In-room cytologic evaluation by trained endosonographer for determination of procedure end in endoscopic ultrasound-guided fine needle biopsy of solid pancreatic lesions: a prospective study in Taiwan
Article information
Abstract
Background/Aims
Endoscopic ultrasound-guided fine needle biopsy (EUS-FNB) is an essential tool for tissue acquisition in solid pancreatic tumors. Rapid on-site evaluation (ROSE) by cytologists ensures diagnostic accuracy. However, the universal application of the ROSE is limited by its availability. Therefore, we aimed to investigate the feasibility of determining the end of the procedure based on the results of in-room cytological evaluation by trained endosonographers (IRCETE).
Methods
A training course focusing on the cytological interpretation of common pancreatic tumors was provided to the three endosonographers. After training, the decision to terminate EUS-FNB was made based on IRCETE results. The diagnostic accuracy, concordance rate of diagnostic categories, and sample adequacy were compared with those determined by board-certified cytologists and macroscopic on-site evaluation (MOSE).
Results
We enrolled 65 patients with solid pancreatic tumors, most of whom were malignant (86.2%). The diagnostic accuracy was 90.8% when the end of the procedure was determined based on IRCETE, compared to 87.7% and 98.5% when determined by MOSE and cytologists, respectively (p=0.060). Based on the cytologists’ results, the accuracy of IRCETE in diagnostic category interpretation was 97.3%.
Conclusions
In the absence of ROSE, IRCETE can serve as a supplementary alternative to MOSE in determining the end of tissue sampling with a high accuracy rate.
INTRODUCTION
The prognosis of pancreatic cancer, specifically ductal adenocarcinoma, is poor. Globally, pancreatic cancer ranks as the seventh leading cause of cancer-related deaths, with its incidence increasing worldwide.1 Both the incidence and mortality rates of pancreatic cancer in Asia are expected to increase by 88.8% and 90.8%, respectively, by 2040.1 Early diagnosis and prompt treatment can significantly enhance outcomes,2 tissue acquisition (TA) plays a crucial role in diagnosing pancreatic cancer. This process is essential for distinguishing pancreatic cancer from mimics, such as pseudotumor of chronic pancreatitis or autoimmune pancreatitis.3 Endoscopic ultrasound (EUS)-guided fine needle aspiration (FNA) and biopsy (FNB) are widely employed methods for sampling pancreatic tissue.4 Nevertheless, the diagnostic yield of needle sampling can be hampered by the necrotic5,6 and fibrotic areas7 within pancreatic tumors. In a meta-analysis, the summarized sensitivity of EUS-FNA in distinguishing malignant from benign solid pancreatic masses was reported to be 84%.8 The development of rapid on-site evaluation (ROSE) by cytologists aims to reduce the false-negative rate.9,10 This approach involves prompt cytological assessment immediately after TA, and if the specimen is considered inadequate for diagnosis, an additional biopsy is performed. However, the widespread implementation of this service in clinical practice faces challenges owing to financial and human resource constraints. A global survey revealed that ROSE was accessible to only 55% and 48% of endosonographers in Asia and Europe, respectively.11 Hence, the aim of this study was to assess the feasibility of ending the procedure based on the results of in-room cytologic evaluation by trained endosonographer (IRCETE) following a structured training program.
METHODS
Study design and patients
This prospective study was conducted at a single high-volume tertiary medical center where an estimated 2,000 EUS procedures are performed annually. Among these, approximately 600 (30.0%) were dedicated to biliopancreatic disease, with over 300 procedures involving EUS-FNA or FNB. Patients with solid pancreatic masses requiring TA to confirm their diagnoses were screened between January 2022 and July 2022. Exclusion criteria included individuals below 20 years of age; pregnant individuals; those who refused or were unable to provide consent; patients who were incompatible with EUS-FNB due to factors such as surgically altered anatomy, uncorrectable coagulopathy, or high anesthesia risk; and cases where EUS results indicated that tissue sampling was unnecessary, such as benign features or pancreatic cystic lesions. Additionally, patients who experienced complications before the initiation of EUS-FNB and those who did not adhere to the protocol were excluded from the analysis.
