The n e w e ng l a n d j o u r na l of m e dic i n e

Original Article

Effect of Colonoscopy Screening on Risks

of Colorectal Cancer and Related Death
M. Bretthauer, M. Løberg, P. Wieszczy, M. Kalager, L. Emilsson, K. Garborg,
M. Rupinski, E. Dekker, M. Spaander, M. Bugajski, Ø. Holme, A.G. Zauber,
N.D. Pilonis, A. Mroz, E.J. Kuipers, J. Shi, M.A. Hernán, H.-O. Adami, J. Regula,
G. Hoff, and M.F. Kaminski, for the NordICC Study Group*​​

A BS T R AC T

BACKGROUND
Although colonoscopy is widely used as a screening test to detect colorectal cancer, The authors’ full names, academic de-
its effect on the risks of colorectal cancer and related death is unclear. grees, and affiliations are listed in the
Appendix. Dr. Bretthauer can be contact-
ed at ­michael​.­bretthauer@​­medisin​.­uio​.­no
METHODS or at the Clinical Effectiveness Research
We performed a pragmatic, randomized trial involving presumptively healthy men Group, University of Oslo, Postbox 1089,
and women 55 to 64 years of age drawn from population registries in Poland, Blindern, N-0318 Oslo, Norway.
Norway, Sweden, and the Netherlands between 2009 and 2014. The participants *The members of the NordICC Study
were randomly assigned in a 1:2 ratio to either receive an invitation to undergo a Group are listed in the Supplementary
Appendix, available at NEJM.org.
single screening colonoscopy (the invited group) or to receive no invitation or
screening (the usual-care group). The primary end points were the risks of colorec- This article was published on October 9,
2022, at NEJM.org.
tal cancer and related death, and the secondary end point was death from any
cause. DOI: 10.1056/NEJMoa2208375
Copyright © 2022 Massachusetts Medical Society.

RESULTS
Follow-up data were available for 84,585 participants in Poland, Norway, and Swe-
den — 28,220 in the invited group, 11,843 of whom (42.0%) underwent screening,
and 56,365 in the usual-care group. A total of 15 participants had major bleeding
after polyp removal. No perforations or screening-related deaths occurred within
30 days after colonoscopy. During a median follow-up of 10 years, 259 cases of
colorectal cancer were diagnosed in the invited group as compared with 622 cases
in the usual-care group. In intention-to-screen analyses, the risk of colorectal
cancer at 10 years was 0.98% in the invited group and 1.20% in the usual-care
group, a risk reduction of 18% (risk ratio, 0.82; 95% confidence interval [CI], 0.70
to 0.93). The risk of death from colorectal cancer was 0.28% in the invited group
and 0.31% in the usual-care group (risk ratio, 0.90; 95% CI, 0.64 to 1.16). The
number needed to invite to undergo screening to prevent one case of colorectal
cancer was 455 (95% CI, 270 to 1429). The risk of death from any cause was
11.03% in the invited group and 11.04% in the usual-care group (risk ratio, 0.99;
95% CI, 0.96 to 1.04).
CONCLUSIONS
In this randomized trial, the risk of colorectal cancer at 10 years was lower among
participants who were invited to undergo screening colonoscopy than among
those who were assigned to no screening. (Funded by the Research Council of
Norway and others; NordICC ClinicalTrials.gov number, NCT00883792.)

n engl j med  nejm.org 1

The New England Journal of Medicine
Downloaded from nejm.org by JOHN OTREMBA on October 16, 2022. For personal use only. No other uses without permission.
Copyright © 2022 Massachusetts Medical Society. All rights reserved.
 The n e w e ng l a n d j o u r na l of m e dic i n e

