Aims The cell block technique for assessing endoscopic ultrasound-guided fine needle aspiration (EUS-FNA) specimens from pancreatic mucinous cystic lesions (MCLs) was systematically evaluated for the first time, including comparisons with three traditional methods of assessing such specimens.
Methods The prospective arm comprised EUS-FNA specimens from EUS-suspected pancreatic MCLs. The retrospective arm comprised EUS-FNA specimens from pancreatic MCLs surgically resected before the study start. For each specimen, these data points were collected: macroscopic likelihood of mucin, cyst fluid carcinoembryonic antigen (CEA) level and presence of mucin in air-dried, direct smears and in cell block preparations.
Results The prospective and retrospective arms of the study comprised 80 and 30 EUS-FNA specimens, respectively. Seven prospective cases led to surgical resections during the study, and therefore, 37 EUS-FNA specimens were confirmed to have originated from MCLs. In the prospective arm, macroscopic mucin was suspected, cyst fluid CEA level exceeded 192 ng/mL, mucin was detected in direct smears and cell block preparations in 78%, 30%, 39% and 73% of cases, respectively. Of the 37 specimens confirmed to originate from MCLs, macroscopic mucin assessment, cyst fluid CEA level, direct smear mucin assessment and cell block mucin assessment had sensitivities for diagnosing MCL of 87%, 45%, 45% and 81%, respectively.
Conclusions Cell block preparations are as likely to identify mucin from pancreatic MCLs as macroscopic assessment but are twice as likely to diagnose MCL than direct smears and fluid CEA biochemistry. The cell block technique is easy for sample collection and processing especially because these are identical for solid and cystic pancreatic lesions.
- fine needle aspiration
- cell block
Statistics from Altmetric.com
Endoscopic ultrasound-guided fine needle aspiration (EUS-FNA) is recognised as the optimal method of sampling tissue in or around the pancreas. Previous studies of how best to process these EUS-FNA specimens have concentrated on sampling of solid masses/nodes.1 By contrast, there has been scant research into the optimal way of handling EUS-FNA specimens from pancreatic cysts. In particular, the optimal way of processing specimens from possible intraductal pancreatic mucinous neoplasm (IPMN) or mucinous cystic neoplasm—collectively referred to hereafter as mucinous cystic lesions (MCLs)—remains controversial. Current methods of assessing an EUS-FNA specimen for an MCL include: macroscopic assessment by the EUS operator (for mucoid fluid, eg, using the ‘string sign’2); cyst fluid biochemistry; and microscopic assessment of the fluid for mucin and cells.3 4 There have also been recent suggestions that KRAS and GNAS mutation testing can help diagnose IPMN though such genetic analyses are not universally available.3 It is well recognised that cellular assessment of EUS-FNA specimens from pancreatic cysts has a low sensitivity and frequent non-diagnostic yield due to low cellularity.3 Therefore, the following paper has focused on mucin assessment. Regarding the processing of EUS-FNA specimens from pancreatic cysts, recent guidelines have mentioned the use of direct smears and cytospin or liquid-based cytology (LBC) techniques3 but have not included cell block preparation. Based on observational studies of sampling pancreatic or subepithelial masses or lymph nodes, preparing EUS-FNA specimens as cell blocks is now recognised as superior to direct smears and LBC, in terms of sensitivity, accuracy and negative predictive value.1 However, whether the cell block technique has an equivalent role in assessing pancreatic cysts is uncertain. Chai and colleagues used a cell block technique among several different assessments of pancreatic cyst EUS-FNA specimens.5 However, only 26 MCLs were studied (of which only four were resected), there was no comment on how often mucin was detected in their cell block preparations, and the latter were not directly evaluated against other cytological preparations.
For assessing pancreatic cysts, cell block preparation has several advantages. First, the preparation of one or, at most, two cell blocks from one EUS-FNA specimen concentrates all its contents. There is no need to divide the specimen for various analyses, and the same cell block can be used for histochemistry, immunohistochemistry and genetic analyses. Second, because some pancreatic cysts are actually cystic variants of solid neoplasms (eg, neuroendocrine tumour), diagnosis of these neoplasms is best made with immunohistochemistry, and cell blocks are optimal for such analysis. Finally, the previously suggested techniques for processing EUS-FNA specimens from pancreatic cysts have disadvantages. It is difficult to perform ancillary studies on direct smears. Cytospin preparations do not concentrate the specimen and still split the specimen contents across different analyses. LBC techniques may concentrate the specimen contents but are designed to remove mucus/mucin.6 7
The following study is therefore the first known to systemically evaluate whether cell block preparation is suitable for the processing and assessment of pancreatic MCL EUS-FNA specimens. Specifically, the cell block technique will be compared against direct smears (traditionally regarded as the gold standard for the microscopic assessment of mucin), biochemical analysis and the EUS operator's macroscopic assessment for mucin.