Training course
Before study initiation, a three-month training course conducted by a board-certified cytologist was provided to the endosonographers. The cumulative training duration was 24 hours. Three endosonographers—HPW, YTK, and WFW–participated in the study. Among them, HPW and YTK had over ten years of experience practicing EUS-guided tissue sampling, whereas WFW, who completed a two-year fellowship, was relatively newer to the field. Prior to training, the endosonographers had limited exposure to specimen handling and cytological interpretation. The training program focused on three key aspects: specimen handling, microscopic operation, and cytology interpretation. Endosonographers were educated on cytology interpretation, specifically focusing on assessing sample adequacy, assigning a diagnostic category, and providing a preliminary diagnosis. In cases where the specimens yielded no diagnostic or valuable information, they were considered inadequate. The diagnostic category was simplified into four modified categories based on the 2019 World Health Organization international classification system for reporting pancreaticobiliary cytology.12,13 These simplified categories included inadequate, no evidence of malignancy, indeterminate, and positive for malignancy (detailed definitions are provided in Supplementary Table 1).13 The training course focused particularly on the cytological characteristics of common pancreatic tumors, including ductal adenocarcinoma, neuroendocrine tumors, and pancreatic cancer mimics. For example, the characteristic features of ductal adenocarcinoma include nuclear size enlargement, nuclear crowding, anisonucleosis, and focal nuclear membrane irregularity (Fig. 1).14 Well-differentiated neuroendocrine tumors exhibit classic features such as single and dyscohesive clusters of relatively uniform tumor cells with eosinophilic cytoplasm, eccentrically located nuclei, and speckled chromatin (Fig. 2).15
The typical manifestations of pancreatic adenocarcinoma as revealed by Liu’s stain include nuclear overlapping (area surrounded by dotted line), enlarged nuclei (white arrowhead), anisonucleosis, which has variation in nuclear size greater than four times within the same cluster (black arrowhead), and focal nuclear membrane irregularity (black arrow) (×400).
Pancreatic neuroendocrine tumor cells stained with Liu’s stain appear as monotonous, small, and dyscohesive cells with round-oval nuclei (×400).
EUS-guided TA
Prior to the examinations, the endosonographers conducted a thorough review of the patient profiles, imaging studies, and blood tests. All procedures were performed or supervised by two experienced endosonographers: HPW and YTK. Conscious sedation was administered by anesthesiologists during the examinations. Linear array echoendoscopes (GF-UCT 260; Olympus Medical Systems) were used for the EUS. Following localization of the tumor and avoidance of intervening vessels using Doppler ultrasound, tissue sampling was carried out with a 22-gauge biopsy needle (SonoTip Topgain; MediGlobe). Irrespective of the IRCETE results, three passes were executed for each lesion, with 16 to 20 back-and-forth movements performed for each pass, as recommended.16 Additional sampling techniques, such as the fanning technique, suction method, or contrast-guided method, were left to the discretion of the endosonographers.
Specimen handling
Following each tissue sampling, one of the three endosonographers managed the specimen while the operator continued with TA. The specimen was processed according to the following steps (Fig. 3). The needle stylet was reintroduced to push the specimen onto a glass slide. The lengths of the white tissue cores were measured, and a positive macroscopic on-site evaluation (MOSE) was considered when the length exceeded 4 mm.17 The time from examination onset to a positive MOSE was documented. Tissues obtained at each pass were preserved in separate formalin bottles until pathological inspection. Pathologists interpreted the three samples separately and provided a final conclusion for each lesion. A glass slide containing the residual tissue specimen and liquid material was smeared onto another glass slide to disperse cells. One of the glass slides was promptly fixed in 95% alcohol for Papanicolaou staining after examination, whereas the others were air-dried for Liu’s staining of IRCETE in the endoscopic unit. Liu’s staining solution, initially designed for blood smear staining, comprises two solutions, Liu A and Liu B, and is known for its rapid and simple use in ROSE.18,19 Air-dried samples were stained with Liu A for 30 seconds and then with Liu B for 90 seconds. Subsequently, the excess dye was washed off under running water. Once the glass slide was air-dried again, it was independently reviewed by one of the three endosonographers (HPW, YTK, and WFW). The interpretation results, including sample adequacy, diagnostic category, and preliminary diagnosis, were recorded. The time from examination onset to malignancy confirmation by endosonographers was also documented in the confirmed malignancy cases. Stained and alcohol-fixed slides were read by board-certified cytologists at the end of the examination. The cytologists were blinded to the interpretation results of both the endosonographers and pathologists.