A
s the third most common type of Me thods
cancer and the second leading cause of
death from cancer worldwide, colorectal Trial Design
cancer is an attractive target for population The pragmatic NordICC trial was conducted in
screening.1 Multiple screening options are avail- Poland, Norway, Sweden, and the Netherlands.
able, but high-quality evidence to indicate the The trial design and rationale have been de-
best strategies is limited.2 The most commonly scribed in detail previously.7,8 Eligible partici-
used screening tests are fecal testing for occult pants were men and women 55 to 64 years of
blood and endoscopic screening with sigmoidos- age who had not previously undergone screening
copy or colonoscopy.3 and who lived in one of the four countries where
In randomized trials, the relative risk of the trial was conducted. Exclusion criteria were
death from colorectal cancer was approximately death or the diagnosis of colorectal cancer be-
15% lower among persons who were assigned fore trial entry, as assessed in national registries
to undergo screening with guaiac fecal testing before randomization.7,8 Participants were iden-
than among those who were assigned to no tified directly from the population registries in
screening; however, screening with this test the four countries and were randomly assigned
had little or no effect on the risk of colorectal in a 1:2 ratio to either invitation to undergo
cancer.3 Because most colorectal cancers de- colonoscopy screening (the invited group) or to
velop from benign polyps that can be detected no invitation and no screening (the usual-care
and removed during endoscopy, endoscopic group). Independent organizations in each par-
screening may prevent colorectal cancer. In a ticipating country randomly assigned partici-
pooled analysis of three randomized trials, the pants with the use of a computer-generated al-
incidence of colorectal cancer was up to 25% location algorithm, stratified according to age,
lower after 10 to 12 years of follow-up among sex, and municipality.8 Screening was performed
persons who had been invited to undergo sig- between June 8, 2009, and June 23, 2014, as re-
moidoscopy screening than among those who ported previously.8
had not been invited.4 At the beginning of the trial, Poland had an
Colonoscopy is considered to be more ef- opportunistic screening program for colorectal
fective than sigmoidoscopy because it can be cancer in some geographic areas but not in the
used to examine the entire large bowel.3,5 area where the trial was conducted. In the other
Thus, sigmoidoscopy has largely been replaced countries, no organized colorectal cancer screen-
by colonoscopy, which is the predominant ing of any kind was available at the beginning of
screening test for colorectal cancer in the the trial. During the last 4 years of trial follow-
United States and is recommended to be per- up, colorectal cancer screening was gradually
formed every 10 years.5 In contrast, colonos- introduced according to region and age group in
copy has not been adopted in many other the participating countries.
parts of the world, partly because evidence The integrity of the trial was preserved
from randomized trials regarding the benefits through collaboration with the screening pro-
of this test is lacking.6 grams in two ways. First, screening programs
A balance among benefits, harms, and cost- were introduced earlier in geographic areas
effectiveness of various colorectal cancer screen- where the trial was not enrolling participants,
ing tests is important because colonoscopy is and second, the trial participants were too old to
more invasive and burdensome for patients than be eligible for the new screening programs by
fecal testing and sigmoidoscopy, and it requires the time the programs were introduced in the
more clinical resources. Here, we report the re- areas where our trial was being conducted.
sults of the Nordic-European Initiative on Colorec- Thus, none of the participants who were en-
tal Cancer (NordICC), a large, multicenter, ran- rolled in the trial were eligible for any colorectal
domized trial that investigated the effects of cancer screening programs outside the trial dur-
population-based colonoscopy screening on the ing screening or follow-up.8 Throughout the
risks of colorectal cancer and related death at 10 trial, we monitored opportunistic colonoscopy
years. screening activity in the trial areas and did not

2 n engl j med  nejm.org

The New England Journal of Medicine

Downloaded from nejm.org by JOHN OTREMBA on October 16, 2022. For personal use only. No other uses without permission.
Copyright © 2022 Massachusetts Medical Society. All rights reserved.
 Colonoscopy Screening and Colorectal Cancer

identify additional colonoscopy procedures be- Trial End Points

yond what would have been expected for clinical The primary end points were the risks of
indications.9 colorectal cancer and death from colorectal can-
This report is based on follow-up data from cer after a median follow-up of 10 to 15 years
all 84,585 participants in Poland, Norway, and (with the first analysis planned after 10 years).7
Sweden (89.1% of all 94,959 participants, includ- The secondary end point was death from any
ing those from the Netherlands, who were cause. A diagnosis of colorectal cancer was de-
originally included in the trial)8 (Fig. S1 in the fined, according to the International Statistical
Supplementary Appendix, available with the full Classification of Diseases and Related Health
text of this article at NEJM.org). Of the remain- Problems, 10th revision, as cancer in the colon
ing 10,374 participants, 594 had been excluded, or rectum (topography codes C18 to C20, com-
and data from the remaining 9780 participants, bined with International Classification of Dis-
all from the Netherlands, could not be included eases for Oncology morphology codes for adeno-
because Statistics Netherlands could not provide carcinoma) (see the Supplementary Appendix).
follow-up data from the usual-care group owing The stage of colorectal cancer was classified as
to a new Dutch law based on the recently intro- early-stage (Dukes’ stage A or B), late-stage
duced European Union General Data Protection (Dukes’ stage C or D), or unknown. Tumors with
Regulation. To ensure timely reporting of pre- a histopathological diagnosis other than adeno-
specified end-point analyses, we decided to sub- carcinoma were not counted as events. Colorec-
mit this report for publication without data from tal cancer–related deaths were defined as those
the Netherlands. that were listed as such in the cause-of-death
The trial was funded by research grants in registries in the participating countries.
the participating countries. The authors vouch
for the accuracy and completeness of the data Follow-up
and for the fidelity of the trial to the protocol, Almost complete long-term follow-up of all par-
available with the statistical analysis plan at ticipants who underwent randomization was
NEJM.org. made possible through the use of unique per-
sonal identification numbers, which were linked
Interventions to cancer registries and cause-of-death regis-
Participants were randomly assigned to either tries, for all trial participants in each country.7
invitation to one-time screening colonoscopy All participants who underwent randomization
or to no invitation to screening, as previously were followed for all end-point events through
described.7 All screening colonoscopies were these registries, regardless of whether they un-
performed at dedicated centers.7,8 A quality- derwent screening.
assurance and training program was imple-
mented for the trial.7 All lesions detected dur- Ethics and Consent
ing colonoscopy were removed if feasible, and This randomized trial followed a pragmatic de-
all tumors were biopsied. Participants in sign; the participants underwent randomization
whom cancer was detected on screening were before they were asked whether they wanted to
referred from the trial centers to the public participate in the trial (in the invited group) or
health service and treated in accordance with not asked to participate (the usual-care group).7,11
national policies. Dedicated histopathologists All the participants who underwent colonoscopy
assessed all polyps and cancers according to screening provided written informed consent.
the classification of the World Health Organ­ With the exception of a subsample of 6900 par-
ization.10 Data from all screening examina- ticipants in Norway, the participants in the
tions were registered in an online electronic usual-care group were not informed about their
case-report form and stored at a central data- enrollment in the trial at inclusion or during
base. Patients were referred for surveillance follow-up. During follow-up, the subsample of
of polyps after screening in accordance with participants in Norway received a questionnaire
national guidelines (see the Supplementary related to lifestyle and general health.12 The trial
Appendix).7 was approved by the ethics committees at all

