Keywords Endoscopy Small Bowel - Capsule endoscopy - Small intestinal bleeding
Introduction
Melena is generally considered to result from bleeding in the upper gastrointestinal
tract until it is proven otherwise [1 ]. Consequently, esophagogastroduodenoscopy (EGD) is the preferred procedure for evaluating
patients who present with melena. However, EGD fails to identify the bleeding source
in approximately one-quarter of patients [2 ]
[3 ]. When EGD is nondiagnostic, evaluation of the mid to lower gastrointestinal bleeding
should be considered.
Guidelines recommend colonoscopy before small bowel evaluation in patients with melena
and negative EGD results [4 ]
[5 ]. These recommendations are based on studies reporting high detection rates of bleeding
sources—23% to 35%—on colonoscopy [6 ]
[7 ]
[8 ]. However, these studies are retrospective and have small sample sizes. In contrast,
a recent large retrospective study involving 1743 patients with melena and negative
EGD findings revealed that colonoscopy identified bleeding sources in only 4.7% of
cases [3 ].
Moreover, after both EGD and colonoscopy, the cause of overt gastrointestinal bleeding
remains undetermined in 4% to 15% of cases. These patients are considered to have
potential small bowel bleeding [4 ]. Guidelines recommend video capsule endoscopy (VCE) for these patients because of
its high diagnostic yield and noninvasiveness [4 ]
[5 ]. However, VCE has limitations, including missed proximal small bowel lesions due
to rapid transit and inability to perform therapeutic interventions.
Push enteroscopy is another valuable tool for patients with obscure gastrointestinal
bleeding. It is a straightforward endoscopic technique that most gastroenterologists
can perform without special instruments. In a single retrospective study, the diagnostic
yield of push enteroscopy in patients with melena was reported to be 40% [9 ]. Therefore, selecting push enteroscopy for patients with melena and negative EGD
findings may be reasonable. This approach could help avoid unnecessary colonoscopy
and VCE, potentially reducing treatment costs. We conducted a prospective study to
evaluate diagnostic yields of push enteroscopy and colonoscopy in patients presenting
with melena without hematemesis and with negative EGD results.
Patients and methods
Study design and setting
We conducted a prospective, multicenter cohort study from July 2019 to October 2022
at four referral centers in Thailand. They were the Faculty of Medicine Siriraj Hospital,
Mahidol University, Bangkok; Golden Jubilee Medical Center, Mahidol University, Bangkok;
Hatyai Hospital, Songkhla; and Ratchaburi Hospital, Ratchaburi.
Patient selection
Consecutive patients presenting with melena (black, tarry stools) without hematemesis
were invited to participate. Inclusion criteria were age 18 years or older and underwent
EGD without bleeding source identified. Exclusion criteria were the following: 1)
unstable hemodynamics defined by blood pressure lower than 90/60 mm Hg despite receiving
an inotropic agent prior to push enteroscopy; 2) respiratory compromise defined by
desaturation requiring high flow oxygen or mechanical ventilator; 3) uncorrectable
bleeding disorders; and 4) pregnancy.
Data collection and procedures
Baseline demographic data, comorbidities, and bleeding characteristics and severity
were recorded. Patients underwent push enteroscopy followed by colonoscopy. Time from
presentation to each endoscopic procedure and the findings were documented. Culprit
lesions—identified as the most likely cause of bleeding—were detailed and included
ulcers, angiodysplasia, Dieulafoy’s lesions, tumors, diverticula, and other abnormalities.
Colonoscopy was not performed for patients whose push enteroscopy detected culprit
lesions and were deemed at moderate to high risk for developing colonoscopic complications,
including significant anesthetic risk or multiple comorbidities. In this group, the
colonoscopic results were presumed to be negative, and a 3-month follow-up without
rebleeding was applied to confirm absence of additional bleeding sources beyond those
identified by push enteroscopy. For patients with negative results from both push
enteroscopy and colonoscopy, further investigations were conducted at the discretion
of the treating physicians. Diagnostic and therapeutic yields of push enteroscopy
and colonoscopy were compared.