Materials and methods
The study only considered EUS-FNA specimens collected from suspected (by EUS using previously published criteria8) or histologically confirmed pancreatic MCLs. All these specimens had been collected at the Queens Day Unit of the Bristol Royal Infirmary (Bristol, UK). In the prospective arm of the study, most specimens included a pair of air-dried, direct smears of the aspirated fluid and then the remaining fluid submitted in 30–40 mL of formalin in a single container. Infrequently, when only a small amount of cyst fluid was aspirated, all the fluid was formalin fixed, and no direct smears were prepared. In the retrospective arm, most specimens only comprised formalin-fixed material without direct smears. When the formalin-fixed sample arrived at the cellular pathology department, if there were any tissue fragments that could be picked out with forceps, this tissue was processed as a separate cassette (using the conventional protocol for biopsies) to produce a paraffin embedded block. The remainder of the formalin-fixed sample was then spun down and processed as a paraffin embedded cell block. Up until May 2016, cell block preparation used an agar-embedding method. After May 2016, cell block preparation used a thrombin clot method.
For each paraffin embedded block, three levels of H&E-stained sections (figure 1D) and then an Alcian blue diastase periodic acidic Schiff (ABDPAS) section were prepared. These sections were routinely assessed and reported by one or more of the consultant pathologists listed as study authors (SB, ZM and NACSW) especially for the presence of free-floating mucin and its subtype, that is, neutral or acidic or mixed (figure 1E,F). Our earlier experience of cell block preparations had been that mucin was often not identified and, at most, might be suspected from microscopic assessment of H&E-stained sections but could only be confirmed with ABDPAS staining. While cell block preparations of EUS-FNA specimens can uncommonly include more granular or solid material which is AB or DPAS positive, only material with a ‘soft’, wispy morphology as well as such positivity (figure 1E,F) was diagnosed as mucin. While the cell block sections were especially assessed for mucin for the purposes of this study, any cellular component was also assessed for, described and interpreted in the diagnostic report. Therefore, any specimens with features of pseudocyst (eg, inflammation, bile or haematoidin pigment, macrophages with or without pigment contents and necrotic fat), serous cystadenoma or neuroendocrine tumour were excluded from the prospective arm of the study.
If air-dried, direct smears had been submitted, these were May Grunwald Giemsa-stained (figure 1A,B) and assessed for mucin (figure 1C) by the reporting pathologist before assessing the cell block sections. Mucin was diagnosed on a direct smear when the material showed a wispy and/or fibrillar texture with or without strands and metachromasia (figure 1C), but its nature (eg, watery vs thick colloid) was not subclassified further.
The prospective arm of the study comprised EUS-FNA specimens collected after February 2017. For each specimen, the following data were recorded: site of the lesion aspirated (pancreatic body/tail or head); CEA level of the aspirated fluid (submitted if there was spare fluid available after preparation of direct smears and/or formalin fixed specimen); and EUS operator’s macroscopic diagnosis of mucin (based on the ‘string sign’2 and categorised as likely, possible or unlikely). The criteria used for the latter were as follows: ‘likely’ if the string of fluid was at least 1 cm long and lasted at least 1 s; ‘possible’ if the string of fluid was less than 1 cm long and/or lasted less than 1 s; and ‘unlikely’ if no string was formed.
The retrospective arm comprised EUS-FNA specimens that were collected between January 2014 and February 2017 and were followed by surgical resection of an MCL. These retrospective cases were identified from a list of all pancreatic operations performed since 2013, provided by the Hepatobiliary Surgical Unit at the Bristol Royal Infirmary; MCLs with adenocarcinoma were excluded from these retrospective cases because many malignant cases showed EUS features of a solid nodule or thickened cyst wall that may have been preferentially sampled (rather than cyst fluid) by FNA and because the possibility of a mucinous carcinoma would confound this study’s analysis of mucin from MCLs alone. Cell block preparation of pancreatic EUS-FNA specimens had been used in Bristol from January 2014 onwards, but several of these retrospective cases had not been sampled with direct smears or had a recorded macroscopic assessment for mucin.
Eighty pancreatic cyst EUS-FNA specimens met inclusion criteria for the prospective arm, and 30 specimens were identified in the retrospective arm. While this study therefore recruited a total of 110 specimens, a varying proportion of these specimens had data for each of the techniques assessed (tables 1 and 2).