n engl j med  nejm.org 3

The New England Journal of Medicine
Downloaded from nejm.org by JOHN OTREMBA on October 16, 2022. For personal use only. No other uses without permission.
Copyright © 2022 Massachusetts Medical Society. All rights reserved.
 The n e w e ng l a n d j o u r na l of m e dic i n e

participating centers, the Swedish National Coun- soring events. Bootstrapping was used to calcu-
cil on Medical Ethics, and the Health Council of late 95% confidence intervals. The number need-
the Netherlands.7 ed to invite to undergo screening to prevent one
case of colorectal cancer was calculated as the
Statistical Analysis reciprocal of the between-group difference in
The sample-size calculation was based on inten- risk with respect to the risk of colorectal cancer
tion-to-screen analyses and has been described at 10 years.
in detail previously.7 We estimated that event We estimated the per-protocol effect of screen-
rates based on colorectal cancer–related mortal- ing, which was defined as the effect of screening
ity after 15 years would provide enough power to if all the participants who were randomly as-
also assess the risk of colorectal cancer.7 We signed to the invited group had undergone
assumed a 25% difference in colorectal cancer– screening.14 The risk of colorectal cancer may
related mortality between the invited group and have differed between the participants who un-
the usual-care group, a 50% participation rate, derwent screening and those who were invited to
and 50% screening efficacy.7 With 80% power at undergo screening but declined to do so; there-
a two-sided significance level of 5%, we calcu- fore, our analyses adjusted for the baseline co-
lated that at least 22,800 participants in the in- variates of the participants (Table S2). We esti-
vited group and 45,600 participants in the usual- mated standardized risks with the use of a
care group would be needed. In Poland, because pooled logistic model with the following covari-
of the availability of resources and lower-than- ates: age at randomization (with the use of re-
anticipated participation in screening, we en- stricted cubic splines with knots at the 5th,
rolled a higher number of participants than the 27.5th, 50th, 72.5th, and 95th percentiles), sex
number we had planned in order to maintain (male or female), country (Poland, Norway, or
statistical power. Sweden), group assignment (invited or usual
The primary analysis was conducted in ac- care), duration of follow-up (with the use of re-
cordance with the intention-to-screen principle. stricted cubic splines with knots at 3-month
Follow-up time was measured from the date of periods at 2, 4, 6, and 8 years), and product
randomization to the date of emigration, the terms representing interactions between group
diagnosis of colorectal cancer (for analyses of assignment and duration of follow-up.14 We did
the risk of colorectal cancer), death from colorec- not observe opportunistic screening of any
tal cancer (for analyses of death from colorectal meaningful extent in the usual-care group and
cancer), or death from causes other than colorec- thus did not adjust for it in the per-protocol
tal cancer or to the end of follow-up after 10 years, analyses. In a sensitivity analysis, we used the
whichever came first. We did not use Cox pro- approach that was proposed by Cuzick et al.16
portional-hazards models for analyses, as origi- and previously used in other trials of colorectal
nally planned, because of the nonproportional cancer screening (see Table S6).17,18
hazards of the risk of colorectal cancer during
follow-up.13,14 We used the Kaplan–Meier estima-
R e sult s
tor to calculate the cumulative 10-year risks of
colorectal cancer and colorectal cancer–related Participants
death in the invited group and the usual-care The trial included 54,927 eligible participants
group, and we compared risks using risk ratios, from Poland, 26,588 from Norway, and 3664
risk differences, and annual incidence rate ra- from Sweden. After randomization and before
tios. We performed analyses in which competing the beginning of the intervention, 175 partici-
events (i.e., death from causes other than colorec- pants who were assigned to the invited group
tal cancer) were considered to be censoring and 419 of those who were assigned to the
events, and we performed additional analyses in usual-care group were excluded because they
which competing events were not treated as cen- had died or had received a diagnosis of colorec-
soring events.15 The competing events were un- tal cancer at randomization but had not yet been
likely to affect the results, and the current report identified as such in the registries at the time.
focuses on analyses in which deaths from causes Thus, the current analyses involved 84,585 par-
other than colorectal cancer were treated as cen- ticipants (28,220 in the invited group and 56,365

4 n engl j med  nejm.org

The New England Journal of Medicine

Downloaded from nejm.org by JOHN OTREMBA on October 16, 2022. For personal use only. No other uses without permission.
Copyright © 2022 Massachusetts Medical Society. All rights reserved.
 Colonoscopy Screening and Colorectal Cancer

Table 1. Characteristics of the Trial Participants.

