Introduction
Colorectal cancer (CRC) is the third most common cancer and the second leading cause
of cancer-related deaths worldwide [1]. Most sporadic CRCs arise from pre-existing adenomas [2], and removal of these precancerous lesions has been shown to effectively reduce
both the incidence and mortality of CRC [3]
[4]. Therefore, the effectiveness of colonoscopy in protecting against CRC hinges on
the detection and removal of adenomas, and the adenoma detection rate (ADR) is the
most important quality indicator of colonoscopy. Previous research showed that a 1%
increase in ADR can reduce CRC incidence and mortality by 3% and 5%, respectively
[5]. Therefore, various modalities have been developed to improve ADR, including image-enhancing
technologies [6], chromoendoscopy [7], and devices enhancing exploration of the mucosa [8]
[9].
Despite improvements in ADR conferred by those modalities, post-colonoscopy colorectal
cancer (PCCRC) remains a concern [10]. The incidence of PCCRC has been reported at 8.6% within 3 years [11], with more than 80% of PCCRCs being attributed to missed adenomas [12]
[13]. Notably, flat and proximal adenomas are independently associated with the development
of PCCRC and are particularly difficult to detect, posing significant challenges to
improving the ADR [12]
[14]
[15].
Three-dimensional (3D) endoscopy provides 3D visualization with superior depth perception
over conventional two-dimensional (2D) endoscopy and may thereby enhance detection
of flat/superficial lesions and subtle mucosal changes. 3D endoscopy has shown promise
in enhancing the detection of superficial gastric neoplasms and accuracy in assessing
morphology [16]. It has also been proposed that 3D endoscopy enhances the detection of colonic adenomas,
showing a 25% increase in adenoma detection in a study using simulated 3D colonoscopy
in a synthetic colon model [17]
[18]. However, whether 3D colonoscopy could improve ADR and facilitate detection of flat
polyps compared with standard 2D colonoscopy in clinical colonoscopy practice remains
to be studied.
The MonoStereo 3D endoscopic visualization system (MedicalTek Co. Ltd, Taichung, Taiwan)
is a novel 3D endoscopy system that performs real-time conversion of standard 2D images
to realistic 3D visualization during endoscopy and has been approved for clinical
use [19]
[20]. We hypothesized that the 3D endoscopic visualization system could enhance polyp
detection during colonoscopy, especially for flat/superficial polyps. Therefore, we
conducted a randomized controlled trial (RCT) to investigate whether 3D colonoscopy
improved adenoma detection compared with standard 2D colonoscopy.
Methods
Study design
This was a prospective, multicenter, randomized, open-label, single-blind trial conducted
in one referral center and two regional hospitals in Taiwan. Complying with the principles
of the Declaration of Helsinki and Good Clinical Practice guidelines, this trial was
approved by the institutional review board of National Taiwan University Hospital
(No. 202109112DIPB). An independent data and safety monitoring committee monitored
the progress of the trial, with regular assessment of safety outcomes, overall trial
integrity, and trial performance.
Participants
Individuals aged 40 years or older who were scheduled for colonoscopy for screening,
surveillance, or symptoms at outpatient clinics in the participating institutions
were consecutively assessed for eligibility. Individuals with a contraindication to
colonoscopy or polypectomy, or with a history of inflammatory bowel disease or hereditary
polyposis syndrome were excluded.
Randomization and masking
In this study, randomization was conducted centrally at the coordinating hospital
by a research assistant using a computer-generated randomization sequence with a block
size of 20, assigning participants from all three hospitals in a 1:1 ratio without
stratification. Allocation concealment was ensured by storing the group allocation
in ordered, sealed, and opaque envelopes. The patients and research assistants who
assessed the outcomes were blinded to the group allocation to avoid bias.