Push enteroscopy procedure
Push enteroscopy was performed using PCF-Q180AL or PCF-HQ190L/I pediatric colonoscopes
(Olympus Medical Systems Corporation, Tokyo, Japan). Patients fasted for 6 to 8 hours
before the procedure to optimize gastrointestinal visualization. Sedation with intravenous
propofol ensured patient comfort and safety. If necessary, hyoscine butyl-bromide
was administered to reduce small bowel motility. The endoscope was inserted orally
and advanced through the esophagus, stomach, and duodenum to reach the proximal small
intestine. Minimal insufflation was used to reduce discomfort and maintain clear mucosal
visualization. Air was used for insufflation during the procedure. CO2 was not used to ensure uniformity of practice and consistency of care across all
participating centers. Careful inspection for abnormalities was conducted during scope
withdrawal. Patients were monitored during recovery and provided with clear recuperation
instructions.
Statistical analysis
Descriptive statistics summarized patient characteristics. Continuous variables are
expressed as means and standard deviations or as medians and interquartile ranges
(IQRs); categorical variables are presented as counts and percentages. Standard two-group
comparison methods were used, including an independent t-test (for normally distributed
data) or Wilcoxon rank-sum test (for non-normally distributed data) for continuous
data, and a chi-square test or Fisher exact test (depending on sample size) for categorical
data. The McNemar test was used to assess differences in diagnostic yield among push
enteroscopy, colonoscopy, and their combination. Statistical analyses were performed
via SAS OnDemand for Academics (SAS Institute, Cary, North Carolina, United States).
A two-tailed P < 0.05 was considered statistically significant.
Ethical considerations
All patients provided written informed consent after receiving detailed information
about the diagnostic procedures. The study was approved by the local institutional
human research review committee (approval number Si-484/2019) and conformed to good
clinical practices. The research was registered with ClinicalTrials.gov (NCT06574542).
Results
Patient characteristics
A total of 221 patients presented with melena without hematemesis and underwent diagnostic
EGD, which identified bleeding sources in 141 patients (63.8%). Three patients were
excluded owing to respiratory compromise. The remaining 77 patients (34.8%) with nondiagnostic
EGD results were included in our analyses. The mean age of these patients was 67.8
years and 51.9% were men ([Table 1 ]).
Table 1 Baseline characteristics of the study population.
Parameters
N = 77
AF, atrial fibrillation; BUN, blood urea nitrogen; Cr, creatinine; INR, international
normalized ratio; IQR, interquartile range; NSAID, nonsteroidal anti-inflammatory
drug; PRC, packed red cell; SD, standard deviation.
Demographic data
67.8 ± 14.8
40 (51.9%)
Comorbidities, n (%)
50 64.9%)
26 (33.8%)
20 (26.0%)
15 (19.5%)
26 (33.8%)
10 (13.0%)
3 (3.9%)
Medications
7 (9.1%)
21 (27.3%)
19 (24.7%)
5 (6.5%)
Clinical presentations
9 (11.7%)
15 (19.5%)
23 (30.3%)
Previous bleeding
18 (23.4%)
PRC transfusions
2.8 ± 2.0
Laboratory investigations
6.5 ± 1.9
1.6 ± 1.4
40.5 ± 31.6
1.1 (0.9 - 2.1)
26.5 ± 20.0
The most common comorbidities were hypertension (64.9%), diabetes mellitus (33.8%),
chronic kidney disease (33.8%), atrial fibrillation or valvular heart disease (26.0%),
and coronary artery disease (19.5%). In addition, 9.1% of patients had a history of
nonsteroidal anti-inflammatory drug use, 27.3% had used antiplatelets, 24.7% had used
anticoagulants, and 6.5% had experienced warfarin overdose. Almost one-quarter (24.7%)
had a history of previous bleeding. No patients with surgically altered upper gastrointestinal
anatomy were enrolled.
Endoscopic procedures and timing
All 77 patients underwent push enteroscopy, and 59 patients proceeded to colonoscopy.
Median times from hospital presentation to EGD, push enteroscopy, and colonoscopy
were 3.0 days (interquartile range [IQR] 1.0–5.0), 3.0 days (IQR 2.0–6.0), and 5.0
(IQR 3.0–7.0), respectively.
Mean procedure time was 27.9 ± 13.9 minutes for push enteroscopy and 31.4 ± 28.7 minutes
for colonoscopy. No serious complications such as bleeding or perforation occurred
during either push enteroscopy or colonoscopy.
Diagnostic yield of push enteroscopy and colonoscopy
As shown in [Fig. 1 ]
a and [Table 2 ], 27 of 77 patients (35.0%) had a culprit lesion identified during push enteroscopy,
including ulcers (25.9%), angiodysplasia (29.6%), Dieulafoy’s lesion (11.1%), tumor
(14.8%), polyps (3.7%), bleeding diverticulum (3.7%), and other conditions (11.1%).
