Introduction
Training for advanced endoscopic procedures such as endoscopic retrograde cholangiopancreatography
(ERCP) remains an important challenge for most training programs across the world.
In order to address this issue, standards for training duration, procedure volume,
and competence assessment have been proposed by several societies [1], and key performance measures for independent ERCP practice have been defined [2]. However, it has been widely recognized that many endoscopists in training fail
to reach the required procedure volume or the predefined competence threshold during
their training period [3]
[4]. To compound this issue, there are relatively limited and conflicting data pertaining
to the safety of ERCP procedures with trainee involvement [5]
[6]. We hypothesize that in a teaching setting, trainee participation in ERCP procedures
is safe and does not compromise the technical success of the procedure. This is very
important because, as the incidence and severity of adverse events during ERCP, especially
in low-volume centers, is not negligible, assuring patient safety remains of paramount
importance.
In our study, we aimed to prove that ERCP procedures with trainee involvement do not
differ significantly from procedures completed without trainee involvement, both in
terms of technical success and procedure-related adverse events.
Methods
We conducted an investigator-driven, multicenter, international study of ERCP procedures
conducted in both high-volume (> 1000 ERCPs/year) and low-volume (< 1000 ERCPs/year)
centers in Southeastern Europe (Romania, Croatia, Italy, and Serbia). Invitations
to participate in the study were issued to several institutions across Southeastern
Europe that had collaborated previously on research projects in ERCP training, with
the aim of canvassing a wide variety of endoscopic practices, including differences
in training methods for novice endoscopists.
Endoscopists performing ERCPs at the participating centers were invited to prospectively
document patient and procedure-related data using a standard report form (see Appendix 1s in the online-only supplementary material), which was designed to capture the technical
aspects of the procedure as well as patient-related outcomes, including procedure-related
adverse events and their outcomes. All consecutive ERCP procedures were documented,
irrespective of trainee involvement; procedures without any trainee involvement served
as a control group for the study.
The main study outcomes were technical success of the procedure and patient safety
as assessed by the incidence of overall procedure-related adverse events.
Definitions
For the purpose of the study, trainees were defined as endoscopists working under
direct supervision and/or endoscopists with fewer than 200 ERCP procedures performed
independently (whichever applied). Technical success of the procedure was defined
as the ability to achieve the planned diagnostic and/or therapeutic procedure in each
individual case (e. g. correct positioning of stents, complete clearance of bile duct
stones) and was assessed by the attending endoscopist at the end of each procedure.
Technical difficulty of the procedure was graded according to the Schutz scale ([Table 1]) [7]. Procedure-related adverse events were defined as any procedure-related complications
that prolonged hospital stay and/or required additional medical or surgical interventions
(e. g. surgery, additional endoscopic interventions, admission to the intensive care
unit), and their severity was assessed using the Cotton scale, as previously described
[8]
[9]. Notably, although not usually considered as an adverse event per se, we included
technical failure of the procedure resulting in prolonged hospital stay for the purpose
of an endoscopic reintervention or a different therapeutic procedure (i. e. surgical
or radiological drainage) as a procedure-related adverse event, to allow a more accurate
assessment of the actual impact of trainee involvement.
Table 1
Modified Schutz scale for grading procedure difficulty [7].
|
Grade
|
Diagnostic
|
Therapeutic
|
|
1
|
CBD cannulation, brushing
|
Sphincterotomy, stones < 10 mm, stents for leaks and extrahepatic malignant strictures
|
|
2
|
Billroth II diagnostic, minor papilla cannulation
|
Stones > 10 mm, hilar stenting, benign biliary strictures
|
|
3
|
Manometry, Whipple, Roux-en-Y, intraductal endoscopy
|
Billroth II therapeutics, intrahepatic stones, pancreatic therapy
|
CBD, common bile duct.
Endoscopic procedures
ERCP was performed according to the standard procedure in each participating center,
with respect to both the endoscopic procedure itself and the periprocedural medical
care and surveillance. All centers used a wire-first approach with a triple-lumen
sphincterotome for initial attempts at selective cannulation of the desired duct,
with the notable exception of one center, which used contrast-guided cannulation with
a double-lumen sphincterotome as the standard initial approach for selective cannulation.
