Introduction
Endoscopic ultrasound (EUS) enables minimally invasive diagnostic evaluation and therapeutic
interventions for patients with gastrointestinal (GI) and pancreaticobiliary disorders.
However, EUS exposes patients to adverse events such as bleeding, infection, pancreatitis,
perforation, and death [1]
[2]
[3]
[4]
[5]
[6]. Accurate data regarding the risks of adverse events and the modifiable factors
associated with these events can enable patients and their caregivers to make informed
decisions regarding their care and minimize procedure-related harm following EUS [7].
Although numerous original studies and several meta-analyses have reported on the
risks of EUS-associated adverse events, estimates to date have important limitations
[1]
[8]. Most original studies only report data on intraprocedural and early post-procedural
adverse events without follow-up to 30 days to capture delayed events. In a 2011 review
on this topic, Wang et al. [1]. reported that only two of the 41 studies in their meta-analysis followed patients
to assess for delayed adverse events [9]
[10]. In addition, most studies have been designed with diagnostic yield as a primary
outcome and thus may not provide precise estimates of adverse event risks.
Data on lower EUS for rectal and pelvic lesions are scant. Among the few studies that
have reported on lower EUS adverse events, risks have been as high as 15 %–21 % [11]
[12]. Many participants, however, underwent other procedures concurrently with their
lower EUS with fine-needle aspiration (FNA), including colonoscopy and polypectomy
[11]. Moreover, few studies in the upper and lower EUS literature report on adverse events
related to EUS alone, with the majority reporting on EUS with FNA.
The objectives of this study were to evaluate the risk of adverse events following
EUS and to assess risk factors for EUS-associated events. In addition, we differentiated
EUS-associated adverse events following upper vs. lower EUS and following EUS with
and without FNA.
Patients and methods
Study design and setting
This population-based retrospective cohort study examined patients who underwent EUS
while residing in the Calgary Zone (CZ) from 2007 to 2013. The CZ provides publicly
funded health care services to approximately 1.4 million people in the Canadian province
of Alberta. All EUS procedures were performed at the Foothills Medical Centre, the
largest tertiary care facility in Alberta, by one of six consultant endoscopists who
had formal training in EUS or by a therapeutic endoscopy trainee under direct supervision
from the consultant. The study received institutional review board approval by the
University of Calgary’s Conjoint Health Research Ethics Board (REB14–2312).
Study cohort
All adult CZ patients who underwent one or more outpatient EUS procedure were eligible
for inclusion. We restricted the cohort to CZ patients to eliminate the possibility
of not being able to conduct chart reviews among non-CZ patients who traveled to Calgary
for their procedure, but later presented to an outside hospital after returning home.
Patients were initially identified through the electronic reporting program endoPRO
(Pentax Medical, Montvale, New Jersey, USA) used to document all endoscopic procedures
in Calgary. Patients with no procedure reports, and those under the age of 18 years
were excluded. Additionally, we excluded pancreatic cyst ablation with ethanol and
cyst gastrostomy procedures due to low sample sizes. We further excluded patients
who underwent EUS-guided celiac plexus block or neurolysis (CPB/CPN), as the indication
for these procedures was abdominal pain and the majority of presentations to hospital
after the procedure were for ongoing abdominal pain; we could not have determined
whether the EUS procedure contributed to their pain or was simply ineffective. Patients
who underwent EUS were administered conscious sedation by nursing staff under the
direction of the attending physician.
Data sources and variables
Study investigators (R.R., N.F.) abstracted demographic (age, sex), clinical (procedure
indication), and procedural data (sedation used, presence of a trainee, EUS approach
[upper vs. lower], type of echoendoscope used [radial vs. linear], needle sampling
method and number of passes when applicable, and concurrence of other endoscopic procedures
[endoscopic retrograde cholangiopancreatography (ERCP)] or colonoscopy) through review
of each of the procedure reports. Esophagogastroduodenoscopy and flexible sigmoidoscopy
were not considered additional procedures as these are routinely performed along with
upper and lower EUS, respectively.
