Introduction
Prevalent worldwide, Helicobacter pylori infection can increase risk of gastric cancer and is associated with the occurrence
of a variety of human diseases, including lymphoma and autoimmune disorders [1]
[2]. H. pylori infection is most common in the stomach, and patients with it may have heartburn,
epigastric pain, dyspepsia, and other nonspecific symptoms, which are easily overlooked
in a majority of cases [3]. Diagnosis of H. pylori infection can be made based on judgment of the clinical manifestation, detection
of a specific antibody in serum, and based on results of a C13 breath test [4]
[5]. In addition, endoscopic examination may provide useful evidence for confirming
the diagnosis, and tissue from biopsy performed under endoscopy can be sent for a
rapid urease test and pathology.
Inflammation, atrophy, and intestinal metaplasia of the gastric mucosa caused by presence
of H. pylori have different appearances under endoscopy [6]
[7]. However, changes in mucosa observed under white light endoscopy (WLE) are not distinctive,
and the stages of H. pylori infection in the stomach rarely have been reported. Linked color imaging (LCI) is
a color-enhanced technique, which enables easier identification of mucosal lesions
during endoscopy by significantly improving color contrast between lesions and normal
tissue [8]
[9]. Our previous data validated that LCI can greatly improve endoscopic diagnostic
accuracy of gastrointestinal mucosal lesions [10]
[11], while the role of LCI endoscopy in evaluating and staging H. pylori infection in the stomach has not been clearly investigated. In the current study,
we conducted a randomized controlled clinical trial (RCT) to compare the diagnostic
efficiency of WLE and LCI for H. pylori infection and further analyzed the endoscopic characteristics of different stages.
The findings will help optimize current management of patients with H. pylori infection.
Patients and methods
Patients
Consecutive adult patients who had indications and underwent gastroduodenoscopy between
September 2016 and February 2017 were enrolled from five endoscopic centers: Affiliate
Hospital to Academy of Military Medical Science (307 Hospital; the Fifth Clinical
Center of Chinese PLA General Hospital), the People’s Hospital of Guangxi Zhuang Autonomous
Region, Shanghai Tenth People’s Hospital, the Second Affiliated Hospital of Soochow
University and the Sixth Affiliated Hospital of Sun Yat-sen University. Patients who
had thrombocytopenia (platelet count < 50,000 /µL) or elevated International Normalized
Ratio (INR > 1.5), hemodynamic instability, pregnancy and lactation, had ever undergone
H. pylori eradication therapy, or were unable or unwilling to give an informed consent were
excluded. Demographic and clinical data were retrieved from the computerized database.
All patients gave informed consent. The study was approved by the Ethics Committee
of Affiliate Hospital to Academy of Military Medical Science. This study was registered
at ClinicalTrials.gov (ClinicalTrials.gov ID: NCT02724280). We had access to the study
data and had reviewed and approved the final manuscript.
Study design
The flow chart of this RCT is shown in Supplemental Fig. 1. McNemar test was used to calculate the sample size by using SPSS software. The sample
size was calculated based on a probability of 0.8 and α error of 0.05. The actual
expected sensitivity of LCI and WLE for diagnosing H. pylori was around 80 % and 45% according to our previous study and experience [10]
[12]. The required sample size was set to 250 patients. 253 patients were enrolled. All
patients were randomized into Group A (n = 127) or Group B (n = 126) at a 1:1 ratio
using the random number method. The random sequence was generated by Excel and concealed.
In Group A, patients received WLE followed by LCI endoscopic examination. In Group
B, patients received LCI, followed by WLE mode for gastroscopy. This study was double-blind,
and patients and endoscopists were not informed of the grouping information, which
was achieved by the participation of a third endoscopist who intubated and withdrew
the endoscopy.
Procedures
All endoscopic procedures were performed under anesthesia. EG-L590WR endoscopes equipped
with the LASEREO endoscopic system (FUJIFILM Co., Tokyo, Japan) were used. WLE and
LCI examinations were conducted by two different endoscopists who had comparable skills
and experience with endoscopic examinations. No magnification was applied. The biopsies
were completed by a third endoscopist for the purposes of conducting a rapid urease
test and histological examination. According to the Sydney System [13], two biopsies from lesser and greater curvature (around 8 cm from the cardia), two
biopsies from the pre-pyloric antrum (3 – 5 cm from the pylorus), and one biopsy from
the gastric angle were taken in each case. H. pylori infection was manifested as diffuse or spotty red with nodularity and enlarged/tortuous
folds under WLE, and red mixed with heterogeneous purple under LCI [12]
[13]. To minimize the inconsistency of diagnostic accuracy among different endoscopists,
typical endoscopic images of LCI and WLE were distributed to train the participating
endoscopists.
