INTRODUCTION
Pancreatic and biliary malignant tumors are asymptomatic and are often detected as advanced forms at small sizes, and owing to the difficulty of their diagnosis and treatment, they are cancers with a poor prognosis [
1,
2]. In recent years, percutaneous fine needle aspiration (FNA), using endoscopic ultrasound (EUS), has been a common approach when a pancreatic mass is suspected, while brush cytology through endoscopic retrograde cholangiopancreatography (ERCP) is commonly used when malignancy differentiation is required for lesions in bile ducts.
The obtained cells are subjected to Papanicolaou (or Diff-Quik) staining after being wet-fixed or air-dried, followed by observation under a microscope [
3]. Recently, EUS-FNA for diagnosis of pancreatic solid masses has been reported to have a sensitivity of 78% to 95%, specificity of 75% to 100%, and accuracy of 78% to 95%, a great improvement compared to previous methods, but its negative predictive values (NPV) are still low [
4,
5]. Similarly, biliary brush cytology makes malignancy differentiation challenging because of its low sensitivity of 33% to 68% [
6-
8], and continuous efforts have been made to improve its sensitivity.
Liquid-based cytology (LBC) was developed in 1991, because of the efforts to improve the quality of samples and effectiveness of cytological tests, and since then, it has been widely utilized in various organs and specimens as well as for cervical cytology. With its effectiveness proven for some diseases, it has tended to replace conventional smears (CS) [
9-
11]. However, rarely have studies applied LBC to cytologically analyze the pancreatobiliary tumors and no studies have ever been carried out using CellPrepPlus (CP; Biodyne, Seongnam, Korea), one of the commercially available LBCs.
In this comparison study, CS, CP, and cell block (CB) tests were carried out on cell specimens obtained via EUS-FNA and brushing cytology from patients with pancreatobiliary tumors to investigate the diagnostic accuracy of these tests and to evaluate the diagnostic usefulness of CP versus CS, which is the main objective of this study.
DISCUSSION
Patients diagnosed with advanced pancreatic cancer have a less than 5% 5-year survival rate, and surgical resection is possible in only around 15% of pancreatic cancer cases. On the contrary, patients with early pancreatic cancer less than 2 cm and 1 cm have a 30% to 60% surgical resection rate and a greater than 75% survival rate, respectively; indicating that early detection of pancreatic cancer greatly influences its treatment and prognosis [
1]. Recently, the early detection rate has increased because of an increased frequency of health examinations. With diagnostic accuracy and usefulness, remaining paramount, cytological diagnoses mostly rely on CS.
Similarly, cytological and histological diagnoses are essential for diagnostic confirmation and treatment of cholangiocarcinoma, and early diagnoses have a significant effect on treatment and prognosis. The only radical treatment for cholangiocarcinoma is complete surgical resection. The 5-year survival rate after radical resection of hilar cholangiocarcinoma is 30% to 50%, while it is only 0.4% in patients who do not undergo resection, showing it is difficult to expect long-term survival without surgical resection [
2,
3].
In general, specimens are obtained via a percutaneous or endoscopic approach to the lesions. Pancreatic specimens are obtained using an endoscopic approach, such as EUS-FNA, while biliary specimens are obtained using brush cytology or endobiliary forceps biopsy while performing ERCP. Previous studies suggest that EUS-FNA for diagnosis of a pancreatic solid mass has been improved to a sensitivity of 78% to 95%, specificity of 75% to 100%, and accuracy of 78% to 95%, but its NPV are still low [
4,
5]. Similarly, biliary brush cytology has a high specificity of 95% to 100%, but a low sensitivity of 35% to 68% [
6-
8,
13,
14], which causes difficulties in malignancy differentiations; hence, researchers are continuously striving for its improvement. Because EUS-FNA for diagnosing pancreatic malignant lesions or ERCP for biliary cytology is limited in repetitive testing because of its invasive nature, obtaining adequate specimens through effective tests within a short time period is important [
12].
Making a diagnosis from pancreatobiliary lesion specimens is affected by diverse variables, mainly the experience of the endoscopy operators, nature of the lesions, patient condition, size and shape of the fine needle or brush, tissue treatment procedures, quality of sample slides, and experience of the pathologists. Causes of decreased sensitivity are largely divided into sample collection errors and analysis errors [
13].
Since its first development in 1939, CS have been usefully employed in many fields as a screening test for malignant lesions. However, several limitations were pointed out, including cell overlaps due to the non-uniform smear, insufficient number of cells, interference by inflammatory cells and blood cells, and inadequate specimens from dryness.
To improve sample quality and test efficiency, LBC was developed in 1991 and has been commonly used since then. LBC is largely divided into filtration methods (ThinPrep, CP, E-Prep) and precipitation methods (SurePath, Liqui-PREP). CP, used in this study, is an automated LBC using a filtration method developed in Korea, which utilizes a vacuum filtration system for cell filtration [
11]. Collected samples are suspended in a liquid fixing solution and placed on the test device, followed by attaching the cells suspended in the fixing solution to a filter of adequate pressure. By contacting this to a slide, a thin cell layer slide 20 mm in diameter is fabricated [
15]. Such automated LBC substitutes for the CS, and its usefulness has been reported particularly for obtaining exfoliative cells during cervicovaginal cytology, and it has been applied to various cytological specimens including body fluid, urine, and sputum, showing favorable outcomes.