Procedures
Three-dimensional colonoscopy
[Fig. 1]
a illustrates the MonoStereo 3D endoscopic visualization system. 2D images ([Fig. 1]
b and right screen in [Fig. 1]
d) are converted in 80 milliseconds to 3D images ([Fig. 1]
c and left screen in [Fig. 1]
d) visualized through polarized glasses worn by the endoscopist, providing real-time
3D imagery without perceptible time lag ([Fig. 1]
d, [Video 1]). The system offers three pupillary distance selections to mitigate eye strain,
and endoscopists are recommended to identify the optimal personal selection before
first use by finding the selection yielding the most vivid 3D image. The system does
not require calibration before examination; endoscopists are advised to place the
3D screen at eye level and stand in front of the screen at a distance tailored to
individual preference (generally 100–150 cm for a 31”/32” screen). Instantaneous switch
between 3D and standard 2D displays is achieved by pressing a button. As the polarized
glasses do not change the visual perception of the surrounding environment or standard
2D endoscopic images, the endoscopist does not need to remove the glasses when not
using the 3D display.
Fig. 1 Use of the novel three-dimensional (3D) endoscopy system during colonoscopy. a Schematic representation of the MonoStereo 3D endoscopic
visualization system (MedicalTek Co. Ltd, Taichung, Taiwan). b,c The endoscopic display can be switched from standard 2D images (b) to reconstructed images, which transform into real-time fully
immersive 3D images when viewed with polarized 3D glasses (c). d Employing 3D colonoscopy during routine
colonoscopic examinations. Source for Figure 1a: MedicalTek Co., Ltd.
Two- and three-dimensional colonoscopy. 2D images are converted to left/right images,
which yield 3D images using 3D monitors and polarized glasses. Deliberate endoscope
movement demonstrates minimal time lag. The apparent difference in polyp shape on
3D images disappears with the use of 3D monitors and polarized glasses.Video 1
Intervention and colonoscopy
Study colonoscopies were performed by three junior colonoscopists (experience <5000
colonoscopies) and one senior colonoscopist (experience ≥5000 colonoscopies). Before
the commencement of the study, the participating colonoscopists received an introduction
on the 3D technology and equipment, and performed 3D colonoscopy using a colonoscopy
simulator. Each colonoscopist was then requested to use 3D colonoscopy in conjunction
with standard 2D colonoscopy for mucosa inspection during colonoscope withdrawal in
at least 10 colonoscopic procedures [Fig. 1]
d).
For the RCT, high-definition colonoscopes (290 series; Olympus, Tokyo, Japan) and
video processors (EVIS Lucera Elite; Olympus) were used for colonoscopy. Bowel preparation
and image-enhanced endoscopy were performed in the same way in both groups. Standard
2D colonoscopy was used for colonoscope insertion, as in routine clinical practice,
in both groups. The use of distal attachment devices, such as cap or cuff, was prohibited.
After the cecum was intubated, colonoscope withdrawal was performed exclusively with
2D or 3D images as per allocation. A standardized protocol for photo documentation
of individual colonic segments and a withdrawal time of 6 minutes or longer were required
during colonoscope withdrawal. During withdrawal, image-enhanced endoscopy (narrow-band
imaging or chromoendoscopy with indigo carmine) was routinely used for suspicious
lesions, and adenomas were removed/resected. The size, morphology, and location of
each polyp were recorded, and specimens were sent for histological examination.
Therapeutic time was defined as the time for optical diagnosis and polyp removal.
Mucosa inspection time was defined as withdrawal time minus therapeutic time. Incomplete
colonoscopic examination was defined as failure to intubate the cecum or poor bowel
preparation, and resulted in exclusion from the analysis. In line with the established
clinical workflow of the participating institutions, bowel preparation was assessed
with the modified Aronchick bowel preparation scale [21]
[22].
Outcomes
The primary outcome was ADR, defined as the proportion of patients with at least one
adenoma detected during colonoscopy. Secondary outcomes were detection rates for flat
(Paris classification 0-IIa, 0-IIb, or 0-IIc), sessile (Paris classification 0-Is),
right-sided (cecum and ascending colon), left-sided (transverse colon to rectum),
proximal (cecum to splenic flexure), distal (descending colon to rectum), and advanced
adenomas, and sessile serrated lesions (SSLs). ADR stratified by size (<5 mm, 5–9
mm, ≥10 mm), polyp detection rate, mean adenoma number per patient, and mean polyp
number per patient were also recorded. Advanced adenoma was defined as adenomas with
size ≥10 mm, villous component, or high grade dysplasia according to World Health
Organization classification [23].