Outcomes
Diagnosis of H. pylori infection was made by positivity in a rapid urease test and/or histological examination
of any of the biopsies in one patient. If rapid urease test and the histological examination
were all negative in all the biopsies of one patient, it was judged as H. pylori infection-negative. Overall diagnostic efficacy evaluation was performed and the
receiver operating characteristic (ROC) curve was drawn. The sensitivity, specificity,
negative predictive value (NPV), positive predictive value (PPV), and Youden index
of WLE and LCI in Groups A and B were calculated in the diagnostic test, respectively.
H. pylori infection in the stomach was comprehensively determined by histological analysis
or rapid urease test, which was administered for all patients. Overall diagnostic
efficacies of WLE and LCI in diagnosing H. pylori infection in the stomach were compared. H. pylori infection in different parts of the stomach was analyzed for profiling the stages.
Kappa values were calculated for the correlation analysis of pathological staging
with WLE staging and LCI staging.
Statistical analysis
All statistical analyses were conducted using SPSS 17.0 software (SPSS Inc., Chicago,
Illinois, United States). Continuous and categorical data were presented as means
(range) and percentages (%), respectively. Data were analyzed using independent t-test and Chi-square or Fisher’s exact probability test, if applicable. Two-tailed
P values less than 0.05 were considered statistically significant.
Results
Demographic and clinical characteristics
There were 127 patients in Group A and 126 patients in Group B. There were no significant
differences in age, gender, or indications for endoscopy between the two groups (all
P > 0.05, [Table 1]). Rapid urease test, histology, WLE, and LCI were applied for evaluating H. pylori infection in the stomach. The detection rate for H. pylori infection, which was around 27.0 % to 50.4 %, was comparable in Group A and Group
B (all P > 0.05, [Table 1]). A total of 107 patients with H. pylori infection were finally diagnosed by rapid urease test and/or histology.
Table 1
Demographic and clinical characteristics.
|
Group A (n = 127)
|
Group B (n = 126)
|
t/x2
|
P value
|
|
Age, years, mean (range)
|
47.20 (19 – 76)
|
49.66 (19 – 72)
|
1.540
|
0.125
|
|
Gender, n (%)
|
|
|
0.102
|
0.750
|
|
Male
|
66 (52.0)
|
68 (54.0)
|
|
|
|
Female
|
61 (48.0)
|
58 (46.0)
|
|
|
|
Indications, n (%)
|
|
|
7.178
|
0.127
|
|
Heartburn
|
19 (15.0)
|
17 (13.5)
|
|
|
|
Abdominal distension
|
21 (16.5)
|
32 (25.4)
|
|
|
|
Dyspepsia
|
15 (11.8)
|
14 (11.1)
|
|
|
|
Epigastric pain
|
23 (18.1)
|
31 (24.6)
|
|
|
|
Other
|
49 (38.6)
|
32 (25.4)
|
|
|
|
H. pylori infection evaluation, n (%)
|
|
Rapid urease test
|
|
|
0.005
|
0.941
|
|
Positivity
|
54 (42.5)
|
53 (42.1)
|
|
|
|
Negativity
|
73 (57.5)
|
73 (57.9)
|
|
|
|
Histological examination
|
|
|
0.001
|
0.970
|
|
H. pylori infection
|
34 (26.8)
|
34 (27.0)
|
|
|
|
No H. pylori infection
|
93 (73.2)
|
92 (73.0)
|
|
|
|
WLE diagnosis
|
|
|
0.708
|
0.400
|
|
H. pylori infection
|
40 (31.5)
|
46 (36.5)
|
|
|
|
No H. pylori infection
|
87 (68.5)
|
80 (63.5)
|
|
|
|
LCI diagnosis
|
|
|
0.674
|
0.412
|
|
H. pylori infection
|
64 (50.4)
|
57 (45.2)
|
|
|
|
No H. pylori infection
|
63 (49.6)
|
69 (54.8)
|
|
|
WLE, white light endoscopy; LCI, linked color imaging
Diagnostic efficacy for H. pylori infection
H. pylori infection in the stomach was independently evaluated by WLE followed by LCI in Group
A, and by LCI followed by WLE in Group B ([Fig. 1]). The overall diagnostic accuracy of WLE and LCI in Group A (n = 127) were 31.5 %
(n = 40) and 50.4 % (n= 64), respectively (P = 0.001), while the overall diagnostic accuracy of WLE and LCI were 36.5 % (n = 46)
and 49.2 % (n = 62) in Group B (n = 126; P = 0.029). In addition, the ROC curve of LCI and WLE for H. pylori infection was analyzed (Supplemental Fig. 2). The results were consistent in indicating that LCI had higher diagnostic efficacy
for H. pylori infection than did WLE. In both groups, LCI had higher sensitivity, specificity,
NPV, PPV, and Youden index than did WLE. The AUC for WLE and LCI was 0.542 and 0.850
in Group A, and 0.559 and 0.891 in Group B ([Table 2]).
