice. In the present study a clear mucinous/non-mucinous diagnosis was recorded on the EUS report in the majority of procedures 111/ 133 (83.4%). This diagnosis was based on a combination of cyst morphology and aspirate appearance. On an intention-to-diagnose basis, the sensitivity of the clinical EUS diagnosis was modest (76.4%), with poor specificity (50%) and modest accuracy (67.7%), but these performance characteristics compare favourably to those reported previously [2,9,19,22e24] for EUS morphology alone. Cyst FNA in addition to enabling cytology and fluid analysis (if sufficient volume is aspirated) allows a visual assessment of cyst content which can be a useful adjunct with a clear mucoid/viscid aspirate suggesting a mucinous cyst. However, a thin non-mucoid aspirate does not exclude a mucinous neoplasm [18,28]. Previous investigators have suggested that viscosity may be a useful marker for differentiating mucinous and non-mucinous lesions [29,30]. Measuring viscosity using commercially available tests is not straightforward as a significant volume of aspirate is needed [31]. Debris, blood, and tissue in pancreatic-cyst fluid can also interfere with the measurement of fluid viscosity [31]. Visual assessment of aspirate viscosity utilising the “string sign” has been suggested as a measure of cyst fluid mucin content [10]. In the present study, aspirate appearance was recorded in a high proportion of cases (89.4%) and accuracy on an intention-to-diagnose basis was 67%.
CEA is the most widely investigated and used tumour marker for cyst classification. A number of different cut-off values have been reported in the literature including 192 ng/ml [2], 109 ng/ml [3], 800 ng/ml [6], 50 ng/ml [32] and 30 ng/ml [8]. Interpretation of the literature for use in everyday practice is further complicated by the fact that some studies have assessed diagnostic performance in differentiating mucinous (MCA, MCAC and IPMN) from nonmucinous [2,3,8,9] whilst others have excluded IPMN [6,30]. No CEA assay is validated for use with aspirated cyst fluid, in addition a recent study found that reproducibility of CEA analysis was variable [33]. In the present study, in which IPMN lesions were not excluded and cystic PDAC were included with mucinous cysts, we found a cut-off of 7 ng/ml to provide the largest area under the ROC curve with an accuracy of 87.1%. This is one of the lowest optimal CEA cutoffs reported, further increasing the heterogeneity of results in the literature, and reflects the fact that a significant number of mucinous cysts, primarily IPMNs, had a low CEA value. The discriminative ability of our internally-derived cut-off was similar to that of 110 ng/ml, derived from the largest single centre series [3], but superior to that of 192 ng/ml. In our study sufficient fluid for CEA analysis was obtained in 58.6% of cases, compared to 49% in a recent Dutch study [4], and 78e79% in the studies by Cizginer et al. [3] and Khashab et al. [25] the performance of CEA on an intention-to-diagnose basis in our cohort was therefore less impressive with a cut-off of 7 ng/ml having an accuracy of 55.9% and a cut-off of 110 ng/ml 47.3%. In the present series, we also documented significantly higher CEA levels in high-risk mucinous cysts compared to benign ones. Previous studies of this issue have reported inconsistent findings [3,8,30,34,35]. The variable rate of obtaining fluid for analysis, lack of clarity in the literature as to whether higher values are seen in high risk cysts and the possible impact of differing assays are all factors that limit the utility of a CEA cut-off used in isolation. The diagnostic performance of cytology for cysts is recognised to be lower than for solid pancreatic lesion [36]. The modest performance of cytology in the diagnosis of mucinous cysts is primarily due to the relatively high proportion of acellular aspirates [3]. A number of studies have reported high rate of acellular aspirates, with two recent series documenting rates of 53e69% [7,37]. Our inadequate aspirate rate of 23.3% was similar to that reported in another recent large series [25]. A recent meta-analysis of 18 studies reported a pooled sensitivity for cytology of 54% and specificity of 93% [38]. This compared with an intention-todiagnose sensitivity of 58.4% and specificity of 59.1% in the present study. We found a significant difference in median amylase levels between pseudocysts and other cysts (32,800 IU/L vs. 207; p < 0.01). Utilising the recommended [6,39] amylase cut-off of 250 IU/L gave a sensitivity of 100%, specificity of 51.5% and an accuracy of 57.1% in line with the findings of previous studies [6,8]. High amylase levels were recorded not only in IPMN as would be expected but also in MCN, supporting the notion of microscopic communication between the pancreatic ductal system and some MCN [8]. We found
