Optimizing cytological specimens of EUS‐FNA of solid pancreatic lesions: A pilot study to the effect of a smear preparation training for endoscopy personnel on sample quality and accuracy

Abstract

Background

In the absence of rapid on‐side pathological evaluation, endoscopy staff generally “smears” endoscopic ultrasound guided fine needle aspiration (EUS‐FNA) specimens on a glass slide. As this technique is vulnerable to preparation artifacts, we assessed if its quality could be improved through a smear‐preparation‐training for endoscopy staff.

Methods

In this prospective pilot study, 10 endosonographers and 12 endoscopy nurses from seven regional EUS‐centers in the Netherlands were invited to participate in a EUS‐FNA smear‐preparation‐training. Subsequently, post training slides derived from solid pancreatic lesions were compared to pre‐training “control” slides. Primary outcome was to assess if the training positively affects smear quality and, consequently, diagnostic accuracy of EUS‐FNA of solid pancreatic lesions.

Results

Participants collected and prepared 71 cases, mostly pancreatic head lesions (48%). Sixty‐eight controls were selected from the pretraining period. The presence of artifacts was comparable for smears performed before and after training (76% vs 82%, P = .36). Likewise, smear cellularity (≥50% target cells) before and after training did not differ (44% (30/68) vs 49% (35/71), P = .48). Similar, no difference in diagnostic accuracy for malignancy was detected (P = .10).

Conclusion

In this pilot EUS‐FNA smear‐preparation‐training for endoscopy personnel, smear quality and diagnostic accuracy were not improved after the training. Based on these results, we plan to further study other training programs and possibilities.

1. INTRODUCTION

Since its introduction in 1992, endoscopic ultrasound‐guided fine needle aspiration (EUS‐FNA) is increasingly popular, due to its ability to sample difficult‐to‐reach target lesions at a low complication rate. Although the technique has gained global ground, diagnostic accuracy rates still vary from 68% to 98%, ¹ , ² , ³ , ⁴ , ⁵ depending on patient characteristics, sampling techniques, and tissue handling and processing. ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ , ¹¹ , ¹² , ¹³ , ¹⁴

In our regional EUS‐working group in the Netherlands, we also noticed significant inter‐center differences in diagnostic outcome for EUS‐FNA of solid pancreatic lesions. This group decided to organize structured meetings to provide centers with feedback on their results and discuss the methods and techniques that were used. Surprisingly, this small intervention (forming a working group and discussing each other's outcomes and techniques) significantly decreased the inter‐center variation between the centers. ¹⁵ A major finding was the suboptimal FNA smear quality, which motivated the group to focus on its improvement. Historically, EUS‐FNA specimens has been collected by spreading cytological material on a glass slide, the so called “smear technique.” Although this technique is fast and cheap, its diagnostic value is easily hampered by contamination and preparation artifacts. ¹⁶ , ¹⁷ In the absence of a (cyto)pathological assistant in the room (rapid on‐side pathological evaluation [ROSE]), smears are prepared by the endoscopy staff, generally without formal training. There is limited data on their performance as compared to a specialized (cyto)pathologist. Although it seems that endoscopy staff is capable of assessing smear sufficiency for diagnostic purposes, ¹⁸ , ¹⁹ , ²⁰ , ²¹ , ²² reports on their ability to prepare the smears themselves are conflicting. ²³ , ²⁴ , ²⁵ , ²⁶ We hypothesized that a smear‐preparation‐training for endoscopy staff can improve smear quality and, thus, diagnostic accuracy of EUS‐FNA.

2. METHODS

2.1. Study design

In this prospective pilot study, endosonographers and endoscopy nurses of seven regional EUS‐centers in the Netherlands were invited to participate in a one‐day EUS‐FNA‐smear preparation training, if they had not undergone formal smear preparation training before. To assess the impact of the training, quality and diagnostic accuracy of smears were compared before and after the training. For this, all study smears were sent to the Erasmus MC University Medical Center Rotterdam for expert review. Although live patients were included in the study, the study did not intervene with routine patient care as the EUS‐guided sampling procedure was not adjusted. The Medical Ethics Committee of the Erasmus University Medical Center of Rotterdam waived the need to comply to the Medical Research Involving Human Subjects Act (MEC‐2016‐022). Consequently, informed consent was not required. This committee also specifically approved for the use of any tissue and fluid samples as a model, as the training location was restricted to a controlled area (biohazard) at the department of Pathology in the Erasmus University Medical Center in Rotterdam.