Training protocol
Trainees worked under the direct supervision of an experienced endoscopist who supplied
verbal as well as hands-on assistance and who could take over the procedure when deemed
necessary. Accordingly, the degree of involvement of any trainee in any given procedure
was entirely at the discretion of the supervisor, who acted on a case-by-case basis,
according to best judgment and in the interest of ensuring patient safety throughout
the procedure. Trainee involvement and technical success were divided and reported
in the following categories, as detailed in the study form (Appendix 1s): 1) failed cannulation attempts; 2) cannulation of the desired duct; 3) partially
successful procedure (including cannulation), but required hands-on assistance from
the supervisor; 4) completed the procedure without hands-on assistance from the supervisor.
Data collection
Data were collected prospectively, using standard report forms completed by each supervising
endoscopist at the end of the procedure. Follow-up of each patient was done by the
attending endoscopist, with the aim of capturing procedure-related adverse events
and patient outcome up to 30 days after the procedure. Although patients were subject
to the individual post-procedure diagnostic and treatment protocols that were in place
at each participating institution at the time of the study, as a general rule all
patients undergoing ERCP were admitted and monitored for a minimum of 24 hours (mainly
for reasons related to healthcare system reimbursement policies). Investigators were
instructed to follow up on all patients for 30 days, using either patient chart review,
telephone interviews, or both, in accordance with local protocols, regulations, and
available infrastructure.
Statistical analysis
Data from all participating centers were collected in a central database and analyzed
using SPSS v.16 (SPSS Inc. Chicago, Illinois, USA).
Data were analyzed in a two-step fashion. Bivariable analysis using the appropriate
tests (chi-squared test, Student t test where continuous variables had a normal distribution, and Mann-Whitney U test where continuous variables had a non-normal distribution) was first carried
out to identify potential risk factors for procedure-related adverse events and technical
failure of the procedure. Multivariable analysis with logistic regression (using the
enter method) was then carried out for the main study outcomes, including all variables
that had a P level of < 0.2 in the bivariable analysis. Based on clinical judgment and previously
reported data, we also analyzed the potential interaction between procedure-related
parameters in the model. Briefly, it was expected that trainee involvement would result
in longer cannulation times (because the operator is inexperienced and requires more
time to achieve deep duct cannulation), which, in turn, would mean that more procedures
would be classified as difficult cannulation in the training group compared with the
control group, as cannulation time > 5 minutes is one of the defining criteria for
difficult cannulation. Furthermore, prolonged cannulation time has been linked to
more use of precut sphincterotomy. With regard to papilla anatomy, native papilla
was expected to correlate with longer cannulation times, difficult cannulation, and
use of precut sphincterotomy, as pre-existing sphincterotomy allows easy access to
the desired ducts. As a result, we included all of these potential interactions in
the multivariable models.
Sample size
The sample size was calculated for a 5 % margin of equivalence in adverse events in
the trainee group compared with the control group, from an estimated 10 % overall,
under the general assumptions of an equivalence trial (780 × 2 arms = 1560 procedures,
Sealed Envelope Ltd. 2012. Power calculator for binary outcome equivalence trial.
Available from: https://www.sealedenvelope.com/power/binary-equivalence/ Accessed 8 September 2019). This also allowed for evaluation of a 5 % margin of equivalence
in cannulation success (using a standard cannulation success estimated at 95 % in
the control group) at a 0.05 significance level, with a beta of 0.9.
Ethical considerations
The study protocol was approved by the local ethics committee at each participating
center, in accordance with the local and national regulations as well as the Declaration
of Helsinki.
Results
A total of 16 independent operators and 21 trainees working at the 6 participating
centers reported on 1843 ERCPs performed between October 2016 and October 2018. Trainees
were involved in 822 procedures (44.6 %), including 565 native papilla cases. Only
4 of the 21 trainees (19.0 %) had been involved in > 100 hands-on procedures prior
to study inception, with the remaining trainees having limited or no experience in
ERCP. The distribution of cases per center, as well the percentage of procedures with
trainee involvement in each center, are illustrated in [Table 2].