Changing the cell block protocol from an agar-embedding to a thrombin clot method did not alter the likelihood of detecting mucin in the resulting slides (data not shown). With all the cell block preparations assessed, the presence or absence of mucin could be unequivocally determined with ABDPAS staining, and therefore, a pathologist was never uncertain whether mucin was demonstrated. By contrast, 5 of the 68 direct smears of the prospective cohort were only considered to possibly contain mucin. These five cases were still classified as demonstrating mucin for this study to compare cell block assessment against the maximum diagnostic potential of direct smears. The presence of mucin was considered possible (rather than likely) by macroscopic assessment in 40% and 23% of the prospective and retrospective cases, respectively (tables 1 and 2). Again to consider the maximum diagnostic potential of macroscopic assessment, cases with both possible and likely macroscopic mucin were combined for the following analyses (tables 1 and 2).
Table 1 compares the findings of the four different assessment techniques among the specimens of the prospective arm. While mucin seemed less commonly detected in direct smears from head as opposed to body/tail lesions, this difference was not statistically significant (p=0.08, corrected χ2 test). There were no statistical differences in the findings of the other assessments between head versus body/tail lesions (data not shown). Of the 18 cases where macroscopic mucin was not suspected, seven did show free floating mucin in the cell block preparations. Of the 21 cases where free floating mucin was not seen in cell block preparations, macroscopic mucin was considered unlikely, possible and likely in nine, eight and four cases, respectively.
Surgical resection of a pancreatic MCL followed 37 of the 110 specimens (7 from the prospective cohort and all 30 of the retrospective cohort). table 2 compares the findings of the four different assessment techniques among these 37 specimens. Because they all had histological confirmation of MCL, the data in this table represented sensitivity of each technique for diagnosing pancreatic MCL. While mucin seemed more commonly detected in cell block preparations from head as opposed to body/tail lesions, this difference was not statistically significant (p=0.21, two-tailed Fisher’s exact test). There were no statistical differences in the findings of the other assessments between head versus body/tail lesions (data not shown). Of the three cases where macroscopic mucin was not suspected, one did show free floating mucin in the cell block preparations. Of the seven cases where free floating mucin was not seen in cell block preparations, the likelihood of macroscopic mucin was not commented on in three cases, and macroscopic mucin was considered unlikely, possible and likely in two, one and one case, respectively.
This study has demonstrated that the cell block technique is more likely to demonstrate mucin from EUS suspected or histologically confirmed pancreatic MCLs than direct smears and is more sensitive at diagnosing these lesions than fluid CEA biochemistry. One limitation of this study (and inevitably of all studies comparing different analyses from one specimen) is that each EUS-FNA specimen had to be divided among 2–3 different analyses and therefore the diagnostic power of each analysis from an undivided EUS-FNA specimen could not be assessed. However, it is noted that in the prospective arm protocol where there was enough material for both preparations, direct smears were always prepared before the cell block specimen. In our experience, the first needle passes/aspirations are more representative of cyst contents without, in particular, contamination from procedure-related bleeding. Therefore, if prepared from the first pass aspirates, cell block specimens might have shown an even higher proportion with mucin than direct smears. Because the EUS-FNA specimens for the prospective arm of this study were already being split between up to three analyses, a comparison against other techniques previously proposed for pancreatic cysts (ie, cytospin and LBC) was not practically feasible. However, the very purpose of LBC techniques is to retain cells but remove debris,6 7 and such techniques are therefore expected to remove mucin from an EUS-FNA specimen. A direct two-way comparison between cell block and cytospin preparations could be considered for the future. However, whereas the cell block technique concentrates an EUS-FNA specimen into one or two blocks and allows mucin staining to be performed on the same block(s), cytospin preparations would need to divide a specimen for additional studies including mucin staining. Finally, this study had focused on microscopic assessment for mucin (rather than of aspirated epithelium) for diagnosing MCLs because of the frequent low cellularity of EUS-FNA specimens of these lesions. However, such an assessment of aspirated epithelium and mucin subtypes is now part of an ongoing study.
It is acknowledged that the prospective arm of this study recruited from cases of pancreatic MCL suspected by EUS and, up to the time of writing this paper, only 7 of the 80 cases of this arm had led to surgical resection (which confirmed the MCL diagnosis in all seven cases). However, this study also included a retrospective cohort of 30 MCL cases confirmed from resection specimens. The fact that macroscopic mucin, mucin in cell block preparations and elevated fluid CEA were all less frequent in the prospective cohort than the retrospective cohort might mean that some of the prospective cases were not MCLs. Furthermore, it is a common issue with all studies where the diagnosis of MCL is made only on clinical/radiological grounds and/or when only resected MCLs are considered, that it is difficult to calculate specificity rates of preoperative assessments for MCL. However, the differences between these assessments for our retrospective and prospective cohorts were only minor (eg, mucin was present in 73% vs 80% of prospective and retrospective cell block preparations, respectively). By the fact that surgery was recommended, resected MCLs would be generally larger than those left in situ, and therefore, the prospective cohort might include smaller cysts that were more difficult to sample. Finally, there were still similar relative trends in the data from both prospective and retrospective cohorts. In particular, both cohorts showed a similar degree of superiority of cell block preparation in demonstrating mucin compared with direct smears and fluid CEA biochemistry.
