Keywords ERCP - ultrasound - x-ray - abdomen - post-ERCP pancreatitis
Introduction
Endoscopic retrograde cholangiopancreatography (ERCP) is the primary therapeutic modality
for biliary and pancreatic ductal diseases [1 ]. Due to the technically demanding examination and the anatomical proximity between
the bile and the pancreatic ducts, there is still a 3.5–9.7% risk of developing post-ERCP
pancreatitis (PEP) and an overall mortality rate of 0.1–0.7%, despite sophisticated
preventive measures [2 ]
[3 ].
Accordingly, PEP prophylaxis has high clinical relevance. The prophylactic placement
of a pancreatic stent (PS) has been shown to significantly reduce PEP by several large
meta-analyses (odds ratio 0.22–0.39) [4 ]
[5 ]. Therefore, international guidelines recommend the placement of prophylactic plastic
stents in the pancreatic duct to secure drainage in the case of accidental cannulation
or the application of contrast agent in the pancreatic duct [6 ]
[7 ]
[8 ]
[9 ]. 5-Fr stents that remain for at least 12 to 24 hours after the ERCP procedure are
used as the standard [10 ]
[11 ]
[12 ].
Although a significant number of stents dislodge spontaneously in the first days after
placement, endoscopic removal of retained PSs is currently recommended after at least
five to ten days by international guidelines to prevent complications [9 ]
[11 ]
[13 ]
[14 ]
[15 ]. To avoid unnecessary esophagogastroduodenoscopy (EGD), imaging before stent removal
is recommended to visualize retained stents [9 ]
[13 ]. In most centers a fluoroscopic image is performed in the ERCP unit [9 ]. Accordingly, this procedure takes up pivotal resources of the endoscopy department.
Furthermore, radiation exposure poses a risk to patients and staff [16 ].
Recently, in a pilot trial with 41 patients, we investigated the feasibility and technical
aspects of detecting prophylactic PSs by ultrasound. In this pilot study, all patients
underwent ultrasound and fluoroscopic imaging, and a positive predictive value of
above 90% for the detection of PSs by ultrasound was reported [16 ].
The aim of this trial was to evaluate a novel ultrasound-based algorithm as primary
directive imaging before endoscopic stent removal.
Methods
Study design
The present study is a prospective single-center study to evaluate a novel algorithm
for the extraction of PSs placed to prevent PEP. All patients provided written informed
consent before study participation.
All PSs were placed in patients at risk of undergoing ERCP to prevent PEP if indicated
by current guidelines [9 ]. The placed stents were straight 6cm, 5-Fr (0.035 inch) polyurethane stents with
a single external flap and no internal one (Pancreatic Stent, Optimed, Ettlingen,
Germany).
Inclusion criteria were I) placement of a prophylactic PS to prevent PEP during ERCP,
II) patients older than 18 years of age, and III) written informed consent.
Excluded were all patients with I) a condition prohibiting ultrasound examination,
X-ray, or EGD, II) patients unable to give informed consent or III) PSs for indications
other than PEP prophylaxis.
Patients included in the trial underwent a bedside ultrasound examination in the ward
before being transferred to the high-end ultrasound device for a second examination
with optimal external conditions (e.g., darkened room, avoidance of disruption). All
ultrasound examinations were performed using a transducer with a frequency of 4.5
to 5 MHz. [Table 1 ] summarizes the used devices. Both ultrasound examinations were performed by independent examiners who were required
to have undergone training of at least 250 examinations in the sonographic department.
If there was no sufficiently experienced examiner in the ward, the examination was
performed by one of the study authors. The results were blinded. The number of ultrasound
examinations and years of experience of the examiners were documented. The time of
the first study examination was used to determine the time of the definition.
Table 1 Summary of the ultrasound devices and transducers used in the study.
