Introduction
Colorectal cancer (CRC) is one of the leading causes of cancer-related death worldwide
[1]. Colonoscopy remains the most important tool in the detection and removal of the
precursor lesions of CRC; however, there is a reported adenoma miss rate of about
20 %–30 % by conventional colonoscopy using white-light imaging (WLI) [2]
[3]. In particular, sessile serrated lesions (SSLs) are more difficult to detect than
conventional adenomas, because they are often flat, inconspicuous, and covered with
mucus [4]
[5]. It has been reported that SSLs may contribute to approximately 20 %–30 % of sporadic
CRCs [4], and missed SSLs during colonoscopy may play an important role in the development
of interval CRC [6]. These findings have underlined the importance of improving the detection efficacy
for SSLs.
The developments in image-enhanced endoscopy technologies have allowed better diagnostic
capability and improved identification of gastrointestinal lesions [7]. Among them, narrow-band imaging (NBI) is the most widely used modality, which exploits
blue and green wavelengths to emphasize the capillary pattern and the surface of the
mucosa, while linked-color imaging (LCI) is a novel technology that relies on wavelength
optimization of three colors (red, green, and blue) to make the lesions appear fuller
[8]. Previous studies have investigated the effectiveness of NBI or LCI in the detection
of colorectal lesions in comparison with WLI, some of which have favored NBI or LCI
as being superior to WLI [9]
[10]. Our previous study suggested that LCI could significantly improve colorectal SSL
detection compared with WLI [11]. There has however been no direct comparison between LCI and NBI for colorectal
SSL detection. Whether LCI could serve as a more appropriate modality than NBI for
colorectal SSL detection remains unclear.
Herein, we report the results of our prospective, parallel, randomized controlled
trial (RCT) investigating the superiority between LCI and NBI for colorectal SSL detection.
We hoped to provide reliable evidence to guide clinical practice in the future.
Methods
Study design
This was a single-center, prospective RCT with parallel arms of LCI and NBI colonoscopy
conducted in the Digestive Endoscopy Center of the Shanghai Tenth People's Hospital
in China. The study was conducted according to the Consolidated Standards of Reporting
Trials, and was approved by the ethics committees of the Shanghai Tenth People's Hospital.
Written informed consent was obtained from all participants. All authors had access
to the data and approved the final manuscript.
Sample size calculation and randomization
The sample size calculation was performed using PASS software (version 15.0; NCSS,
USA). According to a recent study that investigated LCI versus NBI for colorectal
polyp detection, the serrated lesion detection rate was 22.1 % by LCI and 34.6 % by
NBI [12]. The required sample size was calculated in an inequality test for two independent
proportions using a two-sided Z test (unpooled), with a significance level of 0.05 and a statistical power of 80 %.
As a result, 212 patients were needed in each group, allowing for a dropout rate of
5 %.
Eligible patients were randomly allocated to the NBI or LCI groups in a 1:1 ratio.
Randomization was carried out using the sealed envelope method.
Patient enrollment
The indications for colonoscopy complied with current guidelines [13]
[14]
[15]. Generally, the indications included screening, diagnostic, and surveillance purposes
when the patients were referred to the endoscopy center. After fulfilling the inclusion
and exclusion criteria, a total of 424 participants were enrolled in the present study
between June and October 2021. The inclusion criteria were: age between 35 and 75
years; no prior history of colonoscopy; agreement to participate in the trial and
to provide signed informed consent. The exclusion criteria were: poor condition and
unable to tolerate or cooperate with the examination; known inflammatory bowel disease,
colorectal cancer or familial adenomatous polyposis; refractory constipation; anticoagulation
or antiplatelet medication that could not be suspended; severe anemia, uncontrolled
infection, or another high risk condition; pregnancy.
Bowel preparation
A split-dose regimen of high volume polyethylene glycol solution plus simethicone
was normally taken by the patients for bowel preparation prior to the colonoscopy.