2.2. Smear training program

The specifically designed training program comprised of a 2‐hour theoretical and 2‐hour practical “hands‐on” part. The training was provided by an expert pathologist and a group of cytotechnicians from the Erasmus University Medical Center in Rotterdam. During the theoretical part, participants were educated on pancreas pathology, including solid and cystic pancreatic neoplasms, chronic pancreatitis, and focal inflammation. Furthermore, several examples of normal pancreas cytology and histology were discussed, as was the Papanicolaou Society System and common diagnostic pitfalls in pancreas (cyto)pathology. Next, participants were lectured on the different FNA cell preparation techniques, including smears, and commonly encountered pitfalls. ²⁷ The main focus of the training was optimal smear preparation. To prepare a good smear, participants were taught to apply the collected cytological specimens on glass slides by gently placing the needle tip on the glass slide, holding the bevel down, 1 cm from the labeled field of the glass slide. Then, they were told to place a second glass slide on top of the first glass slide that contained the drop of FNA specimens, and try to evenly distribute the cells using the so‐called sandwich method. In addition, participants were explained to limit the amount of cells per glass slide (only 1 drop!) to prevent thick cells layers or overlapping cells, and to avoid crushing artifacts by pressing the two glass slides too firmly. Last, they were instructed on the importance and timing of on‐site fixation, staining and drying of the material. During the hands‐on workshop, participants learned how to optimally smear and stain FNA‐specimens, and how to avoid common pitfalls during preparation. Porcine pancreatic tissue was used as training specimens.

2.3. FNA‐smear selection

After the training day, each participating center prospectively included all consecutive cases, scheduled for EUS‐FNA of solid pancreatic lesions between April 2016 and September 2017. Subsequently, an equal number of historical controls (prior to the training date) was selected for each center. We did not match our controls based on needle type or size or the sampling technique used, as there is there is limited evidence on the impact of these variables on diagnostic accuracy of EUS‐FNA. ²⁸ Smears that were prepared by (cyto)pathologists and/or cytotechnicians were excluded.

2.4. EUS‐guided tissue sampling and smear handling

EUS‐guided tissue sampling was performed according to a standard protocol, using a convex array echoendoscope (Pentax EG‐3870 UTK, Pentax EG‐3270 UK, Olympus UTC 140/180, Olympus linear GF‐UCT180). Tissue sampling was done by endosonographers, who performed between 25 and 100 EUS‐guided tissue sampling procedures annually. The optimal sampling position was determined by scanning the target lesion and its environment with color and pulsed Doppler. Patients were punctured using a 19‐, 22‐ or 25‐gauge FNA needle (EchoTip; Cook Medical, or Expect; Boston Scientific). Per target lesion, the trainees performed two smears from a single pass. All residual material was processed according to the standard protocols of the laboratories involved (Table 1). Furthermore, the number of passes, sampling strategy, and use of additional sampling techniques (eg, applying negative suction with a syringe) was left at the discretion of the endosonographers. If available, on‐site pathological assistance was allowed, but only after the trainee had performed the study smears. The on‐site pathological assistance was not allowed to comment on in the glass slide preparation of the trainee.

TABLE 1.