All Participants Participants in Norway Participants in Poland Participants in Sweden

Characteristic (N = 84,585) (N = 26,411) (N = 54,258) (N = 3646)
Group — no.
Invited 28,220 8,815 18,184 1221
Usual care 56,365 17,596 36,344 2425
Sex — no.
Female 42,186 13,194 27,330 1662
Male 42,399 13,217 27,198 1984
Age at randomization — no.
55–59 yr 43,100 12,524 28,792 1784
60–64 yr 41,485 13,887 25,736 1862
Screening participation
— no./total no. (%)
Women and men 11,843/28,220 (42.0) 5354/8815 (60.7) 6003/18,184 (33.0) 486/1221 (39.8)
Women 5,724/14,066 (40.7) 2580/4390 (58.8) 2918/9117 (32.0) 226/559 (40.4)
Men 6,119/14,154 (43.2) 2774/4425 (62.7) 3085/9067 (34.0) 260/662 (39.3)
55–59 yr 5,877/14,369 (40.9) 2497/4174 (59.8) 3173/9599 (33.1) 207/596 (34.7)
60–64 yr 5,966/13,851 (43.1) 2857/4641 (61.6) 2830/8585 (33.0) 279/625 (44.6)
Screening performance
— no./total no. (%)
Good or very good bowel 110,610/11,635 (91.2) 4739/5174 (91.6) 5445/5999 (90.8) 426/462 (92.2)
preparation*
Cecum intubation 11,470/11,843 (96.8) 5130/5354 (95.8) 5868/6003 (97.8) 472/486 (97.1)
Adenoma detection 3,634/11,843 (30.7) 1453/5354 (27.1) 2111/6003 (35.2) 70/486 (14.4)
Screening-related adverse events
— no./total no. (%)
Perforation† 0 0 0 0
Major bleeding‡ 15/11,843 (0.13) 8/5354 (0.15) 7/6003 (0.12) 0

* Data in this category were missing for 208 participants, including some for whom data on the bowel preparation regimen were missing and
others for whom data on the quality of bowel preparation were missing.
† One perforation that occurred in a participant in the Netherlands was not included in these analyses.8
‡ Major bleeding was defined as bleeding that warranted treatment. All cases of major bleeding were treated endoscopically, and further
therapy was not warranted.

in the usual-care group); 42,399 (50.1%) were younger participants (Table 1). The cecum was
men and 42,186 (49.9%) were women (Table 1). intubated in 96.8% of the colonoscopies per-
The median age at randomization was 59 years formed, and the quality of bowel preparation
(Table S3). The median follow-up for the current was adequate in more than 90% of the colonos-
analysis was 10.0 years in both groups (inter- copies.8
quartile range, 9.9 to 10.0; maximum follow-up, Colorectal cancer was diagnosed at screening
10.0 years). in 62 participants (0.5% of those who underwent
screening). These 62 cases included 2 cases in
Colonoscopy Screening Poland that had been classified as adenomas in
The percentage of participants who underwent our previous analysis (see the Supplementary
screening varied among the countries (from Appendix).8 Adenomas were detected and re-
33.0% in Poland to 60.7% in Norway) and was moved at screening in 3634 participants (30.7%
higher overall among men than among women of those who underwent screening). A total of 15
and among older participants than among participants (0.13%) had polypectomy-related

n engl j med  nejm.org 5

The New England Journal of Medicine
Downloaded from nejm.org by JOHN OTREMBA on October 16, 2022. For personal use only. No other uses without permission.
Copyright © 2022 Massachusetts Medical Society. All rights reserved.
 The n e w e ng l a n d j o u r na l of m e dic i n e

group and 0.50% in the usual-care group re-

100 ceived a diagnosis of late-stage (stage C or D)
90 1.75
colorectal cancer (Table S7). Analyses in which
Cumulative Risk of Colorectal Cancer (%) competing events were not treated as censoring
80 1.50 events showed results that were similar to those
70 1.25 in the main analysis (Table S1).
60 1.00
Invited group Colorectal Cancer–Related Death and Death
50 0.75 0.98 (95% CI, 0.86–1.09) from Any Cause
40 0.50 The risk of colorectal cancer–related death at 10
0.25 Usual-care group
years was 0.28% (72 deaths) among participants
30
1.20 (95% CI, 1.10–1.29) in the invited group and 0.31% (157 deaths)
20 0.00 among those in the usual-care group (risk ratio,
0 2 4 6 8 10
10
0.90; 95% CI, 0.64 to 1.16) (Table 2 and Fig. 3).
During the 10-year follow-up period, 3036 par-
0 ticipants in the invited group (11.03%) died from
0 2 4 6 8 10
any cause, as compared with 6079 (11.04%) in
Years since Randomization
the usual-care group (risk ratio, 0.99; 95% CI,
No. at Risk
Invited group 28,220 27,684 27,111 26,461 24,000 18,748 0.96 to 1.04) (Table 2).
Usual-care group 56,365 55,375 54,192 52,819 47,769 37,313
Adjusted Per-Protocol Analyses
Figure 1. Cumulative Risk of Colorectal Cancer at 10 Years in Intention-to- In adjusted analyses to estimate the effect of
Screen Analyses.
screening if all the participants who were ran-
The inset shows the same data on an enlarged y axis. I bars indicate 95%
domly assigned to screening had actually under-
confidence intervals.
gone screening, the risk of colorectal cancer at
10 years was decreased from 1.22% to 0.84%,
corresponding to an estimated risk ratio of 0.69
major bleeding; all cases of bleeding were treat- (95% CI, 0.55 to 0.83) (Table S4 and Fig. S3). The
ed endoscopically and did not warrant further results were similar in the sensitivity analysis
interventions (Table 1). No perforations or screen- (Table S6). The corresponding risk ratio was 0.55
ing-related deaths occurred within 30 days after (95% CI, 0.38 to 0.74) in Norway and 0.85 (95%
screening. CI, 0.63 to 1.12) in Poland (Table S5).
The risk of death from colorectal cancer was
Incidence of Colorectal Cancer 0.15% in the invited group and 0.30% in the
The risk of colorectal cancer at 10 years was usual-care group. The estimated risk ratio was
0.98% (259 cases) in the invited group and 0.50 (95% CI, 0.27 to 0.77) (Fig. S4). In the sen-
1.20% (622 cases) in the usual-care group, for a sitivity analysis, the risk ratio was 0.72, but the
risk ratio of 0.82 (95% confidence interval [CI], estimate was imprecise (95% CI, 0 to 3.70).
0.70 to 0.93) (Fig. 1 and Table 2). Table S5 shows
the risks and risk ratios of colorectal cancer in Discussion
Norway (risk ratio, 0.76) and Poland (risk ratio,
0.84). Figure 2 shows the yearly incidence rate In our large, population-based, randomized tri-
ratios in the invited group as compared with the al, the risk of colorectal cancer at 10 years was
usual-care group. The number needed to invite 0.98% among participants who were invited to
to undergo screening to prevent one case of undergo colonoscopy screening, as compared
colorectal cancer within 10 years was 455 (95% with 1.20% among those who were assigned to
CI, 270 to 1429). Among the participants with receive usual care. Colonoscopy screening was
cases of colorectal cancer for which staging in- performed in only 42% of the participants who
formation was available, 0.38% in the invited were invited to undergo screening. In adjusted
group and 0.44% in the usual-care group re- analyses to estimate the effect of screening if all
ceived a diagnosis of early-stage (stage A or B) the participants who were randomly assigned to
colorectal cancer, whereas 0.40% in the invited screening had actually undergone screening, the

