Specialized clinical cytology may improve the results of EUS (endoscopic ultrasound)-guided fine-needle aspiration (FNA) from pancreatic tumors

Abstract

Background and study aims: A variety of factors (needle type, needle passes, tumor location, cytological assessment, etc.) may influence the diagnostic accuracy of endoscopic ultrasound-guided fine-needle aspiration cytology (EUS-FNAC) from pancreatic tumors. Whereas most published studies report a diagnostic accuracy of > 80 % for EUS-FNAC, the results in routine settings are often considerably lower. This retrospective study aimed to define the effect of switching microscopic assessment from a standard pathology department to a highly specialized institute of cytology.

Patients and methods: A total of 63 patients underwent EUS-FNAC of solid or semisolid pancreatic masses. Specimens of the first consecutive 20 cases (Phase 1) were assessed by the local department of pathology. Then in Phase 2, involving another 43 subsequent cases, a specialized cytology laboratory examined all aspirates. All EUS-FNACs were performed in the same manner, using a 22-gauge needle. After cytological evaluation, all patients either underwent surgery or were followed up for at least 6 months.

Results: Of the tumors, 56 were solid and 7 semisolid; the mean size was 30 mm. Sensitivity (sens.), specificity (spec.), positive predictive value (PPV), and negative predictive value (NPV) of EUS-FNAC were 38.5 % (95 %CI [confidence interval] 13.9 – 68.4 %), 100 % (59.0 – 100 %), 100 % (47.8 – 100 %), and 46.7 % (21.3 – 73.4 %) during Phase 1 versus 91.4 % (95 %CI 76.9 – 98.2 %), 100 % (63.1 – 100 %), 100 % (89.1 – 100 %), and 72.7 % (39.0 – 94.0 %) during Phase 2.

Conclusion: These results emphasize the considerable impact of a dedicated cytological evaluation on the results of EUS-FNAC.

Introduction

Endoscopic ultrasound-guided fine-needle aspiration (EUS-FNA) is an established means for the diagnostic workup of pancreatic masses ^{1 2}. Since its introduction into clinical medicine, the technique has been refined and different needle types have been compared with regard to their diagnostic gain. Numerous studies report accuracy and sensitivity of EUS-FNA cytology to reach up to 95 %. However, there is a large variation in these statistical results, ranging from 68.1 to 95 % ^{3 4 5 6 7 8 9 10}. The reasons for this discrepancy remain ambiguous, although the reality in many endoscopy departments is that many physicians are dissatisfied with the accuracy of EUS-FNAC in their clinical routines. The technique of tissue sampling via EUS-guided FNA is well defined and largely standardized. The question of preference for small (25-gauge), intermediate (22-gauge), or large-size (19-gauge) needles – or whether to use core biopsy needles, with a chance of obtaining a histological diagnosis – seems to remain a matter of ongoing debate ^{3 11 12 13 14 15}.

Once successful aspiration has been achieved, further processing of the specimen by endoscopists and assessment by cytopathologists seem highly variable: Some physicians prefer formalin fixation to enable histopathological embedding, whereas the majority prepare a varying number of slides that undergo either air drying or chemical fixation. Additionally, cytopathologists employ a wide range of staining methods, depending on individual preferences.

During past years, we have modified several of the abovementioned aspects to improve the sensitivity of EUS-FNAC in our department. However, none of these merely technical modifications yielded a detectable increase in sensitivity.

We therefore agreed on switching the microscopic assessment of the slides from our in-house department of pathology to a dedicated laboratory of clinical cytology. The present study compares the results of EUS-FNAC of solid or semisolid pancreatic masses performed at our department before and after this modification.

Patients and methods

The study population comprised all patients undergoing EUS-FNA at our department for diagnostic workup of solid or semisolid (cystic with solid fractions) pancreatic masses between January 2009 and September 2012. Patients were identified retrospectively by using the electronic medical information database of the hospital. Most patients underwent EUS-FNA for cytological confirmation of suspected unresectable adenocarcinoma of the pancreas. In contrast, patients with highly suspicious but potentially resectable tumors were sent directly for surgery without pursuing a preoperative microscopic diagnosis. Individual courses of disease were followed for a minimum of 6 months, particularly to rule out false-negative results of cytology where pancreatic cancer might evolve despite nonmalignant cytology. In the latter case, lesions were considered malignant either if a carcinoma was histologically confirmed after resection, if the size of the lesion increased during follow-up, or if its size decreased during chemotherapy.