EUS‐guided tissue sampling and cytological processing specifics per center

Center	EUS‐scope type	Annual EUS‐FNA per endosonographer	ROSE available	Additional techniques	SMEAR preparation	Liquid cytology medium	Thin‐layer cytology technique	Cellblock technique
Albert Schweitzer Hospital, Dordrecht	Olympus linear GF‐UCT180	25	Yes	Slow pull or Suction	Air dry, Hemocolor	Cytolyt	ThinPrep	Cellient Hologic
Reinier de Graaf Hospital Delft	Olympus linear GF‐UCT180	30	No	Slow pull	Air dryNo stain	CytoLyt, or Polytransportbuffer*	ThinPrep	Agar
Erasmus MC University Medical Center Rotterdam	Pentax EG‐3870 UTK Olympus UTC 140/180	50	Yes	Slow pull or Suction	Air dry Diff quick	Cytolyt	ThinPrep	Cellient Hologic
Haga Hospital, The Hague	Olympus linear GF‐UCT180	25	Yes	Slow pull	Air dry Diff quick	Formalin	None	Paraffin cellblock
Ijsselland Hospital, Rotterdam	Olympus linear GF‐UCT180	25	Yes	Slow pull	Air dry Diff quick Giemsa	CytoRichRed	None	Agar
Maasstad Hospital, Rotterdam	Pentax EG‐3270 UK Olympus linear GF‐UCT180	30	No	Slow pull	Air dry Diff quick	CytoRichRed	None	Aalfix cellblock a
Sint Franciscus Hospital, Rotterdam	Pentax EUS‐scope	20	No	Slow pull or Suction	Air dry No stain	CytoRichRed	None	Agar

Abbreviations: EUS‐FNA, endoscopic ultrasound guided fine needle aspiration; ROSE, rapid on‐side pathological evaluation.

^aMedium/technique developed locally.

2.5. Sample reviewing

All smears were anonymized and reviewed an expert cytopathologist and two cytotechnicians from the Erasmus MC University Medical Center in Rotterdam. The reviewers were specialized in pancreaticobiliary diseases and blinded for the final clinical and pathological outcome. Smear assessment was done individually and case discussion was not allowed. Each reviewer assessed smear quality, but the cytopathologist determined the smear diagnosis. After the smear assessment, slides were returned to the hospital of origin.

2.6. Outcome measures and definitions

The primary outcome measure was to assess if this one day “hands‐on” EUS‐smear‐preparation‐training improved the diagnostic accuracy of smears, in the absence of an on‐site (cyto)pathologist. Diagnostic accuracy for malignancy was calculated from the correct number of cases that were defined as atypical/suspect for malignancy or malignant. In addition, accuracy for the Papanicolaou Society System was calculated from the number of cases that were correctly classified into the categories; nondiagnostic, benign, atypical/suspect for malignancy or malignancy, according to the formula: (true positive + true negative)/all patients. Gold‐standard diagnosis was based on surgical resection specimens, or a clinical follow‐up period of at least 1 year for nonoperated patients.

Secondly, we assessed if the training improved smear quality, which was defined as smear artifacts (fixation, thick smear/clots, obscuring blood or inflammation, cytolysis, contamination, other) and cellularity (presence of </≥50% cells, either from the target organ or surrounding area). Poor fixation included drying artifacts due to delayed covering of fixated cells. Thick smears were defined as overlapping cells that hamper individual cell assessment, this is generally caused by placing too much cells on the glass slide or inadequate smearing of the cells. Contamination included environmental causes of contamination, such as postoperative stitching material, presence of fungus caused by contaminated room‐air, or foreign bodies (dust, insects, and so on).

2.7. Statistics

Outcome measures were expressed as means ± SD or as medians with ranges. Statistical significance was assessed with the use of Student t test for normally distributed continuous data; either the chi‐square test for categorical data (with Yates' correction when appropriate) or Fisher exact test for categorical data; and the median test for nonnormally distributed continuous data. Smear quality and diagnostic accuracy were compared between cases and controls using a logistic mixed effect model with a random intercept for participating center. ²⁹ The latter was done to consider the clustering structure of this multi‐center trial, that is, that observations from the same site may be correlated. Statistical significance was established as P < .05 (two‐tailed). Analyses were carried out using SPSS version 21, Statistical Package for the Social Sciences, SPSS Inc., Chicago, Illinois, and R (version 3.4.2).