Fig. 1 Typical endoscopic images for H. pylori infection in the stomach.
Table 2
Diagnostic efficacy evaluation of WLE and LCI for H. pylori infection.
|
Group A (n = 127)
|
Group B (n = 126)
|
|
WLE
|
|
AUC (P value)
|
0.542 (0.416)
|
0.559 (0.256)
|
|
Sensitivity, % (95 % CI)
|
35.8 (27.5 – 44.1)
|
43.4 (34.7 – 52.1)
|
|
Specificity, % (95 % CI)
|
72.6 (64.8 – 80.4)
|
68.5 (60.4 – 76.6)
|
|
NPV, % (95 % CI)
|
48.7 (40.0 – 57.4)
|
50.0 (41.3 – 58.7)
|
|
PPV, % (95 % CI)
|
60.9 (52.4 – 69.4)
|
62.5 (54.0 – 71.0)
|
|
Youden index, % (95 % CI)
|
8.4 (3.6 – 13.2)
|
11.9 (6.2 – 17.6)
|
|
LCI
|
|
AUC (P value)
|
0.850 (< 0.001)
|
0.891 (< 0.001)
|
|
Sensitivity, % (95 % CI)
|
90.6 (85.5 – 95.7)
|
90.6 (85.5 – 95.7)
|
|
Specificity, % (95 % CI)
|
79.5 (72.5 – 86.5)
|
87.7 (82.0 – 93.4)
|
|
NPV, % (95 % CI)
|
76.2 (68.8 – 83.6)
|
84.2 (77.8 – 90.6)
|
|
PPV, % (95 % CI)
|
92.1 (87.4 – 96.8)
|
92.8 (88.3 – 97.3)
|
|
Youden index, % (95 % CI)
|
70.1 (62.1 – 78.1)
|
78.3 (71.1 – 85.5)
|
WLE, white light endoscopy; AUC, area under the curve; CI, confidence interval; NPV,
negative predictive value; PPV, positive predictive value; LCI, linked color imaging
LCI could determine stage of H. pylori infection in the stomach
Distribution of H. pylori infection and morphological changes in the stomach mucosa may evolve during the course
of disease. This kind of evidence may help determine the stage, about which little
is known. Thus, we analyzed data from 107 patients with confirmed H. pylori infection and investigated that infection in the gastric body and antrum together
with pathological changes ([Table 3]). According to previous reports [14]
[15]
[16], H. pylori infection commonly starts in the gastric antrum, migrates into the gastric body,
and then spreads throughout the entire stomach (gastric body and antrum). During the
course of infection, incidences of atrophy and intestinal metaplasia in gastric antrum
were noticeably increased.