Please cite this article in press as: Oppong KW, et al., EUS and EUSeFNA diagnosis of suspected pancreatic cystic neoplasms: Is the sum of the parts greater than the CEA?, Pancreatology (2015), http://dx.doi.org/10.1016/j.pan.2015.08.001
K.W. Oppong et al. / Pancreatology xxx (2015) 1e8
amylase levels to be significantly higher in benign (11,429 IU/L) compared to high-risk mucinous cysts (113 IU/L). Similar results were reported from the Mayo clinic [8]. Intention-to-diagnose analysis reveals the true utility of a diagnostic test. On this basis a CEA cut-off of 110 ng/ml was significantly less accurate (42.8%) than EUS diagnosis (67.7%), cytology (58.6%) or aspirate appearance (66.9%) However, combination testing improved diagnostic performance, yielding very similar results to those reported by the largest single centre series to date [3]. The combination of EUS diagnosis, cytology and CEA resulted in a sensitivity, specificity and accuracy of 91%, 75%, 85.7% and this was more accurate than EUS, cytology, CEA 110 ng/ml or aspirate appearance individually. Potential limitations to this study include the fact that it is a retrospective single tertiary centre study. In addition, as with all such studies, there is a selection bias because the study group is restricted to those with a definitive histological diagnosis. Strengths of the study are that analysis was performed on an intention-to-diagnose basis and the selection process for EUS/ EUSeFNA was by the multi-disciplinary team. Adjuncts to the EUS/ EUSeFNA procedure including molecular analysis of fluid and needle-based confocal laser endomicroscopy (nCLE have been studied. A recent study found no difference in accuracy between CEA/cytology and molecular analysis [40], initial results with nCLE [41,42] are promising, but data is relatively sparse and the rate of pancreatitis appears higher than with EUSeFNA. Further studies are required to delineate its role. In conclusion, in our series a CEA cut-off of 110 ng/ml used in isolation performed poorly in part due to the significant proportion of cases in which sufficient fluid for analysis was not obtained. This in addition to the heterogeneity of cut-off values would suggest caution in using CEA in isolation. However combination with the other results from the EUS/EUSeFNA procedure resulted in good diagnostic performance on an intention-to-treat basis. References [1] de Jong K, Nio CY, Hermans JJ, Dijkgraaf MG, Gouma DJ, van Eijck CHJ, et al. High prevalence of pancreatic cysts detected by screening magnetic resonance imaging examinations. Clin Gastroenterol Hepatol 2010;8:806e11. [2] Brugge WR, Lewandrowski K, Lee-Lewandrowski E, Centeno BA, Szydlo T, Regan S, et al. Diagnosis of pancreatic cystic neoplasms: a report of the cooperative pancreatic cyst study. Gastroenterology 2004;126:1330e6. [3] Cizginer S, Turner BG, Turner B, Bilge AR, Karaca C, Pitman MB, et al. Cyst fluid carcinoembryonic antigen is an accurate diagnostic marker of pancreatic mucinous cysts. Pancreas 2011;40(7):1024e8. ndez-Del Castillo C, Adsay V, Chari S, Falconi M, Jang J-Y, et al. [4] Tanaka M, Ferna International consensus guidelines 2012 for the management of IPMN and MCN of the pancreas. Pancreatology 2012;12:183e97. vy P, Ruszniewski P, Bernades P. Diagnostic [5] Hammel P, Voitot H, Vilgrain V, Le value of CA 72-4 and carcinoembryonic antigen determination in the fluid of pancreatic cystic lesions. Eur J Gastroenterol Hepatol 1998;10:345e8. [6] Waaij LA, Van Der Dullemen HM, Van Porte RJ. Cyst fluid analysis in the differential diagnosis of pancreatic cystic lesions: a pooled analysis. Gastrointest Endosc 2005;62:383e9. [7] de Jong K, Poley J-WJ, van Hooft JE, Visser M, Bruno MJ, Fockens P. Endoscopic ultrasound-guided fine-needle aspiration of pancreatic cystic lesions provides inadequate material for cytology and laboratory analysis: initial results from a prospective study. Endoscopy 2011;43:585e90. [8] Park WG-U, Mascarenhas R, Palaez-Luna M, Smyrk TC, O’Kane D, Clain JE, et al. Diagnostic performance of cyst fluid carcinoembryonic antigen and amylase in histologically confirmed pancreatic cysts. Pancreas 2011;40:42e5. [9] Frossard JL, Amouyal P, Amouyal G, Palazzo L, Amaris J, Soldan M, et al. Performance of endosonography-guided fine needle aspiration and biopsy in the diagnosis of pancreatic cystic lesions. Am J Gastroenterol 2003;98:1516e24. [10] Leung KK, Ross WA, Evans D, Fleming J, Lin E, Tamm EP, et al. Pancreatic cystic neoplasm: the role of cyst morphology, cyst fluid analysis, and expectant management. Ann Surg Oncol 2009;16:2818e24. [11] Hammel P, Levy P, Voitot H, Levy M, Vilgrain V, Zins M. Preoperative cyst fluid analysis is useful for the differential diagnosis of cystic lesions of the pancreas. Gastroenterology 1995;108:1230e5. ndez-del Castillo C. Pancreatic cystic neoplasms: management [12] Farrell JJ, Ferna and unanswered questions. Gastroenterology 2013;144:1303e15.
