Introduction
Iron-deficiency anemia (IDA) occurs in 5 % of premenopausal women and 1 – 2 % of men
and postmenopausal women. It is a common reason for referral to gastroenterologists,
accounting for 4 – 13 % of outpatient visits [1]
[2]. IDA may be due to: blood loss (gastrointestinal, hemoptysis, traumatic, menorrhagia,
hematuria, regular donors, post-surgery, iatrogenic repeated studies, etc.), malabsorption
(celiac disease, gastrectomy or other enteropathies), or low iron intake (very low
frequency in adults in our area) [1]
[2]. In premenopausal women, the most common cause is gastrointestinal and/or gyneco-obstetric
blood loss and in men and postmenopausal women, the most common cause is gastrointestinal
blood loss. In most cases, the cause of bleeding is accessible to standard endoscopy
(gastroscopy and colonoscopy) [1]
[2]. However, in 2 – 10 %, the lesions are located in the small intestine, so that these
studies are negative [1]
[2].
Small-bowel capsule endoscopy (CE) is a diagnostic tool that allows painless and safe
assessment of the small bowel, leading to an etiologic diagnosis of small-bowel bleeding
in approximately 60 % of cases [3]
[4]. Two meta-analyses [4]
[5] have shown that CE is superior to push enteroscopy and radiological imaging for
diagnosing clinically significant small-bowel pathology in patients with “suspected
small-bowel bleeding (SBB)”, previously called obscure gastrointestinal bleeding (OGIB).
Furthermore, CE and double-balloon enteroscopy have comparable diagnostic yields in
SBB [6]
[7].
In an attempt to assess the relevance of findings of CE, Saurin et al. classified
the lesions as highly relevant (P2), of uncertain relevance (P1), or of low relevance
(P0) [8]. P2 lesions include angiodysplasia, tumors, ulcerative lesions, and varices, P1
include red spots and small or isolated erosions, and P0 include visible mucosal veins,
diverticula without the presence of blood, or nodules without mucosal breaks. In a
subsequent study, a P3 category was included for active bleeding [9].
Several studies have evaluated the effectiveness of CE in SBB, but few have focused
on IDA alone [10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]. Furthermore, in those few reports, the exact significance of the lesions identified
has been poorly studied [14]
[15]. Therefore, we aimed to assess the diagnostic yield of CE in patients with IDA who
had undergone previous thorough work-up. We also aimed to determine whether sex, age,
NSAIDs, blood transfusions, hemoglobin (Hb), or ferritin levels could be predictors
of findings.
Patients and methods
This is a retrospective cohort study of all consecutive patients undergoing CE for
the evaluation of unexplained IDA conducted at the Gastroenterology Department of
the “Hospital de Clínicas”, a tertiary referral center in Uruguay. All patients with
unexplained IDA who had undergone at least one negative upper and lower endoscopy
between January 2008 and May 2016 were included. IDA was defined, according to the
laboratory indices, as hemoglobin (Hb) less than 13 g/dL in men and 12 g/dL in women,
along with a serum ferritin level less than 50 ng/mL. Negative endoscopy was considered
when no lesions that explained IDA were found or when found, they did not explain
the severity of IDA.
Exclusion criteria included patients aged below 18 years, Crohn’s disease, pregnancy,
with gynecological causes for IDA or celiac disease. Gynecological causes were excluded
by clinical history and gynecological exam. Celiac disease was defined as the presence
of at least one positive antibody (EmA or TTG) and/or any grade of villous atrophy
and increased intraepithelial lymphocytes in duodenal biopsies.
The following variables were collected in a standardized data sheet: age, sex, NSAID
consumption, blood transfusion requirement, hemoglobin, ferritin, and results of upper
and lower endoscopy. A positive diagnostic yield was considered to be one or more
lesions that were believed to be a definitive cause of IDA (P2 or P3 of the Saurin
classification).