Remedial bowel preparation with a further 1–3 liters of polyethylene glycol solution
was suggested in patients with poor bowel preparation, and colonoscopy was performed
afterwards. The quality of the bowel preparation was evaluated according to the Boston
Bowel Preparation Scale (BBPS). Scores of < 6, 6–7, and 8–9 were defined as “inadequate”,
“good,” and “excellent” bowel preparation, respectively.
Colonoscopy procedures
The colonoscopy procedures were performed with high definition colonoscopes using
a LASEREO 7000 or ELUXEO 7000 system (Fujifilm) in the LCI arm and an EVIS LUCERA
ELITE 290 system (Olympus) in the NBI arm. Once cecal intubation had been confirmed,
the endoscope was withdrawn and endoscopists inspected the mucosa under the specified
mode: LCI or NBI. The same modality, LCI or NBI, was adopted in both the insertion
and withdrawal phase.
The withdrawal procedures were at the discretion of the endoscopists. Cold snare resection
was immediately performed for diminutive lesions (diameter ≤ 5 mm) and a limited number
of lesions (usually < 5). Patients with larger lesions (diameter > 5 mm) or more than
five lesions were recommended for hospitalization for scheduled endoscopic resection.
All of the colonoscopies were performed by seven endoscopists, four of whom were experienced,
having performed > 500 colonoscopies, and three of whom were non-experienced, having
performed < 200 colonoscopies.
Histopathology
All resected specimens were fixed in 10 % formalin and the histopathological evaluation
adhered to a standardized process. The assessments were conducted by two individual
pathologists who were specialists in colorectal lesion diagnosis and blinded to the
patient allocation. The diagnosis of SSLs complied with the criteria of the most recent
World Health Organization guidelines [16]. By definition: one unequivocal architecturally distorted crypt was sufficient for
the diagnosis of an SSL. Training for SSL diagnosis was also performed before initiation
of the study. If there was doubt with regards to a lesion, a multidisciplinary discussion
including endoscopists and pathologists was held to determine the final diagnosis.
End points and definitions
The primary end point was the SSL detection rate (SDR), defined as the proportion
of colonoscopies with at least one SSL detected. Other end points included the polyp
detection rate (PDR), adenoma detection rate (ADR), the number of polyps or adenomas
per patient, the detection rate of diminutive (diameter ≤ 5 mm) or flat lesions, the
number of diminutive (diameter ≤ 5 mm) or flat lesions per patient, the advanced ADR,
SSL with dysplasia (SSLD) detection rate, and adenoma miss rate. The adenoma miss
rate was defined as the proportion of patients with an adenoma detected during the
treatment colonoscopy than was not identified during the first colonoscopy. The withdrawal
time was defined as the total inspection time, excluding the mucosa rinsing, diagnostic,
and therapeutic time.
Statistical analysis
Continuous variables were expressed as mean (SD) and analyzed using Student’s t test. The comparison of categorical variables was conducted using the chi-squared
test. Multivariate analysis was conducted using the stepwise logistic regression model
for variable selection and independent factor identification. Statistical significance
was defined as a two-sided P value ≤ 0.05. All statistical analyses were performed using SPSS software (version
23.0; SPSS Inc.).
Results
Study population and clinical characteristics
The flow diagram for patient selection is shown in Fig. 1s, see online-only Supplementary
material. After patients with bowel obstruction due to advanced CRC or in whom the
cecum was not reached for other reasons had been excluded, along with those who had
no histology specimens either from biopsy or resection, a total of 406 patients were
finally included, with 204 in the LCI arm and 202 in the NBI arm. The detailed clinical
characteristics are presented in [Table 1]. There was no significant difference between the two arms in terms of age, sex,
body mass index (BMI), indication for colonoscopy, family history, or operator experience.
Table 1
Comparison of the clinical characteristics of the patients in the linked-color imaging
(LCI) and narrow-band imaging (NBI) groups.
