Introduction
Endoscopic detection and resection of colorectal lesions can decrease colorectal cancer
mortality [1]
[2]. Due to the high prevalence of disease, slow progression from polyp to invasive
cancer, and the possibility to resect its precursor lesions, population-based screening
for colorectal cancer has been shown successful and cost-effective in reducing cancer
mortality [3]
[4]
[5].
Biennial screening with a fecal immunochemical test (FIT), followed by a colonoscopy
in case of a positive FIT, has been adopted in several European countries [6]
[7]. Although suitable for population screening, the diagnostic accuracy of FIT-based
screening is far from perfect.[8] Several adjustments for the pre-selection of high-risk participants in add to FIT-only
testing, therefore, have been suggested [9]
[10].
Research from the last decade showed that not just adenomas but also a subset of serrated
polyps can progress to colorectal cancer. The World Health Organization classifies
serrated polyps into the subgroups hyperplastic polyps, sessile serrated adenomas/polyps
without dysplasia, sessile serrated adenomas/polyps with dysplasia, and traditional
serrated adenomas [11]. Diminutive and/or small hyperplastic polyps located in the rectosigmoid are generally
considered benign, whereas larger and/or proximally located hyperplastic polyps, sessile
serrated adenomas/polyps, and traditional serrated adenomas are considered to possess
a higher neoplastic potential [12]
[13]. The latter do so via the alternative serrated neoplasia pathway [12]
[14]. These serrated polyps may be held responsible in the development of a relatively
large amount of interval cancers [15]
[16]. Based on these considerations, the European Society for Gastrointestinal Endoscopy
has recommended that people with large (≥ 10 mm) or dysplastic serrated polyps (referred
to as advanced serrated polyps) should be classified as at high risk of developing
colorectal cancer [17]. Accordingly, it has been suggested that population-based screening should not only
aim to detect participants with advanced neoplasia, defined as advanced adenomas or
colorectal cancer, but also those with large serrated polyps [10].
Expanding the target lesions and the definition of people at risk will potentially
have a major impact on the performance of the different methods of colorectal cancer
screening. A recent study showed that FIT had no value in the detection of serrated
polyps in a screening setting, whereas stool DNA testing might have this diagnostic
potential [18]. However, DNA testing is expensive and its diagnostic performance does not seem
seems good enough yet to accurately detect advanced serrated polyps. Clinical risk
factors for advanced serrated polyps may offer a noninvasive and inexpensive means
of detection.
The aim of this study was to evaluate whether the well-known clinical risk factors
for the detection of advanced adenomas and CRC, such as smoking, also act as risk
factors for advanced serrated polyps.
Patients and methods
Study design
We analyzed data collected in the colonoscopy arm of a multicenter randomized trial
conducted in the Netherlands from June 2009 until July 2010, comparing colonoscopy
with computed tomography (CT) colonography for primary population screening (COCOS).
In 2010, population screening for CRC had not yet been introduced in the Netherlands.
Detailed information about the study protocol has been described previously [19]. The trial was registered in the Dutch Trial Register: NTR1829 (www.trialregister.nl). Ethical approval was obtained from the Dutch National Health Council. (2009/03WBO,
The Hague, The Netherlands)
Study population
A total of 6 600 asymptomatic individuals from the greater Amsterdam and Rotterdam
regions were invited to undergo colonoscopy as primary CRC screening. Individuals
who had undergone a complete colonic examination (colonoscopy, CT colonography and/or
double contrast barium enema) within the 5 years prior to the invitation were excluded
from participation in the trial. Individuals who were already under colonoscopic surveillance
(e. g., personal history of CRC, adenomas or inflammatory bowel disease) and/or had
end-stage disease and a life expectancy < 5 years were also excluded.
Fecal immunochemical testing
Participants willing to undergo a colonoscopy were invited to complete a one-sample
FIT (OC-Sensor, Eiken Chemical CO., LTD., Japan) prior to the colonoscopy. Informed
consent was obtained from all individuals who agreed to participate. All participants
were verbally instructed how to properly perform the FIT; they could do so either
at home or in one of the screening centers. Participants were instructed to perform
the FIT at home within 48 hours before the colonoscopy, but before the start of their
bowel preparation, and to bring the test to the screening center on the day of colonoscopy.