Small-bowel capsule endoscopy procedure
Informed consent was obtained from all participants, and CE was carried out using
the Capsule Endoscopy System (Medtronic, United States). PillCamSB, SB2, and SB3 capsules
were used, and videos were analyzed using Rapid Reader, version 5, 7, and 8, respectively
(Medtronic, United States). As none of the patients included were suspected of having
a small-bowel stricture or known or suspected Crohn’s disease, a patency capsule (dissolvable
radio-opaque capsule) was not performed before CE. On the afternoon before the study,
all patients had only liquids and some patients (mostly the first studies performed)
also received a prep solution (Precolsur, Celsius, Uruguay) containing sodium sulfate,
sodium bicarbonate, sodium chloride, and potassium chloride. All patients fasted the
night before the procedure. A sensor array was worn externally and the capsule was
swallowed. Patients were allowed to drink clear fluids 2 hours after swallowing the
capsule and to eat a light lunch 4 hours after. The capsule passed with a normal bowel
movement and was discarded. The sensor recorded images for 8 hours with SB type capsules,
and until the capsule reached the colon (real time viewer) with SB2 or SB3 type capsules.
The images were uploaded as a video and interpreted by a single experienced endoscopist
(CO).
Statistical analysis
A descriptive analysis was performed, with the presentation of absolute frequencies
and relative measures. Our unit of analysis was the endoscopic study, and the same
patient may have needed more than one endoscopic study at different times. For quantitative
variables, measures of central tendency and dispersion were used. The chi-squared
test was used to contrast between nominal and/or ordinal variables.
Significance tests were used to compare continuous variables, and previous verifications
of the conditions of application of the tests. Normality tests were performed to study
the distribution of these variables. The significance level in all cases was 0.05. When
the variables did not follow a normal distribution, non-parametric tests were used
to study their association with angiodysplasia and the presence of findings (Mann-Whitney
U test). The analysis of age was made as a continuous variable and also in two groups
(older and younger than 50 years). This cut-off was selected following Muhammad et
al. [13]. A multivariate analysis of the variables whose P value ≤ 0.25 was performed. A logistic regression was performed using as dependent
variables: presence of findings and the presence of angiodysplasia.
Results
In total, 120 CE were performed in 118 patients. CE completion rate was 96.7 %, defined
by visualization of the cecum on the recording. Four studies (3.3 %) were defined
as incomplete, in one of them CE was stopped after 3 hours due to an ulcerated stenosis
that was later crossed through; in another patient CE stopped at a submucosal lesion
and stayed there for 5 hours but was defecated 24 hours later. In the remaining two
cases, the sensor was removed 8 hours after being swallowed (misreading of cecal image
at data recorder screen). There were no complications. Of the total, 51 CE were performed
with PillCamSB, 51 CE with PillCamSB2, and 18 with PillCamSB3.
The mean patient age was 58.5 years, ranging from 18 to 86 years. Eighty-seven patients
(72.5 %) were older than 50 years. Eighty-two were women (68.3 %). Fifty-four patients
(45 %) used NSAIDs and 54 patients (45 %) had received blood transfusions ( [Table 1]). Only 10 patients (8.3 %) were receiving anticoagulant therapy. The mean Hb level
was 9 g/dL (range 4 – 11.9 g/dL). Twenty-one patients (17.5 %) had severe anemia (Hb ≤ 7 g/dL)
and 99 patients (82.5 %) had mild to moderate anemia. The mean ferritin level was
15.7 ng/dL (range 1.5 – 48.0 ng/dL).
Table 1
Characteristics of the total population with IDA and those with and without positive
findings.
|
Total n (%)
|
Positive findings
(n = 60)
|
Without positive findings
(n = 60)
|
P value
|
|
Sex: male
|
38 (32 %)
|
27 (45 %)
|
11 (18 %)
|
0.02
|
|
Age > 50 years
|
87 (72.5 %)
|
50 (83 %)
|
37 (62 %)
|
NS
|
|
Anemia, g/dL
|
|
|
99 (83 %)
|
45 (75 %)
|
54 (90 %)
|
NS
|
|
|
21 (17 %)
|
15 (25 %)
|
6 (10 %)
|
NS
|
|
Ferritin, ng/dL
|
|
|
68 (57 %)
|
35 (58 %)
|
33 (55 %)
|
NS
|
|
|
37 (31 %)
|
18 (30 %)
|
19 (32 %)
|
NS
|
|
|
15 (12 %)
|
7 (12 %)
|
8 (13 %)
|
NS
|
|
NSAID: yes
|
54 (45 %)
|
30 (50 %)
|
24 (40 %)
|
NS
|
|
Transfusion: yes
|
54 (45 %)
|
33 (55 %)
|
21(35 %)
|
0.03
|
|
Anticoagulant therapy: yes
|
10 (8 %)
|
4 (7 %)
|
6 (10 %)
|
NS
|
|
Capsule reached cecum
|
116 (97 %)
|
57 (95 %)
|
59 (98 %)
|
NS
|
IDA, iron deficiency anemia; NS, not significant.
