Endoscopy 2018; 50(11): 1059-1070
DOI: 10.1055/a-0607-2534
Original article
© Georg Thieme Verlag KG Stuttgart · New York

Reliability and accuracy of a novel classification system using peroral cholangioscopy for the diagnosis of bile duct lesions

Carlos Robles-Medranda
Gastroenterology and Endoscopy Division, Instituto Ecuatoriano de Enfermedades Digestivas (IECED), Guayaquil, Ecuador
,
Manuel Valero
Gastroenterology and Endoscopy Division, Instituto Ecuatoriano de Enfermedades Digestivas (IECED), Guayaquil, Ecuador
,
Miguel Soria-Alcivar
Gastroenterology and Endoscopy Division, Instituto Ecuatoriano de Enfermedades Digestivas (IECED), Guayaquil, Ecuador
,
Miguel Puga-Tejada
Gastroenterology and Endoscopy Division, Instituto Ecuatoriano de Enfermedades Digestivas (IECED), Guayaquil, Ecuador
,
Roberto Oleas
Gastroenterology and Endoscopy Division, Instituto Ecuatoriano de Enfermedades Digestivas (IECED), Guayaquil, Ecuador
,
Jesenia Ospina-Arboleda
Gastroenterology and Endoscopy Division, Instituto Ecuatoriano de Enfermedades Digestivas (IECED), Guayaquil, Ecuador
,
Haydee Alvarado-Escobar
Gastroenterology and Endoscopy Division, Instituto Ecuatoriano de Enfermedades Digestivas (IECED), Guayaquil, Ecuador
,
Jorge Baquerizo-Burgos
Gastroenterology and Endoscopy Division, Instituto Ecuatoriano de Enfermedades Digestivas (IECED), Guayaquil, Ecuador
,
Carlos Robles-Jara
Gastroenterology and Endoscopy Division, Instituto Ecuatoriano de Enfermedades Digestivas (IECED), Guayaquil, Ecuador
,
Hannah Pitanga-Lukashok
Gastroenterology and Endoscopy Division, Instituto Ecuatoriano de Enfermedades Digestivas (IECED), Guayaquil, Ecuador
› Author Affiliations
TRIAL REGISTRATION: single-center, observational, case-crossover, prospective Study NCT02794987 at clinicaltrials.gov
Further Information

Corresponding author

Carlos Robles-Medranda, MD
Instituto Ecuatoriano de Enfermedades Digestivas – IECED
OMNI Hospital
Av. Abel Romeo Castillo y Av. Juan Tanca Marengo
Torre Vitalis, Mezanine 3
090505, Guayaquil
Ecuador   
Fax: +593-4-2109180    

Publication History

submitted 03 September 2017

accepted after revision 12 March 2018

Publication Date:
28 June 2018 (eFirst)

 

Abstract

Background The aim of this study was to propose a novel, comprehensive, macroscopic classification for bile duct lesions.

Methods A two-stage protocol was designed. In Stage I, a retrospective study (September 2013 to September 2015) of patients with bile duct lesions detected by peroral cholangioscopy (POCS) was performed. A total of 315 images with at least 6 months of follow-up were recorded, analyzed, and correlated to histology, and were classified as non-neoplastic (one of three types, 1 – 3) or neoplastic (one of four types, 1 – 4) based on morphological and vascular patterns. In Stage II, a prospective, nonrandomized, double-blind study was performed from December 2015 to December 2016 to validate the proposed classification. Sensitivity, specificity, positive and negative predictive values (PPV and NPV, respectively), and positive and negative likelihood ratios (LR + and LR – , respectively) were calculated (gold standard: 6-month follow-up). Inter- and intraobserver agreement (kappa value, κ) among experts and non-experts were calculated.

Results 171 patients were included (65 retrospective; 106 prospective). In Stage I, 28/65 cases were neoplastic and 37 /65 were non-neoplastic, according to the final diagnosis. In Stage II, 56/106 were neoplastic with a sensitivity, specificity, PPV, NPV, LR + , and LR – for neoplastic diagnosis of 96.3 %, 92.3 %, 92.9 %, 96 %, 12.52, and 0.04, respectively. The proposed classification presented high reproducibility among observers, for both neoplastic and subtypes categories. However, it was better for experts (κ > 80 %) than non-experts (κ 64.7 % – 81.9 %).

Conclusion The novel classification system could help physicians to distinguish non-neoplastic from neoplastic bile duct lesions.


#

Introduction

Endoscopic retrograde cholangiopancreatography (ERCP) has an important imaging limitation in the diagnosis of bile duct lesions, leading to the risk of missed pathology and sampling errors.

Recently, the development of a new single-use scope for cholangioscopy, called SpyGlass (Boston Scientific, Marlborough, Massachusetts, USA), has enabled the direct visualization of the pancreaticobiliary system for the evaluation of intraductal lesions [1]. It has been demonstrated that the use of SpyGlass and the SpyBite forceps changes clinical management decisions in 64 % of patients [2]. The system has a sensitivity of 76.5 % for indeterminate stricture diagnosis, compared with 29.4 % and 5.9 % sensitivity using blind biopsy or a brushing catheter, respectively [3].