Statistical analysis
A previous simulation study suggested that 3D colonoscopy could increase the ADR by
60% (from 42.7% to 67.7%) compared with standard colonoscopy [18]. Following international guidelines, we set the ADR with standard 2D colonoscopy
at 25% [24]. To detect a 60% increase in ADR between 3D and standard 2D colonoscopy (40% vs.
25%) with an 80% statistical power and a two-sided significance level of 0.05, a minimum
of 150 participants per group was needed. Accounting for potential exclusions or dropouts
of approximately 10%, the enrollment target was at least 165 participants for each
group. The analysis was by intention-to-treat.
Categorical variables were summarized using frequencies and percentages, and continuous
variables as means and SDs. Statistical significance for categorical variables was
tested using the Pearson chi-squared test, and differences between groups for continuous
variables were tested using the independent sample t test. Univariable and multivariable logistic regression analyses were conducted to
identify factors predictive of adenoma detection. Variables with a P value of <0.05 in the univariable analysis were included in the multivariable analysis,
and variance inflation factor was used to detect multicollinearity. Post hoc analysis
of the temporal changes in ADR and mucosa inspection time was conducted to explore
the learning curve of 3D colonoscopy.
All analyses were performed using STATA software (StataCorp, College Station, Texas,
USA). All tests were two-tailed, and differences were considered significant if P < 0.05.
Results
Patients
From February 2022 through June 2023, 348 individuals were screened for eligibility
([Fig. 2]), and 339 consented to participate. Finally, 334 participants underwent colonoscopy
and were randomly allocated to either the 2D or 3D group (n = 167 each). After excluding
cases with incomplete colonoscopy and inadequate bowel preparation, 158 and 160 participants
in the 2D and 3D groups, respectively, were analyzed. There was no crossover between
the two groups.
Fig. 2 Screening, recruitment, randomization, and analysis of the study participants.
Baseline characteristics
The baseline characteristics and clinical information are summarized in [Table 1]. Among the 318 enrolled participants, the mean age was 61.9 (SD 10.6) years and
150 (47.2%) were men. Most (69.8%) of the participants were asymptomatic, and the
major indication for colonoscopy among the asymptomatic patients was positive fecal
immunochemical test (FIT) or surveillance colonoscopy. The groups were comparable
in age, sex, family history of CRC, smoking status, alcohol consumption, antithrombotic
agent use, underlying diseases, colonoscopy indications, and bowel preparation status.
There was no significant difference between the two groups in mucosa inspection time
for all colonoscopies (2D vs. 3D: 9.4 [SD 3.1] vs. 9.8 [SD 2.6] minutes; P = 0.21).
Table 1 Demographics and clinical characteristics of study participants.
|
|
2D colonoscopy (n = 158)
|
3D colonoscopy (n = 160)
|
P
|
|
2D/3D, two-dimensional/three-dimensional; CRC, colorectal cancer; FIT, fecal immunochemical
test.
1Participants rating poor or inadequate bowel preparation were excluded from the study.
2Withdraw time = the total time from cecum to anus.
3Inspection time = withdraw time – time for observing and removing polyps.