Table 2
Total number of procedures and percentage of trainee involvement per center.
|
Center (country)
|
Total no. of procedures, n
|
Procedures with trainee involvement, n (%)
|
Total number of endoscopists (experts/trainees), n
|
|
Colentina Hospital (Romania)
|
811
|
384 (47.3)
|
10 (4/6)
|
|
Policlinico Gemelli (Italy)
|
356
|
191 (53.7)
|
14 (5/9)
|
|
UHBC Zagreb (Croatia)
|
252
|
21 (8.3)
|
5 (4/1)
|
|
Cantacuzino Hospital (Romania)
|
201
|
201 (100)
|
2 (1/1)
|
|
Zadar Hospital (Croatia)
|
76
|
22 (28.9)
|
2 (1/1)
|
|
Belgrade University Hospital (Serbia)
|
147
|
4 (2.7)
|
2 (1/1)
|
The most common indication for ERCP was the presence of common bile duct (CBD) stones
(46.8 %), followed by malignant strictures of the biliary tract (30.5 %). Most patients
were male (51.3 %) and mean age in the patient population was 66.8 years (standard
deviation 14.6). Native papilla cases were evenly distributed between the trainee
and control groups (565 vs. 754; P = 0.07 chi-squared); however, there were significantly more grade 2 and 3 procedures
in the control group than in the trainee group (25 % vs. 17.7 %; P = 0.002). General data about the patient population are further detailed in [Table 3].
Table 3
General data about the patient population and the indication for endoscopic retrograde
cholangiopancreatography (n = 1843).
|
Sex, male/female, n (%)
|
946 (51.3)/897 (48.7)
|
|
Age, mean (SD), years
|
66.8 (14.6)
|
|
Indication for the procedure, n (%)
|
|
|
863 (46.8)
|
|
|
563 (30.5)
|
|
|
99 (5.4)
|
|
|
57 (3.1)
|
|
|
168 (9.1)
|
|
|
40 (2.2)
|
|
|
41 (2.2)
|
|
Procedure difficulty (modified Schutz scale), n (%)
|
|
|
1441 (78.2)
|
|
|
337 (18.3)
|
|
|
65 (3.5)
|
|
Type of sedation used, n (%)
|
|
|
16 (0.9)
|
|
|
136 (7.4)
|
|
|
1691 (91.8)
|
SD, standard deviation; CBD, common bile duct.
Trainee involvement and patient safety
In 270 cases (14.7 %) there was at least one procedure-related adverse event reported
by the attending endoscopist. The most frequent adverse event reported was the technical
failure of the procedure (89 cases, 4.8 %), followed by post-ERCP pancreatitis (PEP;
52 cases, 2.8 %) and cholangitis (45 cases, 2.4 %) ([Fig. 1]). Most adverse events were either mild or moderate, but there were 19 severe adverse
events (1.0 %) and 5 deaths occurring in the 30-day follow-up period (0.3 %); all
of the fatalities occurred in patients with severe co-morbidities (American Society
of Anesthesiologists score of 3 or more).
Fig. 1 Comparison of procedure-related adverse events across study groups.
The overall incidence of adverse events was similar between the trainee group and
the control group (14.7 vs. 14.6 %; P > 0.99 chi-squared) ([Table 4]), with a difference of 0.1 percentage points between study groups (95 % confidence
interval [CI] – 3.12 % to 3.38 %, within the 5 % predefined equivalence margin). However,
there were significantly more moderate or severe adverse events in the control group
compared with the trainee group (63 [6.2 %] vs. 28 [3.4 %]; P = 0.01 chi-squared).
Table 4
Comparison of patient and procedure-related characteristics between the two study
groups.
|
Parameter
|
Trainee group (n = 822)
|
Control group (n = 1021)
|
P value
|
|
Sex, male/female, n
|
398/424
|
548/473
|
0.04
|
|
Age, mean (SD), years
|
66.8 (14.1)
|
66.7 (14.9)
|
0.30
|
|
Technical success, %
|
760 (92.4)
|
957 (93.7)
|
0.30
|
|
Procedure difficulty, n (%)
|
|
|
676 (82.2)
|
765 (74.9)
|
0.002
|
|
|
122 (14.8)
|
215 (21.1)
|
|
|
24 (2.9)
|
41 (4.0)
|
|
Adverse events, n (%)
|
|
|
121 (14.7)
|
149 (14.6)
|
> 0.99
|
|
|
93 (11.3)
|
86 (8.4)
|
0.01
|
|
|
18 (2.2)
|
49 (4.8)
|
|
|
10 (1.2)
|
14 (1.4)
|
|
Failure of the initial procedure and/or early reintervention required, n (%)
|
61 (7.4)
|
72 (7.1)
|
0.09
|
|
PEP, n (%)
|
25 (3)
|
27 (2.6)
|
0.58
|
|
Cholangitis, n (%)
|
16 (1.9)
|
29 (2.8)
|
0.29
|
|
Bleeding, n (%)
|
6 (0.7)
|
14 (1.4)
|
0.26
|
|
Perforation, n (%)
|
3 (0.4)
|
6 (0.6)
|
0.74
|
|
Other, n (%)
|
11 (1.3)
|
11 (1.1)
|
0.26
|
SD, standard deviation; PEP, post-endoscopic retrograde cholangiopancreatography.