Regardless of the instrumentation chosen, the macroscopic features used to determine malignancy of bile duct lesions during peroral cholangioscopy (POCS) are the presence of an irregular surface, with bleeding and oozing or tortuous vessels; in contrast, benign lesions typically have a smooth surface without vessels or mass [4].

However, despite these well-defined features distinguishing neoplastic from non-neoplastic lesions, many misdiagnoses are made, owing to the lack of a correlation between these macroscopic aspects and histology [4] [5]. Moreover, masses are sometimes benign. The aim of this study was thus to propose a new, more detailed macroscopic classification of bile duct lesions based on the morphological and vascular patterns observed during POCS.


#

Methods

Study design

This was an observational, analytical, case-crossover, ambispective, diagnostic study performed at the Instituto Ecuatoriano de Enfermedades Digestivas (IECED). The study protocol and consent form were approved by the Institutional Review Board, registered at ClinicalTrials.gov (ID: NCT02794987), and the study was conducted according to the Declaration of Helsinki. All patients signed an informed consent form.

A two-stage protocol was designed. Both stages included patients with bile duct lesions detected by POCS, as described below.

In Stage I, an observational, retrospective study with case collection from September 2013 to September 2015 was conducted. Patients in Stage I underwent cholangioscopic bile duct imaging and video recording at the start of the study and at 6-month follow-up, for macroscopic analysis.

Malignancy of lesions was confirmed after targeted biopsy, surgical pathology or autopsy assessment. Patients with a non-neoplastic or indeterminate lesion had a minimum follow-up duration of 6 months and underwent a repeat POCS at 6 months post-procedure, which served to re-evaluate the lesion to confirm the diagnosis. Using the histopathology results and follow-up, 65 lesions were classified as non-neoplastic or neoplastic.

Once malignancy was determined, an experienced endoscopist (C.R.-M.), who had experience of more than 140 POCS examinations, analyzed 315 images together with the histopathology results and, based on morphological and vascular patterns from the 65 bile duct lesions, developed a POCS macroscopic classification to differentiate non-neoplastic from neoplastic lesions using macroscopic features.

In Stage II, a prospective, nonrandomized, double-blind study was performed from December 2015 to December 2016. Images and video recordings of bile duct lesions from included patients were analyzed using the proposed macroscopic classification to determine whether the lesions were non-neoplastic or neoplastic, and to assign them to a particular subtype compared with histology from the targeted biopsy obtained during POCS, surgical pathology or autopsy assessment. The sensitivity, specificity, positive and negative predict values (PPV and NPV, respectively), and positive and negative likelihood ratios (LR + and LR-, respectively) using the classification were calculated in order to validate the classification; histopathology, surgery, and/or confirmatory 6-month follow-up were considered the gold standard.

A collection of 40 photographs of bile duct lesions from Stage II were randomly selected by C.R-M. Three experienced endoscopists (M.V., M.S.-A., and H.A.-E.) and three general practitioners (J.O.-A., J.B.-B., and R.O.) were enrolled to apply the proposed classification to the photographic dataset at two different time points, 3 weeks apart. Participants were blind to clinical records and histopathology.


#

Population selection

The eligibility criteria were the same for both stages. Patients with bile duct lesions detected by POCS who were aged ≥ 18 years and agreed to participate in the study were included. They had been referred to the unit for a suspected bile duct lesion (including mass or stenosis) based on physical symptoms (weight loss, jaundice with high levels of bilirubin), elevated serological tumor markers (carcinoembryonic antigen, carbohydrate antigen 19 – 9, carbohydrate antigen 125) and/or presence of a bile duct lesion on previous cross-sectional imaging such as ultrasonography, computed tomography (CT), magnetic resonance cholangiopancreatography (MRCP), endoscopic ultrasound, and/or ERCP, with no evidence of an extrinsic lesion.

Patients presenting with lesions but with no subsequent histological confirmation or with less than 6 months of follow-up were excluded from the analysis. Other exclusion criteria were: severe uncontrolled coagulopathy; esophageal/gastric/duodenal stenosing tumors or prior history of esophageal/gastric surgery with no possibility of scope passage; contrast allergy; and pregnancy.


#

Endoscopic technique

Stage I

The included patients were evaluated using a standard duodenoscope (Pentax ED 3670TK; Pentax Medical, Hoya Corp., Tokyo, Japan), the Pentax video processor EPK-i and EPK-i5010, and SpyGlass cholangioscopes (first-generation SpyGlass DVS direct visualization system [fiberoptic], and second-generation-SpyGlass DS digital system) in a mother–baby manner. Procedures were performed by a single experienced endoscopist who had conducted more than 300 ERCP procedures per year (C.R.-M.).

Patients were placed in a supine position, under general anesthesia, and all received antibiotic prophylaxis. SpyGlass biliary cannulation was achieved in all cases after sphincterotomy with an over-the-wire approach. The SpyScope was passed proximally, suction was used to clear bile and contrast material, sterile saline solution was infused to optimize imaging, and the cholangioscope was slowly withdrawn, allowing systematic inspection of the ductal mucosa. Images were captured, and video footage was recorded using a high definition image capture system.