|
|
Age, mean (SD), years
|
62.4 (11.2)
|
61.4 (9.9)
|
0.40
|
|
Male sex, n (%)
|
79 (50.0)
|
71 (44.4)
|
0.32
|
|
Body weight, mean (SD), kg
|
65.4 (12.2)
|
66.4 (13.2)
|
0.48
|
|
Body height, mean (SD), cm
|
164.2 (9.7)
|
162.9 (7.8)
|
0.19
|
|
Body mass index, mean (SD), kg/m2
|
24.2 (4.2)
|
24.9 (4.1)
|
0.13
|
|
Family history of CRC, n (%)
|
17 (10.8)
|
27 (16.9)
|
0.11
|
|
Ever smoking, n (%)
|
39 (24.7)
|
32 (20.0)
|
0.32
|
|
Alcohol consumption, n (%)
|
14 (8.9)
|
11 (6.9)
|
0.51
|
|
Antithrombotic agent use, n (%)
|
21 (13.3)
|
27 (16.9)
|
0.37
|
|
Diabetes mellitus, n (%)
|
24 (15.2)
|
25 (15.6)
|
0.91
|
|
Hypertension, n (%)
|
52 (32.9)
|
64 (40.0)
|
0.19
|
|
Indication, n (%)
|
0.35
|
|
|
44 (27.8)
|
50 (31.3)
|
|
|
|
59 (37.3)
|
45 (28.1)
|
|
|
|
43 (27.2)
|
53 (33.1)
|
|
|
|
12 (7.6)
|
12 (7.5)
|
|
|
Modified Aronchick bowel preparation scale, n (%)1
|
0.09
|
|
|
100 (63.3)
|
87 (54.4)
|
|
|
|
58 (36.7)
|
73 (45.6)
|
|
|
Withdrawal time, mean (SD), minutes2
|
11.0 (5.2)
|
12.5 (5.0)
|
0.009
|
|
Mucosa inspection time, mean (SD), minutes3
|
|
Entire cohort
|
9.4 (3.1)
|
9.8 (2.6)
|
0.21
|
|
|
9.6 (2.6)
|
11.1 (2.6)
|
0.01
|
|
|
10.0 (3.6)
|
10.1 (2.5)
|
0.90
|
|
|
9.5 (3.4)
|
9.9 (2.1)
|
0.54
|
|
|
8.3 (2.6)
|
8.0 (2.2)
|
0.64
|
Outcomes
The 3D colonoscopy function was successfully implemented in all cases allocated to
the 3D group without temporary equipment dysfunction during the colonoscopic procedures.
For the two groups combined (n = 318), polyp detection rate and ADR were 54.4% and
45.9%, respectively. ADR was significantly higher in the 3D group compared with the
2D group (53.1% vs. 38.6%; difference 14.5% [95%CI 3.7 to 25.4]; odds ratio [OR] 1.80
[95%CI 1.15 to 2.82]; P = 0.009) ([Table 2]). Regarding the secondary outcomes, compared with the 2D group, the 3D group had
higher detection rates for flat adenomas (35.0% vs. 21.5%; difference 13.5% [95%CI
3.7 to 23.3]; OR 1.96 [95%CI 1.19 to 3.24]; P = 0.008), right-sided adenomas (26.3% vs. 15.2%; difference 11.1% [95%CI 2.2 to 19.9];
OR 1.98 [95%CI 1.14 to 3.48]; P = 0.02), proximal adenomas (38.1% vs. 23.4%; difference 14.7% [95%CI 4.7 to 24.7];
OR 2.02 [95%CI 1.24 to 3.28]; P = 0.005), and small (5–9 mm) adenomas (45.0% vs. 31.0%; difference 14.0% [95%CI 3.4
to 24.5]; OR 1.82 [95%CI 1.15 to 2.88]; P = 0.01). The median number of adenomas per patient was also higher in the 3D group
(1 [interquartile range (IQR) 1–2] vs. 0 [IQR 0–1]; P = 0.03). As all individuals with adenomas had at least one left-sided adenoma (adenomas
at transverse, descending, sigmoid colon, or rectum), the left-sided ADR was equivalent
to the overall ADR in both groups. There was no significant difference in the detection
rates for sessile adenomas, distal adenomas, advanced adenomas, and SSLs.
Table 2 Comparison of primary and secondary outcomes between two- and three-dimensional colonoscopy.
|
|
Detection during colonoscopy
|
Difference in detection rate (95%CI), %
|
OR (95%CI)
|
P
|
|
2D
(n = 158)
|
3D
(n = 160)
|
|
2D/3D, two-dimensional/three-dimensional; IQR, interquartile range; OR, odds ratio.