6 n engl j med  nejm.org

The New England Journal of Medicine

Downloaded from nejm.org by JOHN OTREMBA on October 16, 2022. For personal use only. No other uses without permission.
Copyright © 2022 Massachusetts Medical Society. All rights reserved.
 Colonoscopy Screening and Colorectal Cancer

3.5

0.82 (0.70 to 0.93)

0.99 (0.96 to 1.04)

0.90 (0.64 to 1.16)
Risk Ratio
(95% CI) 3.0

2.5

Incidence Rate Ratio

2.0
−0.22 (−0.37 to −0.07)

−0.01 (−0.47 to 0.44)

−0.03 (−0.11 to 0.05)
1.5
percentage points
Risk Difference
(95% CI)

1.0

0.5

0.0
11.04 (10.78 to 11.30)

0 1 2 3 4 5 6 7 8 9 10
0.31 (0.26 to 0.35)
1.20 (1.10 to 1.29)

Years since Randomization

10-Yr Risk
(95% CI)

percent

Figure 2. Incidence Rate Ratios for Colorectal Cancer in the Invited Group
Usual-Care Group

as Compared with the Usual-Care Group in Intention-to-Screen Analyses.

The shaded area indicates 95% confidence intervals. The usual-care group
served as the reference group (red horizontal line).
Participants

number

6079
622

157

100

90 1.75
11.03 (10.66 to 11.40)
0.98 (0.86 to 1.09)

0.28 (0.21 to 0.34)

80 1.50
Cumulative Risk of Death (%)
10-Yr Risk
(95% CI)

percent

70 1.25

60 1.00 Usual-care group

Invited Group

0.31 (95% CI, 0.26–0.35)

50 0.75

0.50 Invited group

40
0.28 (95% CI, 0.21–0.34)
30 0.25
Participants

number

72
3036
259

20 0.00
0 2 4 6 8 10
Table 2. Primary and Secondary End Points.

10

0
0 2 4 6 8 10
Years since Randomization
From colorectal cancer

No. at Risk
Invited group 28,220 27,768 27,224 26,591 25,273 18,856
Usual-care group 56,365 55,469 54,362 53,086 50,356 37,604
From any cause
Colorectal cancer

Figure 3. Cumulative Risk of Death from Colorectal Cancer at 10 Years in In-

tention-to-Screen Analyses.
End Point

The inset shows the same data on an enlarged y axis. I bars indicate 95%
Death

confidence intervals.

n engl j med  nejm.org 7

The New England Journal of Medicine
Downloaded from nejm.org by JOHN OTREMBA on October 16, 2022. For personal use only. No other uses without permission.
Copyright © 2022 Massachusetts Medical Society. All rights reserved.
 The n e w e ng l a n d j o u r na l of m e dic i n e