Table 3
H. pylori infection in different parts of the stomach.
|
No H. pylori infection (n = 146)
|
H. pylori infection in gastric antrum (n = 64)
|
H. pylori infection in gastric body (n = 9)
|
H. pylori infection in gastric body and antrum (n = 34)
|
|
Gastric body, n (%)
|
|
Atrophy
|
16 (11.0)
|
0 (0.0)
|
0 (0.0)
|
0 (0.0)
|
|
Intestinal metaplasia
|
26 (17.8)
|
3 (4.7)
|
1 (11.1)
|
2 (5.9)
|
|
Gastric antrum, n (%)
|
|
Atrophy
|
13 (8.9)
|
3 (4.7)
|
2 (22.2)
|
4 (11.8)
|
|
Intestinal metaplasia
|
31 (21.2)
|
11 (17.2)
|
2 (22.2)
|
11 (32.4)
|
We further hypothesized that H. pylori infection may have four stages: 1) Stage 1, H. pylori infection in the antrum without intestinal metaplasia; 2) Stage 2, H. pylori infection in the antrum and body without intestinal metaplasia; 3) Stage 3, intestinal
metaplasia in the antrum and H. pylori infection in the body; and 4) Stage 4, both intestinal metaplasia and H. pylori infection in the antrum and body. However, in some patients, infection could not
be staged and they were thus classified as Stage X (indeterminable). Different stages
had different endoscopic appearances ([Fig. 2]). LCI staging yielded greater consistency with pathological staging than did WLE
staging (Kappa value 0.772 vs. 0.516, [Table 4]). Of 16 patients with LCI Stage X, 12 were pathological Stage X and four were pathological
Stage 4. Under LCI mode, the inflammation appeared as diffusive red, mucosal atrophy
appeared as white, and intestinal metaplasia appeared as purple (Supplemental Table 1). The LCI observations were closely correlated with the pathology ([Fig.3]). During LCI endoscopy, purple was associated with intestinal metaplasia (P = 0.025) and diffusive red was associated with inflammation (P = 0.016). As intestinal metaplasia is a typical mucosal change in H. pylori infection, presence of red ringed with purple under LCI can predict H. pylori infection in the stomach (P = 0.022). This provides a rapid and convenient evaluation for screening the stomach.
However, no such correlations were found under WLE mode.
Fig. 2 Typical endoscopic images of H. pylori infection-associated gastritis at different stages. In Stage 1, H. pylori infection was observed mainly in the gastric antrum; it can gradually spread into
the gastric body in Stage 2. In Stage 3, along with H. pylori-associated inflammation in the mucosa, intestinal metaplasia was found in the antrum.
Intestinal metaplasia characterized by presence of goblet cells in the epithelial
layer was the main feature of the mucosa of the gastric body in Stage 4; the inflammation
can be subtle.
Table 4
Stages in the course of H. pylori infection in the stomach.
|
WLE staging, n
|
LCI staging, n
|
Total, n
|
|
Stage 1
|
Stage 2
|
Stage 3
|
Stage 4
|
Stage X
|
Stage 1
|
Stage 2
|
Stage 3
|
Stage 4
|
Stage X
|
|
Pathological staging, n
|
Stage 1
|
30
|
2
|
0
|
0
|
0
|
28
|
4
|
0
|
0
|
0
|
32
|
|
Stage 2
|
8
|
15
|
2
|
1
|
0
|
4
|
22
|
0
|
0
|
0
|
26
|
|
Stage 3
|
0
|
1
|
9
|
2
|
0
|
0
|
0
|
10
|
2
|
0
|
12
|
|
Stage 4
|
3
|
1
|
1
|
7
|
11
|
0
|
0
|
3
|
16
|
4
|
23
|
|
Stage X
|
0
|
2
|
1
|
5
|
6
|
0
|
0
|
0
|
2
|
12
|
14
|
|
Total, n
|
41
|
21
|
13
|
15
|
17
|
32
|
26
|
13
|
20
|
16
|
107
|
|
Kappa = 0.516, P < 0.001
|
Kappa = 0.772, P < 0.001
|
|
WLE, white light endoscopy; LCI, linked color imaging
Fig. 3 LCI findings were closely associated with the pathology.
