Professional Documents
Culture Documents
Interval Colorectal Cancer Incidence Among Subject
Interval Colorectal Cancer Incidence Among Subject
Interval Colorectal Cancer Incidence Among Subject
CLINICAL AT
1
Cancer Center Amsterdam, Department of Gastroenterology and Hepatology, and 3Clinical Epidemiology, Biostatistics and
Bioinformatics, Academic Medical Centre, University of Amsterdam, Amsterdam, The Netherlands; 2Department of
Gastroenterology and Hepatology, and 7Department of Public Health, Erasmus University Medical Centre, Rotterdam, The
Netherlands; 4Department of Research, Netherlands Comprehensive Cancer Organisation (IKNL), Utrecht, The Netherlands;
5
Foundation of Population Screening Mid-West, Amsterdam, The Netherlands; and 6Regional Organization for Population
Screening South-West Netherlands, Rotterdam, The Netherlands
LIMITATIONS
Characteristics
The study did not perform a head-to-head comparison Date of birth, sex, and postal codes of all invitees were
with gFOBT. collected from the municipal population register. Socioeco-
IMPACT nomic status was based on the “social status scores” that are
available through the Netherlands Institute of Social Research
The results support the superior effectiveness of FIT over
gFOBT. (www.scp.nl). The postal code of the invitee was used to
evaluate social status. The social status score of a postal code
area was based on the unemployment rate, educational level,
positive FOBT.7 Other cancer cases will be detected in those average income, and position on the labor market. The average
who did not participate in screening. score is 0 and the 2006 standard deviation in the Netherlands
Monitoring the incidence of interval cancers is regarded was used to assign invitees to 1 of 3 categories: high (status
scores >0.96), average (status scores between 0.96 and 0.96),
as an indicator of program sensitivity and FIT sensitivity. So
and low socioeconomic status (SES) (status scores <0.96).
far, most studies have reported interval cancers of programs
The first available postal code of the invitee was used to
using guaiac-based FOBT, showing high proportions of in-
categorize invitees.
terval cancers. In such programs, the proportion of interval
cancer cases ranged from 48% to 55%.5,6,8,9 These studies
have shown better survival rates in screen-detected cancers Invitations and Stool Tests
than in interval cancers.10 Because of organizational and logistical issues, the time
As fecal immunochemical tests have a better diagnostic interval between rounds differed (median time between in-
accuracy than guaiac-based FOBTs (gFOBTs), it is conceiv- vitations was 2.37 years; interquartile range [IQR] 2.01 to
able that interval cancer proportions are lower than those 2.76). Invitations were sent between June 2006 and December
reported in gFOBT-based programs.11 In this study, we 2014. Two different brands of stool tests were used in our
report FIT interval cancers over multiple rounds of FIT cohort. In the first, second, and third round, all invitees received
screening. an OC-Sensor (Eiken Chemical Co, Tokyo, Japan). In the fourth
round (executed between March 2014 and February 2015), all
invitees were randomly allocated to receive either an OC-
Methods/Materials Sensor or an FOB-Gold (Sentinel Diagnostics SpA, Milan,
Italy).14 A hemoglobin (Hb) value of 10 mg Hb/g feces was used
Population and Design as the positivity threshold; this corresponds to 50 ng/mL buffer
Since 2006, 2 pilot programs of biennial FIT-based CRC for OC-Sensor and 58 ng/mL buffer for FOB-Gold.
screening have been conducted in the southwest and northwest
regions of the Netherlands. These 2 cohorts were combined for
a fourth pilot round of screening in 2014. Details about the Test Results
design of these CRC screening programs have been reported Screenees with a negative FIT result received a letter
previously.12,13 In 2006, we selected the initial 2 cohorts based explaining that no blood had been detected in the stool sample
on postal code areas within our regions, which were both rural and no follow-up was needed at that time. It was emphasized
and urban areas with a known average uptake in breast cancer that the FIT is not 100% sensitive and that vigilance for
screening. Since 2006, we used the same inclusion and exclu- symptoms of CRC remained important. Participants were
sion criteria for both cohorts. In short, demographic data of all instructed to contact their general practitioner in case of
invitees between 50 and 74 years living in the target areas symptoms, despite the negative test result.