risk of colorectal cancer was decreased from 1.22% to be higher than the number of colorectal can-
to 0.84%, and the risk of colorectal cancer–­ cer–related deaths during the early part of the
related death was decreased from 0.30% to follow-up period.17,18,21 The lack of a significant
0.15%. Our results may serve to quantify the screening benefit with respect to colorectal can-
effectiveness of screening colonoscopy for the cer–related death in intention-to-treat analyses
prevention of colorectal cancer and thus enable should therefore be interpreted in this context.
decision makers to properly prioritize resources Optimism related to the effects of screening on
for cancer screening and health care services. colorectal cancer–related death may be war-
The overall aim of our trial was to quantify ranted in light of the 50% decrease observed in
the benefits of colonoscopy in population screen- adjusted per-protocol analyses.
ing. Therefore, we chose a pragmatic trial design Although we observed appreciable reductions
with consent after randomization; this design in relative risks, the absolute risks of the risk of
best mimics a population screening program. A colorectal cancer and even more so of colorectal
downside of this approach is that participation cancer–related death were lower than those in
in screening cannot be artificially enriched as in previous screening trials and lower than what
some sigmoidoscopy trials,4 and thus the bene- we anticipated when the trial was planned.7,21
fits of screening may be underestimated.7 To This finding may reflect both a declining risk
address this issue, we performed per-protocol of colorectal cancer observed in many countries
analyses with adjustments for potential con- in recent years and an appreciable improvement in
founders.14 Our estimates of a 31% decrease in the prognosis of colorectal cancer owing to bet-
the risk of colorectal cancer and a 50% decrease ter treatment options. Thus, our estimation of
in colorectal cancer–related death (if all the par- the number needed to invite to screening to
ticipants who were eligible for colonoscopy prevent one case of colorectal cancer was higher
screening had undergone screening) probably than that in the older sigmoidoscopy trials, al-
underestimated the benefit because, as in most though the relative effects were similar.21-24
other large-scale trials of colorectal cancer These findings underscore the importance of
screening, we could not adjust for all important absolute risks and effects when planning cancer
confounders in all countries.17,18 In fact, even screening programs. Comparative absolute bene­
though a 31% decrease in the risk of colorectal fits as well as harms and the burden of colonos-
cancer is a clinically relevant benefit, it is lower copy, sigmoidoscopy, and other screening tests
than that anticipated in clinical guidelines based should be discussed with patients with the use
on observational and modeling studies5,19,20 and of shared decision making to find the best test
similar to the estimates in trials of sigmoidos- on the basis of personal values and preferences.20
copy screening, which have shown a decrease Owing to the small number of events at the
in the risk of colorectal cancer of 23 to 24% in 10-year follow-up, we did not include analyses of
intention-to-treat analyses and 33 to 40% in distal as compared with proximal cancer, sex, or
adjusted per-protocol analyses after a similar age at screening. Continued follow-up in our
follow-up time.21-24 Thus, these results suggest trial and analyses of other ongoing trials may
that colonoscopy screening might not be sub- provide clarity regarding differences with re-
stantially better in reducing the risk of colorectal spect to distal as compared with proximal can-
cancer than sigmoidoscopy. Future analyses of cers, as well as benefits of screening colonos-
our trial results may provide more precise esti- copy after 10 years in women and men.25-27
mates of the per-protocol effects of colonoscopy In the countries with the most participants in
screening for comparison purposes with other the trial, the percentage of participants in the
screening tests. invited group who underwent screening was
Owing to the nature of colonoscopy as a pre- higher in Norway (61%) than in Poland (33%). In
ventive screening test and the nature of colorec- intention-to-screen analyses, the screening bene­
tal cancer, the benefits of endoscopic screening fit was similar in the two countries (risk ratios,
with respect to the risk of colorectal cancer are 0.76 and 0.84, respectively), but in adjusted per-
expected to be apparent earlier than those with protocol analyses, this benefit was estimated to
respect to death related to this disease. Also, the be greater in Norway than in Poland (risk ratios,
number of cases of colorectal cancer is expected 0.55 and 0.85, respectively). Confidence intervals

8 n engl j med  nejm.org

The New England Journal of Medicine

Downloaded from nejm.org by JOHN OTREMBA on October 16, 2022. For personal use only. No other uses without permission.
Copyright © 2022 Massachusetts Medical Society. All rights reserved.
 Colonoscopy Screening and Colorectal Cancer

were overlapping, and event rates were small; and a lack of information about adherence to
thus, no firm conclusions can be drawn. How- recommendations regarding surveillance for
ever, two differences between Norway and Po- polyps. The previously reported variation in
land warrant attention in interpreting our re- quality indicators among endoscopists may have
sults. First, the absolute risk of colorectal cancer resulted in differences in the detection of cancer
in the usual-care group was higher in Norway after screening, but event rates were too small to
than in Poland. The results of trials of sigmoid- investigate further.8 Admittedly, enrollment of a
oscopy screening suggest that relative screening population-based sample probably entails lower
benefits may be smaller with a smaller risk of participation than a trial with randomization
colorectal cancer.4 Second, although the risk of preceded by informed consent. However, our
colorectal cancer in the usual-care group was design should produce a more realistic estimate
lower in Poland than in Norway in our trial, we of benefits and harms in real-life screening pro-
observed a high rate of detection of colorectal grams. Although we adhered to the protocol by
cancer at screening in Poland (Fig. S2C), find- reporting the first results at this time, longer
ings that indicate a strong tendency of partici- follow-up may be needed to capture the full ef-
pants with a high risk of colorectal cancer in the fect of colonoscopy screening.
invited group to undergo screening. In Norway, Supported by research grants from the Research Council of
Norway (197309), Nordic Cancer Union, Norwegian Cancer So-
such a tendency was not observed, which ex- ciety, and Health Fund of South-East Norway (5135); bowel
plains the larger effect of adjustment with re- preparation free of charge for colonoscopies in Norway from Dr.
spect to adherence in Norway than in Poland. Falk Pharma; grants from the National Center for Research and
Development of Poland (N R13 0024 04), Polish–Norwegian
The strengths of our trial are its originality, Research Program (Pol-Nor/204233/30/2013), Medical Center for
its randomized design and considerable size, the Postgraduate Education (501-1-09-12-12/22), the Polish Founda-
fact that participants had not previously under- tion of Gastroenterology, the Dutch Ministry of Health and
Health Care Prevention, Program–Implementation (ZonMw 2008),
gone screening, and the minimal-to-nil screen- the Netherlands Organization for Health Research and Develop-
ing contamination of the control group. Training ment of the Dutch Ministry of Health (ZonMw 120720012), the
programs for endoscopists were implemented, Center for Translational Molecular Medicine (CTMM DeCoDe-
project), and the Swedish Cancer Foundation (2010/345 and
and quality indicators were monitored through- CAN 2013/553); a Distinguished Professor Award from the Karo-
out the trial, as previously reported.8 Finally, linska Institutet, Regional forskningsfond i Uppsala–Örebro
follow-up was virtually complete and the accu- regionen (2368/10-221, to Dr. Adami); and a grant from Afa
(130072).
racy of classification of causes of death was Disclosure forms provided by the authors are available with
considered to be high in all the participating the full text of this article at NEJM.org.
countries.8 A data sharing statement provided by the authors is available
with the full text of this article at NEJM.org.
The limitations of our trial include lower- We thank the NordICC data and safety monitoring board,
than-expected participation in some countries which consisted of Jean Faivre, Jack Mandel, and Stephen Duffy.