1Right-sided adenoma: adenoma at cecum or ascending colon; left-sided adenoma: adenoma
at transverse colon, descending colon, sigmoid colon, or rectum; proximal adenoma:
adenoma at cecum, ascending colon, or transverse colon; distal adenoma: adenoma at
descending colon, sigmoid colon, or rectum; flat adenoma: Paris classification 0-IIa,
0-IIb, or 0-IIc; sessile adenoma: Paris classification 0-Is.
|
|
Primary outcome, n (%)
|
|
|
61 (38.6)
|
85 (53.1)
|
14.5 (3.7 to 25.4)
|
1.80 (1.15 to 2.82)
|
0.009
|
|
Secondary outcomes, n (%)1
|
|
|
34 (21.5)
|
56 (35.0)
|
13.5 (3.7 to 23.3)
|
1.96 (1.19 to 3.24)
|
0.008
|
|
|
44 (27.8)
|
47 (29.4)
|
1.6 (–8.4 to 11.5)
|
1.08 (0.66 to 1.75)
|
0.76
|
|
|
24 (15.2)
|
42 (26.3)
|
11.1 (2.2 to 19.9)
|
1.98 (1.14 to 3.48)
|
0.02
|
|
|
61 (38.6)
|
85 (53.1)
|
14.5 (3.7 to 25.4)
|
1.80 (1.15 to 2.82)
|
0.009
|
|
|
37 (23.4)
|
61 (38.1)
|
14.7 (4.7 to 24.7)
|
2.02 (1.24 to 3.28)
|
0.005
|
|
|
48 (30.4)
|
53 (33.1)
|
2.7 (–7.5 to 12.8)
|
1.11 (0.68 to1.82)
|
0.66
|
|
|
8 (5.1)
|
11 (6.9)
|
1.8 (–3.4 to 7.0)
|
1.38 (0.54 to 3.54)
|
0.50
|
|
|
11 (7.0)
|
15 (9.4)
|
2.4 (–3.6 to 8.4)
|
1.38 (0.61 to 4.11)
|
0.43
|
|
Adenoma, n (%)
|
|
|
14 (8.9)
|
13 (8.1)
|
0.7 (–6.9 to 5.4)
|
1.10 (0.50 to 2.42)
|
0.81
|
|
|
49 (31.0)
|
72 (45.0)
|
14.0 (3.4 to 24.5)
|
1.82 (1.15 to 2.88)
|
0.01
|
|
|
15 (9.5)
|
25 (15.6)
|
6.1 (–1.1 to 13.4)
|
1.77 (0.89 to 3.49)
|
0.10
|
|
Polyps, n (%)
|
73 (46.2)
|
100 (62.5)
|
16.3 (5.5 to 27.1)
|
1.94 (1.24 to 3.04)
|
0.004
|
|
Adenomas per patient, median (IQR), n
|
0 (0–1)
|
1 (1–2)
|
–
|
–
|
0.03
|
|
Polyps per patient, median (IQR), n
|
0 (0–1)
|
1 (0–2)
|
–
|
–
|
<0.001
|
Factors associated with adenoma detection
In the univariable logistic regression analysis, age, hypertension, FIT positivity,
bowel preparation (excellent/good vs. fair), mucosa inspection time, and 3D colonoscopy
were significantly associated with adenoma detection ([Table 3]). The multivariable analysis showed that 3D colonoscopy was independently associated
with adenoma detection (adjusted OR [aOR] 1.76 [95%CI 1.09 to 2.83]) after adjusting
for FIT positivity, mucosa inspection time, and other confounders. Age (aOR 1.03 [95%CI
1.01 to 1.06]) and mucosa inspection time (aOR 1.16 [95%CI 1.06 to 1.28]) were also
independently associated with adenoma detection.