Variables
|
Total (n = 406)
|
LCI group (n = 204)
|
NBI group (n = 202)
|
P value
|
Age, years, mean (SD)
|
53.6 (11.1)
|
53.8 (11.7)
|
53.4 (10.5)
|
0.73
|
Sex, n (%)
|
0.85
|
|
173 (42.6)
|
86 (42.2)
|
87 (43.1)
|
|
233 (57.4)
|
118 (57.8)
|
115 (56.9)
|
BMI, mean ± SD
|
23.8 (3.1)
|
23.6 (3.0)
|
24.0 (3.1)
|
0.13
|
Indication for colonoscopy, n (%)
|
0.92
|
|
210 (51.7)
|
105 (51.5)
|
105 (52.0)
|
|
196 (48.3)
|
99 (48.5)
|
97 (48.0)
|
Family history, n (%)
|
0.37
|
|
350 (86.2)
|
179 (87.7)
|
171 (84.7)
|
|
56 (13.8)
|
25 (12.3)
|
31 (15.3)
|
Operator, n (%)
|
0.38
|
|
224 (55.2)
|
117 (57.4)
|
107 (53.0)
|
|
182 (44.8)
|
87 (42.6)
|
95 (47.0)
|
BMI, body mass index.
1 The indication “screening” included both primary screening colonoscopy and programmed
screening colonoscopy after positive fecal occult blood test or elevated serum tumor
biomarkers (CEA, CA724, CA199, etc.) among patients without any symptoms.
Colonoscopy outcomes
The detailed colonoscopy outcomes are presented in [Table 2]. There were no significant differences in the intubation time, withdrawal time,
or in bowel preparation between the two arms. The total PDR, ADR, and SDR were 54.2 %,
38.7 %, and 10.8 %, respectively. [Fig. 1] shows two cases of typical SSLs that were detected by LCI and NBI, respectively.
Table 2
Comparison of colonoscopy outcomes between the linked-color imaging (LCI) and narrow-band
imaging (NBI) groups.
Variables
|
Total
|
LCI group (n = 204)
|
NBI group (n = 202)
|
P value
|
Intubation time, mean (SD), minutes
|
4.5 (4.1)
|
4.6 (3.4)
|
4.3 (4.7)
|
0.53
|
Withdrawal time, mean (SD), minutes
|
7.7 (2.2)
|
7.9 (2.3)
|
7.6 (2.1)
|
0.18
|
BBPS, mean (SD)
|
6.5 (1.2)
|
6.5 (1.1)
|
6.5 (1.3)
|
0.84
|
Polyp detection rate
|
54.2 % (220/406)
|
53.9 % (110/204)
|
54.5 % (110/202)
|
0.91
|
Adenoma detection rate
|
38.7 % (157/406)
|
39.7 % (81/204)
|
37.6 % (76/202)
|
0.67
|
SSL detection rate
|
10.8 % (44/406)
|
12.3 % (25/204)
|
9.4 % (19/202)
|
0.36
|
SSL dysplasia detection rate
|
1.5 % (6/406)
|
1.5 % (3/204)
|
1.5 % (3/202)
|
> 0.99
|
Advanced adenoma detection rate
|
9.9 % (40/406)
|
10.3 % (21/204)
|
9.4 % (19/202)
|
0.76
|
Diminutive lesion detection rate
|
40.4 % (164/406)
|
43.6 % (89/204)
|
37.1 % (75/202)
|
0.18
|
Flat lesion detection rate
|
45.1 % (190/406)
|
46.6 % (95/204)
|
43.6 % (88/202)
|
0.54
|
Polyps per patient, n (SD)
|
1.7 (3.4)
|
1.8 (3.8)
|
1.6 (2.9)
|
0.52
|
Adenomas per patient, n (SD)
|
0.9 (2.1)
|
1.1 (2.4)
|
0.8 (1.7)
|
0.24
|
Diminutive lesions per patient, n (SD)
|
0.9 (2.0)
|
1.0 (2.2)
|
0.7 (1.8)
|
0.17
|
Flat lesions per patient, n (SD)
|
1.2 (2.6)
|
1.4 (3.0)
|
1.1 (2.2)
|
0.32
|
Additional adenoma detection rate during treatment colonoscopy
|
25.7 % (19/74)
|
22.0 % (9/41)
|
30.3 % (10/33)
|
0.41
|
BBPS, Boston bowel preparation scale; SSL, sessile serrated lesion.