Using each patient’s regional health record number, the procedural dataset was then
linked to provincial emergency department and hospital discharge databases to identify
all visits to the emergency department and/or unplanned inpatient admissions occurring
anywhere in the CZ within 30 days of the EUS procedures. The electronic health record
for each of these encounters was individually reviewed by a trained research assistant
(C.M.). Data abstracted included the presenting complaint, treatment disposition (discharge
home, admission), final diagnosis, length of hospital stay, need for interventions
(endoscopic procedure, surgery, antibiotics, admission to intensive care, blood transfusion),
and vital status.
Classification of adverse events
We defined a potential adverse event a priori as any emergency department visit or
unplanned hospitalization that involved a main diagnosis that has been previously
reported as EUS related, including abdominal pain, fever, pancreatitis, bleeding,
infection, perforation, cardiopulmonary event, or death. Planned hospitalizations
were excluded from the analysis.
Using a previously reported approach to evaluating adverse events related to endoscopic
procedures [13]
[14]
[15], three investigators (R.R., P.D.J., and S.J.H.) independently determined the causal
relationship of each clinical encounter to the index EUS procedure by applying predefined
criteria: 1) no other exposure was more likely than the EUS procedure to be related
to the event, 2) the event was a new clinical presentation for the patient (i. e.
not due to a pre-existing condition prior to the EUS exposure), and 3) the event was
a previously reported adverse event of EUS. Events that met all three criteria were
classified as “definitely related”; those meeting none of these criteria were classified
as “not related.” Finally, events that met at least one but not all criteria were
classified as “possibly related.” Disagreements were resolved by consensus.
In cases where patients had more than one definitely or possibly related adverse event
following an EUS procedure, investigators identified and prioritized the most clinically
significant and closely timed encounter relative to the EUS. One adverse event was
permitted per procedure.
Data analysis
The primary outcome was the 30-day risk of adverse events categorized as definitely
or possibly related to the EUS procedure. We also examined associations among patient,
procedure, and endoscopist characteristics and adverse events through univariable
and multivariable analyses.
Categorical variables were reported as proportions and comparisons were made using
Fisher’s exact tests. We conducted univariable analyses on a priori selected risk
factors previously shown to be associated with EUS adverse events, including patient
age (≥ 65 vs. < 65 years) [16]
[17]
[18], sex [19], presence of a trainee [13], and the use of FNA [20]. To account for the possibility of differential risks of adverse events between
patients who only had one EUS procedure and those who underwent multiple procedures
during the study period, we included the performance of multiple EUS procedures, compared
with a single procedure, as a binary variable. Finally, given that the completion
of additional endoscopic procedures concurrently with EUS could be associated with
higher adverse event risk [11], we also included “additional procedures” (same-day ERCP or colonoscopy) as a potential
risk factor. These risk factors were used for univariable and multivariable logistic
regression, with estimates provided using odds ratios (ORs) with 95 % confidence intervals
(CIs). Given that adverse events after EUS are considered a rare event, we used Firth’s
logistic regression to reduce the bias in maximum likelihood estimates. All statistical
analyses were completed using STATA (StataCorp. 2017. Stata Statistical Software:
Release 15. College Station, Texas, USA).
Results
Among the 2913 patients assessed for eligibility, those with missing procedure reports
(n = 1) and those aged under 18 years (n = 3) were excluded. We also excluded one
patient who underwent one pancreatic cyst ablation with ethanol, three who underwent
one cyst gastrostomy procedure each, and 10 who underwent 37 EUS-guided celiac plexus
interventional procedures ([Fig. 1]). Five patients who underwent EUS-CPN or CPB presented to the emergency department
within 30 days of their procedure. Two of these patients had abdominal pain considered
to be related to the EUS-CPN procedure; one of these patients had undergone concurrent
EUS-FNA and one had not. None of the patients had perforation, infection, or bleeding.
Owing to the uncertainty of the three post-procedure presentations with abdominal
pain and their relation to the procedure vs. their underlying pancreatic cancer, this
group was excluded from the analyses. Thus, the final study cohort comprised 2895
patients who underwent a total of 3552 procedures, including those with (n = 733)
and without (n = 2819) FNA.
Fig. 1 Study flow diagram. EUS, endoscopic ultrasound; FNA, fine-needle aspiration.