Positive findings were present in 60 patients with IDA(50 %). The distribution of
positive findings is shown in [Fig. 1] and images of their spectrum are shown in [Fig. 4]. Two patients (3.3 %) had more than one lesion. Active small-bowel bleeding was
found in 5 patients (8.3 %), 3 of them without identification of a clear lesion. Three
patients (5 %) had positive findings outside the small bowel (colonic active bleeding,
pyloric ulcer, and cecal angiodysplasia) and were included as positive findings for
statistical analysis.
Fig. 1 Distribution of positive findings in 60 patients with IDA.
In the 3 patients with bleeding lesions outside the small bowel, the following interventions
were performed: a second colonoscopy was performed in the patient with colonic active
bleeding and a previously missed colon cancer was diagnosed; upper gastrointestinal
endoscopy was performed to take gastric and pyloric ulcer biopsies and a colonoscopy
was performed to treat cecal angiodysplasias with argon plasma coagulation (APC).
Demographic, clinical, and laboratory data of patients with and without positive findings
are shown in [Table 1].
Male sex (P = 0.02) and transfusion requirement (P = 0.028) were statistically significantly associated with positive findings.
In a multivariable analysis using a logistic regression model, it was verified that
male sex (OR 3.93; 95 %CI 1.57 – 9.86), age (OR 1.03; 95 %CI 1.00 – 1.06), and hemoglobin
(OR 0.73; 95 %CI 0.57 – 0.94) were the variables having an independent effect on the
probability of obtaining positive findings in patients with IDA ([Fig. 2] and [Fig. 3]). Age older than 50 years (OR 14.05; 95 %CI 1.69 – 116.23) and male sex (OR 3.63;
95 %CI 1.29 – 10.17) were the variables which increased the risk of diagnosing angiodysplasia.
Fig. 2 Box plots of Hb levels in patients with and without positive findings.
Fig. 3 Box plots of age in patients with and without positive findings.
Fig. 4 Spectrum of positive findings. a Active bleeding; b malignant tumor-like lesion; c angiodysplasia; d ulcerated stenosis; e ulcerated submucosal lesion; f multiple phlebectasia.
Discussion
In this study, the diagnostic yield of CE for IDA was 50 %, which is in agreement
with data from previous studies [17]. Our study was a phase IV clinical study, confirming the high diagnostic yield of
CE found in previous phase II clinical trials [18]
[19]. Most studies on CE had been performed in the USA, Europe, and the Far East. Our
study is the largest on CE in IDA in South America. Therefore, our data also help
define the epidemiological and etiological characteristics of IDA in additional geographic
areas.
The majority of positive findings in our study were due to angiodysplasia and non-specific
inflammation/ulceration, which is consistent with those found in similar studies from
USA and Europe but not those from Asia. In Asia, reports of small-bowel tumors, inflammatory
lesions, and multifocal stenosis enteritis are more common [20]
[21]
[22]
[23]. It has been recognized that CE is more sensitive than other modes of small-bowel
imaging for the detection of vascular malformations [24].
All lesions included were considered to be highly relevant lesions (P2 of the Saurin
classification). The “Saurin classification of small-bowel lesions on CE” allows risk
stratification of patients [8]
[9]. Highly relevant lesions (P2) had previously been validated by two blinded endoscopists
in 100 % of cases, compared with 73 % and 27 % of cases for intermediate- (P1) and
low-relevance (P0) lesions, respectively [25]. There was a necessity for therapy in a significantly higher number of P2 lesions
(61 %), compared with P1 or P0 lesions (23 %) [25].