Patients’ clinical records were reviewed to obtain case data both at the time of the endoscopy and at a follow-up session 6 months after the initial procedure.


#

Stage II

Enrolled patients were evaluated using a standard duodenoscope (Pentax ED 3670TK and ED 3490TK; Pentax Medical), the Pentax video processor EPK-i7010, and the second-generation SpyGlass DS Digital System in a mother – baby manner. Procedures were performed by three experienced endoscopists who had conducted more than 300 ERCP procedures per year (C.R.-M., M.V., and M.S.-A.) using the same technique described above. The principal investigator (C.R.-M.) was an experienced endoscopist in POCS (> 140 cases), whereas the other two endoscopists had less experience (< 20 cases of POCS) at the start of this stage of the protocol.

In both Stage I and Stage II, biopsies were taken using the SpyBite forceps (four biopsies per lesion) for histopathological analysis.


#
#

POCS macroscopic classification system

The POCS macroscopic classification was performed using the following new definitions, as established in Stage I of the study ([Table 1], [Fig. 1], [Video 1], [Video 2], [Video 3]):

  • non-neoplastic lesions

    • Type 1 “villous pattern”

      • A – micronodular, or

      • B – villous pattern without vascularity

    • Type 2 “polypoid pattern”

      • A – adenoma, or

      • B – granuloma pattern without vascularity

    • Type 3 “inflammatory pattern” – regular or irregular fibrous and congestive pattern with regular vascularity.

  • Neoplastic lesions

    • Type 1 “flat pattern” – flat and smooth, or irregular surface with irregular or spider vascularity without ulceration

    • Type 2 “polypoid pattern” – polypoid with fibrosis and irregular or spider vascularity

    • Type 3 “ulcerated pattern” – irregular ulcerated and infiltrative pattern with or without fibrosis and with irregular or spider vascularity

    • Type 4 “honeycomb pattern” – fibrous honeycomb pattern with or without irregular or spider vascularity.

Table 1

Peroral cholangioscopy macroscopic classification system of non-neoplastic and neoplastic common bile duct lesions.

Non-neoplastic lesions

Type 1

Villous pattern

A Micronodular, or
B Villous pattern without vascularity

Type 2

Polypoid pattern

A Adenoma, or
B Granuloma pattern without vascularity

Type 3

Inflammatory pattern

Regular or irregular fibrous and congestive pattern with regular vascularity

Neoplastic lesions

Type 1

Flat pattern

Flat and smooth or irregular surface with irregular or spider vascularity and without ulceration

Type 2

Polypoid pattern

Polypoid with fibrosis and irregular or spider vascularity

Type 3

Ulcerated pattern

Irregular ulcerated and infiltrative pattern with or without fibrosis and with irregular or spider vascularity

Type 4

Honeycomb pattern

Fibrous honeycomb pattern with or without irregular or spider vascularity

Zoom Image
Fig. 1 Macroscopic classification of common non-neoplastic bile duct lesions using SpyGlass peroral cholangioscopy (Boston Scientific, Marlborough, Massachusetts, USA). The lesions were classified as non-neoplastic based on their morphological and vascular patterns. The left column shows images with fiberoptic SpyGlass DVS (direct visualization system) and SpyGlass DS (digital system) of the lesions, and the right column shows illustrations depicting key features of each lesion type. In the illustrations, red lines represent vascularity and the yellow lines represent fibrosis. a Non-neoplastic lesions: Type 1 “villous pattern” (A – micronodular; B – villous pattern without vascularity); Type 2 “polypoid pattern” (A – adenoma; B – granuloma pattern without vascularity); Type 3 “inflammatory pattern” (regular or irregular fibrous and congestive pattern with regular vascularity). b Neoplastic lesions: Type 1 “flat pattern” (flat and smooth or irregular surface with irregular or spider vascularity); Type 2 “polypoid pattern” (polypoid or mass shape with fibrosis and irregular or spider vascularity); Type 3 “ulcerated pattern” (irregular ulcerated and infiltrative pattern with or without fibrosis and with irregular or spider vascularity); and Type 4 “honeycomb pattern” (fibrous honeycomb pattern with or without irregular or spider vascularity). Yellowish areas illustrate fibrosis, while reddish areas represent vascular pattern.

Video 1 Peroral cholangioscopy showing a non-neoplastic Type 1B lesion. Histological analysis showed an inflammatory lesion.

Georg Thieme Verlag. Please enable Java Script to watch the video.

Video 2 Peroral cholangioscopy showing a neoplastic lesion Type 1 (6 o’clock position) associated with a Type 4 (12 o’clock position). Histological analysis confirmed a cholangiocarcinoma.

Georg Thieme Verlag. Please enable Java Script to watch the video.

Video 3 Peroral cholangioscopy showing a neoplastic lesion Type 2. Histological analysis confirmed a cholangiocarcinoma.

Georg Thieme Verlag. Please enable Java Script to watch the video.