The entire specimen was expelled onto a glass slide. The tissue core was preserved in formalin for pathology. The residual tissue fragments and liquid material were smeared with another glass slide. One of the glass slides was air-dried for Liu’s stain for in-room cytology evaluation by trained endosonographers. The other one was fixed in 95% alcohol for Papanicolaou stain. The cytologists read both of the glass slides after the end of tissue sampling.
Outcomes and final diagnosis
The primary outcome was diagnostic accuracy when the endpoint of EUS-FNB was determined by IRCETE, MOSE, and cytology. Diagnostic accuracy was calculated by dividing the sum of true-positive and true-negative cases of malignancy by the total number of patients, based on the final diagnosis. Secondary outcomes included the procedure time associated with different termination criteria and the concordance rate of sample adequacy and diagnostic category between endosonographers and cytologists for each slide. Final diagnoses were established using the several methods: (1) surgical specimens in resectable cases; (2) a compatible clinical course in cases positive for malignancy; (3) spontaneous resolution or lack of progression in cases negative for malignancy. A minimum follow-up period of six months was required to assess the final diagnoses.
Statistical analysis
Continuous variables are expressed as means with standard deviations, while categorical data are presented as numbers with percentages, unless otherwise specified. Differences between groups were compared using the chi-square or Fisher's exact tests. The interobserver agreement was assessed using Cohen’s kappa statistics, Fleiss’ kappa statistics (slight, 0.0–0.20; fair, 0.21–0.40; moderate, 0.41–0.60; substantial, 0.61–0.80; and perfect, 0.81–1.00).20 Assuming an accuracy rate of 80% for endosonographers and 90% for cytologists, the non-inferiority margin for endosonographers compared to cytologists was set at a 10% absolute difference. With an α-error level of 0.05, a power of 0.80, and a drop-out rate of 10%, the estimated sample size was 87 patients. Statistical analyses were performed using IBM SPSS Statistics for Macintosh ver. 25.0 (IBM Corp.). All tests were 2-sided, and a p-value <0.05 was considered statistically significant.
Ethics approval statement
This study was approved by the Ethics Committee of the National Taiwan University Hospital (IRB/REC No. 202107108RINB) and was conducted in accordance with the 1964 Declaration of Helsinki and its later amendments. Written informed consent was obtained from each patient prior to inclusion in the study.
RESULTS
Patient characteristics and details of EUS-FNB
From January 2022 to July 2022, 90 patients were initially assessed for eligibility, but 25 patients were subsequently excluded (reasons for exclusion are detailed in Supplementary Table 2). Finally, 65 patients were included in this study. The average age of the patients was 66 years, and 32 were women (48.5%). The median carbohydrate antigen 19-9 (CA 19-9) level was 139.6 (interquartile range, 15.9–3,112.7) IU/mL. Fifty-six patients (86.2%) had malignant pancreatic tumors. The most common diagnosis was ductal adenocarcinoma (45 patients, 69.2%), followed by neuroendocrine tumors (7 patients, 10.8%) and immunoglobulin G4-related autoimmune pancreatitis (5 patients, 7.7%) (Table 1). Tissue sampling was prematurely terminated in eight patients, five due to complications and three due to risks associated with anatomical problems, such as proximity to major vessels. A total of 186 slides were analyzed. A total of 181 specimens (97.3%) were sampled using the fanning technique. Contrast-guided, Doppler-guided, and suction methods were employed in four specimens, one specimen, and one specimen, respectively.