Table 3 Factors associated with detection of adenoma.
|
|
Univariable analysis
|
Multivariable analysis
|
|
OR (95%CI)
|
P
|
aOR (95%CI)
|
P
|
|
3D, three-dimensional; (a)OR, (adjusted) odds ratio; BMI, body mass index; CRC, colorectal
cancer; FIT, fecal immunochemical test.
|
|
Age, per 1-year increment
|
1.04 (1.02 to 1.06)
|
<0.001
|
1.03 (1.01 to 1.06)
|
0.008
|
|
Male sex
|
1.44 (0.92 to 2.24)
|
0.11
|
|
|
|
BMI, per 1-kg/m2 increment
|
1.04 (0.98 to 1.10)
|
0.16
|
|
|
|
Ever smoking
|
1.70 (1.00 to 2.90)
|
0.05
|
|
|
|
Alcohol consumption
|
1.30 (0.57 to 2.93)
|
0.54
|
|
|
|
Antithrombotic agent use
|
1.79 (0.96 to 3.34)
|
0.07
|
|
|
|
Diabetes mellitus
|
1.40 (0.76 to 2.58)
|
0.28
|
|
|
|
Hypertension
|
1.90 (1.20 to 3.02)
|
0.006
|
1.32 (0.79 to 2.20)
|
0.29
|
|
Family history of CRC
|
0.63 (0.33 to 1.21)
|
0.17
|
|
|
|
FIT positivity
|
1.90 (1.17 to 3.10)
|
0.01
|
1.46 (0.86 to 2.47)
|
0.16
|
|
Good/excellent bowel preparation
|
0.60 (0.38 to 0.94)
|
0.03
|
0.89 (0.54 to 1.47)
|
0.67
|
|
Mucosa inspection time, per 1-minute increment
|
1.20 (1.10 to 1.31)
|
<0.001
|
1.16 (1.06 to 1.28)
|
0.001
|
|
3D colonoscopy use
|
1.80 (1.15 to 2.81)
|
0.009
|
1.76 (1.09 to 2.83)
|
0.02
|
Temporal changes in ADR and mucosa inspection time
Compared with the 2D group, the mean mucosa inspection time in the 3D group was significantly
longer in the first 40 examinations (11.1 [SD 2.6] vs. 9.6 [SD 2.6] minutes; P = 0.01) but became comparable thereafter ([Table 1], [Fig. 3]
a). Similar trends were observed for each endoscopist, with interendoscopist variations.
The learning curve, as inferred by the difference in mucosa inspection time between
3D and 2D colonoscopy, seemed shortest for the senior colonoscopist, with the time
difference reduced from 2.8 minutes for procedures 1–10 to 0.5 minutes for procedures
11–20. By contrast, one junior endoscopist appeared to have the longest learning curve
(time difference 1.9, 0.9, and 0.5 minutes for procedures 1–10, 11–20, and 21–30,
respectively). On the other hand, ADR in the 3D group was consistently higher than
that in the 2D group by approximately 15% throughout the study, even among the first
40 examinations ([Fig. 3]
b). All endoscopists achieved numerically higher ADRs with 3D colonoscopy (difference
in ADRs [3D minus 2D]: senior endoscopist 12%; junior endoscopists 12.5%, 21.6%, and
50.0%, respectively). However, per-endoscopist analyses on differences in mucosa inspection
time and ADR were post hoc and included a limited sample size, and thus should be
considered exploratory.
Fig. 3 Temporal changes between two- and three-dimensional colonoscopy. a Mean mucosa inspection time. b Adenoma detection rate. 2D/3D, two-dimensional/three-dimensional.
Discussion
This RCT conducted in individuals aged 40 years or older showed that 3D colonoscopy
resulted in a significant 15% increase in ADR, as well as increases in the detection
rates of small, flat, right-sided, and proximal neoplasms, which are commonly overlooked
by standard 2D colonoscopy. Notably, 3D colonoscopy enhanced polyp detection without
increasing the mucosa inspection time and could be used in conjunction with other
image-enhancing modalities such as narrow-band imaging and chromoendoscopy.