#

Prospective diagnosis during cholangioscopy

For validation of the proposed classification, one of the three experienced endoscopists who was blinded to clinical charts, laboratory results, and CT scan, MRCP or cholangiogram, gave the diagnosis 1 day after the POCS procedure by using the images recorded by the image capture system of the institution.


#

Inter- and intraobserver agreement

After finishing the prospective enrollment of patients, a dataset containing 40 random photographs of bile duct lesions from Stage II patients was collected by C.R.-M, which allowed an ≥ 80 % kappa value and a 20 % relative error. Randomization of pictures was possible through R v3.4.3 function “sample.” This photographic dataset was presented to three experienced endoscopists (M.V., M.S.-A., and H.A.-E.) and three general practitioners (J.O.-A., J.B.-B., and R.O.) who classified the images according to the proposed classification system at two different time points, 3 weeks apart. All participants were blinded to medical charts and histopathology results. The same images were shown to the participants at both sessions in a different order. Intraobserver agreement was measured through the comparison of the classification results at the two different times. Interobserver agreement was assessed based on the image classification results among the six participants (experienced endoscopists and general practitioners).


#

Statistical analysis

In Stage I, the sample size corresponded to all patients who successfully matched the selection criteria during the recruitment period. In Stage II, sample size was estimated by considering a 95 % confidence interval (CI), a 10 % margin of error, and the proportion of neoplastic cases calculated in Stage I. Baseline characteristics were fully described per stage through frequency (percentage) and mean (SD) or median (range), according to statistical distribution (Kolmogórov–Smirnov or Shapiro–Wilk test). Characteristics were compared between neoplastic and non-neoplastic bile duct lesions using a Pearson’s chi-squared or Fischer’s exact test for categorical variables, and using Student’s t test, Welch’s t test or Mann–Whitney U test for continuous variables.

Diagnostic accuracy to determine malignancy in POCS-guided biopsy of bile duct lesions evaluated through POCS visual impression using the proposed classification was estimated with sensitivity, specificity, PPV, NPV, likelihood ratios (LR + and LR –), observed agreement and inter-rater agreement, which were calculated with a 95 % CI. Intra and interobserver agreement were determined with Cohen’s kappa and Fleiss’s kappa, respectively. Kappa values below 40 % indicated “poor agreement”; from 40 % to 60 %, “moderate agreement”; from 60 % to 80 %, “good agreement”, and greater than 80 %, “excellent agreement”. A P value of < 0.05 was considered statistically significant. Data analysis was performed using R v3.4.3 (R Foundation for Statistical Computing, Vienna, Austria).


#
#

Results

A total of 171 patients were included (65 retrospective; 106 prospective) ([Fig. 2]).

Zoom Image
Fig. 2 Flow chart describing patient selection.

Stage I

A total of 65 patients (median age 67 years [range 29 – 89 years], 34 female [52.3 %]) were recruited to Stage I ([Table 2]). In 28 patients, the indication for POCS was suspicion of common bile duct (CBD) tumor, based on symptoms, elevated serological tumor markers or cross-sectional imaging; 21/28 (75.0 %) lesions were neoplastic and 7/37 (18.9 %) were non-neoplastic (P < 0.001) following histopathology and 6-month follow-up. 

Table 2

Baseline characteristics of patients in Stage I (retrospective analysis) who underwent SpyGlass peroral cholangioscopy.

Final diagnosis[1]

P value

Neoplastic (n = 28)

Non-neoplastic (n = 37)

Total (n = 65)

Age, median (range), years

68 (32 – 89)

64 (29 – 86)

67 (29 – 89)

0.367[2]

Sex, female, n (%)

15 (53.6)

19 (51.4)

34 (52.3)

0.859[3]

Procedure, n (%)

0.102[3]

  • Spyglass DVS (fiberoptic)

20 (71.4)

19 (51.4)

39 (60.0)

  • Spyglass DS (digital)

8 (28.6)

18 (48.6)

26 (40.0)

Main POCS indication, n (%)

< 0.001[3]

  • Suspicion of CBD tumor

21 (75.0)

7 (18.9)

28 (43.1)

  • Indeterminate CBD stricture

7 (25.0)

17 (45.9)

24 (36.9)

  • Post-surgical CBD stricture

2 (5.4)

2 (3.1)

  • CBD stones

11 (29.7)

11 (16.9)

Proposed macroscopic classification system, n (%)

< 0.001 [3]

  • Neoplastic lesions

28 (100)

2 (5.4)

30 (46.2)

    • Type I

5/28

1/2

    • Type II

10/28

1/2

    • Type III

11/28

    • Type IV

2/28

  • Non-neoplastic lesions

35 (94.6)

35 (53.8)

    • Type I

11/35

    • Type II

11/35

    • Type III

13/35

Histopathology result, n (%)

< 0.001 [3]

  • Neoplastic lesions

27 (96.4.)