As a second option, participants could call their screening center directly after
performing the FIT, so that the test could be collected at home within 48 hours. FIT
samples were immediately stored at – 20°C and analysed within 6 weeks. Detailed information
about this procedure has been previously described [8].
Risk questionnaire
Participants willing to undergo colonoscopy were also invited to complete a questionnaire
with questions about a set of putative CRC risk factors. These risk factors were selected
based on a review of the existing literature [20]. Only those variables that could be obtained without additional testing were eligible
for the questionnaire. The questionnaire consisted of 10 independent items; it was
developed based on three pre-existing validated risk questionnaires: the Prevention
Compass, the Municipal Health Agency, and the Interheart questionnaire [21]
[22]
[23]. Selected risk factors were: gender, age, familial history of CRC, adiposity, smoking
behavior, sleeping behavior, amount of exercise, alcohol consumption, fiber intake,
calcium intake, red meat consumption and aspirin use. Detailed information about the
risk factor selection procedure and development of the risk questionnaire is available
elsewhere [20]. The risk questionnaire was handed out to the participants at the day of the screening
colonoscopy and collected directly before this procedure.
Colonoscopy
All colonoscopies were performed in one of the two participating academic screening
centers by highly experienced gastroenterologists, each having performed more than
1 000 colonoscopies at the beginning of the trial. Colonoscopists were blinded to
the results of the FIT as well as to response to the risk questionnaire. All colonoscopies
were performed according to the standard quality indicators defined by the American
Society for Gastrointestinal Endoscopy and recorded on DVD [24]. Pre-colonoscopy bowel cleansing was obtained by a low-fiber diet, 2 L of hypertonic
polyethylene glycol solution (Moviprep; Norgine bv, Amsterdam, the Netherlands) and
2 L of clear fluids. Colonoscopy was performed under conscious sedation using intravenous
midazolam (Dormicum, Actavis, Baarn, the Netherlands) and fentanyl (Bipharma, Weesp,
the Netherlands), at the discretion of both the participant as well as the endoscopist.
Cecal intubation was confirmed by still images of the ileocecal valve as well as the
appendiceal orifice or by intubation of the terminal ileum. Quality of bowel preparation
was measured using the validated Ottawa bowel preparation score [25]. In case of inadequate bowel preparation after fecal residue and fluid suction,
the colonoscopy was interrupted and rescheduled. The withdrawal time was measured
by an endoscopy-nurse. The net withdrawal time (time of mucosal inspection) was set
at a minimum of 6 minutes. All detected lesions were immediately resected and obtained
for histologic assessment. When resection of detected lesions was not directly possible,
biopsies were taken to provide a histopathologic diagnosis. Polyp characteristics
(size, location, morphology and optical diagnosis) were obtained for all lesions.
Pathology
All tissue specimens were assessed according to the Vienna criteria by one of two
expert gastrointestinal pathologists, one in each center [26]. The diagnosis was based on the morphologic features on hematoxylin and eosin staining.
Lesions were classified as either adenomatous, serrated, or carcinoma. Adenomatous
lesions were subdivided based on the grade of dysplasia (low-grade or high-grade)
as well as the presence of a villous component (tubular, tubulovillous or villous).
Serrated polyps (SP) were subdivided into hyperplastic polyps (HPs), sessile serrated
adenomas/polyps (SSA/Ps) without dysplasia, SSA/Ps with dysplasia or traditional serrated
adenomas (TSAs). A SSA/P was predominantly defined based on the presence of (hyper-)serration
up to the base of the crypts as well as distortion of the crypts, presenting with
abnormal shapes like the common L-shape or inverted T-shape [11]. A TSA was predominantly defined based on the presence of a complex and distorted
tubulovillous or villous configuration, prominent serration, aberrant crypt formation
and diffuse cytoplasmic eosinophilia [11]. Advanced adenomas were defined as all adenomas ≥ 10 mm, with a tubulovillous or
villous histology and/or with high-grade dysplasia. Advanced neoplasia was defined
as either CRC or advanced adenoma. The pathologist of the other center revised all
lesions that were primarily classified as advanced neoplasia, as well as a random
sample of 10 % of the other lesions. This demonstrated no structural discrepancies
between both pathologists. In case of inconsistency, the slides were evaluated by
both pathologists together to obtain a consistent diagnosis. Advanced SPs were not
reevaluated for the purpose of this study.