Enhancing the ADR is crucial for reducing the incidence of PCCRC and associated mortality
[5]. Despite the multitude of advanced image-processing technologies that have been
developed to improve adenoma detection [6], the incidence of PCCRC remains as high as 8% in Asia and Europe and is mainly attributed
to missed neoplasms during colonoscopy [11]
[25]
[26]. Neoplasms with flat morphology, particularly those located in the proximal colon,
are more likely to be overlooked [27]. The larger colonic folds in the proximal colon, where neoplasms are more often
flat, further compound adenoma detection [28]. This study corroborated the notion that 3D colonoscopy enhances anatomical details
and depth perception and thereby facilitates identification of those hard-to-detect
neoplasms. Our finding that 3D colonoscopy improved ADR and detection of flat, right-sided,
and proximal adenomas, respectively, support its potential to reduce PCCRCs, warranting
further long-term follow-up research. Multicenter clinical trials and real-world studies,
advocacy by gastroenterology societies and opinion leaders, regulatory approval, and
education/training are crucial for the dissemination of 3D colonoscopy.
The finding that 3D colonoscopy mainly enhanced the detection of small polyps (5–9
mm) might be due to these polyps being near the threshold of detection with 2D colonoscopy;
therefore, enhanced depth perception conferred by 3D colonoscopy significantly increased
the ability to detect these polyps. By contrast, polyps 1–5 mm might remain difficult
to detect despite enhanced depth perception and thus 3D colonoscopy did not significantly
improve detection. In line with this notion, studies on chromoendoscopy using indigo
carmine found no or minimal improvement in detecting adenomas 1–5 mm [29]
[30]. On the other hand, polyps >10 mm could be easily detected by 2D colonoscopy, with
limited room for further improvement by 3D colonoscopy.
It is worth noting that while high ADRs (ADR 38.6%, right-sided ADR 15.2%, proximal
ADR 23.4%, flat ADR 21.5%) were achieved by standard 2D colonoscopy with a mean mucosa
inspection time of approximately 9 minutes, 3D colonoscopy could further increase
the ADRs by approximately 15% (ADR 53.1%, right-sided ADR 26.3%, proximal ADR 38.1%,
flat ADR 35.0%). The ADRs of the 2D group in our study were in line with a recent
RCT by Zhao et al., which showed that 2D white-light colonoscopy with a mucosa inspection
time of 9 minutes achieved ADR, proximal ADR, and flat ADR of 36.6%, 21.4%, and 27.4%,
respectively [31]. An OR of 1.76 for detecting adenomas after adjusting for mucosa inspection time
and other confounders firmly supported that 3D colonoscopy provided a distinct advantage
over 2D colonoscopy in adenoma detection that cannot be provided by alternative means
such as increasing the mucosa inspection time. Whether 3D colonoscopy could provide
greater benefit over standard 2D colonoscopy in real clinical settings where the mucosa
inspection time is shorter than 9 minutes warrants further study.
Our exploratory analysis confirmed that 3D colonoscopy has a short learning curve
and consistently confers an improvement in ADR even during the learning phase. The
findings suggested a learning curve of between 10 and 20 procedures for 3D colonoscopy,
with interendoscopist variation. Taken together, the consistent benefit in ADR and
short learning curve suggest that 3D colonoscopy could be easily adopted by endoscopists
in routine colonoscopy practice.
A recent crossover RCT including patients younger than 40 years compared 2D then 3D
vs. 3D then 2D colonoscopy (i.e. tandem colonoscopy) and showed that ADR in the first
examination was comparable between 3D and 2D colonoscopy (24.7% vs. 23.8%), whereas
in the second examination ADR was significantly higher with 3D compared with 2D (13.8%
vs. 9.9%) [32]. However, the tandem colonoscopy design could introduce bias, because the diagnostic
performance of the second examination is influenced by the findings of the first one.