The median age of the 2895 patients in the final cohort was 59.7 years and 53 % were
male. There were 3034 upper EUS procedures, of which 710 included FNA (23 %), and
518 lower EUS procedures, of which 23 (4 %) involved FNA. The most common indications
for EUS were for assessment of hepatopancreaticobiliary lesions (pancreatic cyst,
pancreatic mass, bile or pancreatic duct stone, bile duct stricture, and chronic pancreatitis)
(n = 1773, 50 %), gastrointestinal subepithelial lesions (n = 1334, 38 %), and other
indications (n = 373, 11 %), which included hepatopancreaticobiliary lesions other
than those mentioned above and nonluminal soft tissue or cystic lesions. Linear and
radial echoendoscopes were used for 960 (27 %) and 2592 (73 %) procedures, respectively.
Conscious sedation was used during 3062 procedures (86 %) and general anesthesia in
only 64 procedures (2 %). No sedation was used for the remaining 426 procedures, 93 %
of which were lower EUS procedures. A fellow was present for 2211 procedures (62 %).
A complete colonoscopy or ERCP was performed concurrently in 67 (2 %) and 33 (1 %)
procedures, respectively. Among the 733 procedures that involved FNA, a 22- or 25-gauge
needle was used for 514 procedures (70 %) and a 19-gauge needle was used for 75 procedures
(10 %). The endoscopist performed 1–3 passes for 547 procedures (75 %), and > 3 passes
for 103 (14 %). Additional procedural details are available in [Table 1].
Table 1
Procedure characteristics
Characteristics
|
EUS only (n = 2819)
|
EUS + FNA (n = 733)
|
EUS approach, n (%)
|
|
2324 (82.4)
|
710 (96.9)
|
|
495 (17.6)
|
23 (3.1)
|
EUS indication, n (%)
|
|
1286 (45.6)
|
487 (66.4)
|
|
1212 (43.0)
|
122 (16.6)
|
|
18 (0.6)
|
54 (7.4)
|
|
303 (10.8)
|
70 (9.6)
|
Type of endoscope, n (%)
|
|
227 (8.1)
|
733 (100)
|
|
2592 (91.9)
|
0 (0.0)
|
Sedation, n (%)
|
|
410 (14.5)
|
16 (2.2)
|
|
2359 (83.7)
|
703 (95.9)
|
|
50 (1.8)
|
14 (1.9)
|
|
1772 (62.9)
|
439 (59.9)
|
Additional procedures performed, n (%)
|
|
65 (2.3)
|
2 (0.3)
|
|
31 (1.1)
|
2 (0.3)
|
FNA details, n (%)
|
|
|
NA
|
514 (70.1)
|
|
NA
|
75 (10.2)
|
|
NA
|
144 (19.7)
|
Number of passes, n (%)
|
|
NA
|
547 (74.6)
|
|
NA
|
103 (14.1)
|
|
NA
|
83 (11.3)
|
EUS, endoscopic ultrasound; FNA, fine-needle aspiration; HPB, hepatopancreaticobiliary;
ERCP, endoscopic retrograde cholangiopancreatography; NA, not applicable.
Adverse events
From a total of 3552 procedures, patients had unplanned presentations to the emergency
department or were admitted to hospital within 30 days of EUS after 504 (14 %) of
the procedures. Adverse events that were definitely or possibly related to EUS occurred
after 69 procedures (2 %), while 435 presentations were considered unrelated. There
were 64 adverse events related to upper EUS (2 % of all upper EUS procedures), 41
of which occurred following upper EUS-FNA (6 % of all upper EUS-FNA). Five adverse
events occurred following a lower EUS procedure (1 %), two of which involved EUS-FNA
(9 % of the 23 lower EUS procedures that involved FNA). With respect to EUS-only procedures
(without FNA), there were 23 adverse events (1 %) among 2324 upper EUS procedures,
and 3 events (0.6 %) among 495 lower EUS procedures.
Among the 69 definitely or possibly related adverse events, 36 (52 %) were emergency
department visits alone, 5 (7 %) were direct inpatient admissions without an emergency
department visit, and 28 (41 %) were emergency department visits that led to an inpatient
admission. The adverse event was gastrointestinal in nature for 61 procedures (88 %)
and cardiopulmonary for 8 procedures (12 %) ([Table 2]).
Table 2
Associated risk of emergency room visit and/or hospitalization within 30 days of endoscopic
ultrasound.