Statistical analysis
We evaluated associations between the putative CRC risk factors and the presence of
advanced subtypes of SP. An advanced SP was defined as either a SP ≥ 10 mm or a SP
with (any) dysplasia, based on the European Society of Gastrointestinal Endoscopy
(ESGE) post-polypectomy colonoscopy surveillance guideline [17]. We focused on the most advanced lesion per participant in the analysis. Advanced
adenomas and CRC were considered as lesions more advanced than advanced SP, while
non-advanced adenomas were classified as less advanced. To evaluate risk factors for
SP, we compared participants with advanced SP, but without more advanced lesions,
with the group of participants without lesions or with non-advanced adenoma and/or
non-advanced SP as most advanced lesion.
For each risk factor, the univariate association with the presence of at least one
advanced SP was calculated and presented as odds ratio (OR) with 95 % confidence interval
(CI). Multivariable logistic regression was used to evaluate the conditional, adjusted
associations between the risk factors and the presence of at least one advanced SP.
Stepwise backward elimination was used to select risk factors included in the multivariate
analysis. All risk factors that showed a P value of < 0.2 in the univariate analysis were included in the multivariable analysis,
as well as gender and age. Multiple imputation was used to handle the missing values
from the individual covariates [27].
Results
Participants
Of the 6 600 invitees, 1 426 participated in this trial, and 1 236 also completed
the risk questionnaire. CRC was detected in seven participants, advanced adenoma in
105, and advanced SP in 53; 1 086 participants did not have an advanced lesion. Thirteen
participants with an advanced SP also had an advanced adenoma detected, resulting
in 40 participants with an advanced SP as most advanced detected lesion. Those 40
participants were included in the analysis, and compared with the 1 086 participants
who did not have an advanced lesion ([Fig.1]). [Table 1] summarizes key characteristics of these 1,126 participants. The mean age was 60.1
years (SD 6.2); 575 were male (51 %).
Fig. 1 Study flowchart
Table 1
Diagnostic risk factors for advanced serrated lesions as most advanced lesion in an
average-risk screening population
|
Overall
|
No advanced lesion
|
Advanced
SP
|
Univariate OR
(95 % CI)
|
P value
|
Multivariable OR
(95 % CI)
|
P value
|
|
Cohort (n)
|
1126
|
1086
|
40
|
|
|
|
|
|
Sex
Male, n (%)
Female, n (%)
|
575 (51.1)
551 (48.9)
|
553 (50.9)
533 (49.1)
|
22 (55.0)
18 (45.0)
|
1.18 (0.63 – 2.25)
|
0.61
|
1.16 (0.61 – 2.24)
|
0.65
|
|
Age in years, mean (SD)[1]
≤ 60, n (%)
> 60, n (%)
|
60.1 (6.2)
595 (52.8)
531 (47.2)
|
60.1 (6.2)
572 (52.7)
514 (47.3)
|
60.4 (6.6)
23 (57.5)
17 (42.5)
|
1.01 (0.96 – 1.06)