Locations of lesions were at the duodenum reached by EGD in five patients, the duodenum
not reached by EGD in two patients, and the jejunum in 22 patients ([Fig. 2 ]). Five patients (18.5%) had lesions missed during previous EGD. Lesions comprised
duodenal ulcers, duodenal angiodysplasia, and hemosuccus pancreaticus. Colonoscopy
was performed in 59 cases (76.6%) and bleeding sources were identified in 10 patients
(12.9%). Culprit lesions found during colonoscopy were ulcers (3 patients), Dieulafoy’s
lesions (3 patients), cancer (1 patient), polyp (1 patient), diverticulum (1 patient),
and colitis (1 patient).
Fig. 1
a Diagnostic yield and b therapeutic intervention success rate for push enteroscopy and colonoscopy in melena
patients with negative esophagogastroduodenoscopy results.
Fig. 2 Locations of lesions identified by push enteroscopy in melena patients with negative
esophagogastroduodenoscopy results. EGD, esophagogastroduodenoscopy.
Table 2 Types and locations of lesions detected by push enteroscopy and colonoscopy.
Characteristics
Total (N = 37)
Push enteroscopy (N = 27)
Colonoscopy (N = 10)
Type of lesions
Ulcers
10 (27.0%)
7 (25.9%)
3 (30.0%)
Angiodysplasia
8 (21.6%)
8 (29.6%)
0 (0.0%)
Dieulafoy’s lesion
6 (16.2%)
3 (11.1%)
3 (30.0%)
Tumor
5 (13.5%)
4 (14.8%)
1 (10.0%)
Polyp
2 (5.4%)
1 (3.7%)
1 (10.0%)
Diverticulum
2 (5.4%)
1 (3.7%)
1 (10.0%)
Other
4 (10.8%)
3 (11.1%)
1 (10.0%)
Diagnostic yield of push enteroscopy was significantly greater than that of colonoscopy
(P = 0.005). Combining push enteroscopy and colonoscopy increased the overall diagnostic
yield to 48% (37 patients), which was significantly greater than the yield of push
enteroscopy alone (P = 0.002) or colonoscopy alone (P < 0.0001).
Univariate analysis of predictive factors for positive findings on push enteroscopy
Univariate analysis was performed to identify predictive factors associated with positive
findings on push enteroscopy ([Table 3 ]). Variables analyzed included age, sex, underlying comorbidities, medication use,
and blood urea nitrogen/creatinine ratio. No significant associations were found between
these factors and positive findings on push enteroscopy.
Table 3 Univariate analysis of predictors for positive findings in push enteroscopy.
Parameters
Odds ratio (95% CI)
AF, atrial fibrillation; BUN, blood urea nitrogen; CI, confidence interval; CKD, chronic
kidney disease; Cr, creatinine; NSAID, nonsteroidal anti-inflammatory drug.
Age
1.00 (0.97–1.04)
Sex, male
1.25 (0.49–3.20)
CKD
1.25 (0.47–3.33)
AF or valvular heart disease
1.33 (0.47–3.81)
NSAIDs
2.72 (0.56–13.19)
Aspirin
0.73 (0.24–2.20)
Antiplatelets
0.66 (0.22–1.98)
Warfarin
0.91 (0.27–2.99)
Time to push enteroscopy
1.06 (0.94–1.19)
Hypotension
0.91 (0.21–3.99)
Hemoglobin
1.02 (0.80–1.31)
BUN/Cr
1.00 (0.98–1.03)
Therapeutic interventions and outcomes
Therapeutic interventions were conducted during push enteroscopy in 13 patients (16.9%)
and during colonoscopy in five patients (6.5%). Successful interventions were achieved
in 12 patients (15.6%) during push enteroscopy and in all five patients (6.5%) during
colonoscopy. As shown in [Fig. 1 ]
b , the successful therapeutic intervention rate was greater for push enteroscopy than
for colonoscopy, although this difference was not statistically significant (P = 0.089). Compared with either method alone, combined use of push enteroscopy and
colonoscopy significantly increased the successful therapeutic intervention rate.
The 30-day rebleeding rate was 3.7% for patients who received therapeutic intervention
via push enteroscopy and 0% for those treated via colonoscopy.