Appendix
The authors’ full names and academic degrees are as follows: Michael Bretthauer, M.D., Ph.D., Magnus Løberg, M.D., Ph.D., Paulina
Wieszczy, Ph.D., Mette Kalager, M.D., Ph.D., Louise Emilsson, M.D., Ph.D., Kjetil Garborg, M.D., Ph.D., Maciej Rupinski, M.D., Ph.D.,
Evelien Dekker, M.D., Ph.D., Manon Spaander, M.D., Ph.D., Marek Bugajski, M.D., Ph.D., Øyvind Holme, M.D., Ph.D., Ann G.
Zauber, Ph.D., Nastazja D. Pilonis, M.D., Ph.D., Andrzej Mroz, M.D., Ph.D., Ernst J. Kuipers, M.D., Ph.D., Joy Shi, Ph.D., Miguel A.
Hernán, M.D., Dr.Ph., Hans‑Olov Adami, M.D., Ph.D., Jaroslaw Regula, M.D., Ph.D., Geir Hoff, M.D., Ph.D., and Michal F. Kamin-
ski, M.D., Ph.D.
The authors’ affiliations are as follows: the Clinical Effectiveness Research Group (M. Bretthauer, M.L., P.W., M.K., K.G., M. Buga-
jski, Ø.H., H.-O.A., M.F.K.) and the Department of General Practice, Institute of Health and Society (L.E.), University of Oslo, the
Clinical Effectiveness Research Group, Department of Transplantation Medicine, Oslo University Hospital (M. Bretthauer, M.L., M.K.,
K.G., Ø.H., H.-O.A.), and the Cancer Registry of Norway (G.H.), Oslo, the Institute of Clinical Medicine, University of Tromsø,
Tromsø (M. Bretthauer), and the Department of Research and Development, Telemark Hospital, Skien (G.H.) — all in Norway; the
Department of Gastroenterological Oncology, Maria Sklodowska-Curie Memorial Cancer Center and Institute of Oncology (P.W., M.R.,
M. Bugajski, N.D.P., A.M., J.R., M.F.K.), and the Departments of Pathology (A.M.) and Gastroenterological Oncology (P.W., M.R., M.
Bugajski, N.D.P., A.M., J.R., M.F.K.), Medical Center of Postgraduate Education — both in Warsaw, Poland; Vårdcentralen Värmlands
Nysäter and the Center for Clinical Research, County Council of Värmland, Karlsdad (L.E.), and the Department of Medical Epidemiol-
ogy and Biostatistics, Karolinska Institutet, Solna (L.E., H.-O.A.) — both in Sweden; the Department of Gastroenterology and Hepatol-
ogy, Amsterdam University Medical Centers, and the University of Amsterdam, Amsterdam (E.D.), and the Department of Gastroenter-
ology and Hepatology, Erasmus University Medical Center, Rotterdam (M.S., E.J.K.) — all in the Netherlands; the Department of Epi-
demiology and Biostatistics, Memorial Sloan Kettering Cancer Center, New York (A.G.Z.); and CAUSALab, Department of Epidemiol-
ogy, Harvard T.H. Chan School of Public Health, Boston (J.S., M.A.H.).

n engl j med  nejm.org 9

The New England Journal of Medicine
Downloaded from nejm.org by JOHN OTREMBA on October 16, 2022. For personal use only. No other uses without permission.
Copyright © 2022 Massachusetts Medical Society. All rights reserved.
 The n e w e ng l a n d j o u r na l of m e dic i n e