Group
Ultrasound device
Transducer
Company and origin
Bedside ultrasound
Acuson ×300
CH5–2 transducer (frequency: 5.0 MHz, range: 1.4–5.0 MHz, field of view: 66°)
Siemens, Munich, Germany
High-quality ultrasound
Aplio 500
PVT-375SC transducer (frequency: 5.0 MHz, range: 1.5–6.0 MHz, field of view: 70°)
Toshiba, Tokyo, Japan
Aplio i800
i8CX1 transducer (frequency: 5.0 MHz, range: 1.8–6.2 MHz, field of view: 70°)
Canon, Ōtawara, Japan
Hi Vision Ascendus
EUP C715 transducer (frequency: 5.0 MHz, range: 1.0–5.0 MHz, field of view: 70°)
Hitachi, Tokyo, Japan
Acuson S2000
a 4C1 transducer (frequency: 4.5 MHz, range: 1.0–5.0 MHz, field of view: 66°)
Siemens, Munich, Germany
Acuson Sequoia
5C1 transducer (frequency: 5 MHz, range: 1.4–5.0 MHz, field of view: 70°)
Siemens, Munich, Germany
All patients were fasting on the day of the examination. The examination was performed
in the supine position. If visibility of the pancreatic head region was limited, the
patient was turned into the left lateral position. The stomach was never filled with
fluid, as this could pose a risk for the subsequent EGD. Poor examination conditions
were defined as those where reliable visualization of the target structures (pancreatic
head, pancreatic duct, common hepatic duct and confluence of the lienal and superior
mesenteric veins) was not possible (e.g., due to overlaying gas, abdominal fatty tissue).
Results of the ultrasound examinations were PS visualized in the pancreatic duct or
stent not being visualized. All patients underwent at least one of the two possible
ultrasound examinations. If patients underwent both examinations, high-end ultrasound
defined the subsequent process.
[Fig. 1 ] shows the implemented algorithm. If the stent was classified as dislodged, a fluoroscopic
image was obtained in the ERCP unit to rule out retained stents by false-negative
sonographic results. In the case of stents being classified as retained, fluoroscopic
imaging was omitted, and the stent was removed directly via EGD.
Fig. 1 Sonography-favoring algorithm to remove pancreatic stents*. All pancreatic stents
being placed to prevent post-ERCP pancreatitis (PEP) are primarily visualized by sonography.
Stents being retained in the pancreatic duct were then directly removed by esophagogastroduodenoscopy
(EGD) omitting further imaging. To prevent false-negative results, stents being described
as dislodged by sonography underwent subsequent X-ray imaging. Only stents then described
as retained were removed via EGD. No further intervention was needed in the case of
pancreatic stents being identified as dislodged by both sonography and X-ray. EGD
= esophagogastroduodenoscopy; ERCP = endoscopic retrograde cholangiopancreatography;
PEP = post-ERCP pancreatitis. *The algorithm was first published by Michael et al.
[17 ] during a first pilot trial.
Patient transport between the ward, the ultrasound room, and the endoscopy unit was
performed on a bed and in quick succession to reduce the minor risk of stent dislodgment
between examinations. In this single-arm study, neither randomization nor blinding
was required.
Outcomes
There were three ultrasound groups: I) ultrasound procedure overall (defined as either
the high-end ultrasound procedure or the bedside ultrasound examination in the absence
of a high-end ultrasound examination), II) high-end ultrasound, and III) bedside ultrasound.
The primary outcome of the study was to calculate the positive predictive value of
sonography for the detection of pancreatic and biliary stents in each group. The fluoroscopic
image or the EGD result was used as the reference method. Secondary outcomes were
to calculate the negative predictive value, sensitivity, and specificity. Contingency
coefficients were determined to compare the agreement between the different types
of ultrasound examinations and X-ray or EGD.
Statistical analysis was performed to determine associations between baseline characteristics
and success of sonographic stent detection in high-end and bedside ultrasound examination
and PS dislodgement.
Statistical analysis
Data collection, data management, and statistical analyses were performed using the
SPSS software package, release 21 (IBM, Armonk, USA).
The sample size was calculated on the basis of the data from a previous pilot trial
on the same topic [16 ]. Assuming a sensitivity of 93.5%, a desired confident P for the confidence interval
of 95%, and a desired length of the confidence interval of 12%, a case number of 64
patients was obtained. This refers to patients in whom a PS could be visualized sonographically
in the duct. According to the previous study, this was the case in approximately 73.1%
of patients on the day of removal. As a result, 88 subjects were identified as meeting
the study goal.