Further investigations and definitive diagnoses
Further investigations were performed for patients whose bleeding sources were not
identified or managed. These were VCE in 10 patients, computed tomography in six patients,
and balloon-assisted enteroscopy (BAE) in four patients.
A definitive source of bleeding was ultimately identified in 53 patients (68.8%) through
combined use of initial procedures and further investigations.
Discussion
Melena results from degradation of hemoglobin and implies bleeding from the upper
gastrointestinal tract. Therefore, EGD should be the first-line investigation in patients
presenting with melena [2 ]
[3 ]. Our findings support this recommendation: In 64% of patients with melena, bleeding
sources were revealed via EGD. A previous study reported that approximately three-quarters
of these patients had lesions detected on EGD [10 ].
After negative EGD, guidelines recommend performing colonoscopy in patients presenting
with melena before proceeding to small bowel evaluation [4 ]
[5 ]. Our study is the first prospective analysis to evaluate diagnostic yield of colonoscopy
after negative EGD in patients with melena. We found that the diagnostic yield was
14%, which is lower than the previous retrospective reports of 23% to 35% [6 ]
[7 ]
[8 ]. Although the diagnostic yields were not high, 5% to 10% of the lesions detected
in previous studies and our study were colon cancers [10 ]
[11 ]. Furthermore, colonoscopy can be used to perform therapeutic interventions. In our
study, therapeutic interventions were performed on five of 10 lesions detected during
colonoscopy and all the treatments were successful. Given the benefits of cancer detection,
therapeutic capability, and accessibility, we believe that colonoscopy should still
be considered in patients with melena despite its fair diagnostic yield. However,
more investigations are likely to be required.
VCE is a noninvasive modality that standard guidelines recommend as the first-line
investigation after negative EGD and colonoscopy; the condition is defined as potential
small bowel bleeding [4 ]
[5 ]. VCE has a high diagnostic yield of 55% to 65% for overall potential small bowel
bleeding. However, data regarding its yield specifically for patients with melena,
not hematochezia, are limited. Only one study by Mussetto et al reported the VCE yield
in patients with melena and negative EGD results using a panenteric capsule in 12
patients. Bleeding sources were found in 10 patients (83.3%). Besides limited data
on VCE in this specific setting, VCE has several limitations. First, it can miss proximal
small bowel lesions due to rapid transit; the major papilla in the second part of
the duodenum is visualized in only 10.4% to 43.6% of patients [12 ]
[13 ]. Second, VCE cannot perform therapeutic interventions. Some lesions, such as Dieulafoy’s
lesion and angiodysplasia, may not actively bleed and cannot be identified during
subsequent direct enteroscopy. Early direct enteroscopy could be beneficial for these
vascular lesions. Interestingly, small bowel angiodysplasia has been reported to be
predominantly located in the proximal small bowel. Specifically, studies have reported
that 67% of small bowel angiodysplasia lesions are located in the duodenum or ligament
of Treitz [14 ]. Therefore, an antegrade approach for small bowel enteroscopy is a reasonable option.
This technique can overcome the limitations of VCE in missing proximal small bowel
lesions and the inability to provide therapeutic intervention for vascular lesions.
Given the substantial possibility of bleeding in the proximal small bowel in the setting
of melena, push enteroscopy should play a role. Limited evidence has demonstrated
the performance of push enteroscopy in patients with melena. Most previous trials
included all potential cases of small bowel bleeding—either melena or hematochezia,
reporting a diagnostic yield of 19% to 35% [15 ]
[16 ]
[17 ]
[18 ]. Only a study by Lepère et al [9 ] reported the diagnostic yield specifically in patients with melena. In their retrospective
study, the overall diagnostic yield of push enteroscopy for all potential small bowel
bleeding cases was 34%. The yield increased to 40% in patients presenting with melena
and decreased to 17% in patients presenting with hematochezia. Our study is the first
prospective analysis to evaluate diagnostic yield of push enteroscopy in patients
presenting with melena and negative EGD results. We demonstrated that diagnostic yield
of push enteroscopy was 35% in these patients.
Another benefit of push enteroscopy is its ability to perform therapeutic interventions.
In our study, 13 of 27 patients whose bleeding sources were detected received interventions
and 12 of these interventions were successful. Furthermore, push enteroscopy is safe;
no serious complications were observed in our study. Push enteroscopy is also a simple
endoscopic technique that does not require special instruments and most endoscopists
can perform it. Thus, push enteroscopy should be considered in patients with melena
and negative EGD findings. The American College of Gastroenterology clinical guidelines
also recommend push enteroscopy as an option in patients with gastrointestinal bleeding
who have negative EGD and colonoscopy results before VCE [4 ]. Another advantage of using push enteroscopy as the initial investigation is that
it may reduce need for bowel preparation, which is typically required for colonoscopy.