Using bivariable analysis, bilirubin levels, presence of a native papilla, use of
precut, difficult cannulation, procedure difficulty, indication for ERCP, time to
cannulation, and participating center were selected as potential risk factors for
the occurrence of a procedure-related adverse event ([Table 5]) and included in the multivariable analysis model using logistic regression. We
also adjusted for interaction between the relevant variables such as time to cannulation,
difficult cannulation, native papilla, trainee involvement, and use of precut to exclude
potential confounders in the model.
Table 5
Bivariable analysis of potential risk factors for adverse events and technical failure.
|
Potential risk factors
|
Any adverse event
|
No adverse event
|
P value
|
Technical success
|
Technical failure
|
P value
|
|
Sex, % female
|
49.4
|
48.3
|
0.74
|
48.6
|
45.6
|
0.52
|
|
Age, mean (SD), years
|
67.4 (13.9)
|
66.7 (14.7)
|
0.40
|
66.7 (14.6)
|
67.5 (13.6)
|
0.57
|
|
Native papilla, n (%)
|
216 (80.2)
|
1089 (69.4)
|
< 0.001[1]
|
1195 (70.2)
|
104 (83.2)
|
0.001[1]
|
|
Use of precut, n (%)
|
93 (34.5)
|
183 (11.6)
|
< 0.001[1]
|
223 (13.1)
|
53 (42.4)
|
< 0.001[1]
|
|
Time to cannulation (< 5/5 – 10/> 10 min), n
|
93/38/136
|
1094/257/205
|
< 0.001[1]
|
1164/285/240
|
16/7/101
|
< 0.001[1]
|
|
Indication for ERCP, n (%)
|
|
|
91 (10.6)
|
770 (89.4)
|
< 0.001[1]
|
823 (96.3)
|
31 (3.7)
|
< 0.001[1]
|
|
|
14 (14.1)
|
85 (84.9)
|
88 (88.9)
|
11 (11.1)
|
|
|
124 (22.1)
|
438 (77.9)
|
491 (87.8)
|
69 (12.2)
|
|
|
1 (9.1)
|
10 (90.9)
|
11 (100)
|
0 (0)
|
|
|
12 (21.1)
|
45 (78.9)
|
50 (87.8)
|
7 (12.2)
|
|
|
13(7.7)
|
155 (92.3)
|
164 (98.2)
|
3 (1.8)
|
|
|
6 (15)
|
34 (85)
|
36 (92.3)
|
3 (7.7)
|
|
|
8 (20)
|
32 (80)
|
39 (97.5)
|
1 (2.5)
|
|
Procedure difficulty level (1/2/3), n
|
191/59/15
|
1216/277/50
|
0.03[1]
|
1316/307/52
|
83/29/11
|
0.001[1]
|
|
Trainee involvement, n (%)
|
120 (44.6)
|
701 (44.6)
|
> 0.99
|
751 (44.1)
|
61 (48.8)
|
0.30
|
|
Participating center, n (%)
|
|
|
150 (18.5)
|
660 (81.5)
|
< 0.001[1]
|
737 (90.8)
|
74 (9.2)
|
< 0.001[1]
|
|
|
36 (18)
|
164 (82)
|
186 (93)
|
14 (7)
|
|
|
38 (15.7)
|
214 (84.3)
|
233 (92.4)
|
19 (7.6)
|
|
|
6 (7.9)
|
70 (92.1)
|
73 (96.1)
|
3 (3.9)
|
|
|
25 (7)
|
331 (93)
|
339 (98.5)
|
4 (1.5)
|
|
|
14 (9.6)
|
132 (90.4)
|
137 (93.2)
|
10 (6.8)
|
|
Difficult cannulation, n (%)
|
152 (43.1)
|
364 (23.2)
|
< 0.001[1]
|
417 (24.5)
|
96 (77.4)
|
< 0.001[1]
|
|
Bilirubin, median (range), mg/dL
|
5.7 (0.1 – 27.5)
|
2.2 (0.1 – 93.6)
|
< 0.001[2]
|
2.38 (0.1 – 93.6)
|
7.95 (0.1 – 36)
|
< 0.001[2]
|
SD, standard deviation; ERCP, endoscopic retrograde cholangiopancreatography; CBD,
common bile duct.