In contrast, the parallel design of the current study minimized bias, better reflected
clinical reality, and used ADR, the surrogate for PCCRC, as the primary outcome. Notably,
the ADR of the first colonoscopy in the previous study did not differ between 2D and
3D and seemed lower than that in the current study, probably due to the shorter withdrawal
time (<6 minutes) and the inclusion of younger patients (aged 18–40 years) in that
study. In contrast, the current study enrolled individuals aged over 40 years, and
thus the results should be more generalizable to the examinees of clinical colonoscopy
practice, and the ability to further improve ADR where colonoscopy quality assurance
measures were rigorously implemented highlighted the benefit of 3D colonoscopy in
enhancing adenoma detection. The use of different 3D endoscopy systems could have
also contributed to the differences between the two studies, as the vividness of 3D
visualization might differ between systems, depending on the image reprocessing algorithms
employed.
This study had several notable strengths. This RCT is the first to demonstrate the
ability of 3D imaging to improve ADR and enhance detection of flat and proximally
located adenomas, which are challenging to detect with standard 2D colonoscopy. Second,
this study ensured high-quality colonoscopy through measures such as attention to
bowel cleansing, photo documentation, and maintaining a withdrawal time exceeding
6 minutes, in accordance with the international benchmarks. Third, this study enrolled
individuals aged over 40 years to align the study population with the examinees in
general colonoscopy practice, enhancing the relevance and generalizability of the
results. Finally, this study conducted stratified comparisons according to polyp morphologies
and location, revealing the advantage of 3D colonoscopy in enhancing detection of
flat and proximal adenomas.
This study also had limitations. Given the apparent differences between 2D and 3D
colonoscopy, it was not possible for the colonoscopists to be blinded to group allocation.
However, the quality assurance program, including standardized photo documentation
in participating institutions, ensured that the mucosa inspection time was comparable
between the two groups and >6 minutes, refuting the possibility that colonoscopists
tried harder to find polyps in the 3D group. Therefore, nonblinding of endoscopists
should not have introduced significant bias. While the endoscopists’ ADRs might have
been affected by study participation (i.e. Hawthorne effect), the potential influence
should occur in both 2D and 3D groups to a similar degree; therefore, the observed
differences in ADR should be little influenced by the Hawthorne effect and remain
valid. The comparability in other procedural factors and randomization minimized the
possibility of confounding, and regression analysis adjustment for potential confounders
further confirmed that the observed improvement in adenoma detection was attributed
to 3D colonoscopy. Second, given the limited availability of the newly developed 3D
colonoscopy equipment, this RCT included only a limited number of institutions and
colonoscopists. A larger trial including more institutions/colonoscopists and diverse
patient populations is warranted to further ascertain the potential benefit conferred
by wide implementation of 3D colonoscopy. Third, this study did not evaluate the endoscopists’
burden such as eye strain because of the lack of a well-established objective evaluation
tool/method. However, none of the participating endoscopists reported fatigue or eye
strain after performing 3D colonoscopy, probably because this 3D endoscopy system
uniquely considers pupillary distance. Tailoring the 2D to 3D conversion process according
to pupillary distance is crucial for mitigating visual discomfort when watching 3D
imagery [33]. The finding that a significant increase in ADR with 3D colonoscopy was not accompanied
by an increase in the mucosa inspection time compared with 2D colonoscopy also suggest
that processing the 3D images did not significantly increase endoscopist burden. Whether
more prolonged use of this 3D system for colonoscopy might increase endoscopist burden
remains to be evaluated. Finally, given the relatively low prevalence of SSLs and
advanced adenomas, this study was not powered to detect potential differences in the
rate of SSLs and advanced adenomas between 3D and 2D colonoscopy. The numerically
higher detection rates for SSLs and advanced adenomas with 3D colonoscopy observed
in this study warrants confirmation by further research with larger sample sizes.
In conclusion, this RCT demonstrated that for individuals aged 40 years and above,
3D colonoscopy significantly increased the detection rates for adenomas, particularly
small, flat, and proximal adenomas, compared with standard 2D colonoscopy. The sizable
increases in ADR suggest that implementing 3D colonoscopy in clinical practice might
deliver significant improvement in patient outcomes.