|
0.71
|
1.02 (0.97 – 1.08)
|
0.46
|
|
Smoking status
Current smoker, n (%)
Former/no smoker, n (%)
Missing, n (%)
|
162 (14.4)
951 (84.4)
13 (1.2)
|
148 (13.6)
926 (85.3)
12 (1.1)
|
14 (35.0)
25 (62.5)
1 (2.5)
|
3.50 (1.74 – 6.80)
|
< 0.001
|
4.50 (2.23 – 8.89)
|
< 0.001
|
|
Fiber intake in g/day, mean (SD)[1]
≤ 41, n (%)
> 41, n (%)
|
41.4 (21.8)
665 (59.1)
461 (41.9)
|
41.2 (21.7)
646 (59.5)
440 (40.5)
|
47.5 (24.5)
19 (47.5)
21 (52.5)
|
1.01 (1.00 – 1.02)
|
0.07
|
1.02 (1.00 – 1.03)
|
0.01
|
|
BMI in kg/m2, mean (SD)[1]
< 25, n (%)
25 – 30, n (%)
> 30, n (%)
Missing, n (%)
|
26.6 (4.1)
417 (37.0)
506 (44.9)
185 (16.4)
18 (1.7)
|
26.6 (4.1)
404 (37.2)
489 (45.0)
175 (16.1)
18 (1.7)
|
27.6 (4.3)
13 (32.5)
17 (42.5)
10 (25.0)
|
1.06 (0.98 – 1.13)
|
0.11
|
1.07 (0.99 – 1.15)
|
0.06
|
|
Alcohol in units/week, median (IQR)[1]
≤ 15, n (%)
> 15, n (%)
Missing, n (%)
|
5 (1 – 10)
934 (82.9)
136 (12.1)
56 (5.0)
|
5 (1 – 10)
902 (83.0)
129 (11.9)
55 (5.1)
|
4 (1 – 14)
32 (80.0)
7 (17.5)
1 (2.5)
|
1.01 (0.97 – 1.03)
|
0.78
|
|
|
|
No. of relatives with CRC, median (IQR)[1]
0, n (%)
1, n (%)
> 1, n (%)
|
0 (0 – 0)
964 (85.6)
138 (12.3)
24 (2.1)
|
0 (0 – 0)
932 (85.8)
130 (12.0)
24 (2.2)
|
0 (0 – 0)
32 (80.0)
8 (20.0)
0
|
1.15 (0.55 – 2.02)
|
0.67
|
|
|
|
Intensive exercise
< 1 hour/week, n (%)
≤ 1 hour/week, n (%)
|
693 (61.5)
433 (38.5)
|
669 (61.6)
417 (38.4)
|
24 (60.0)
16 (40.0)
|
1.07 (0.55 – 2.02)
|
0.84
|
|
|
|
Sleep behavior in h/day, mean (SD)[1]
≤ 7, n (%)
> 7, n (%)
Missing
|
7.3 (1.1)
597 (53.0)
463 (41.1)
66 (5.9)
|
7.3 (1.1)
576 (53.0)
446 (41.1)
64 (5.9)
|
7.3 (1.0)
21 (52.5)
17 (42.5)
2 (5.0)
|
1.04 (0.80 – 1.41)
|
0.78
|
|
|
|
NSAID/Aspirin
Non-user
User
|
880 (78.2)
246 (21.8)
|
848 (78.1)
238 (21.9)
|
32 (80.0)
8 (20.0)
|
0.89 (0.38 – 1.87)
|
0.77
|
|
|
|
Calcium intake in mg/day, median (IQR)[1]
≤ 1200
> 1200
|
750 (505 – 960)
992 (88.1)
134 (11.9)
|
750 (510 – 960)
957 (88.1)
129 (11.9)
|
750 (486 – 909)
35 (87.5)
5 (12.5)
|
1.00 (1.00 – 1.00)
|
0.63
|
|
|
|
Red meat in units/week, median (IQR)[1]
≤ 3
> 3
Missing
|
3 (1 – 4)
705 (62.6)
380 (33.7)
41 (3.7)
|
3 (1 – 4)
681 (62.7)
365 (33.6)
40 (3.7)
|
2 (2 – 4)
24 (60.0)
15 (37.5)
1 (2.5)
|
1.01 (0.85 – 1.21)
|
0.91
|
|
|
|
FIT value in ng/ml, median (IQR)[1]
≤ 50
> 50
Missing
|
0 (0 – 5)
943 (83.7)
65 (5.8)
118 (10.5)
|
0 (0 – 5)
912 (84.0)
61 (5.6)
113 (10.4)
|
0 (0 – 11)
31 (77.5)
4 (10.0)
5 (12.5)
|
1.00 (1.00 – 1.00)
|
0.99
|
|
|
BMI, body mass index; CRC, colorectal cancer; NSAID, nonsteroidal anti-inflammatory
drug; FIT, fecal immunochemical test; IQR, interquartile range
1 Quantitative variables were treated as such in the analyses
Of the 40 participants with at least one advanced SP, 26 (65 %) had at least one large
SP detected, and 16 (40 %) individuals had at least one SP containing any dysplasia.