2.8. Power calculation

To determine the power needed for this study, we assessed the impact of the introduction of ROSE in one of the participating centers as a substitute intervention for our smear‐preparation‐training. To determine if the diagnostic accuracy of smears improved, an expert pathologist reviewed 20 smears from the period before and 18 smears from the period after ROSE was introduced in that center. Smear accuracy improved with 30% since the implementation of ROSE. Based on this assumption, a two‐group continuity corrected chi‐squared test with a 0,050 two‐sided significance level will have 80% power to detect the difference between a group 1 proportion (results before training), π1 of 0,400 and a group 2 proportion (results after training), π2, of 0,670 (odds ratio of 3045) when the sample size in each group is 60 cases. ³⁰

3. RESULTS

3.1. Endoscopy staff characteristics

A total of 10 endosonographers and 12 endoscopy nurses attended the EUS‐smear‐preparation‐training. Participants were selected by the principal investigators of the participating centers, during a meeting in February 2016. If they had not received a formal EUS‐smear‐preparation‐training previously, the study coordinator invited the participants by e‐mail. Table 2 demonstrates the participants' characteristics. Majority of the trainees was female, with a median age of 38 (range 22‐49). As only one of the centers was an academic hospital, most were working at a community hospital (77%). Experience with EUS‐FNA ranged from several months to years. We consider our study population to be representative for, at least, the other regions in the Netherlands, since most regions in the Netherlands comprise an academic and several smaller hospitals. Furthermore, majority of today's medical staff comprises young to middle‐aged women, and exposure to EUS‐FNA varied greatly, which corresponds well with exposure in the academic and non‐academic centers.

TABLE 2.

Characteristics of EUS‐smear training participants

Hospital	Profession	Age (years)	Female	Experience with EUS‐FNA (years)	No. of EUS‐FNA procedures annually
1	Doctor	42	No	12	100
1	Doctor	39	Yes	4	30
2	Nurse	24	Yes	2	300
2	Nurse	33	Yes	6	300
2	Nurse	22	Yes	2	300
2	Nurse	23	Yes	0	25
2	Nurse	30	Yes	0	30
3	Doctor	38	No	3	10
3	Doctor	35	Yes	1	25
3	Nurse	48	Yes	3	25
4	Doctor	44	Yes	10	50
4	Doctor	42	Yes	8	50
4	Nurse	48	Yes	11	92
5	Nurse	37	Yes	8	60
5	Doctor	49	No	7	50
5	Nurse	31	Yes	7	50
6	Nurse	29	Yes	5	40
6	Nurse	29	Yes	5	40
6	Nurse	47	Yes	0	45
6	Doctor	36	Yes	2	50
7	Doctor	39	No	1	60
7	Doctor	44	Yes	10	25

Abbreviations: EUS‐FNA, endoscopic ultrasound guided fine needle aspiration; No., number.

3.2. Target lesion characteristics

Seventy‐one cases and 68 controls were assessed (see Table 3 for target lesion characteristics), with a mean lesion size of 31 mm (SD ± 1.37 mm). Pancreatic corpus and tail lesions were somewhat over‐represented in the control group (P < .01, Table 3). Most case lesions were sampled with a 25G needle (61%), while controls were mostly targeted with a 22G needle.

TABLE 3.

Characteristics of included cases and controls

Variables	Controls (n = 68)	Cases (n = 71)	P‐value
Center of inclusion, n (%)
Albert Schweitzer	6 (9)	6 (9)	n.s.
Reinier de Graaf	12 (18)	15 (22)
Erasmus MC	28 (41)	28 (39)
Haga Hospital	3 (4)	3 (4)
Ijsselland Hospital	6 (9)	6 (9)
Maasstad Hospital	6 (9)	6 (9)
Sint Franciscus Hospital	7 (10)	7 (10)
Target lesion location, n (%)
Head	39 (57)	34 (48)	<.01
Uncinate process	5 (7)	6 (9)
Neck	9 (13)	4 (6)
Corpus	9 (13)	14 (20)
Tail	0 (0)	13 (18)
Missing	6 (8)	0 (0)
Target lesion size (mm), mean ± SD	28.7 ± 9.63	31.0 ± 1.37	n.s.
FNA needle size, n (%)
19‐gauge	3 (6)	1 (1)	.02
22‐gauge	31 (57)	27 (38)
25‐gauge	20 (37)	43 (61)
Number of passes, median (IQR)	3.00 (2.00‐3.00)	3.00 (2.00–3.00)	n.s.