2 (5.4)

29 (44.6)

    • Cholangiocarcinoma

17/27

2 (100.0)

19/29

    • Intraductal papillary neoplasm

3/27

3/29

    • Hepatocarcinoma

1/27

1/29

    • CBD adenoma

4/27

4/29

    • Hemangioma

1/27

1/29

    • Papillomatosis

1/27

1/29

  • Non-neoplastic lesions

1 (3.6)

34 (91.9)

35 (53.8)

  • Chronic inflammation[4]

1/1

31/34

32/35

  • Acute inflammation

1/34

1/35

  • Mucosal hyperplasia

2/34

2/35

Not enough sample, n (%)

1 (2.7)

1 (1.5)

Presence of parasites, n (%)

2 (7.1)

3 (8.1)

5 (7.7)

0.632[5]

Parasites, n (%)

n/a

  • Ascaris lumbricoides

1/3

1/5

  • Enterobius vermicularis

1/2

1/5

  • Fasciola hepatica

1/3

1/5

  • Helminths

1/3

1/5

  • Nematodes and monilias

1/2

1/5

Complications, n (%)

1 (3.6)

2 (5.4)

3 (4.6)

0.605[5]

  • Acute cholangitis

1/1

1/2

2/3

  • Acute pancreatitis

1/2

1/3

POCS, peroral cholangioscopy; CBD, common bile duct.

1 Final diagnosis was based on histopathology and 6-month follow-up.


2 Mann–Whitney U test.


3 Pearson’s chi-squared test.


4 Chronic inflammation secondary to CBD stones.


5 Fischer’s exact test.


Histopathological assessment showed 29/65 lesions (44.6 %) to be neoplastic; cholangiocarcinoma (CCA) was the most commonly observed cancer (17/29 [58.6 %]). Two lesions had a positive biopsy for CCA but with a survival of more than 2 years, which made the diagnosis less probable (false positive). A total of 35/65 lesions (53.8 %) were non-neoplastic, with chronic inflammation secondary to bile duct stones being the most common cause of lesion development. There was 1/35 case that was considered an inflammatory lesion (non-neoplastic), but the patient underwent liver transplantation 3 months later (false negative).

The biopsy sample was inadequate in one patient in Stage I; however, the patient was alive at 3 years’ follow-up and showed no signs of malignancy, and the lesion was thus classified as non-neoplastic.

[Table 2] summarizes the main POCS indications. There were 21 (75.0 %) cases in which the POCS was indicated for a suspected neoplastic bile duct lesion based on the physical symptoms, elevated serological tumor markers, or cross-sectional imaging. The remaining 25 % of the suspected neoplastic lesions had an indeterminate bile duct stricture based on a previous negative ERCP with brushing and/or biopsy. By contrast, for non-neoplastic bile duct lesions, an indeterminate stricture on ERCP was the main POCS indication (45.9 %).

The POCS macroscopic classification does not describe pathognomonic features related to parasite infections; however, there were five cases of bile duct lesions caused by parasites, four of which were initially classified as neoplastic lesions by POCS macroscopic classification and histopathology. In 2/4 cases, lesions had completely disappeared after antiparasitic drug treatment, as assessed at the 6-month follow-up ([Fig. 3]). In the remaining 2/4 cases, papillomatosis and hemangioma were reported.

Zoom Image
Fig. 3 Peroral cholangioscopy images. a Image showing a Type 2B non-neoplastic lesion without vascularization. The biopsy of this lesion showed a granuloma by Ascaris lumbricoides. b Image captured at follow-up, 6 months after albendazole treatment, showing the lesion has disappeared.

#

Stage II

Using the proportion of neoplastic cases in Stage I (43.1 %), a sample size of 95 patients was estimated for Stage II. A total of 106 patients were successfully enrolled ([Fig. 2]): median age 65.5 years (range 15 – 95 years), 54 female (50.9 %, [Table 3]). A total of 45/106 patients (42.5 %) had a main POCS indication of indeterminate CBD stricture: 20/54 (37.0 %) were finally diagnosed as neoplastic lesions and 25/52 (48.1 %) as non-neoplastic (P < 0.001). A total of 44/106 lesions (41.5 %) were confirmed as neoplastic by histopathology. Follow-up confirmed CCA as the most frequent diagnosis (38/44 [86.3 %]); 5/44 (11.4 %) were CBD adenoma, and a further non-neoplastic lesion was confirmed as CBD adenoma by histopathology.

Table 3

Baseline characteristics of patients in Stage II (prospective analysis) who underwent SpyGlass peroral cholangioscopy.

Final diagnosis[1]

P value

Neoplastic (n = 54)

Non-neoplastic (n = 52)

Total (n = 106)

Age, median (range), years

71.5 (31 – 95)

58 (15 – 87)

65.5 (15 – 95)

< 0.001[2]

Sex, female, n (%)

29 (53.7)

25 (48.1)

54 (50.9)

0.562[3]

Procedure, n (%)

n/a

  • Spyglass DS (digital)

54 (100)

52 (100)

106 (100)

Main POCS indication, n (%)

< 0.001[3]

  • Suspicion of CBD tumor

30 (55.6)

7 (13.5)

37 (34.9)

  • Indeterminate CBD stricture

20 (37.0)

25 (48.1)

45 (42.5)

  • Post-surgical CBD stricture

1 (1.9)

2 (3.8)

3 (2.8)

  • CBD stones

3 (5.6)

18 (34.6)