Basic characteristics of the study population
Diagnostic accuracy
Based on the final diagnosis, the diagnostic accuracy was 90.8% when the end of the procedure was determined using IRCETE (Table 2, Supplementary Table 3). The diagnostic accuracies were 87.7% and 98.5% when MOSE and the interpretation results of the cytologist were used as the criteria for ending the procedure, respectively (p=0.060). In comparison to MOSE, IRCETE required more time to establish the diagnosis (4.0±1.7 vs. 14.3±4.7 minutes, p<0.001) (Table 2). The median cumulative pass number needed to achieve the correct diagnosis in malignant cases was one pass, regardless of the termination criterion applied. However, in the case of neuroendocrine tumors, cytologists were unable to make a diagnosis despite three passes of the sample. Final confirmation was obtained by pathological examination.
Comparison of diagnostic accuracy under each termination criterion
Sample adequacy and diagnostic category
In total, 182 (97.8%) slides were considered adequate by the endosonographers, and only one slide was considered inadequate by the cytologists (Table 3). According to IRCETE, 163 slides (87.6%) were positive for malignancy. Interpretation of results by cytologists and IRCETE agreed on 153 slides (93.9%). However, 10 (6.1%) slides were reclassified as indeterminate (5 slides), benign (4 slides), or inadequate (1 slide) by the cytologists. Among the 6 (3.2%) slides interpreted as benign by IRCETE, none were classified as malignant by cytologists. Additionally, three slides with malignant cells were incorrectly categorized as inadequate using IRCETE. Cohen’s kappa coefficient for the diagnostic category between the cytologists and IRCETE was 0.570 (p<0.001). Compared with cytologists, IRCETE demonstrated a sensitivity of 95%, specificity of 60%, positive predictive value of 94%, negative predictive value of 65%, and accuracy of 97% for malignancy interpretation (Table 3).
Diagnostic category between IRCETE and the cytologists
DISCUSSION
ROSE performed by cytologists increases the diagnostic yield of EUS-guided needle sampling; however, it is not widely available due to limitations in human and financial resources. Therefore, this study aimed to evaluate the feasibility of in-room cytological evaluation by trained endosonographers (IRCETE) following a structured and comprehensive training program to determine the end of the procedure. Savoy et al.21 observed that endosonographers had a limited ability to detect inadequate samples (56%–58%) and lower accuracy in diagnosing malignancies than cytotechnicians (69%–72% vs. 89%, p<0.001). This poor performance might be attributed to insufficient training time and the complexity of involving multiple organs during the procedure (e.g., lymph nodes, liver, and pancreas). Li et al.22 implemented a formal cytological training curriculum for four junior EUS trainees focusing on solid pancreatic masses. The accuracy of specimen adequacy and atypical grade assessments improved over the course of the program. However, more than 90 training cases were required to achieve competency in atypical assessment according to their model, and the study lacked comparative arms, such as cytopathologists or cytotechnicians. Moreover, it remains uncertain whether excellent training results in specimen adequacy assessment can be translated into a high diagnostic yield rate in real-world practice.