Fig. 1 Examples of typical sessile serrated lesions (SSLs) detected by: a–c narrow-band imaging (NBI); d–f linked-color imaging (LCI); as seen on: a, d white-light imaging; b, e image-enhanced distant view; c, f image-enhanced close observation.
The SDR was numerically higher in the LCI arm when compared with that in the NBI arm;
however, the difference was not statistically significant (12.3 % vs. 9.4 %; P = 0.36). Similarly, although LCI detected higher rates of diminutive lesions and
flat lesions and larger per-patient numbers of these lesions than NBI did, none of
the differences between the two modalities reached statistical significance. There
were also no significant differences between the two arms in terms of PDR, ADR, SSLD
detection rate, advanced adenoma detection rate, or the number of polyps or adenomas
per patient.
Among the 74 patients who were hospitalized for a treatment colonoscopy, 19 had other
adenomas detected during the treatment colonoscopy, resulting in a total adenoma miss
rate of 25.7 %. The adenoma miss rate was numerically lower in the LCI arm than that
in the NBI arm, but the difference was not significant (22.0 % vs. 30.3 %; P = 0.41).
Factors associated with the SDR
Univariate and multivariate analysis showed that neither LCI or NBI were independent
factors associated with the SDR, whereas the BBPS (odds ratio [OR] 1.35, 95 %CI 1.03–1.76;
P = 0.03), withdrawal time (OR 1.13, 95 %CI 1.00–1.26; P = 0.04), and operator experience (OR 3.73, 95 %CI 1.67–8.32; P = 0.001) were independent factors associated with the SDR ([Table 3]).
Table 3
Multivariate analysis for factors associated with the sessile serrated lesion (SSL)
detection rate.
Factors
|
SSL detection rate
|
Univariate
|
Multivariate
|
P
|
Odds ratio
|
95 %CI
|
P
|
Sex: male vs female
|
0.47
|
NA
|
NA
|
NA
|
Age, years
|
0.35
|
NA
|
NA
|
NA
|
Body mass index
|
0.44
|
NA
|
NA
|
NA
|
Indication: screening vs. diagnostic
|
0.47
|
NA
|
NA
|
NA
|
Family history: yes vs. no
|
0.18
|
NA
|
NA
|
NA
|
Operator: experienced vs. non-experienced
|
< 0.001
|
3.73
|
1.67–8.32
|
0.001
|
Intubation time, minutes
|
0.22
|
NA
|
NA
|
NA
|
Withdrawal time, minutes
|
0.02
|
1.13
|
1.00–1.26
|
0.04
|
BBPS
|
0.01
|
1.35
|
1.03–1.76
|
0.03
|
Group: LCI vs. NBI
|
0.356
|
NA
|
NA
|
NA
|
BBPS, Boston bowel preparation scale; LCI, linked color imaging; NBI, narrow band
imaging; NA, not applicable.
Subgroup analysis
As shown in [Fig. 2], the SDR was numerically higher by LCI than by NBI when the colonoscopies were performed
by experienced operators (17.9 % [21/117] vs. 14.0 % [15/107]; P = 0.42), when the withdrawal time was ≥ 9 minutes (21.7 % [13/60] vs. 18.4 % [7/38];
P = 0.70), and with good, but not excellent, bowel preparation (12.8 % [16/125] vs.
8.3 % [10/121]; P = 0.25); however, none of the differences reached statistical significance. Additionally,
the SDR was almost equivalent for LCI and NBI when the colonoscopies were performed
by non-experienced operators (4.6 % [4/87] vs. 4.2 % [4/95]; P = 0.90), when the withdrawal time was < 9 minutes (8.3 % [12/144] vs. 7.3 % [12/164];
P = 0.74), and with inadequate (5.6 % [2/36] vs. 5.0 % [2/40]; P > 0.99) or excellent bowel preparation (16.3 % [7/43] vs. 17.1 % [7/41]; P > 0.99). The optical visibility of the colorectal mucosa under LCI or NBI in patients
with good or excellent bowel preparation is shown in [Fig. 3].