The superiority of cell block preparation over direct smears is in at least three ways. First, approximately twice as many cases showed mucin in the cell block preparations than in direct smears. This high false negative rate for direct smears was still seen in pancreatic MCLs confirmed by resection histology (table 2). Second, ABDPAS staining greatly assists in the histological confirmation of mucin but such staining cannot be readily performed on an air-dried, direct smear. Third, ease of assessing a direct smear is very dependent on its initial preparation (a thick smear across the whole face of a slide–figure 1B—takes longer to screen and even then may be unsatisfactory for mucin assessment); the processing of cell block preparations is much more standardised (figure 1D) as is therefore the assessment of the resulting slides.
The sensitivity of cyst fluid CEA biochemistry for diagnosing MCL in our study (and especially in the cohort confirmed by surgical resection, ie, 45%, table 2) first seemed low, especially when compared with earlier reported sensitivity rates of more than 70% with the same threshold of 192 ng/mL.9 However, subsequent studies reported sensitivity rates as low as 52% or even 28%.3 10 It is also worth noting that an elevated cyst fluid CEA level is not specific to a pancreatic MCL. Such an elevated level can be seen especially with lymphoepithelial cysts because squamous epithelium expresses CEA.11
The only assessment for mucin that compared similarly with cell block preparation in this study was macroscopic examination of the aspirated fluid by the EUS operator, though only when categories of ‘likely’ and ‘possible’ macroscopic mucin were combined. The ‘string sign’ for macroscopically assessing for mucin has a reported sensitivity of 58% for diagnosing mucinous pancreatic cysts based on 98 histologically proven cases.2 This is similar to the sensitivity we recorded (ie, 62%) based on 37 histologically proven MCLs and considering only the ‘likely’ macroscopic mucin category (table 2). There were only a few exceptions to the absolute correlation between macroscopic impression of mucin (categorised as ‘likely’ or ‘possible’) and mucin in cell block preparations shown by our study. These exceptions were present in both the prospective and retrospective cohorts and went both ways, that is, where mucin was suspected macroscopically but not seen in cell block preparations and vice versa. Macroscopic assessment does not, of course, deprive any FNA material from being processed as cell blocks. Furthermore, on occasion, a pancreatic MCL is suspected by EUS but then shown to represent another neoplasm by the FNA specimen; for example, during our study’s prospective arm, two cystic neuroendocrine tumours were diagnosed histologically instead of EUS-suspected MCL. Therefore, macroscopic examination for mucin and cell block preparations should be viewed as complementary rather than alternative assessments of EUS-FNA specimens from pancreatic cystic lesions.
In conclusion, cell block preparation is an optimal technique for processing and assessing EUS-FNA specimens from pancreatic MCLs and regardless of their location within the pancreas. It is particularly easy for EUS operators in requiring only a simple sample collection (ie, all aspirated material into a formalin pot) that can be used for all pancreatic lesions, that is, solid or cystic or both. It is also particularly easy for pathology laboratories in requiring a single processing pathway, again for all pancreatic lesions. Finally, with increasing interest in genetic testing of EUS-FNA specimens, all the aspirated material stored as a cell block is potentially available for contemporaneous or later testing, and there is therefore no need to remove material from the diagnostic specimen as a separate genetic sample.
Take home messages
Cell block preparation is an optimal method for processing endoscopic ultrasound-guided fine needle aspiration specimens assessing for mucinous cystic lesions of the pancreas.
The cell block technique is more likely to demonstrate mucin from EUS suspected or histologically confirmed pancreatic mucinous cystic lesions than direct smears and is more sensitive at diagnosing these lesions than fluid CEA biochemistry.
The cell block technique is easy for sample collection and processing especially because these processes are identical for solid and cystic pancreatic lesions.
We would like to thank Mr Jonathan Rees for providing data on surgical resections from the Hepatobiliary Surgical Unit at the Bristol Royal Infirmary.
Handling editor Runjan Chetty.
Contributors NACSW conceived the idea for this study. PG and TM compiled the data. NACSW, ZM and SB assessed the endoscopic ultrasound-guided fine needle aspiration (EUS-FNA) specimens. FG, EA and SN generated and macroscopically assessed the EUS-FNA specimens. All authors contributed to the preparation of the manuscript.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests None declared.
Patient consent for publication Not required.
Provenance and peer review Not commissioned; externally peer reviewed.
Data availability statement All data relevant to the study are included in the article or uploaded as supplementary information.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.