References
1. Sung H, Ferlay J, Siegel RL, et al. Glob- 10. Hamilton SR, Aaltonen LA, eds. Pa- 20. Helsingen LM, Vandvik PO, Jodal HC,
al cancer statistics 2020: GLOBOCAN esti- thology and genetics of tumors of the di- et al. Colorectal cancer screening with
mates of incidence and mortality world- gestive tract: World Health Organization faecal immunochemical testing, sigmoid-
wide for 36 cancers in 185 countries. CA classification of tumors, 3rd ed. Lyon, oscopy or colonoscopy: a clinical practice
Cancer J Clin 2021;​71:​209-49. France:​IARC Press, 2000;​104-19. guideline. BMJ 2019;​367:​l5515.
2. Helsingen LM, Kalager M. Colorectal 11. Løberg M, Kalager M, Bretthauer M. 21. Holme Ø, Løberg M, Kalager M, et al.
cancer screening — approach, evidence, “Randomize, then consent”, or “consent, Effect of flexible sigmoidoscopy screen-
and future directions. N Engl J Med Evid then randomize”? Revitalizing the issue ing on colorectal cancer incidence and
2022. DOI:​10.1056/EVIDra2100035 of pre-consent randomization. Epidemi- mortality: a randomized clinical trial.
3. Lauby-Secretan B, Vilahur N, Bianchini ology 2016;​27:​393-4. JAMA 2014;​312:​606-15.
F, Guha N, et al. The IARC perspective on 12. Gjøstein DK, Huitfeldt A, Løberg M, 22. Atkin WS, Edwards R, Kralj-Hans I,
colorectal cancer screening. N Engl J Med et al. Incentives and participation in a et al. Once-only flexible sigmoidoscopy
2018;​378:​1734-40. medical survey. Tidsskr Nor Laegeforen screening in prevention of colorectal can-
4. Holme Ø, Schoen RE, Senore C, et al. 2016;​136:​1082-7. cer: a multicentre randomised controlled
Effectiveness of flexible sigmoidoscopy 13. Stensrud MJ, Hernán MA. Why test trial. Lancet 2010;​375:​1624-33.
screening in men and women and differ- for proportional hazards? JAMA 2020;​323:​ 23. Schoen RE, Pinsky PF, Weissfeld JL,
ent age groups: pooled analysis of ran- 1401-2. et al. Colorectal-cancer incidence and mor-
domised trials. BMJ 2017;​356:​i6673. 14. Hernán MA, Robins JM. Per-protocol tality with screening flexible sigmoidos-
5. Davidson KW, Barry MJ, Mangione analyses of pragmatic trials. N Engl J Med copy. N Engl J Med 2012;​366:​2345-57.
CM, et al. Screening for colorectal cancer: 2017;​377:​1391-8. 24. Segnan N, Armaroli P, Bonelli L, et al.
US Preventive Services Task Force recom- 15. Young JG, Stensrud MJ, Tchetgen Tchet­ Once-only sigmoidoscopy in colorectal
mendation statement. JAMA 2021;​ 325:​ gen EJ, Hernán MA. A causal framework cancer screening: follow-up findings of
1965-77. for classical statistical estimands in fail- the Italian randomized controlled trial —
6. Basu P, Ponti A, Anttila A, et al. Status ure-time settings with competing events. SCORE. J Natl Cancer Inst 2011;​103:​1310-
of implementation and organization of Stat Med 2020;​39:​1199-236. 22.
cancer screening in the European Union 16. Cuzick J, Edwards R, Segnan N. Ad- 25. Forsberg A, Westerberg M, Metcalfe
Member States — summary results from justing for non-compliance and contami- C, et al. Once-only colonoscopy or two
the second European screening report. Int nation in randomized clinical trials. Stat rounds of faecal immunochemical testing
J Cancer 2018;​142:​44-56. Med 1997;​16:​1017-29. 2 years apart for colorectal cancer screen-
7. Kaminski MF, Bretthauer M, Zauber 17. Atkin W, Wooldrage K, Parkin DM, ing (SCREESCO): preliminary report of a
AG, et al. The NordICC study: rationale et al. Long term effects of once-only flex- randomised controlled trial. Lancet Gas-
and design of a randomized trial on colo- ible sigmoidoscopy screening after 17 troenterol Hepatol 2022;​7:​513-21.
noscopy screening for colorectal cancer. years of follow-up: the UK Flexible Sig- 26. Dominitz JA, Robertson DJ, Ahnen
Endoscopy 2012;​44:​695-702. moidoscopy Screening randomised con- DJ, et al. Colonoscopy vs. Fecal Immuno-
8. Bretthauer M, Kaminski MF, Løberg trolled trial. Lancet 2017;​389:​1299-311. chemical Test in Reducing Mortality From
M, et al. Population-based colonoscopy 18. Senore C, Riggi E, Armaroli P, et al. Colorectal Cancer (CONFIRM): rationale
screening for colorectal cancer: a Euro- Long-term follow-up of the Italian flexi- for study design. Am J Gastroenterol 2017;​
pean randomized trial. JAMA Intern Med ble sigmoidoscopy screening trial. Ann 112:​1736-46.
2016;​176:​894. Intern Med 2022;​175:​36-45. 27. Quintero E, Castells A, Bujanda L, et al.
9. Sykehuset Telemark. Gastronet — 19. Nishihara R, Wu K, Lochhead P, et al. Colonoscopy versus fecal immunochemi-
colonoscopy center. 2021 (https://www​.­sthf​ Long-term colorectal-cancer incidence and cal testing in colorectal-cancer screening.
.­no/​­helsefaglig/​­g astronet/​­g astronet​ mortality after lower endoscopy. N Engl J N Engl J Med 2012;​366:​697-706.
-­resultater​-­koloskopi​-­senter). Med 2013;​369:​1095-105. Copyright © 2022 Massachusetts Medical Society.

10 n engl j med  nejm.org

The New England Journal of Medicine

Downloaded from nejm.org by JOHN OTREMBA on October 16, 2022. For personal use only. No other uses without permission.
Copyright © 2022 Massachusetts Medical Society. All rights reserved.