1 Using chi-squared test.
2 Using Mann – Whitney U test.
Finally, increased bilirubin levels, time to cannulation, and procedure difficulty
level were shown to independently increase the risk of any procedure-related adverse
event ([Table 6]).
Table 6
Risk factors for procedure-related adverse events identified using multivariable analysis.
|
Risk factor
|
OR
|
95 %CI
|
|
Increased bilirubin levels
|
1.01[1]
|
1.00 to 1.03
|
|
Time to cannulation (1)[2]
|
0.76
|
0.32 to 1.77
|
|
Time to cannulation (2)[2]
|
5.17
|
2.14 to 12.40
|
|
Difficulty level (1)[3]
|
1.53
|
1.05 to 2.23
|
|
Difficulty level (2)[3]
|
1.98
|
1.01 to 3.91
|
|
Trainee involvement
|
1.18
|
0.72 to 1.93
|
OD, odds ration; CI, confidence interval.
Covariates for the model included: endoscopy center, indication for endoscopic retrograde
cholangiopancreatography, difficulty level, time to cannulation, difficult cannulation,
use of precut, bilirubin levels, papilla anatomy, trainee involvement, as well as
the following interactions: time to cannulation and trainee involvement; time to cannulation
and use of precut; difficult cannulation and time to cannulation; difficult cannulation
and papilla anatomy;, papilla anatomy and use of precut.
1 Risk increase for every 1 mg/dL increase in bilirubin levels.
2 OR provided for increase in time to cannulation from < 5 to 5 – 10 min (1) and from
5 – 10 min to > 10 min (2), respectively.
3 OR provided for increase in difficulty level from 1 to 2 (1) and from 2 to 3 (2),
respectively.
Trainee involvement and technical success of the procedure
Trainees were involved in 822 procedures and managed to successfully complete their
respective procedures in 480 cases (58.4 %). In 150 cases (18.2 %), trainees required
hands-on assistance from their supervisor to complete their procedure, and in 28 cases
(3.4 %) they could only selectively cannulate the desired duct, with a supervisor
completing the rest of the procedure. In a further 164 cases (20.0 %), trainees were
unable to access the targeted duct, requiring an expert to take over to gain ductal
access and, subsequently, perform the diagnostic or therapeutic procedure required
in each case.
There was no difference in the incidence of failed cannulations between the trainee
group (6.0 %) and the control group (4.6 %; P = 0.21), including also subgroup analysis in native papilla cases only (7.7 % vs.
5.6 %; P = 0.14).
The incidence of technical failure of the procedure, as appreciated by the attending
endoscopist, was also similar between the trainee group and the control group (7.6 %
vs. 6.3 %; P = 0.31), with a calculated difference between the two groups of 1.3 percentage points
(95 %CI – 1 % to 3.7 %, falling within the 5 % predefined equivalence margin).
On bivariable analysis ([Table 5]), bilirubin levels, difficult cannulation, difficulty level of the procedure, use
of precut, presence of a native papilla, time to cannulation, indication for the procedure,
and participating center were all associated with an increased risk of procedure failure.
However, using multivariable analysis with logistic regression, only precut use, difficulty
level, and time to cannulation were identified as risk factors for the technical failure
of the procedure, after adjusting for ERCP indication and endoscopy center ([Table 7]).