EUS-FNA was performed using a curvilinear array echoendoscope (Pentax EG-3870UTK, Pentax Medical, Hamburg, Germany) attached to a Hitachi EUB 8500 ultrasound processor (Hitachi Medical Systems, Wiesbaden, Germany). The procedure was performed by experienced endoscopists (> 30 previous procedures) with patients in the left lateral position under sedation with intravenous propofol. For tumors of the pancreatic head, puncture was done from the duodenal position in the majority of patients. The remaining patients were visualized and biopsied using a transgastric approach. All EUS-FNA were performed with a 22-gauge needle (Sonotip II, Medi-Globe, Achenmühle, Germany). After visualization of the tumor in a stable position of the endoscope, the needle was advanced into the tumor under continuous EUS guidance, then the stylet was entirely removed and suction applied. A minimum of six needle passes to and fro were performed inside of the lesion during two to four separate punctures, depending on the macroscopic appearance of the specimen with respect to cellularity and bloody additives. Further punctures were abandoned when gross inspection augured tissue without relevant bloody admixture. The aspirate was then recovered by reintroduction of the stylet and subsequent sparging of the needle with air from a 5-mL luer lock syringe. Then smears on glass slides were readily prepared from aspirated fluid with corpuscular fractions. Additionally, cellular clots or spools were gently rolled over glass slides to allow for a detachment of cells on the slide surface. The slides were finally air dried and sent for further cytopathological evaluation.

During study Phase 1 from January 2009 through October 2010, dried aspirates were sent for microscopic diagnosis to the hospital’s institute of pathology. There, the slides were stained with Giemsa and analyzed for adequacy of the specimen, cellularity, blood cell additives, and final diagnosis. After this period (Phase 2, November 2010 – September 2012), the slides were sent to a specialized institute of clinical cytology. In contrast, May-Grünwald-Giemsa was the standard staining method of the cytologist, whereas further microscopic assessment of cytological aspirates corresponded to the workflow outlined above. Sensitivity (sens.), specificity (spec.), positive predictive value (PPV), and negative predictive value (NPV) as well as corresponding 95 % confidence interval (CI) were calculated with GraphPad Prism 6 (GraphPad Software Inc., La Jolla, CA, USA). The study was approved by the local ethics committee (Ethics Committee of the Faculty of Medicine, Technical University Munich; approval no. 588/13).

Results

Of 64 consecutive patients who had undergone EUS-FNAC, aspirates of 63 patients (24 male [38 %], 39 female [62 %]; mean age 70, range 28 – 85) showed cytological features of pancreatic parenchyma and were included in the study. During Phase 1, 20 patients underwent EUS-FNAC; in Phase 2, patients undergoing the procedure numbered 43. The tumor sites of the entire group were: 35 tumors (56 %) located in the head; 21 tumors (33 %) in the body; and 7 tumors (11 %) in the tail of the pancreas. When the study population was differentiated according to affiliation (Phase 1/Phase 2 of the study period), the distribution of the sites broke down to: 5/16 (25 /37 %) in the head; 11/22 (55/51 %) in the body; and 4/5 (20/12 %) in the tail. Lesions were classified hypoechoic in 56 cases (89 %) and hypoechoic with cystic areas in 7 cases (11 %). The mean size of the tumors in the entire population was 30 mm (range 12 – 80 mm), with 39 mm (range 12 – 80 mm) in the Phase 1 group, and 30 mm (range 13 – 70 mm) in the Phase 2 group (P = 0.12, Mann-Whitney test). During EUS-FNAC, 2 to 5 separate needle passes provided 4 to 27 slides with smears. Mean follow-up of all patients was 17.8 months (range 6.2 – 34.7 mo).

Surgical resection was done on 9 patients and 54 patients were solely followed up. On final diagnosis, 48 lesions were malignant (44 adenocarcinoma, 3 neuroendocrine tumors [NET], and 1 metastasis of breast carcinoma); and 15 lesions were of benign origin (2 focal post-inflammatory residues after acute pancreatitis, 2 focal inflammatory changes within chronic pancreatitis, 1 autoimmune pancreatitis, 2 intraductal papillary mucinous neoplasms [IPMN], 3 hemorrhagic pancreatic cysts, and 5 undefined).

Accordingly, sens., spec., PPV, and NPV of EUS-FNAC were 38.5 % (95 %CI 13.9 – 68.4 %), 100 % (59.0 – 100 %), 100 % (47.8 – 100 %), and 46.7 % (21.3 – 73.4 %) during Phase 1 versus 91.4 % (95 %CI 76.9 – 98.2 %), 100 % (63.1 – 100 %), 100 % (89.1 – 100 %), and 72.7 % (39.0 – 94.0 %) during Phase 2 (Table 1). Accuracy was 60 and 93 %, respectively.

Table 1. Summary of statistical results.

Phase 1 (n = 20)				Phase 2 (n = 43)
	Real result		Σ		Real result		Σ
	Benign	Malignant			Benign	Malignant
FNAC +	0	5	5	FNAC +	0	32	32
FNAC-	7	8	15	FNAC-	8	3	11
Σ	7	13	20	Σ	8	35	43
Sens.	38.5 % (13.9 – 68.4 %)			Sens.	91.4 % (76.9 – 98.2 %)
Spec.	100 % (59.0 – 100 %)			Spec.	100 % (63.1 – 100 %)
PPV	100 % (47.8 – 100 %)			PPV	100 % (89.1 – 100 %)
NPV	46.7 % (21.3 – 73.4 %)			NPV	72.7 % (39.0 – 94.0 %)

Abbreviations: FNAC-, fine-needle aspiration cytology benign; FNAC + , fine-needle aspiration cytology malignant; NPV, negative predictive value; PPV, positive predictive value; Sens., sensitivity; Spec., specificity.95 % confidence interval.