In one (2.5 %) participant an SP containing high-grade dysplasia was detected. The
median number of detected SPs of any kind was two (range 1 – 8), while in total 14
(35 %) out of these 40 participants were also found to have a non-advanced adenoma
(median 0; range 0 – 3).
In the 1,086 participants without an advanced lesion, there were 249 (23 %) in whom
at least one non-advanced SP was detected; 21 (1.9 %) had at least one SSA/P. There
were 243 (22 %) participants with at least one non-advanced adenoma.
CRC risk factors
In [Table1] associations between the CRC risk factors and the presence of at least one advanced
SP as most advanced lesion are presented. In univariate analysis we observed a strong
association between current smoking and the presence of an advanced SP as most advanced
lesion (OR 3.50; (95 % CI 1.74 – 6.80; P < 0.001). In addition, higher fiber intake (OR 1.24 per 20 gram; 95 % CI 0.98 – 1.58;
P = 0.07) as well as higher body mass index (BMI) (OR 1.06 per point; 95 % CI 0.98 – 1.13;
P = 0.11) also showed a moderate association, although this did not reach statistical
significance. Other clinical risk factors were not associated with the presence of
an advanced SP as most advanced lesion. This also included fecal hemoglobin concentration
(OR 1.00 per 10 ng/mL; 95 % CI 0.97 – 1.03, P = 0.99). The association between current smoking and the presence of an advanced
SP as most advanced lesion remained significant in the multivariable analysis.(OR
4.50; 95 % CI 2.23 – 8.89; P < 0.001). A significant association was also demonstrated for higher fiber intake
(OR 1.36 per 20-g intake; CI 1.07 – 1.73; P = 0.01). Higher BMI was not significantly associated with advanced SPs in the multivariable
analysis (OR 1.07 per point; CI 0.99 – 1.15; P = 0.06).
Discussion
Our analysis shows that current smoking is strongly associated with advanced SPs as
most advanced detected lesion in an asymptomatic screening population, with an odds
ratio of about four. In addition, higher fiber intake was moderately associated with
the presence of advanced SPs. Other CRC risk factors, including fecal hemoglobin level,
did not show a significant association with the presence of at least one advanced
SP.
All available data for this analysis had been prospectively collected in a structured
and transparent manner. All colonoscopies were performed according to the most up-to-date
colonoscopy quality parameters and endoscopists were instructed to resect all detected
lesions, unaware of location or predicted histology [19]. Yet a number of potential limitations have to be acknowledged. The number of subjects
with at least one advanced SP was relatively small, which limits the power to detect
more moderate associations. For this reason we decided not to perform any sub-analyses
in this study. In a recent study we described the prevalence of SP subtypes in the
same cohort, showing that the overall detection rate of HPs as well as SSA/Ps was
sufficient and comparable to other recent reports [28]. Therefore, the low number of advanced SP, as reported in the current study, most
probably is a realistic representation of the prevalence of disease in average-risk
individuals. Associations between clinical risk factors and advanced SPs were based
on a self-completed questionnaire, which has the risk of socially desirable answers.
To limit this effect, participants were told that the results from the questionnaire
would be anonymized and in no way affect options for treatment.
A number of previous studies have assessed the association between smoking behavior
and the occurrence of SPs, all reporting a moderate to strong association [29]
[30]
[31]
[32]
[33]
[34]
[35]
[36]
[37]. However, most studies had a case-control design, were executed in an era during
which the risk of SPs was not yet well defined and/or were not stratified for SP subtype
[29]
[30]
[31]
[32]
[33]
[35]
[37]. Although well designed, the risk information value of smoking behavior for the
detection of advanced SPs in an average-risk asymptomatic screening population, therefore,
could not be reliably assessed in these studies. Two recent studies evaluated the
association between smoking and SPs via a prospective cohort design [34]
[36]. The first study, performed in a cohort of 985 asymptomatic, screening-naïve participants,
reported a significant association between current smoking and the prevalence of SSA/Ps,
with an odds ratio of about five [34]. Male gender and increasing age were also associated with the presence of SSA/Ps
in this study. However, the detection rate of SSA/Ps was reported to be only 2.3 %.