In this series, in addition to angiodysplasia, one case of multiple phlebectasias
was included as vascular lesions. These are rare benign vascular anomalies that may
present with occult or overt small-bowel bleeding. We have already reported two cases
of these lesions, even with active bleeding during the procedure [26].
Due to the clinical relevance, in the 50 % diagnostic yield, we included 4.4 % of
findings outside the small bowel. Our lesion miss rate outside the small bowel is
much lower than those reported in previous studies, which found that patients referred
for small-bowel endoscopic imaging had a significant number of lesions within the
reach of upper and lower endoscopy [27]
[28]
[29]. Fry et al. [30] found 24.3 % of sources of gastrointestinal bleeding outside the small bowel with
double balloon enteroscopy (DBE). Elijah et al. [31] also found that up to 38.8 % of patients undergoing CE for OGIB had significant
lesions of the upper gastrointestinal tract such as gastric antral vascular ectasias
(GAVE), gastric ulcers, and Cameron’s lesions. These lesions were evidently missed
during electrocardiography (EGC) and colonoscopy performed before DBE and/or CE. In
our series, the lesions missed were: a pyloric ulcer whose histology revealed an adenocarcinoma,
a colonic active bleeding due to a small colorectal cancer (CRC), and a cecal angiodysplasia.
We believe that having performed the previous upper and lower endoscopies in our tertiary
center may have resulted in a lower number of missed lesions. Indeed, despite the
reported high percentages of lesions outside the small bowel and the high cost of
the CE, most guidelines still do not routinely support a second-look endoscopy prior
to small-bowel CE [32]
[33]
[34].
There is limited published data in which clinical and laboratory factors predict the
ability of CE to detect pathology in IDA. In a large series of patients with OGIB
of which 74 % had IDA, Siduh et al. [33] found that increasing age, use of warfarin, and liver comorbidity were significant
factors which predicted a higher yield with CE. More recently, in 221 patients with
IDA, Koulaouzidis et al. [17] found that age, prior blood transfusions, use of anticoagulants, and male sex were
factors associated with a higher diagnostic yield for all types of small-bowel pathology.
Similar to Koulaouzidis et al., we found that increasing age was associated with a
higher detection rate of small-bowel angiodysplasia.
All these results are partially in line with ours. Nevertheless, and to the best of
our knowledge, ours is the first study to show low hemoglobin levels as a predictor
of findings in patients with IDA. Furthermore, in contrast to most previous studies,
our study focused exclusively on patients with IDA. We carefully select the IDA group,
excluding celiac disease diagnosis in all patients by antibody detection and/or duodenal
biopsies. Thus, this study demonstrated that, in IDA patients, after an intensive
previous work-up, capsule endoscopy would be a very useful technique to find lesions
in elderly patients with severe anemia.
Our study has several potential limitations. First, it is a retrospective study and
thus prone to the potential deficits of such design. However, we carefully collected
all variables including all videos of the procedures to ensure a complete database.
Second, we are a tertiary center and our findings may not be reproduced by others.
Third, as most of the patients were referred exclusively for CE, some data is missing.
We only had data for H. pylori investigation in 29 %. As H. pylori infection can be a cause of unexplained iron-deficiency anemia, this could be a limitation.
Also, data on iron supplementation was not available. Fourth, we included nonspecific
diagnosis as potentially explaining the bleeding leading to IDA. But until now there
is no clear definition of which lesions result in bleeding and we based our definitions
of bleeding lesions on existing classifications and guidelines. Nonetheless, further
studies will be mandatory to clearly define which lesions results in bleeding IDA.
Finally, patients were referred only for CE from different healthcare centers and
the subsequent medical treatment depended on the attending physician and the availability
of resources in their health service. It was not possible to perform a complete follow-up
of these patients so an analysis of outcomes was not included in the study. It would
be interesting to perform a study on this variable in the future.
In conclusion, CE is a useful technique in patients with IDA. However, as shown by
this study, to improve its yield, it is necessary to select patients carefully. Male
sex, old age, and low hemoglobin levels increased the risk of positive findings in
this group of patients. The risk of diagnosing angiodysplasia increases with age.
Furthermore, large multicenter, phase IV studies are needed to better define the lesions
causing IDA and the outcomes of patients with IDA subjected to CE.