Characteristics
|
EUS only (n = 2819)
|
EUS + FNA (n = 733)
|
Emergency department visit or hospitalization, n (%)
|
374 (13.3)
|
130 (17.7)
|
Relation to EUS procedure
|
|
7 (1.9)
|
32 (24.6)
|
|
19 (5.1)
|
11 (8.5)
|
|
348 (93.0)
|
87 (66.9)
|
Definitely or possibly related adverse events, n (%)
|
26 (0.9)
|
43 (5.9)
|
Disposition
|
|
20 (76.9)
|
16 (37.2)
|
|
0 (0.0)
|
5 (11.6)
|
|
6 (23.1)
|
22 (51.2)
|
Final diagnosis
|
|
21 (80.8)
|
40 (93.0)
|
|
0 (0.0)
|
2 (4.7)
|
|
1 (3.9)
|
4 (9.3)
|
|
11 (42.3)
|
26 (60.5)
|
|
9 (34.6)
|
8 (18.6)
|
|
5 (19.2)
|
3 (7.0)
|
EUS, endoscopic ultrasound; FNA, fine-needle aspiration; GI, gastrointestinal.
For the 33 adverse events (48 % of all definite or possible events) that resulted
in an inpatient admission, the median length of stay was 9 days ([Table 3]). No adverse event resulted in admission to an intensive care unit or patient death
within 30 days of the EUS. Five adverse events (7 % of all definitely or possibly
related events) required use of antibiotics, one (1 %) required endoscopic therapy,
one (1 %) required surgery, and one (1 %) required blood transfusion.
Table 3
Definite or possibly related adverse events leading to inpatient admissions.
Characteristic
|
Number of patients n = 33
|
Indication
|
|
24
|
|
4
|
|
2
|
|
3
|
Age in years
|
|
3
|
|
16
|
|
9
|
|
5
|
Length of stay in days
|
|
11
|
|
12
|
|
4
|
|
0
|
|
3
|
|
3
|
Final diagnosis
|
|
16
|
|
2
|
|
2
|
|
9
|
|
4
|
HPB, hepatopancreaticobiliary.
Risk factors for adverse events
For EUS alone, there were no significant differences with respect to incidence of
definitely or possibly related adverse events for upper vs. lower EUS (P = 0.61), EUS indication (P = 0.63), type of echoendoscope (P = 0.72), and type of sedation (P = 0.47). Fewer adverse events occurred when a fellow was present (n = 9, 0.5 %) vs.
not present (n = 17, 1.6 %) (P = 0.004). For EUS with FNA, there were no significant differences in adverse event
risk for upper vs. lower EUS (P = 0.64), type of sedation (P = 0.39), or the presence of a fellow (P = 0.52) ([Table 4]).
Table 4
Characteristics of adverse events within 30 days of endoscopic ultrasound procedures,
with and without fine-needle aspiration.
Characteristics
|
EUS only
|
EUS + FNA
|
Definitely or possibly related (n = 26)
|
None or not related (n = 2793)
|
P value[*]
|
Definitely or possibly related (n = 43)
|
None or not related (n = 690)
|
P value[*]
|
EUS approach, n (%)
|
|
23 (1.0)
|
2301 (99.0)
|
0.61
|
41 (5.8)
|
669 (94.2)
|
0.64
|
|
3 (0.6)
|
492 (99.4)
|
2 (8.7)
|
21 (91.3)
|
EUS indication, n (%)
|
|
10 (0.8)
|
1276 (99.2)
|
|
35 (7.2)
|
452 (92.8)
|
|
|
12 (1.0)
|
1200 (99.0)
|
|
1 (0.8)
|
121 (99.2)
|
|
|
0 (0.0)
|
18 (100)
|
0.63
|
2 (3.7)
|
52 (96.3)
|
0.022
|
|
10 (0.8)
|
1276 (99.2)
|
|
35 (7.2)
|
452 (92.8)
|
|
Type of endoscope, n (%)
|
|
1 (0.4)
|
226 (99.6)
|
0.72
|
43 (5.9)
|
690 (94.1)
|
NA
|
|
25 (1.0)
|
2567 (99.0)
|
NA
|
NA
|
Sedation, n (%)
|
|
1 (2.0)
|
49 (98.0)
|
|
0 (0.0)
|
14 (100)
|
0.39
|
|
21 (0.9)
|
2338 (99.1)
|
0.47
|
41 (5.8)
|
662 (94.2)
|
|
4 (1.0)
|
406 (99.0)
|
|
2 (12.5)
|
14 (87.5)
|
Fellow present, n (%)
|
|
9 (0.5)
|
1763 (99.5)
|
0.004
|
28 (6.4)
|
411 (93.6)
|
0.52
|
|
17 (1.6)
|
1030 (98.4)
|
15 (5.1)
|
279 (94.9)
|
Additional procedures performed, n (%)
|
|
0 (0.0)
|
96 (100)
|
> 0.99
|
0 (0.0)
|
4 (100)
|
> 0.99
|
|
26 (1.0)
|
2697 (99.0)
|
43 (5.9)
|
686 (94.1)
|
EUS, endoscopic ultrasound; FNA, fine-needle aspiration; HPB, hepatopancreaticobiliary;
NA, not applicable.