Fig. 2 Comparisons of sessile serrated lesion (SSL) detection rates between linked-color
imaging (LCI) and narrow-band imaging (NBI) for: a experienced operators (left) and non-experienced operators (right); b for colonoscopies with a withdrawal time < 9 minutes (left) and ≥ 9 minutes (right);
c colonoscopies with inadequate (left), good (middle), and excellent (right) bowel
preparation according to the Boston bowel preparation scale (BBPS). Error bars show
the 95 %CIs.
Fig. 3 The optical visibility of the colorectal mucosa under: a,c,e,g white-light imaging (WLI); b,d narrow-band imaging (NBI); and f, h linked-color imaging (LCI); in patients with: a, b good bowel preparation under WLI and NBI; c, d excellent bowel preparation under WLI and NBI; e, f good bowel preparation under WLI and LCI; g, h excellent bowel preparation under WLI and LCI.
Discussion
NBI has been investigated extensively for its potential in improving colorectal lesion
detection; however, the results from previous studies have been contradictory. Some
earlier RCTs and meta-analyses showed that NBI had no superiority over WLI in detecting
colorectal neoplasms [17]
[18]
[19]
[20]. One major reason for these results could be the low resolution and dark images
achieved with early-generation NBI systems. The new-generation NBI system has been
modified to achieve high resolution images and increased brightness in the colonic
lumen. Recent studies have demonstrated that new-generation NBI has improved the visibility
and detection of polyps and adenomas when compared with WLI [21]
[22]
[23]
[24].
Furthermore, many studies have reported that NBI can significantly improve the detection
of easily overlooked flat or diminutive lesions [25]
[26]. In particular, several studies have added to the debate on the effectiveness of
NBI in detecting colorectal serrated lesions. Rex et al. reported that new-generation
NBI may increase the detection of serrated lesions in the proximal colon when compared
with WLI [27]. Hazewinkel et al. reported that, in patients with serrated polyposis syndrome,
the miss rates were 29 % for WLI and 20 % for NBI, respectively [28]. Recent meta-analyses also demonstrated that NBI detected significantly more proximal,
flat, non-adenomatous (presumed serrated) lesions than WLI [21]
[29]. Taken together, these studies indicated a potential advantage of new-generation
NBI in detecting colorectal SSLs. Similarly, the effectiveness of LCI in detecting
colorectal neoplasms has also been investigated extensively. Some studies, together
with our previous study, have also suggested a certain superiority of LCI over WLI
for SSL detection [10]
[11]
[30]
[31]
[32]
[33]
[34].
So far, only one study has reported a comparison between NBI and LCI for colorectal
polyp detection [12]. The authors performed a prospective, randomized, tandem colonoscopy study and showed
that NBI resulted in a higher PDR, ADR, and percentage of serrated lesions than LCI.
Multivariable analysis indicated that NBI was an independent factor associated with
the PDR. These results provided an early experience in LCI versus NBI for colorectal
lesion detection; however, the involved study population was not equal in the two
groups, with significantly more symptomatic patients in the LCI group than in the
NBI group. The withdrawal time was also significantly longer in the NBI group than
in the LCI group for both the first and second examination. Moreover, the tandem colonoscopies
were performed using the same modality (NBI followed by NBI and LCI followed by LCI).
The true benefit of tandem colonoscopies with the same modality remains unclear, while
tandem colonoscopies using alternative modalities may be more beneficial. Additionally,
a recent systematic analysis suggested that a tandem study is more likely to yield
positive results than a parallel trial, indicating a higher likelihood of bias in
tandem studies [35]. Therefore, it was suggested the results of tandem studies should be accepted only
when validated by parallel trials. Finally, the serrated lesions in this study included
all three classifications of serrated lesions (traditional serrated adenomas, SSLs,
and hyperplastic polyps), without attention being dedicated to SSLs. Whether LCI or
NBI performs better for SSL detection and is worth widespread recommendation remains
unclear, pending a large-scale parallel trial in an equal average-risk patient population.