3.3. Smear quality

The presence of artifacts was comparable for smears prepared before and after the training session (76% vs 82%, P = .363, Table 4), as were individual types of artifacts. Also, for smear cellularity, there was no difference between cases and controls (P = .480).

TABLE 4.

Diagnostic outcome of smears from cases versus controls

Variables, n (%)	Cases (n = 71)	Controls (n = 68)	P‐value a
Presence of artifacts	54 (76)	56 (82)	.363
Type of artifacts b
Poor fixation	3 (6)	3 (5)	1
Thick smear/clots	45 (83)	42 (75)	.35
Cytolysis	25 (46)	30 (54)	.57
Cellularity
< 50%	36 (51)	38 (56)	.48
≥50%	35 (49)	30 (44)
Sample diagnosis
Impossible to determine	21 (30)	21 (31)	.10
Benign	1 (1)	1 (1)
Atypical/suspect for malignancy	13 (18)	12 (18)
Malignant	36 (51)	34 (50)
Gold standard diagnosis
Benign	4 (6)	2 (3)	.56
Atypical (NET, pancreatitis)	3 (4)	5 (7)
Malignant	64 (90)	61 (90)
Diagnostic accuracy for Papanicolaou Society System % (n/n)	51 (36/71)	47 (32/68)	.67
Diagnostic accuracy for malignancy % (n/n)	66 (47/71)	66 (45/68)	.10

^aGeneralized linear mixed model.

^bMore than one option possible.

3.4. Smear diagnosis and accuracy

After a median follow‐up time of 24 months (range 21‐32), 70 (50%) of the smears were scored as malignant, 25 (18%) as atypical or suspect for malignancy, and 2 (1%) as benign. Smears were considered non‐diagnostic in 42 lesions (30%). Gold standard diagnosis revealed 125 (90%) malignant lesions, 8 (6%) atypical lesions or suspect for malignancy (one IgG‐mediated pancreatitis, two pancreatitis, five neuroendocrine tumors), and 6 (4%) benign lesions (three chronic pancreatitis, one fibrotic lesion, two non‐specified benign lesions). Similar to FNA smear quality, the preparation‐training did not result in a significant increase in the diagnostic accuracy for malignancy (P = .10) or the Papanicolaou Society System (P = .67, Table 4).

4. DISCUSSION

With this pilot study, we aimed to evaluate the efficacy of an EUS‐FNA smear‐preparation training for endoscopy staff, in centers lacking ROSE. Unfortunately, our training did not improve the smear quality or diagnostic accuracy in our regional EUS‐working group. For this, several reasons may be found.

First of all, our training program may have been inadequate to achieve a significant improvement in the performance of the trainees. As official EUS‐smear preparation‐courses do not exist, we had to design our own program. We chose a comprehensive training, combining theoretical and practical hands‐on elements. However, this program may have fallen short. It is, for example, well known that practical skills are better achieved after extensive training, and tend to grow with exposure. Therefore, it may have be more effective to intensify or repeat the training by one or more refresh sessions. In addition to this, the specimen collection period may have been too short to allow trainees to gain sufficient experience, thereby improving their skills.

Second, it has been demonstrated that self‐assessment and standardized feedback improves the learning curve for colonoscopy of Gastroenterologists in training. ³¹ Therefore, implementing standardized self‐assessment forms could have increased the training effect. In addition, we could have implemented frequent multidisciplinary meetings of the trained endosonographers with the (cyto)pathologists. Such an off‐site feedback moment may further improve the learning curve for smear preparation.

Third, our results might be inherent to the nature of the smear technique itself, since it is a manual method that is sensitive to artifacts and is prone to heterogeneous preparations. In contrast, collection of FNA specimens in a liquid medium, liquid‐based cytology, has several advantages including less contamination by red blood cells, less drying artifacts. ⁸