Table 7
Multivariable analysis of risk factors for procedure failure.
|
Risk factor
|
OR
|
95 %CI
|
|
Use of precut
|
12.2
|
1.56 to 95.12
|
|
Difficulty level (1)[1]
|
2.01
|
1.14 to 3.51
|
|
Difficulty level (2)[1]
|
3.29
|
1.32 to 8.19
|
|
Time to cannulation (1)[2]
|
2.35
|
0.36 to 15.14
|
|
Time to cannulation (2)[2]
|
48.40
|
11.19 to 209.36
|
|
Endoscopy center[3]
|
0.25
|
0.10 to 0.71
|
|
Indication for ERCP (1)[4]
|
4.0
|
1.56 to 10.26
|
|
Indication for ERCP (2)[4]
|
6.79
|
2.20 to 20.89
|
|
Trainee involvement
|
1.23
|
0.36 to 4.17
|
OR, odds ratio; CI, confidence interval; ERCP, endoscopic retrograde cholangiopancreatography.
Covariates for the model included: endoscopy center, indication for ERCP, difficulty
level, time to cannulation, difficult cannulation, use of precut, bilirubin levels,
papilla anatomy, trainee involvement, and the following interactions: time to cannulation
and trainee involvement; time to cannulation and use of precut; difficult cannulation
and time to cannulation; difficult cannulation and papilla anatomy; papilla anatomy
and use of precut.
1 OR provided for increase in difficulty level from 1 to 2 (1) and from 2 to 3 (2),
respectively.
2 OR provided for increase in time to cannulation from < 5 to 5 – 10 min (1) and from
5 – 10 min to > 10 min (2), respectively.
3 OR reported for center number 4, which showed significantly higher technical success
rates, using center number 1 as the reference category.
4 OR reported using common bile duct stones as the reference category for this variable;
(1) – benign bile duct stricture; (2) – bile leak or trauma.
Discussion
The main finding of our study is that, although there is still a significant variation
in training methods and, indeed, the practice of ERCP between endoscopy units, the
differences in terms of procedure-related adverse events or technical failure between
procedures with trainee involvement and those without trainee involvement do not seem
clinically significant. In our trial, which was designed as an equivalence study,
differences in the main outcome variables between the two groups did not exceed the
predefined 5 % margin.
Furthermore, although we noted a significant difference in the performance of individual
endoscopy centers, the overall safety data across all participating centers was within
the recommended targets [2] (PEP 2.8 %, cholangitis 2.4 %, postsphincterotomy bleeding 1.1 %, and perforations
0.5 %), with a rate of severe adverse events of 1.0 % and a 30-day mortality of 0.3 %.
Also in accordance with proposed European Society of Gastrointestinal Endoscopy guidelines,
CBD cannulation rates were above the proposed 90 % threshold, across both study groups,
with overall technical success of the procedure also above 90 %. Notably, this included
a large number of procedures that were completed by trainees without any hands-on
assistance from their supervisors (480 procedures out of a total of 822 procedures
with trainee involvement). Our study was not powered to evaluate a potential increase
in specific adverse events such as PEP, bleeding or cholangitis caused by trainee
involvement. However, the consistently low incidence of these complications, well
within the established guideline targets in both study groups, leads us to believe
that any such existing relationship would have limited clinical significance in a
real-life setting.
Training in endoscopy, and particularly in high-risk procedures such as ERCP, has
been the focus of numerous studies, most of which were aimed at defining what represents
competence in ERCP and how to best assess endoscopy trainees during and after their
training period. The findings from these studies have shifted the paradigm in endoscopy
training from a milestone-based approach (i. e. fulfilling a minimum number of procedures
before independent practice is allowed) to a competence-based approach, requiring
trainees to meet certain performance measures before being deemed competent. For example,
the new Joint Advisory Group on Gastrointestinal Endoscopy accreditation system requires
a trainee to show complication rate (death, transfusion requiring hemorrhage or perforation)
of < 5 %, satisfactory completion of intended therapeutic procedure of > 80 %, and
more than 75 procedures performed in the previous 12 months) [10].
However, the question of how these thresholds can be met in real life remains open
to debate. Moreover, there are limited data about the outcome of ERCP procedures with
trainee involvement, both in terms of technical success and, even more importantly,
concerning patient safety. Our study is the first European multicenter trial to address
these key issues prospectively, gathering data from both referral centers and smaller
endoscopy units from four countries in Southern and Eastern Europe.