In the present study, we performed the first prospective parallel RCT to compare LCI
with NBI for the detection of colorectal SSLs among an average-risk population. The
LCI and NBI systems used were both the latest generation, which had high resolution
images and improved brightness. The baseline clinical characteristics and colonoscopy-related
variables showed no significant differences between the two arms. Our main findings
indicated no significant difference in the SDR between LCI and NBI. The differences
in the detection rate and the per-patient numbers of polyps, adenomas, diminutive
lesions, and flat lesions between LCI and NBI were also not statistically significant.
Given the fact that existing evidence has suggested the superiority of both LCI and
NBI over WLI in colorectal lesion detection, our results were rational and consistent
with the published literature.
It bears mentioning that, although not statistically significant, there was a slight
tendency toward a higher SDR in the LCI arm than in the NBI arm, and this tendency
seemed more pronounced when the colonoscopies were performed by experienced endoscopists,
with prolonged withdrawal time (≥ 9 minutes), or in patients with suboptimal bowel
preparation. Consistent with our previous study, the present study also revealed that
endoscopists’ experience and withdrawal time were independent factors associated with
the SDR [11]. Familiarity with new techniques is important in performing an effective procedure.
In our center, LCI and NBI have been routinely used in the evaluation of gastrointestinal
lesions; the experience in LCI and NBI for individual endoscopists accumulated correspondingly
with the total colonoscopy volume. The withdrawal time of at least 9 minutes has also
been reported to improve the ADR and SDR in previous studies [36]
[37].
Interestingly, the present study revealed that the BBPS was also an independent factor
associated with the SDR, which was inconsistent with the results of our previous study.
This discrepancy may likely be attributed to the involvement of NBI in the present
study. The importance of high quality bowel preparation in NBI has been proposed by
abundant numbers of studies. A parallel RCT on NBI versus WLI found that, in patients
with “good” bowel preparation, there was a statistically significant benefit of NBI
over WLI for adenoma detection, whereas there was no difference between NBI and WLI
when preparation was “fair” [38]. A recent meta-analysis of RCTs also showed NBI had a higher ADR than WLI, and this
effect was greater when bowel preparation was optimal [22]. The residual liquids appear dark red under NBI and can significantly impair mucosal
visualization and negate its benefits. In contrast, the color of residual liquids
is less influential under LCI inspection. Therefore, the application of NBI in colorectal
lesion detection may require more rigorous bowel preparation than LCI does.
Taken together, these results may imply a potential superiority of LCI over NBI for
SSL detection in real-world situations where the bowel preparation in the majority
of patients is just “good” but hardly “best”, and when the endoscopists have sufficient
experience of the technique in their own hands and perform the colonoscopy examination
with an adequate inspection time.
There were several limitations of the present study. First, it was conducted in a
single tertiary hospital, which may impact generalization of the results. Second,
it is not possible to blind the endoscopists to the modality (LCI or NBI). Third,
the colonoscopies were performed for both diagnostic and screening purposes. Therefore,
these results should be properly interpreted with attention paid to potential differences
from studies with only screening colonoscopies. Fourth, because we have previously
reported a comparison between LCI and WLI for the detection of colorectal SSLs [11], and the effectiveness of NBI vs. WLI in colorectal SSL detection has also been
reported previously [21]
[27]
[28]
[29], we did not include another WLI arm in the present study. It was more practical
to conduct a direct comparison between the different modalities rather than comparing
them with WLI again. Finally, although we involved a large number of participants
(n = 406), the sample size seemed insufficient in subgroup analyses based on the low
prevalence of colorectal SSLs.
In conclusion, the present study is the first and to date the largest parallel RCT
focusing on comparing LCI with NBI for the detection of SSLs. The results demonstrated
that LCI is comparable to NBI for SSL detection, as well as for polyp and adenoma
detection. Future multicenter RCTs with larger sample sizes are warranted to validate
our results and to further elucidate the superiority of LCI and NBI among stratified
patients.