Discussion
Gastric mucosa infected with H. pylori can undergo a series of pathological changes, including inflammation, atrophy, and
intestinal metaplasia [17]
[18]. A single-center study reported that H. pylori-associated gastritis was connected with endoscopic modifications and histopathology
in a prospective cohort of children [19]. It was also shown that the conventional endoscopy features can be used to diagnose
H. pylori [20]
[21]
[22], but the efficiency of WLE was relatively poor [23]
[24]. Endocytoscopy may facilitate in vivo gastric mucosal histopathology [25], but it is time-consuming and some patients cannot tolerate it. However, whether
endoscopy can be used to predict these kinds of mucosal lesions, a question that has
been rarely explored in a randomized setting, remains unclear [2]. In the current study, we compared the diagnostic efficacy of WLE and LCI in a multicenter
RCT, and for the first time propose that observation of red ringed with purple under
LCI mode could predict diagnosis of H. pylori infection in the stomach.
LCI is a recently developed endoscopic technique that incorporates image post-processing
into the laser endoscopic system [26]. LCI makes it easier for endoscopists to identify color changes in the gastrointestinal
mucosa by enhancing color contrast. Our previous study demonstrated that LCI can improve
the efficiency and accuracy of differentiating gastrointestinal mucosal lesions and
improve the performance of target biopsies [10]. Calculation of pixel brightness based on a red-green-blue color model may be introduced
as an objective evaluator for analyzing typical endoscopic images. Consistently, we
also correlated color features under LCI with pathology. Statistical analysis validated
that a purple color observed under LCI could predict existence of intestinal metaplasia,
and red predicts observation of inflammation. However, analysis of the correlation
between atrophy and a white color under LCI was not conducted, due to the small number
of patents with atrophy. Sensitivity of LCI for diagnosis of H. pylori infection was similar to previous studies, but that of WLI was quite low compared
to the previous studies [8], and the possible reasons may be the different clinical practice in different medical
institutions and the selected sample.
H. pylori infection, which is considered a precancerous lesion, is an independent risk factor
for gastric cancer [27]. Gastroduodenoscopy is well known as an effective strategy for early detection of
gastric mucosal lesions [28]
[29]
[30]. Our staging system was mainly based on comprehensive analysis of endoscopic images
in different locations. Our data profiled four stages of H. pylori infection in the stomach. In Stage 1, H. pylori infection was observed mainly in the gastric antrum; it can gradually spread into
the gastric body in Stage 2. In Stage 3, along with H. pylori-associated inflammation in the mucosa, intestinal metaplasia was found in the antrum.
Intestinal metaplasia characterized by presence of goblet cells in the epithelial
layer was the main feature of the mucosa of the gastric body in Stage 4; the inflammation
can be subtle. We then further calculated incidence of H. pylori infection in different parts of the stomach, and investigated the respective pathological
changes. Of 107 patients with H. pylori infection in the stomach, there were 64 patients (59.8 %) with H. pylori infection only in the gastric antrum, nine (8.4 %) with H. pylori infection only in the gastric body, and 34 31.2 %) with H. pylori infection in both the gastric body and antrum ([Table 3]). Intestinal metaplasia and atrophy occurred more often in the gastric antrum than
in the gastric body. Typical color changes under LCI were highly consistent with the
pathology (purple for intestinal metaplasia, P = 0.025; diffusive red for inflammation, P = 0.016). In addition, we validated LCI staging by comparing WLE staging in terms
of its correlation with pathological staging. LCI had higher consistency with pathological
staging than did WLE staging. However, 16 patients could not be staged by LCI (LCI
Stage X), 12 of whom could not be staged by pathology due to atypical pathological
changes (pathological Stage X). All of the patients were older than age 30 year, and
no other special demographical and clinical characteristics were detected.
There were limitations in our study. First, patients who had ever undergone H. pylori eradication therapy were excluded. Thus, our staging method could not be used for
them. Second, our conclusion was obtained based on the findings in a patient cohort,
which may be further validated in a large-scale clinical trial. The main limitation
in this study was that the role of LCI in surveillance of patients with H. pylori infection was not examined. Follow-ups will be further investigated in future research.
It was expected that the endoscopic and histological features could help screen for
gastric mucosal changes after eradication of H. pylori infection [31]
[32].
Conclusion
From this multicenter RCT, we conclude that LCI has higher diagnostic efficacy for
evaluating H. pylori infection in the stomach than does WLE, and this new endoscopic technique can be
used as an effective method for quick diagnosis. Correlation analysis with pathology
indicated that red ringed with purple observed under LCI could predict presence of
H. pylori infection. In addition, this staging system can help clarify development of H. pylori infection in the stomach, which might have benefits for clinical management of such
disease.