*
P value was calculated using Fisher’s exact tests.
In the univariable logistic regression of a priori selected risk factors for adverse
event risk, there was no association between adverse event risk and patient age (OR = 1.00,
95 %CI 0.61–1.64; P = 0.98), male sex (OR = 1.34, 95 %CI 0.83–2.15; P = 0.23), the performance of multiple EUS procedures (OR = 1.22, 95 %CI 0.74–2.01;
P = 0.44), the performance of other procedures (OR = 0.24, 95 %CI 0.01–3.91; P = 0.31), or the presence of a fellow (OR = 0.69, 95 %CI 0.43–1.12; P = 0.13). The use of FNA was significantly associated with a higher risk of adverse
events (OR = 6.64, 95 %CI 4.07–10.84; P < 0.001).
On multivariable analysis, the only factor associated with an increased risk of adverse
events was the use of FNA (OR = 6.43, 95 %CI 3.92–10.55; P < 0.001) ([Table 5]).
Table 5
Factors associated with adverse events within 30 days of endoscopic ultrasound procedures.
Characteristics
|
Univariable
|
Multivariable
|
Adverse event OR (95 %CI)
|
P value
|
Adverse event OR (95 %CI)
|
P value
|
Patient characteristics
|
|
1.00 (0.61–1.64)
|
0.98
|
0.80 (0.48–1.32)
|
0.38
|
|
1.34 (0.83–2.15)
|
0.23
|
1.28 (0.79–2.07)
|
0.31
|
|
1.22 (0.74–2.01)
|
0.44
|
1.10 (0.67–1.82)
|
0.71
|
|
0.24 (0.01–3.91)
|
0.31
|
0.40 (0.02–6.57)
|
0.52
|
Procedural characteristics
|
|
0.69 (0.43–1.12)
|
0.13
|
0.73 (0.45–1.18)
|
0.20
|
|
6.64 (4.07–10.84)
|
< 0.001
|
6.43 (3.92–10.55)
|
< 0.001
|
OR, odds ratio; CI, confidence interval; EUS, endoscopic ultrasound; FNA, fine-needle
aspiration; ERCP, endoscopic retrograde cholangiopancreatography.
1 Patients who had multiple EUS procedures vs. patients who only had one EUS procedure
during the study period.
2 Patients who had ERCP (n = 34) or colonoscopy (n = 67) on the same day as EUS vs.
patients who only had EUS.
Discussion
In this population-based retrospective cohort study we found that unplanned emergency
department visits and hospitalizations within 30 days that were definitely or possibly
related to EUS without FNA occurred in 1 % and 0.6 % of upper and lower procedures,
respectively. However, for EUS procedures that involved FNA, the risk of adverse events
increased to 6 % and 9 %, respectively. Performing FNA during EUS increased the odds
of an adverse event by over sixfold in multivariable analysis.
Upper EUS with FNA is generally considered a safe procedure, during which serious
adverse events are rare [14]
[21]
[22]. A systematic review of EUS-FNA complications, which included 10 941 patients, reported
a 1 % risk of morbidity secondary to pancreatitis, abdominal pain, bleeding, or infection
in the immediate post-procedure period [1]; however, the majority of studies in the review involved short-term follow-up, with
only two articles [9]
[10] reporting 30-day events. Our higher reported risk of adverse events is likely due
to a more inclusive definition of adverse event and longer follow-up period. Fortunately,
the incidence of serious adverse events in our cohort was still low, with no event
requiring intensive care or resulting in patient death.