Compared to the CS, LBC has sound sample quality, ease of use owing to automation, shortened analysis time, fewer cellular overlaps and dryness, and fewer effects from blood and inflammation. Moreover, immunohistochemical and molecular studies are possible on the remaining specimens after LBC, if needed [
11,
16,
17]. However, some studies show LBC has a lower sensitivity due to low cellularity compared to CS, and there is controversy as to its feasibility for some specimens [
4,
18].
Few studies applying LBC for pancreatobiliary lesions have been published and no studies have ever been carried out using CP. This study compared the diagnostic outcomes of three cytological methods including CP, CS, and CB, and evaluated the diagnostic usefulness of CP relative to CS.
An interim analysis was carried out on 50 subjects in the EUS-FNA group and 32 in the ERCP group because a number of inadequate specimens were identified throughout the study. In both the EUS-FNA and ERCP groups, our results showed a significantly lower sensitivity of CP compared to CS, as expected. Therefore, we decided to stop recruiting subjects considering the costs and possibility of complications associated with prolonged operation times.
In the EUS-FNA cytology based analysis of pancreatic specimens, specificity and PPV were the same for all three cytological methods at 100%. LBC had a sensitivity of 60.7%, accuracy of 77.1%, and NPV of 64.5%; CS had a sensitivity of 85.7%, accuracy of 91.7%, and NPV of 83.3%; CB had a sensitivity of 58.3%, accuracy of 76.2%, and NPV of 64.3%.
For the biliary brush cytology, specificity and PPV were also the same for all three cytological methods at 100%. However, sensitivity, accuracy, and NPV were 53.1%, 54.5%, and 6.3% for LBC, respectively; 78.1%, 78.8%, and 12.5% for CS; and 40.6%, 42.4%, and 5.0% for CB.
For EUS-FNA cytology, the sensitivity and specificity of CS were found to be similar to that of previous studies whereas the sensitivity and NPV of CP were remarkably decreased. This is primarily because of the substantial number of inadequate specimens on CP, and the inadequate specimens are all attributable to an insufficient number of cells. In a recent study by Lee et al. [
12], eight inadequate specimens were found on CS whereas 20 were found on LBC (ThinPrep), and LBC had a sensitivity of 75%, accuracy of 81%, and NPV of 56%.
For the biliary brush cytology, the sensitivity of CS was 78.1%, which was higher than or similar to previous studies, but CP’s sensitivity was markedly low at 53.1%. The proportion of inadequate specimens was 6.1% on CS but was much higher for CP at 27.3%, which is attributed to the lack of cells, similar to that of pancreatic FNA.
Among the pancreatobiliary samples, using CP, the monolayer preparation technique, cells can be stained better because of the absence of obscuring blood elements and air-drying artifacts that are commonly encountered in CS (
Fig. 2). However, this did not appreciably influence the cytologic diagnosis. Although CP can provide qualified slides and reduce the analysis time compared to CS, with its many inadequate specimens because of the insufficient number of cells, CP is unlikely to substitute for CS as a diagnostic tool for pancreatobiliary diseases.
A significant limitation of our study was that it was not a randomized trial. The study design might have a lack of standardization when it comes to the number of passes for each method and it is also biased in favor of smear methods. Therefore, caution is needed in analyzing the results. Another limitation of the study includes its small sample size as well as the fact that there was no on-site cytopathologist available who could evaluate the sample’s adequacy at the time of collection.
In addition, cellularity issues on CP can be caused by technical issues during sample preparation. A few studies showed that CP is not inferior compared to other LBCs and is even superior in some aspects [
19]. However, there are no studies that compared CP to CS for FNA from solid organs with a relatively small amount of cells or biliary brush cytology and other LBCs. The CP used in this study utilizes a vacuum filter system for cell filtration. This filtration method using a vacuum filters the cell mass unlike the precipitation methods, and may be unfavorable in terms of cell count.
FNAs of solid organs do not have abundant aspirated cells, and cells are diluted when placed in a liquid medium for LBC. Then, once the cells are attached to slides via filtration, the cell count is further reduced. Such phenomenon would be unfavorable for CP, which uses a greater amount of diluents compared to other filtration liquid-based cytological methods [
11].
Therefore, if techniques that reduce the number of inadequate specimens owing to a lack of cell count are developed, LBC utilization for pancreatobiliary cytology could become feasible.
In this study, LBC (CP), CS, and cellblock were performed prospectively to investigate the usefulness of LBC for 48 pancreatic EUS-FNA specimens and 33 biliary brush cytology specimens obtained from 75 pancreatobiliary patients. Most of the slides using LBC had a lower cellularity and a clear background. Consequently, LBC had a lower diagnostic yield and accuracy when using pancreatic EUS-FNA cytology and biliary brush cytology compared with CS.