Discussion

Published data report on a high accuracy of EUS-FNAC for the diagnosis of pancreatic tumors. According to a variety of studies, mean sensitivity, specificity, and accuracy can be expected to be around 83, 100, and 88 % – although these statistical parameters may vary from 54 to 95 %, 71 to 100 %, and 65 to 96 %, respectively ^{16 17 18}.

However, clinical experience among physicians applying the technique in routine diagnostics is often considerably lower than these study results. Some endoscopists may even have abandoned EUS-guided fine-needle aspiration because, due to ambiguous cytology, they lack faith in the results. In this regard, several factors may have an effect on the diagnostic yield of EUS-FNAC.

The endoscopic technique of EUS-FNA has been well defined and even outlined in clinical guidelines. A minimum of two to five needle passes per lesion, depending on a gross macroscopic assessment of the aspirate by the endoscopist, are considered mandatory for adequate microscopic diagnosis ¹⁸. The question of whether to use smaller (25-gauge) or larger size (19-, 22-gauge) or even core biopsy needles is still a matter of debate ^{3 11 12}. However, conflicting results of these studies suggest a minor effect of this issue. Immediate microscopic assessment of aspirates by an on-site cytopathologist may further improve EUS-FNAC and reduce the need for multiple needle passes; however, this is rarely available in the routine setting ^{19 20 21}. If the initial EUS-FNAC is not diagnostic, repeating the procedure may improve results ^{19 20}. Alternatively, percutaneous computed tomography (CT) or ultrasound-guided puncture can be performed, but implies a higher risk for peritoneal tumor seeding. There is no evidence that larger tumors facilitate FNAC assessment ²². This may derive from the fact that the larger the lesion, the more often nondiagnostic aspirates from cystic or necrotic areas are encountered.

Regarding microscopic diagnosis, cytopathologists generally consider pancreatic adenocarcinoma a challenging task because scarce cellularity, inflammation, and desmoplastic changes may mimic both well-differentiated carcinoma in chronic pancreatitis and fibrosing pancreatitis in cancers ^{17 23 24}. Since standardized guidelines defining the cytological workup are lacking, preparation and further processing of slides remain highly individual. Whereas some cytopathologists prefer air-dried samples, others recommend different methods of fixation. Moreover, staining methods largely differ and imply Papanicolaou, hematoxylin eosin, or May-Grünwald-Giemsa ^{3 10 22 23 24 25 26 27}.

During a continuous effort to improve the results of EUS-FNAC in our own clinic, we changed several parameters (e. g., needles of different sizes and shapes, number of needle passes and smears, puncture with and without vacuum, etc.) regarding the technical process of specimen acquisition. None of these measures, however, yielded substantial improvement. As a consequence, we discussed the issue with the local pathologist and agreed upon sending FNAC smears to an external specialized cytologist. Although we continue to obtain excellent histopathological results of core biopsies from the local pathologist, the evaluation of FNAC smears by a specialized cytologist had a striking effect on the diagnostic performance. In particular, sensitivity exceeded 90 % and therefore equaled published data.

In most cases, we perform EUS-FNAC of solid pancreatic lesions to prove malignancy prior to initiating palliative chemotherapy or chemoradiotherapy. For this reason, the pretest probability of adenocarcinoma was quite high in the relatively large tumors (mean size 30 mm) in this group of patients. Nevertheless, previous publications that addressed the statistical issues of EUS-FNAC investigated tumors of similar sizes and are therefore comparable with our study ^{3 4 5 6 7 8 9 10}.

Due to the fact that the technical aspects of EUS-FNAC are largely standardized (or have only a minor impact on sensitivity and specificity), preparation of smears and microscopic assessment remain the most variable factors for EUS-FNAC results. A retrospective study by Alsibai et al. has shown that an expert cytopathologist may considerably improve the sensitivity of EUS-FNAC as compared with local cytopathologists ²⁷. A similar effect has been demonstrated for the training of on-site cytotechnicians present during EUS: Diagnostic accuracy of smear evaluation increased from 74.8 to 90.5 % after 1 year of training, involving more than 100 patients ²⁵. Recently, a German survey among more than 100 EUS centers demonstrated sobering results regarding the outcome of EUS-FNAC – more than half of the institutions reported a success rate of 75 % and lower ²⁸.

Conclusions

Summing up, specialized cytologists should be involved if ambiguous or inadequate results are encountered despite optimization of specimen-acquisition technique. A sensitivity and specificity exceeding 90 % can be achieved, but involves a well-balanced system of cell acqusition, preparation of smears, and further processing in the cytological laboratory. The Papanicolaou Society of Cytopathology has recently formulated guidelines for the terminology and nomenclature in pancreatobiliary cytology ²⁹. In the future, further consensual standards are mandatory for both endoscopists and cytologists to ensure a widespread, consistent quality of EUS-FNAC.