Descriptive statistics were computed to provide frequencies and percentages for categorical
variables and median and 25%/75% quartiles for continuous values. The positive predictive
value, negative predictive value, sensitivity, and specificity were calculated. The
mean contingency coefficient (φ) was calculated to evaluate the correlation between
the ultrasound device and X-ray and/or endoscopy. Univariate and multivariate analyses
were performed to detect risk factors for PS dislodgement and the success of sonographic
stent detection. All reported p-values are two-sided. Statistical significance was
considered if the p-values were below 0.05.
Results
Study characteristics
[Fig. 2 ] shows a flowchart of the patient inclusion. A total of 98 patients were assessed
for eligibility. Ten patients had to be excluded: five patients did not undergo sonography
before endoscopy, two patients died before stent extraction due to end stage cancer,
one PS migrated via naturalis, one patient was pregnant, and one patient refused stent
extraction after screening. Therefore, 88 patients underwent the intended study protocol
and were analyzed for the primary outcome. 86 patients underwent sonography with a
high-end device and 77 patients were examined with bedside ultrasound. The predefined
number of cases was reached. Patient and procedural characteristics are summarized
in [Table 2 ]. [Fig. 3 ] illustrates the methods.
Fig. 2 Study flowchart. ERCP = endoscopic retrograde cholangiopancreatography.
Table 2 Patient and procedural characteristics.
Continuous parameters are expressed as medians with range, nominal parameters as number
of patients with percentage of occurrence.
* Hypotension was treated by IV infusion with no delay of the procedure and no further
complications.
ERCP: endoscopic retrograde cholangiopancreatography; EGD: esophagogastroduodenoscopy
Patient characteristics
Female gender
41 (47%)
Age (years)
62 (52/69)
BMI (kg/m²)
24.5 (21.3/29.0)
Pancreatic disease
18 (20%)
Pancreatic carcinoma
12 (14%)
Pancreatitis
6 (7%)
Pancreas lipomatosis
9 (11%)
Liver disease
40 (45%)
Liver metastasis
10 (11%)
Liver transplantation
10 (11%)
Cholangiocarcinoma
6 (7%)
Sclerosing bile duct disease
4 (5%)
Liver cirrhosis
9 (10%)
Budd-Chiari-Syndrome
1 (1%)
Abdominal surgery
24 (27%)
Procedural characteristics
ERCP indication
Malignant stenosis
35 (40%)
Choledocholithiasis
29 (33%)
Anastomotic stenosis after liver transplantation
8 (9%)
Biliary leakage
5 (6%)
Prophylactic stenting after ampullectomy
4 (5%)
Sclerosing bile duct disease
4 (5%)
Others
3 (3%)
ERCP complications
15 (17%)
Post-ERCP pancreatitis
12 (14%)
Perforation
1 (1%)
Cholangitis
1 (1%)
Hypoxemia*
1 (1%)
Days between pancreatic stent placement and removal
2 (2/3.75)
Ultrasound procedures overall
88 (100%)
Bedside ultrasound procedures
77 (88%)
High-end ultrasound procedures
86 (98%)
Aplio i800 (Canon)
28 (32%)
Hi Vision Ascendus (Hitachi)
24 (28%)
Aplio 500 (Toshiba)
17 (20%)
Acuson Sequoia (Siemens)
8 (9%)
Acuson S2000 (Siemens)
5 (6%)
Undocumented
4 (5%)
Sonographic conditions evaluated as difficult by the examiner
26 (30%)
Number of X-rays performed
23 (26%)
Number of EGDs performed
78 (89%)
Adverse events during EGD
1 (1%)*
Adverse events after EGD
0 (0%)
Fig. 3 Images of the applied procedures. The upper row presents sonographic pictures in B-mode
made by devices of the high-end ultrasound group (left and right picture) and of the
bedside ultrasound group (middle picture). Displayed is the head region of the pancreas
in a longitudinal section tilted slightly clockwise. The left and the middle pictures
are of the same patient on consecutive procedures. The orange arrow points at the
pancreatic stent. The orange x is placed in the vena cava inferior and the orange
* is placed in the portal vein. The green arrow in the right picture indicates a plastic
biliary duct stent. In the lower row, a fluoroscopic image of a pancreas stent and
a biliary duct stent is displayed. The picture in the middle shows an endoscopic image
of a side-viewing scope. On the image, a straight 5-Fr 6cm plastic pancreatic stent
with an external flange and a plastic pig tail stent enter the papilla. On the endoscopic
picture on the right side, a prophylactic pancreatic stent and a biliary stent are
displayed using a diagnostic front-view gastroscope.