21 (19.8)

Proposed macroscopic classification system, n (%)

< 0.001 [3]

  • Neoplastic lesions

52 (96.3)

4 (7.7)

56 (52.8)

    • Type I

6/52

1/4

    • Type II

23/52

2/4

    • Type III

15/52

1/4

    • Type IV

8/52

  • Non-neoplastic lesions

2 (3.7)

48 (92.3)

50 (47.2)

    • Type I

1/2

30/48

    • Type II

8/48

    • Type III

1/2

10/48

Histopathology result, n (%)

< 0.001 [3]

  • Neoplastic disease

43 (79.6)

1 (1.9)

44 (41.5)

    • Cholangiocarcinoma

38/43

38/44

    • CBD adenoma

5/43

1/1

6/44

  • Non-neoplastic disease

9 (16.7)

47 (90.4)

56 (52.8)

    • Chronic inflammation[4]

9/9

42/47

51/56

    • Acute inflammation

5/47

5/56

Not enough sample, n (%)

2 (3.7)

4 (7.7)

6 (5.7)

Parasites, n (%)

7 (13.5)

7 (6.6)

0.005[3]

  • Ascaris lumbricoides

5/7

5/7

  • Nematodes

2/7

2/7

Complications, n (%)

3 (5.6)

3 (2.8)

0.128[5]

  • Acute cholangitis

1/3

1/3

  • Acute pancreatitis

2/3

2/3

POCS, peroral cholangioscopy; CBD, common bile duct.

1 Final diagnosis was based on histopathology and 6-month follow-up.


2 Mann–Whitney U test.


3 Pearson’s chi-squared test.


4 Chronic inflammation secondary to CBD stones.


5 Fischer’s exact test.


A total of 56 /106 lesions (52.8 %) were confirmed as non-neoplastic by histopathology, with chronic inflammation as the most common cause (51/56 [91.1 %]). POCS-guided biopsy did not obtain sufficient sample material in six cases; follow-up confirmed neoplastic and non-neoplastic lesions in 2 /54 (3.7 %) and 4 /52 (7.7 %), respectively. One case had an inadequate biopsy sample on the initial POCS in March 2016, but two subsequent POCS-guided biopsies conducted 3 and 8 months after the first biopsy revealed inflammation that led it to be classified as a non-neoplastic lesion.

There were seven cases of parasite infection in the non-neoplastic group, five (71.4 %) of which were due to Ascaris lumbricoides and two (28.6 %) were due to nematodes (P = 0.005).

Three patients (2.8 %) experienced POCS-related adverse events, including one case of cholangitis requiring a hospital stay of 2 – 3 days with intravenous antibiotic administration. In addition, there were two cases of post-procedure mild acute pancreatitis, and these were hospitalized and managed conservatively with no complications.

Of the whole patient cohort (n = 171), 22 patients (12.8 %) had undergone previous plastic biliary stent placement with or without brush cytology. In the non-neoplastic group, the characteristic of the epithelium showed a villous pattern Type 1 according to the proposed classification ([Table 1]). In the neoplastic group the vascular pattern was not altered by the previous plastic stent.

[Table 4] shows the overall accuracy of POCS for neoplasia diagnosis in Stage II based on histopathological and visual impressions using the proposed classification.

Table 4

Overall accuracy of neoplastic diagnosis in Stage II patients using the proposed macroscopic classification system, considering follow-up as gold standard.

n (%)

95 %CI

Proposed macroscopic classification system (n = 106)

  • Sensitivity

52/54 (96.3)

87.0 – 100.0

  • Specificity

48/52 (92.3)

81.0 – 98.0

  • PPV

52/56 (92.9)

83.0 – 98.0

  • NPV

48/50 (96.0)

86.0 – 100.0

  • LR + 

12.52

4.88 – 32.14

  • LR – 

0.04

0.01 – 0.16

  • Observed agreement

100/106 (94.3)

86.0 – 98.9

  • Inter-rater agreement

100/106 (94.3)

79.9 – 97.5

POCS-guided biopsy (n = 100)[*]

  • Sensitivity

43/52 (82.7)

70.0 – 92.0

  • Specificity

47/48 (97.9)

89.0 – 100.0

  • PPV

43/44 (97.7)

88.0 – 100.0

  • NPV

47/56 (83.9)

72.0 – 92.0

  • LR + 

39.36

5.68 – 277.16

  • LR – 

0.18

0.10 – 0.32

  • Observed agreement

90/100 (90.0)

82.4 – 95.1

  • Inter-rater agreement

90/100 (90.0)

68.5 – 91.7

PPV, positive predictive value; NPV, negative predictive value; LR, likelihood ratio; POCS, peroral cholangioscopy; CI, confidence interval.

* POCS-guided biopsy was not able to classify six cases properly owing to insufficient sample material.


Stability validation of the classification criteria for neoplasia diagnosis and neoplastic types was robust, especially in the expert group subanalysis. For differentiation between neoplastic and non-neoplastic lesions, the proposed classification showed high reproducibility between different observers (interobserver agreement), but also within individuals (intraobserver agreement 3 weeks later). However, this reproducibility was better for endoscopy experts (κ value always higher than 80 % – excellent agreement) than non-experts (κ value from 64.7 % to 81.9 %) ([Table 5], [Table 6]).