7
[13] Robinson SM, Scott J, Oppong KW, White SA. What to do for the incidental pancreatic cystic lesion? Surg Oncol 2014;23(3):117e25. €ppel G, Werner J, McKay C, et al. Eu[14] Del Chiaro M, Verbeke C, Salvia R, Klo ropean experts consensus statement on cystic tumours of the pancreas. Dig Liver Dis 2013;45:703e11. [15] Del Chiaro M, Segersv€ ard R, Pozzi Mucelli R, Rangelova E, Kartalis N, Ansorge C, et al. Comparison of preoperative conference-based diagnosis with histology of cystic tumors of the pancreas. Ann Surg Oncol 2014;21:1539e44. [16] Gress F, Gottlieb K, Cummings O, Sherman S, Lehman G. Endoscopic ultrasound characteristics of mucinous cystic neoplasms of the pancreas. Am J Gastroenterol 2000;95:961e5. [17] Koito K, Namieno T, Nagakawa T, Shyonai T, Hirokawa N, Morita K. Solitary cystic tumor of the pancreas: EUS-pathologic correlation. Gastrointest Endosc 1997;45:268e76. [18] Pitman MB, Lewandrowski K, Shen J, Sahani D, Brugge W, Fernandez-del Castillo C. Pancreatic cysts: preoperative diagnosis and clinical management. Cancer Cytopathol 2010;118:1e13. [19] Ahmad NA, Kochman ML, Lewis JD, Ginsberg GG. Can EUS alone differentiate between malignant and benign cystic lesions of the pancreas? Am J Gastroenterol 2001;96:3295e300. [20] Song MH, Lee SK, Kim M-H, Lee HJ, Kim K-P, Kim HJ. EUS in the evaluation of pancreatic cystic lesions. Gastrointest Endosc 2003;57:891e6. [21] Sahani DV, Lin DJ, Venkatesan AM, Sainani N, Mino-Kenudson M, Brugge WR. Multidisciplinary approach to diagnosis and management of intraductal papillary mucinous neoplasms of the pancreas. Clin Gastroenterol Hepatol 2009;7:259e69. [22] Ahmad NA, Kochman ML, Brensinger C, Brugge WR, Faigel DO, Gress FG. Interobserver agreement among endosonographers for the diagnosis of neoplastic versus non-neoplastic pancreatic cystic lesions. Gastrointest Endosc 2003;58:59e64. [23] Hernandez LV, Mishra G, Forsmark C, Draganov PV, Petersen JM, Hochwald SN. Role of endoscopic ultrasound (EUS) and EUS-guided fine needle aspiration in the diagnosis and treatment of cystic lesions of the pancreas. Pancreas 2002;25:222e8. [24] Sedlack R, Affi A, Vazquez-Sequeiros E, Norton ID, Clain JE, Wiersema MJ. Utility of EUS in the evaluation of cystic pancreatic lesions. Gastrointest Endosc 2002;56:543e7. [25] Khashab M a, Kim K, Lennon AM, Shin EJ, Tignor AS, Amateau SK, et al. Should we do EUS/FNA on patients with pancreatic cysts? The incremental diagnostic yield of EUS over CT/MRI for prediction of cystic neoplasms. Pancreas 2013;00:1e5. [26] de Jong K, van Hooft JE, Nio CY, Gouma DJ, Dijkgraaf MGW, Bruno MJ, et al. Accuracy of preoperative workup in a prospective series of surgically resected cystic pancreatic lesions. Scand J Gastroenterol 2012;8e9:1056e63. [27] de Jong K, Verlaan T, Dijkgraaf MG, Poley JW, van Dullemen H, Bruno MJ, et al. Interobserver agreement for endosonography in the diagnosis of pancreatic cysts. Endoscopy 2011;43:579e84. [28] Al-Haddad M, Schmidt MC, Sandrasegaran K, Dewitt J. Diagnosis and treatment of cystic pancreatic tumors. Clin Gastroenterol Hepatol 2011;9:635e48. [29] Lewandrowski KB, Warshaw AL, Compton CC, Pins MR, Southern JF. Variability in cyst fluid carcinoembryonic antigen level, fluid