This is particularly relevant for hospitalized patients, in whom bowel prep is known
to be more frequently suboptimal compared with outpatient settings. This can contribute
to lower diagnostic yield of colonoscopy in this population [19 ]
[20 ].
The drawback of push enteroscopy is its limited reach: It can access the small bowel
only approximately 0.6 meters distal to the duodenojejunal junction. Melena can originate
from bleeding at more distal parts [21 ]. Therefore, identifying predictors for proximal small bowel bleeding is important
in selecting suitable patients for push enteroscopy. A study by Lepère et al revealed
that chronic renal failure was a predictor associated with positive findings on push
enteroscopy [9 ]. Unfortunately, our univariate analysis did not identify any significant predictive
factors associated with positive findings on push enteroscopy. Variables analyzed
included age, sex, underlying comorbidities, medications used, and blood urea nitrogen/creatinine
ratio.
BAE has a high diagnostic yield comparable to that of VCE in patients with potential
small bowel bleeding [22 ]
[23 ]
[24 ]. Compared with push enteroscopy, deeper small bowel insertion leads to a greater
diagnostic yield [25 ]
[26 ]. Furthermore, BAE can perform therapeutic interventions. Therefore, BAE may be an
appropriate option for investigating patients with melena. However, BAE carries risks
of complications, including sedation-related issues, bleeding, perforation, and pancreatitis
[27 ]
[28 ]. In addition, BAE is unavailable at many centers and requires highly skilled endoscopists,
limiting its widespread use.
Given these considerations, we believe that push enteroscopy should be performed in
patients who present with melena and negative EGD results. It could be considered
before colonoscopy. However, these two modalities complement each other. Our study
revealed that combining push enteroscopy and colonoscopy increased the diagnostic
yield to 48%, which was significantly greater than that of either method alone. Furthermore,
therapeutic intervention rates were significantly higher for the combination than
for each modality. Therefore, if push enteroscopy is negative, colonoscopy should
be performed afterward. After negative push enteroscopy and colonoscopy, VCE should
be performed, and BAE should be considered according to the VCE findings.
This study has several strengths. It is the first prospective investigation of consecutive
patients who presented with melena without hematemesis and negative EGD results. We
assessed diagnostic and therapeutic yields of both push enteroscopy and colonoscopy.
In addition, therapeutic interventions were performed in the same or nearly the same
session, providing timely treatment when necessary.
However, there are several limitations. Colonoscopy was not performed for every patient
because of safety concerns. For patients whose culprit lesion was identified during
push enteroscopy and who had heightened health risks, colonoscopy was sometimes omitted
to avoid unnecessary procedural risks. This omission could have resulted in missing
additional lesions. To address this issue, we conducted 3-month follow-up to ensure
no recurrent bleeding. We believe that clinically significant lesions would likely
have manifested as recurrent bleeding or anemia during this period, enabling timely
detection and management. Although there is no universally established guideline for
optimal duration of follow-up, randomized controlled trials in patients with lower
gastrointestinal bleeding have used a shorter duration of 30-day rebleeding as a secondary
outcome [29 ]
[30 ].
Furthermore, a significant proportion of patients in our study were recruited from
large tertiary referral centers, which may introduce referral bias. These institutions
often care for patients with more severe comorbidities or refractory gastrointestinal
bleeding, which may not reflect the typical case mix encountered in general or community
hospital settings. Consequently, diagnostic yields reported in this study may not
be fully generalizable to broader clinical populations.
Finally, our findings are intended to complement current diagnostic strategies for
patients with suspected small bowel bleeding. Push enteroscopy may serve as a valuable
adjunctive tool that enhances diagnostic and therapeutic outcomes, especially in settings
where it is available, rather than acting as a replacement for established diagnostic
modalities.
Conclusions
In conclusion, this study confirms the essential role of EGD in evaluating patients
who present with melena despite absence of hematemesis. If EGD is negative, push enteroscopy
should be considered before, or in combination with, colonoscopy. Future prospective
studies with larger sample sizes are warranted to allow for more robust statistical
analyses and validation of these findings.
Bibliographical Record