Table 5

Inter and intraobserver agreement in Stage II: neoplastic or non-neoplastic lesions.

Experts

Non-experts

Both groups

Interobserver agreement (Fleiss kappa), %

89.0

81.9

84.3

Intraobserver agreement 3 weeks later (Cohen kappa), %

M.V.

94.7

J.O.-A.

64.7

M.S.-A.

83.9

J.B.-B.

76.5

H.A.-E.

89.4

R.O.

67.1

Table 6

Inter and intraobserver agreement in Stage II: types.

Experts

Non-experts

Both groups

Interobserver agreement (Fleiss kappa), %

90.6

66.5

74.0

Intraobserver agreement three weeks later (Cohen kappa), %

M.V.

93.0

J.O.-A.

63.0

M.S.-A.

82.1

J.B.-B.

75.0

H.A.-E.

85.6

R.O.

66.0


#
#

Discussion

The accurate differentiation of neoplastic from non-neoplastic biliary lesions is imperative because management of neoplastic and non-neoplastic biliary diseases differs greatly [6] [7] [8] [9] [10]. Current POCS diagnosis of neoplastic strictures is based on previously described features, such as the presence of irregularly tortuous, dilated blood vessels caused by neovascularization, easy oozing of blood, an irregular surface or the presence of an intraductal mass protruding into the lumen of the bile duct [4] [9]. However, despite these consistent and well-defined independent features distinguishing neoplastic from non-neoplastic lesions, recent studies using currently accepted criteria demonstrate poor interobserver agreement, with only 45 % accuracy [4].

It should be noted that the presence of tumor vessels (the most well-described indicator of malignancy) is present in only 61 % of cases [7]. Other features such as ulceration and papillary projections (not previously described) were more greatly associated with a neoplastic diagnosis [4]. Moreover, not all masses protruding into the lumen of the bile duct are neoplastic, and all these features could be present in combination. In addition, current studies do not provide a further description of non-neoplastic, as has been performed in our present analysis. Therefore, heterogeneity and interobserver variation must be expected (as has been reported) [4], leading to incorrect diagnosis and treatment of patients. Thus, a standardized POCS classification using SpyGlass is necessary. For example, the presence of papillary projections has been associated with neoplasia [4] (1B subtype of the proposed classification); in our study, however, this morphological characteristic was associated with non-neoplastic lesions. Furthemore, 2A lesions were associated with eosinophilic granuloma secondary to parasites, and 2B lesions were associated with intrapapillary mucinous lesions.

A systematic review evaluated the accuracy of visual cholangioscopic findings in determining malignancy of biliary strictures. In particular, the presence of a mass with dilated tortuous vessels was the macroscopic feature assessed, and most of the studies used the first-generation SpyGlass [10]. The sensitivity and specificity were found to be 84.5 % and 82.6 %, respectively.

Another review evaluated the diagnostic accuracy of the cholangioscopy-guided biopsies in the diagnosis of neoplastic biliary strictures, and sensitivity and specificity of 60.1 % and 98.0 %, respectively, were found [10]. A further subanalysis including only patients who had previous negative imaging, brushings and/or intraductal biopsies found that the sensitivity was higher (74.7 %) but the specificity was lower (93.3 %) [11] [12] [13]. Among the studies discussed in the review that specifically reported the role of POCS in CCA diagnosis, the pooled sensitivity and specificity were 66.2 % and 97.0 %, respectively.

Similarly, Sun et al. [14], in a recent meta-analysis, evaluated the overall accuracy of the visual impression and SpyBite biopsy guided by SpyGlass POCS in the diagnosis of malignant biliary lesions [2] [3] [5] [11]. The overall diagnostic performance did not differ greatly between the SpyGlass visual impression and the SpyBite biopsy. The authors concluded that the SpyGlass system performs well in differentiating neoplastic from non-neoplastic biliary lesions. Visual impression is useful for detecting neoplastic lesions, whereas SpyBite biopsy is better for confirming a neoplastic diagnosis. However, neither visual impression nor SpyBite biopsy is perfect at excluding biliary cancer. Negative results of SpyBite biopsy should be interpreted with caution [14].

Another multicenter study [15], evaluating 44 patients with indeterminate biliary strictures using the new digital POCS system (SpyGlass DS), demonstrated that using macroscopic features to visually diagnose malignancy showed a similar sensitivity (90 %) but higher specificity (95.8 %) compared with previous reports. However, the endoscopists were not blinded, and thus bias may have been introduced in the evaluation [15]. Moreover, 20 patients had a final diagnosis of malignancy (15 CCA and 5 pancreatic adenocarcinoma), 18 of whom had shown POCS findings suggestive of neoplastic lesions (tumor vessels and/or masses). Interestingly, the two patients with no cholangioscopic features of neoplastic lesions had pancreatic cancer. This result is not surprising because biliary strictures caused by extraluminal compression, such as those associated with pancreatic cancer, cannot be detected by visualization of the bile duct unless the disease is at later stages when the tumor has infiltrated and penetrated the bile duct wall [15].