Simultaneous stenting of the common bile duct during ERCP was performed in 79 (90%)
patients. One stent was placed in 73 (92%), and two stents in 6 (8%) patients. Plastic
stents were placed in 75 (95%) patients, and self-expandable metal stents in 4 (5%)
patients. 15 patients (17%) developed an ERCP-related complication, with mild PEP
(12, 14%) being the most common.
PS extraction according to the algorithm was performed between days 1 and 13 (median:
2 days) after stent placement. A total of 67 stents (76%) were retained on the day
of extraction and 21 stents were dislodged. [Table 3 ] shows whether the PS was retained on the particular day of stent visualization and
extraction.
Table 3 Status of pancreatic stent position on the day of removal.
Days after placement of pancreatic stent
Pancreatic stent retained
Total
n = 88
No
n = 21
Yes
n = 67
1
3
12
15
2
10
27
37
3
3
11
14
4
3
7
10
5
1
4
5
6
0
1
1
7
1
1
2
8
0
1
1
10
0
1
1
12
0
1
1
13
0
1
1
Outcome of pancreatic stent detection using ultrasound
The results of PS detection according to the presented algorithm are shown in [Fig. 4 ].
Fig. 4 Contingency table with ultrasound outcomes of pancreatic stent detection. The table
displays the trial outcomes comparing ultrasound results with esophagogastroduodenoscopy
and/or X-ray results given as absolute values and proportion in brackets. Positive
predictive value (1), negative predictive value (2), sensitivity (3), and specificity
(4) are calculated as proportion and 95% confidence interval in brackets. EGD = esophagogastroduodenoscopy.
Ultrasound reported 65 retained stents, 54 stents were confirmed and extracted by
EGD. Thus, the positive predictive value was 83% (95%-CI: 71–91%). The negative predictive
value was 43% (95%-CI: 23–65%), with 10/23 stents correctly being reported to be dislodged.Supplementary Table 2 gives an overview of the examiners’ experience and their performance during the trial.
X-ray was performed in 23 patients only and was avoided in the remaining 65 cases
(74%). EGD was performed in 78 patients (89%). In 11 patients (13%), EGD was performed
although the PS was already dislodged, even though sonography reported a retained
stent (false-positive rate). No complications were recorded during stent removal via
EGD, and no procedure-related pancreatitis was documented. In one case, arterial hypotension
occurred during EGD, which was treated with intravenous fluid.
Outcome of biliary stent detection using ultrasound
The positive predictive value was 97% (95%-CI: 91–100%), with 76/78 biliary stents
being correctly described as retained by ultrasound. Stent dislodgement was assessed
correctly by sonography in 1/3 cases (specificity: 33%, 95%-CI: 8–91%). The sensitivity
and negative predictive value were 100%.
Univariate and multivariate logistic regression analyses
No significant association was observed between PS dislodgement and the time from
ERCP to stent visualization (p > 0.2), BMI > 25 kg/m2 (p = 0.09), BMI > 30 kg/m2 (p = 0.12), pancreatic disease (p > 0.2), liver disease (p > 0.2), previous abdominal
surgery (p > 0.2), indication of ERCP (p > 0.2), coincidental biliary stent placement
(p > 0.2), and PEP (p = 0.09).
The success of sonographic stent detection was significantly associated with the ability
of the examiner to visualize the target structures (OR: 5.27, 95%-CI: 1.37–20.37,
p = 0.016). The association was also significant in the multivariate analysis (OR:
4.91, 95%-CI: 1.14–20.96, p = 0.031). [Table 4 ] shows the results of the risk regression analysis.