Because these associations were not adjusted for concomitant advanced adenomas, they
are difficult to interpret. In the second study a combined analysis of three colonoscopy-based
clinical trials (n = 2915 patients) was conducted to evaluate the smoking-associated
risk of adenomas as well as serrated polyps compared to patients without any polyps
[36]. Smoking appeared to be associated with greater risk of serrated polyps, particularly
in the left-sided colon. Sub-analysis by polyp size showed a moderate association
between smoking behavior and large SPs (current smoking: RR 1.62, 95 % CI 1.27 – 2.07
and former smoking: RR 1.21, 95 % CI 0.98 – 1.49). This association was far more moderate
than found in the current study. The causal pathway between smoking and SPs is still
under debate. One recent study has shown that tobacco smoking leads to extensive genome-wide
changes in DNA methylation, which, to a large extent, returns to the level of non-smokers
when discontinued [38]. This would explain the increased risk for current smokers compared to former smokers
and non-smokers, as found in our study. Unfortunately, sub-analysis for the risk difference
between former smokers and non-smokers could not be performed due to the restrictions
in sample size.
Studies that have assessed the relationship between fiber intake and the prevalence
of SPs are scarce [33]. In a large colonoscopy-based case-control study, a protective trend of higher fiber
intake was found for the detection of SPs, which seems contrary to our findings [33]. In that study, all participants with at least one SP were included in the analysis,
regardless of SP subtype. This means that diminutive HPs in the rectosigmoid were
also included, which makes an informative comparison of the results difficult. The
role of fiber intake in the pathogenesis of SPs, if any, is still speculative. It
has been suggested that the development of advanced lesions through the serrated neoplasia
pathway is induced by alterations in the microbiome of these participants [39]. Fiber intake may have a role in this process.
Two other recent studies have assessed the diagnostic accuracy of FIT for the detection
of SPs larger than 1 cm [10]
[18]. These showed that including large SPs in the target condition leads to a decreased
diagnostic performance of FIT. In a large study comparing FIT-based screening to multitarget
stool DNA testing (including the quantitative molecular assays for KRAS mutations, aberrant NDRG4 and BMP3 methylation, β-actin and fecal haemoglobin levels) for CRC screening, only 5.1 % of participants with
a large SP had a quantitative FIT-score above 100 ng hemoglobin per mL of buffer,
while 42.4 % of participants were detected with DNA testing, at the same specificity
level [10]. In another study, the diagnostic performance of BMP3 methylation alone was compared to FIT for the detection of SPs larger than 1 cm,
showing that at 95 % specificity, the stool assay of BMP3 methylation detected 63 % of large SPs, compared to 0 % by FIT [18].
In our analysis, the performance of fecal hemoglobin in detecting advanced SPs was
very limited. At a cut-off of 50 ng/mL, the sensitivity for the detection of advanced
SPs in those without other, more advanced lesions, was 11 % at a specificity of 94 %.
At a cut-off of 100 ng/mL, sensitivity would be 5.7 % with a specificity of 97 %.
To put this in perspective: the sensitivity of current smoking, considered in isolation,
was 36 % for a specificity of 86 %.
We believe that the low performance of FIT in detecting advanced adenoma and advanced
SPs, and the promising information of CRC risk factors, could and should lead to the
development of better-performing population screening programs. These would be programs
that not only detect CRC, or advanced adenoma, but also advanced SPs. Such programs
may include fecal hemoglobin, DNA testing, and other markers, but they should not
ignore the incremental value of easy-to-collect information on CRC risk factors, such
as smoking behavior. Ideally, the value of fecal hemoglobin levels, other fecal DNA
markers and clinical risk factors such as smoking status will be used in a risk stratification
model in which the individual fitted risk value determines the need for a subsequent
colonoscopy. Future research should focus on the validation of such a model.