Hikichi et al.23 conducted a comparative analysis of diagnostic yields and the number of needle passes between two distinct periods when ROSE was carried out by endosonographers and cytologists, respectively. Although the specimen collection rate exhibited similarities between endosonographers and cytologists (97.4% vs. 97.1%, p=0.95), there was a marginal increase in the number of needle passes when ROSE was conducted by endosonographers (4.0±1.6 vs. 3.4±1.5, p=0.06). It is important to note that this study focused on comparing diagnostic yields in real-world settings. However, the historical control aspect suggests that the observed difference could also be influenced by advancements in equipment or improved proficiency of endosonographers over time. To the best of our knowledge, this is the first study to investigate the diagnostic accuracy and concordance of assessments within diagnostic categories among endosonographers, MOSE, and cytologists in real-world practice following a structured training program. The training curriculum included microscopic operations, specimen processing, and interpretation of cytological features, focusing on the 10 diagnostic cytological features of well-differentiated pancreatic carcinoma as previously outlined by Lin and Staerkel.14 Predominant features included nuclear enlargement (more than two red blood cells), anisonucleosis (variation in nuclear size greater than four times within the same epithelial group), nuclear crowding/overlapping/three-dimensionality, and nuclear membrane irregularity. A study has demonstrated that the implementation of this grading system resulted in a decrease in inconclusive diagnoses compared to gross tissue core inspection (from 26.4% to 8.2%, p=0.004).24 In our current study, the accuracies of IRCETE in differentiating malignant cases and diagnostic category were 90.8% and 97.3%, respectively. In a study involving the training of cytotechnicians to perform ROSE in EUS-FNA of pancreatic lesions, an accuracy of 90% in adequacy assessments was considered satisfactory.25 The outcomes of our study support the adoption of IRCETE in determination of procedural endpoints following appropriate training.
Given that no solid evidence was available to confirm the high accuracy rate of cytological interpretation by endosonographers in real-world practice before this study, we obtained three passes of tissue sampling, regardless of the interpretation result. This approach aimed to prevent compromising diagnostic yield in cases where the training course proved ineffective. Furthermore, subsequent treatments were planned without considering IRCETE results. In the current study, using IRCETE as the termination criterion, the diagnostic accuracy was 90.8%, confirming the appropriateness of this method for achieving a satisfactory diagnostic yield. However, when MOSE was used as the criterion to terminate the procedure, the diagnostic accuracy was 87.7%. This figure is slightly lower than the pooled result in a recently published meta-analysis, which reported a 93% accuracy (95% confidence interval, 87.7%–96.1%) after summarizing 14 studies.26 This discrepancy may be attributed to the assumption in our study that the procedure would end upon reaching a positive MOSE, rather than achieving a positive MOSE with every three sampling passes.
The main limitation of this study is that it was performed in a single high-volume tertiary medical center. All EUS-TA procedures were performed or supervised by experienced endosonographers. It is unclear whether this technique can be applied in local hospitals with limited pancreatic TA. Second, the cytologists performed cytological assessments after the procedure, indicating that they had more interpretation time than endosonographers. Cytologists might also have obtained additional information from Papanicolaou staining, a resource unavailable to endosonographers. Despite these drawbacks, endosonographers achieved a high diagnostic accuracy and a high concordance rate with cytologists in terms of diagnostic category and sample adequacy. Third, most diagnoses in our cases were malignant, and the diagnostic yield of EUS-FNB was high; therefore, the diagnostic accuracy of endosonographers may have been overestimated. Fourth, the drop-out rate was higher than expected (25 of 90, 27.8%), which led to a lower number of enrolled patients than anticipated. Finally, in this proof-of-concept study, the results of IRCETE did not affect the management of patients; therefore, a comparison of adverse event rates was not allowed. A prospective, randomized, parallel study in which IRCETE, MOSE, or ROSE are used as criteria to end the procedure is required to elucidate this point.
In conclusion, IRCETE can serve as a supplementary alternative for determining the end of tissue sampling in EUS-FNB for solid pancreatic tumors when ROSE is unavailable.
Supplementary Material
Supplementary Table 1. Definition of the simplified diagnostic category
Supplementary Table 2. Details of exclusion
Supplementary Table 3. Diagnostic accuracy of malignancy according to the final diagnosis
Supplementary materials related to this article can be found online at https://doi.org/10.5946/ce.2024.143.
Notes
Conflicts of Interest
The authors have no potential conflicts of interest.
Funding
None.
Author Contributions
Conceptualization: YTK, HPW; Data curation: WFW, YTK, WCC, CTS, MLH, CCC; Formal analysis: WFW; Investigation: YTK; Methodology: YTK, HPW; Supervision: HPW; Resources: WCC, CTS, MLH, CCC; Visualization: WFW; Writing–original draft: WFW; Writing–review & editing: all authors.