Table 4 Regression analysis of success of sonographic procedure to detect a pancreatic stent.
Characteristics
Success of ultrasound
Univariate analysis
Multivariate analysis
p-value
OR
95%-CI
p-value
OR
95%-CI
* the target structures: pancreatic head, pancreatic duct, common hepatic duct, and
confluence of the lienal and superior mesenteric veins
Age
0.43
0.98
0.94–1.03
0.66
BMI > 25
0.92
0.93
0.20–4.33
0.92
BMI > 30
0.13
4.07
0.67–24.72
0.70
Pancreatic disease
0.98
1.02
0.24–4,31
0.21
Liver disease
0.64
1.35
0.38–4.72
0.16
Previous abdominal surgery
0.46
0.59
0.14–2.44
0.71
Time from ERCP
0.30
0.63
0.27–1.50
0.97
Indication for ERCP
0.36
0.82
0.54–1.25
0.11
PEP
0.18
0.27
0.04–1.82
0.84
Placement of biliary stent
0.25
0.34
0.05–2.15
0.94
Pancreas lipomatosis
0.32
0.36
0.05–2.68
0.35
No visualization of the target structures*
0.016
5.27
1.37–20.37
0.031
4.91
1–14–20.96
Sonography device
0.76
0.95
0.70–1.30
0.45
Comparison of high-end ultrasound and bedside ultrasound
Supplementary Figure 1 describes the results of PS detection (A–C) and biliary stent detection (D–F) according
to the presented algorithm in the pivot tables for the three ultrasound groups. There
was no clinically relevant difference regarding sensitivity, specificity, positive
predictive value, and negative predictive value between the high-end ultrasound and
the bedside ultrasound groups for pancreatic and biliary stent detection. The correlation
between both ultrasound groups was statistically significant for PSs (φ = 0,3; p <
0.01) and for biliary stents (φ = 0,9; p < 0.001).
No visualization of the target structures was associated with less accurate stent
assessment in the high-end ultrasound group (OR: 8.2, 95%-CI: 1.7–41, p = 0.01) and
the bedside ultrasound group (OR: 268, 95%-CI: 8.7–8315, p = 0.001). Furthermore,
pancreas lipomatosis was beneficial for PS detection only in the high-end ultrasound
group (OR: 0.06, 95%-CI: 0.004–0.88, p = 0.04) and the time from ERCP in the bedside
ultrasound group (OR: 0.41, 95%-CI: 0.19–0.93, p = 0.03). All other baseline variables
had no statistically significant effect on the successful detection of a PS in the
subgroups (as shown in Supplementary Table 1a ). In multivariate logistic regression analysis, only the visualization of the target
structures remained an independent risk factor in the high-end ultrasound group (OR:
5.27, 95%-CI: 1.29–21.59, p = 0.021) but not in the bedside ultrasound group (as shown
in Supplementary Table 1b ).
Discussion
The current trial evaluated a new ultrasound-based algorithm for detecting prophylactic
PSs before their removal. The algorithm reduced the need for an X-ray examination
by 74%. This was even higher than the reported 71% of the previously published pilot
trial [16 ]. Accordingly, the algorithm reduces radiation exposure for the patient and staff.
Furthermore, pivotal resources in the endoscopy unit are saved, leading to increased
flexibility in the organization of examinations.
To the best of our knowledge, there are no further trials evaluating the accuracy
of sonographic detection of PSs besides the feasibility trial performed at our department
[16 ]. The sensitivity in both trials is comparable with the sensitivity of 81% achieved
in the present trial and 85% in the previous trial, but the positive predictive value
of 83% achieved in the present trial is lower than the 97% achieved in the previous
one [16 ]. However, our results seem to be consistent with the sensitivity of 67% to 89% in
trials using transabdominal ultrasound to detect small pancreatic lesions in the head
region [18 ]
[19 ].