viscosity, amylase content, and cytologic findings among multiple loculi of a pancreatic mucinous cystic neoplasm. Am J Clin Pathol 1993;100:425e7. [30] 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. Gastrointest Endosc 2006;64:697e702. [31] Brugge W. The mistletoe and cyst-fluid analysis: a sticky insight. Gastrointest Endosc 2006;64:703e4. [32] Oh H-C, Kang H, Brugge WR. Cyst fluid amylase and CEA levels in the differential diagnosis of pancreatic cysts: a single-center experience with histologically proven cysts. Dig Dis Sci 2014;59(12):3111e6. [33] Boot CS, Mahon BS, Bramhall SR, Clark PM. Validity of carcinoembryonic antigen and carbohydrate antigen 19-9 measurements in pancreatic cyst fluid with a serum-based immunoassay. Clin Chem 2010;56:1351e2. [34] Kucera S, Centeno BA, Springett G, Malafa MP, Chen YA, Weber J, et al. Cyst fluid carcinoembryonic antigen level is not predictive of invasive cancer in patients with intraductal papillary mucinous neoplasm of the pancreas. JOP 2012;13:409e13. [35] Buscaglia JM, Giday SA, Kantsevoy SV, Jagannath SB, Magno P, Wolfgang CL, et al. Patient- and cyst-related factors for improved prediction of malignancy within cystic lesions of the pancreas. Pancreatology 2009;9:631e8. [36] Wadehra V. The challenge of pancreatic endoscopic ultrasound-guided fine needle aspiration cytology. Cytopathology 2013;24:143e9. [37] Lim LG, Lakhtakia S, Ang TL, Vu CKF, Dy F, Chong VH, et al. Factors determining diagnostic yield of endoscopic ultrasound guided fine-needle aspiration for pancreatic cystic lesions: a multicentre Asian study. Dig Dis Sci 2013;58: 1751e7. [38] Thornton GD, McPhail MJW, Nayagam S, Hewitt MJ, Vlavianos P, Monahan KJ Endoscopic ultrasound guided fine needle aspiration for the diagnosis of pancreatic cystic neoplasms: a meta-analysis. Pancreatology 13:48e7. [39] Khalid A, Brugge W. ACG practice guidelines for the diagnosis and management of neoplastic pancreatic cysts. Am J Gastroenterol 2007;102:2339e49. [40] Al-Haddad M, DeWitt J, Sherman S, Schmidt CM, LeBlanc JK, McHenry L, et al. Performance characteristics of molecular (DNA) analysis for the diagnosis of mucinous pancreatic cysts. Gastrointest Endosc 2014;79:79e87.
Please cite this article in press as: Oppong KW, et al., EUS and EUSeFNA diagnosis of suspected pancreatic cystic neoplasms: Is the sum of the parts greater than the CEA?, Pancreatology (2015), http://dx.doi.org/10.1016/j.pan.2015.08.001
8
K.W. Oppong et al. / Pancreatology xxx (2015) 1e8
[41] Konda VJ, Meining A, Jamil LH, Giovannini M, Hwang JH, Wallace MB, et al. A pilot study of in vivo identification of pancreatic cystic neoplasms with needle-based confocal laser endomicroscopy under endosonographic guidance. Endoscopy 2013;45:1006e13.
[42] Lemaistre A, Pujol B, Caillol F, Lucidarme D, Morellon-mialhe B, Fumex F, et al. A novel approach to the diagnosis of pancreatic serous cystadenoma: needlebased confocal laser endomicroscopy. Endoscopy 2015;47(1):26e32.
Please cite this article in press as: Oppong KW, et al., EUS and EUSeFNA diagnosis of suspected pancreatic cystic neoplasms: Is the sum of the parts greater than the CEA?, Pancreatology (2015), http://dx.doi.org/10.1016/j.pan.2015.08.001