A limitation of our study is that our power calculation was based on the training effect in our regional EUS‐working group. Therefore, we could not assess the impact of the training on an individual basis. This prevents us from identifying trainees who did benefit from the training. It is known, that a learning curve can vary greatly between trainees. This has been shown for endoscopy and ERCP learning, ³² and seems to have led to a more competence‐based training schedule rather than a threshold number‐based training for Gastroenterology residents. ²⁸ , ³³ As our group comprised of endoscopy staff (both physicians and nurses) from high, medium and low volume centers with different levels of experience, differences in learning curves seem inevitable. Previous studies found that endosonographers performed equally well as compared to cytopathologists, but endoscopy nurses did not. ²³ , ²⁴ , ²⁵ , ²⁶ We did not power our study to compare the smear quality and accuracy between doctors and nurses. Another limitation, one that hampers most EUS‐FNA studies, is the inter‐center variability in practice protocols. As we report in Table 1, our centers use a variety of sampling and smear preparation protocols. Although this may introduce a bias, today, this is inevitable in multicenter studies, as no consensus exists on the optimal sampling and FNA specimen handling protocol. ¹⁶ , ²⁸ , ³⁴ Furthermore, the endpoints that we used to measure EUS‐FNA specimen or smear quality are not globally harmonized. The most important problem is that there are no uniform guidelines that advise on how to mark FNA smear diagnosis, ³⁵ and there is no consensus on how to describe smear quality. Therefore, quality definitions used in the current study were jointly created by the study group.

Last, our inclusion rate was rather low. This was mainly due to the fact that study‐smears had to be performed by course participants. As only part of each endoscopy was trained to participate in the study, inclusion rates did not match the regular daily EUS volume of the participating centers. Inclusion rate was not affected by the use of ROSE, since ROSE was allowed in all study cases, but only after the participants had performed the study‐slides.

Taken all together, this pilot EUS‐FNA smear‐preparation‐training for endoscopy personnel did not improve EUS‐FNA smear quality or accuracy. Nevertheless, it stands to reason that endoscopy staff could benefit from some form of smear‐preparation‐training, and perhaps an adjusted, more elaborate program will be more effective. However, optimization of smear quality is just one link in the chain towards a higher diagnostic accuracy. Therefore, we also need to explore other strategies to achieve this. For example, by improving the skills of the endosonographer, adjusting the needle type, the sampling technique, or the tissue preparation technique of the harvested material (liquid based cytology or cell block instead of smears). As for the needle type, core biopsy needles (CBN) are designed to collect core biopsy samples, rather than cytological material. Although overall accuracy rates seem to be higher for CBN than FNA, sensitivity and specificity do not reach 100% yet ¹ , ² . Therefore, we believe that ROSE (either using FNA or CBN material) is still needed to ensure their harvest of a diagnostic sample.

CONFLICT OF INTEREST

The authors declare no potential conflict of interest.

AUTHOR CONTRIBUTIONS

Priscilla A. van Riet: Study design and coordination, data analysis, writing manuscript. Rutger Quispel: Data collection and manuscript design and review. Djuna L. Cahen: Data collection and manuscript design and review. Nicole S. Erler: Data analysis. Mieke C. Snijders‐Kruisbergen: Data collection. Petri Van Loenen: Data collection. Jan‐Werner Poley: Data collection and manuscript review. Lydi M.J.W. van Driel: Data collection and manuscript review. Sanna A. Mulder: Data collection and manuscript review. Bart J. Veldt: Data collection and manuscript review. Ivonne Leeuwenburgh: Data collection and manuscript review. Marie‐Paule G.F. Anten: Data collection and manuscript review. Pieter Honkoop: Data collection and manuscript review. Annemieke Y. Thijssen: Data collection and manuscript review. Lieke Hol: Data collection and manuscript review. Mohammed Hadithi: Data collection and manuscript review. Claire E. Fitzpatrick: Data collection and manuscript review. Ingrid Schot: Data collection and manuscript review. Jilling F. Bergmann: Data collection and manuscript review. Abha Bhalla: Data collection and manuscript review. Marco J. Bruno: Study design, supervising study execution, writing manuscript. Katharina Biermann: Study design, supervising study execution, writing manuscript.

van Riet PA, Quispel R, Cahen DL, et al. Optimizing cytological specimens of EUS‐FNA of solid pancreatic lesions: A pilot study to the effect of a smear preparation training for endoscopy personnel on sample quality and accuracy. Diagnostic Cytopathology. 2021;49:295–302. 10.1002/dc.24645

DATA AVAILABILITY STATEMENT

Data available on request from the authors.