Very few studies have examined adverse events related to lower EUS procedures [23]. Studies on diagnostic yield of EUS-FNA for rectal and pelvic lesions have reported
a low incidence of adverse events but are limited by small sample sizes and reported
only immediate adverse events [24]
[25]. One large prospective study found a strikingly high risk of adverse events related
to lower EUS FNA. Levy et al. found that 21 % of 502 participants followed for 2–4
months post-procedure experienced an adverse event [11]. The investigators phoned participants 7–14 days after the procedure and also saw
them in clinic 2–4 months later. We found a lower rate of adverse events in our study,
with only 2 of the 23 participants (9 %) who underwent lower EUS-FNA experiencing
an adverse event that was definitely or possibly related to their procedure. This
difference is likely due to how events were defined and captured in each study. In
the study by Levy et al. adverse events included presentations that did not require
medical intervention or hospitalization, such as post-procedure asymptomatic hypotension
and minor rectal bleeding [11]. The rate of serious adverse events requiring immediate intervention or hospitalization
was 6 %, which is more comparable to our findings. Additionally, 37 % and 17 % of
participants underwent colonoscopy and polypectomy during the same session as their
lower EUS procedure, making it challenging to differentiate the risk of EUS-FNA alone
from the cumulative risk of undergoing multiple procedures on the same day [11]. Our study is the first to examine the risk of lower EUS with and without FNA in
isolation of other lower endoscopic procedures.
Our findings may help to inform health care providers and patients contemplating undergoing
EUS. We have demonstrated that upper and lower EUS without FNA for evaluation, surveillance,
and staging is very safe. These observations do not support individualized counseling
based on EUS indication, the type of echoendoscope used, and patient sex or age, as
these factors were not found to be related to adverse event risk. The presence of
a fellow did not increase adverse event risk in the univariable or multivariable analyses,
an important finding for advanced endoscopy educators who work with trainees. A previous
study found that adverse event risk was higher for EUS procedures performed with a
trainee present, but this was limited to the first 3 months of training [13]. In our study, fewer adverse events occurred when a fellow was present for EUS only
procedures, with a low overall risk (< 1 %). There was no difference in adverse event
risk for EUS-FNA procedures performed with or without a fellow present.
This study has several strengths. First, it is population based and therefore less
prone to referral bias as seen in smaller single-center studies. Second, we utilized
a robust process to identify potential adverse events based on previously reported
methodology [13]
[14]
[15]. The medical record of every possible adverse event was formally reviewed rather
than relying on administrative or survey data to assess outcomes. Finally, relatively
few participants underwent ERCP, colonoscopy, or polypectomy concurrently with EUS. This
allowed us to evaluate the specific risk of adverse events associated with EUS, rather
than the cumulative risk for participants undergoing multiple procedures on the same
day.
Our study also has several limitations. We were unable to capture individuals who
may have experienced an adverse event but visited their primary physician rather than
going to an acute care facility. While such encounters would reflect less serious
events, the excess health care utilization resulting from any unplanned health care
presentation is worthy of consideration. Moreover, symptoms such as post-procedure
rectal bleeding can adversely affect patients even if they are never brought to medical
attention. Our lower EUS data only apply to distal lesions as we did not have any
patients who underwent EUS of the proximal colon. This is consistent with other studies
in this field. The median length of stay of patients admitted for an adverse event
was 9 days. Although a third were discharged in 5 days or less, among those admitted
for longer it is likely that the underlining diseases of some patients (e. g. cancer)
contributed to prolongation of hospitalization. We could not determine the proportion
of a hospitalization stay directly related to the adverse event. In addition, owing
to the low number of endoscopists performing EUS procedures at our institution, we
could not report their individual outcomes. We also could not determine whether a
trainee participated as an observer or an operator during procedures. Finally, our
study may have misclassification bias and limited generalizability, as it was retrospective
and conducted at a single center.
In conclusion, upper and lower EUS were shown to be safe procedures with adverse events
leading to unplanned emergency department visits and hospitalization following 2 %
of procedures. The addition of FNA increased the risk of adverse events sixfold and
events requiring inpatient admission were rare. The rate of adverse events related
to lower EUS procedures was low when they were considered in isolation of other lower
endoscopic procedures.