Because of the heterogeneity and interobserver variation for visual impression, and the differences between histological and macroscopic findings, we propose a new POCS classification in order to standardize the macroscopic characteristic of bile duct lesions. For this reason, we excluded all indeterminate biliary strictures due to pancreatic cancer or others, to avoid bias regarding the macroscopic characteristics.

Our results suggest that by using the new classification, the visual impression could be improved with an increase in sensitivity and an NPV of 96 %. Moreover, the interobserver agreement was excellent in the expert and nonexpert participants; meanwhile the intraobserver agreement was excellent in experienced participants and good for non-experts for defining neoplastic and non-neoplastic lesions. Meanwhile, heterogeneity in the SpyBite biopsy sensitivity rate has been reported, which may be explained by the fact that there is currently no consensus on the optimal number of SpyBite biopsy attempts. Some authors recommend that at least four biopsies need to be performed [5]. Although analyzing the accuracy of the SpyGlass-guided biopsy was not the main objective of our study, the overall sensitivity and specificity were 82.7 % and 97.9 %, respectively. It has to be considered that four targeted biopsies were performed per case, and more tissue was obtained, which probably improved the results. However, these results are similar to other studies where the sensitivity and specificity of POCS-guided biopsies for the diagnosis of malignancy was 85 % and 100 %, respectively, with a median of 1 biopsy pass performed (range 1 – 6) [10].

The present study has some limitations. It was a single-center, nonrandomized, noncontrolled study, and both cholangioscopy systems were used during Stage I, although this does not impact the development of the proposed classification. Patients with postoperative strictures were not included as part of the population evaluated. In addition, potential limitations include the lack of a gold standard, failure to include patients with primary sclerosing cholangitis (uncommon in our country) or previous self-expanding metal stents, and not having evaluated all images for interobserver and intraobserver analysis.

The study also has several strengths and clinical implications. It was performed in an ambispective way, with 6-months’ follow-up for POCS analysis, which would have improved the diagnosis of the lesions evaluated and the correlation with histopathology. A further description of non-neoplastic lesions was included, which took account of inflammatory lesions and, importantly, bile duct lesions caused by parasites, which has not been included in previous studies. Although, we included patients with previous biliary plastic stent and cholangitis, these patients had Type 1 non-neoplastic villous pattern for non-neoplastic lesions or areas. Biliary plastic stents did not affect the vascular pattern compatible with neoplastic lesions in case of neoplasia. Finally, we validated the proposed classification by including experienced endoscopists, POCS operators, and general practitioners.

In conclusion, our findings demonstrate that the novel classification system could help physicians to distinguish non-neoplastic from neoplastic bile duct lesions. Further prospective evaluation of these data is necessary.


#
#

Competing interests

None

Acknowledgment

This study was presented as an oral presentation at EUG Week 2016 (Vienna, Austria; United European Gastroenterol J 2016; 4 (5S): A48).


Corresponding author

Carlos Robles-Medranda, MD
Instituto Ecuatoriano de Enfermedades Digestivas – IECED
OMNI Hospital
Av. Abel Romeo Castillo y Av. Juan Tanca Marengo
Torre Vitalis, Mezanine 3
090505, Guayaquil
Ecuador   
Fax: +593-4-2109180    


Zoom Image
Fig. 1 Macroscopic classification of common non-neoplastic bile duct lesions using SpyGlass peroral cholangioscopy (Boston Scientific, Marlborough, Massachusetts, USA). The lesions were classified as non-neoplastic based on their morphological and vascular patterns. The left column shows images with fiberoptic SpyGlass DVS (direct visualization system) and SpyGlass DS (digital system) of the lesions, and the right column shows illustrations depicting key features of each lesion type. In the illustrations, red lines represent vascularity and the yellow lines represent fibrosis. a Non-neoplastic lesions: Type 1 “villous pattern” (A – micronodular; B – villous pattern without vascularity); Type 2 “polypoid pattern” (A – adenoma; B – granuloma pattern without vascularity); Type 3 “inflammatory pattern” (regular or irregular fibrous and congestive pattern with regular vascularity). b Neoplastic lesions: Type 1 “flat pattern” (flat and smooth or irregular surface with irregular or spider vascularity); Type 2 “polypoid pattern” (polypoid or mass shape with fibrosis and irregular or spider vascularity); Type 3 “ulcerated pattern” (irregular ulcerated and infiltrative pattern with or without fibrosis and with irregular or spider vascularity); and Type 4 “honeycomb pattern” (fibrous honeycomb pattern with or without irregular or spider vascularity). Yellowish areas illustrate fibrosis, while reddish areas represent vascular pattern.
Zoom Image
Fig. 2 Flow chart describing patient selection.
Zoom Image
Fig. 3 Peroral cholangioscopy images. a Image showing a Type 2B non-neoplastic lesion without vascularization. The biopsy of this lesion showed a granuloma by Ascaris lumbricoides. b Image captured at follow-up, 6 months after albendazole treatment, showing the lesion has disappeared.