The targeted structures were not visualized in 30% of the procedures. In another prospective
trial focusing on chronic pancreatitis, visualization of the entire pancreas was achievable
in only 61% of the patients, with the pancreatic tail being the most difficult part
[20 ]. Better technical features have improved imaging of the pancreas [21 ]. However, overlying gas represents an insurmountable technical limitation. In addition,
artifacts often resembling the stent may have led to false-positive results and unnecessary
EGDs. In this trial, EGD for stent removal was performed in 89% of the patients. However,
in 13% of all patients, no PS was retained, although this was reported by ultrasound.
Nevertheless, EGD is a safe procedure according to the literature with a complication
rate between 0.1% and 0.5% [22 ]. In the present trial, sedation-related hypotension occurred in one patient and
was treated with intravenous fluid without further harm to the patient and without
delay to the procedure. No further adverse events were reported during or after endoscopic
stent removal. Notably, there were no pancreatitis events associated with stent removal
in the present trial or in a previously published feasibility trial [16 ]. Moffatt et al. [17 ] reported a 3% pancreatitis rate associated with prophylactic PS removal. ESGE guidelines,
therefore, suggest that stent removal should be performed using side-viewing scopes
[9 ]. However, all stents in this trial and the past feasibility trial were removed by
standard gastroscopes with front-view using either forceps or a snare.
In this trial most PSs were removed within the same hospital stay, thus making it
convenient for the patients. We found that most stents dislodged within the first
four days, and none dislodged after one week.
In the literature, there is still a debate regarding the necessity to remove retained
PSs to avoid pancreatitis [11 ]
[13 ]. In the present trial, no patient returned with delayed pancreatitis, even if the
patient disagreed on stent removal and was excluded. Contrary to the recommendation
of the ESGE guidelines to remove the PS after five to ten days, leaving PSs in place
until the next ERCP could be a safe, cost-reducing, and resource-saving alternative
based on the present data and two previous studies [9 ]
[11 ]
[13 ]. In our study, imaging and EGD could have been reduced by 90%. A prospective safety
study is recommended.
In a subgroup analysis, a high-end ultrasound device group and bedside ultrasound
group were compared. Both groups had clinically comparable outcomes and statistically
significant correlation. Therefore, ultrasound could be performed in the ward, which
is a logistical advantage.
As a secondary endpoint, the detection of bile duct stents was also evaluated. In
96%, bile duct stents were still in place. The sensitivity and negative predictive
value were 100%, and the positive predictive value was 97%. The specificity was poor
at 33%, which might be due to the small number of dislodged stents. In a comparable
trial with 221 patients, the results showed a sensitivity of 77.3%, a positive predictive
value of 93.4%, a specificity of 94.6%, and a negative predictive value of 80.8% [23 ]. Imaging of bile duct stents might be more accurate than that of prophylactic PSs
because of the larger size and extended length of the stent. Even though there might
be difficulties with overlying gas, especially in the hilus area, biliary stents are
usually well detectable.
The main limitation of this trial might be the single-center design. A multicenter
design is required to evaluate the presented algorithm on a higher scale. Another
limitation may be that different sonographic devices and shorter PSs might yield different
results. Nevertheless, we used a variety of sonographic devices from different manufacturers
in the high-end ultrasound group to minimize this effect.
If centers implement our algorithm, we highly recommend a learning phase in which
sonographic results are controlled by X-ray until the accuracy rate is sufficient.
Even though transabdominal ultrasound is a simple and inexpensive alternative to X-ray
without radiation exposure, there are patients in whom reliable imaging of the pancreatic
duct in the head region cannot be achieved even by highly experienced examiners. In
these cases, further imaging should be performed.
A strength of the study is the prospective design in a real-world environment. The
patient population is representative of almost all gastroenterology units.
In conclusion, a new algorithm primarily based on ultrasound was presented. X-ray
was avoided in 74% of examinations. Both high-end and bedside ultrasound procedures
yielded comparable results which implies that sonography can be performed in the ward.
However, to avoid false results, only experienced examiners should perform the examination
after a learning period. As shown by our data, ultrasound experience in general does
not serve as a good predictor of the ability to detect PSs by ultrasound. Instead,
an individualized approach seems to be necessary. When in doubt, fluoroscopy should
be indicated.