With regard to patient safety, we were particularly stringent in defining the main
outcome measure of procedure-related adverse events to include the partial or complete
technical failure of the procedure, leading to prolonged hospital stay and need for
reintervention. On multivariable analysis, we found increased difficulty level of
the procedure and increased bilirubin levels to be risk factors for adverse events.
Increased bilirubin levels are probably a surrogate marker of increased difficulty
reflecting high grade or complex malignant strictures, such as hilar cholangiocarcinoma.
Interestingly, an increase in cannulation time above 10 minutes was also found to
increase the risk of adverse events, although trainee involvement, a recognized factor
for delayed access to the bile duct [11], did not significantly increase the risk of adverse events in the multivariable
analysis model.
Technical success of the procedure was achieved in over 90 % of the procedures in
our study; however, there was a statistically significant difference between the participating
centers, with one center in particular reporting extremely high success rates compared
with the others. This finding is concordant with reports from other real-life reports,
showing the large variation in performance between different centers [12]
[13]. In our study, this finding could be explained by the fact that the outlier center
carried out the highest number of procedures per year and had the largest number of
expert endoscopists among the centers involved, although this was not explicitly analyzed
in the study.
The main limitation of our study is its observational nature, which carries an inherent
risk of selection bias with regard to the cases where trainees were involved compared
with the control group. Although there were no significant differences concerning
patient demographics between the two study groups, significantly more grade 2 and
3 cases were found in the control group compared with the trainee group. In addition,
the degree of trainee and supervisor involvement in each case was not standardized,
which could also potentially constitute a source of bias. However, we believe that
the large number of included cases (> 1800 ERCPs), reflecting a broad spectrum of
endoscopic practice with 6 centers and over 30 endoscopists involved in the study,
minimizes the risk of bias and ensures an accurate reflection of the real-life situation
of trainee involvement in ERCP procedures.
Another potential source of bias in our study is represented by the inherent differences
in patient management, including the assessment and reporting of early and delayed
adverse events, across centers in different countries. As none of the participating
centers has access to a national patient database, it is conceivable that some adverse
events, particularly delayed complications, might have been under-reported in our
study. However, taking into account the large catchment area for each center involved
in the trial, as well as the fact that all patients were admitted for at least 24
hours after the procedure, we believe the risk of missing significant clinical outcomes
to be very low in this particular cohort.
To date, there have been limited data reporting on the involvement of trainees in
ERCP procedures, with somewhat contradictory findings and based on relatively small,
single-center studies. While Frost et al. [14] showed that trainee involvement does not negatively influence either successful
cannulation or time to cannulation, other data suggest that trainee involvement might
actually increase the use of precut in order to facilitate access to the CBD [11]. With respect to patient safety, although available data [11]
[14]
[15] suggest that patient safety is not compromised by trainee involvement, these studies
are usually underpowered for the analysis of complication rates.
Our findings underscore the fact that cannulation time, use of precut, and case complexity
as assessed by the Schutz scale remain the most important risk factors for procedure
failure and procedure-related adverse events, irrespective of trainee involvement
in the procedure. We believe that future studies should focus on how a tailored approach
to training, such as selecting low-risk cases for trainees in their initial phase
of training, might further mitigate the risk of procedure-related adverse events while
maximizing the benefits of hands-on training. Based on the evidence from this study,
we believe trainers could be advised that some high-risk patients (i. e. those with
high levels of bilirubin or those with grade 2 or 3 difficulty according to the Schutz
scale) might not be adequate for inclusion on the procedure lists of less experienced
endoscopists. Standardizing the approach to training has already been advocated by
Kwek et al. [16], who propose some clear-cut criteria for situations in which the expert should take
over from the trainee and continue the procedure, such as time on the papilla or exceeding
a certain number of inadvertent pancreatic duct cannulations, with the aim of limiting
the risk of complications. This approach could reasonably be expanded to incorporate
high-risk features of the patient, as suggested by our study.
In conclusion, we have found that trainee involvement in ERCP procedures is safe,
both in terms of procedure outcome and patient wellbeing. Future studies should focus
on exploring whether procedure and training outcomes can be improved further by applying
selection algorithms based on individual risk factors, to ensure that high-risk procedures
are correctly identified and treated accordingly.