were obtained from municipal population registers. See Screenees with a positive FIT result were invited for a
Supplementary Table 1 for baseline invitee characteristics per consultation at the outpatient clinic to discuss the test result.
region. No national screening program had been implemented During the intake at the outpatient clinic, for all person who
at the start of this pilot program; the target population was were FIT-positive, a colonoscopy was scheduled within 4
screening-naive when first contacted. In the Netherlands, a weeks. In some cases, patient-related circumstances led to
national FIT-based CRC screening program has been gradually postponing the colonoscopy. All endoscopies were performed
initiated from January 2014 onward. Invitees for the study within 6 months and, therefore, a timeframe of 6 months after a
cohort were not invited for the national CRC screening positive FIT was used for the definition of screen-detected
program. cancers.
August 2017 FIT Interval Cancers in Colorectal Cancer Screening 441
CLINICAL AT
FIT before the next FIT was due
after removal of adenomatous polyps, large (10 mm) serrated
2. Colonoscopy interval cancer within a FIT
lesions, or cancer was given to the participant according to the screening program defined as cancers diag-
Dutch guideline colonoscopy surveillance.16 Data on the loca- nosed after negative colonoscopy after a posi-
tion, size, macroscopic aspects, and morphology, as well as tive FIT within the surveillance interval
details on the technique for polypectomy and endoscopic 3. Colorectal cancer in nonparticipants defined as
assessment of radicality were recorded for all colorectal lesions cancers diagnosed in those who did not
detected during colonoscopy. Tumor location was categorized respond to their FIT invitation.
as either proximal or distal to the splenic flexure. Collected General The reference group was defined as all persons
lesions were evaluated by an experienced gastrointestinal population diagnosed with CRC in the Netherlands
population during the same period and in the
pathologist, using the Vienna criteria.17 Cancers were staged
same age range (50–76 years), who were not
per the seventh edition American Joint Committee on Cancer
offered CRC screening.
classification.18
Ethics Approval found for age at diagnosis between participants with SD-
Approval for the study was provided by the Dutch National CRC and individuals with non–SD-CRC (P ¼ .83) (Table 2).
Health Council (WBO 2642467, 2832758, 3049078, and A colonoscopy interval cancer was detected more often in
161536–112008, The Hague, The Netherlands). All authors had women than in men (56% vs 44%), but this difference was
access to the study data and reviewed and approved the final not statistically significant. There were also no significant
manuscript. differences in SES among the groups (P ¼ .76). The age and
sex distribution was comparable with that of the general
population (not tested for statistical significance).
Results Among the 9 participants with a colonoscopy interval
A total of 27,304 members of the target population were cancer, the colonoscopies were reported as complete with
eligible for FIT screening, of whom 18,716 (69%) partici- cecal intubation. In 6 cases, sufficient bowel preparation
CLINICAL AT
pated at least once (Figure 1). Of these, 3005 (16%) had a was reported, in 3 cases these data were missing. Three
positive FIT in 1 of the 4 screening rounds and 2762 sub- participants received a colonoscopy surveillance advice, but
sequently underwent a colonoscopy (92% adherence). In developed a CRC within this interval. One person received
the total cohort of 27,304, CRC was detected in 269 invitees: an advice of a 3-year interval and 1 person a 6-year interval,
in 160 participants (60%) and in 109 who never partici- and both recommendations were in agreement with the
pated (40%). The cancers detected in the 160 participants Dutch Surveillance Guideline of 2008. One person received
were 116 SD-CRCs (72%), 27 FIT interval cancers (17%), 9 an adjusted advice of a 1-year surveillance interval because
colonoscopy interval cancers (6%), 5 CRCs in FIT-positive multiple adenomas were removed at index colonoscopy. Six
participants not adhering to colonoscopy (3%), and 3 participants were advised a 10-year screening interval as
CRCs detected at surveillance (2%). These last 2 groups, per protocol (see Supplementary Table 4).
with a total of 8 CRCs, were excluded from further analyses.
Tumor Location and Stage Distribution
FIT Interval Cancer Proportion and FIT Tumor location and CRC stage distribution are described
CRC Sensitivity in Table 3 and Figure 2. Screen-detected CRC, FIT interval
The FIT interval cancer proportion was 23% (27 FIT in- cancers, and CRC in nonparticipants were mostly located in
terval cancers versus 89 SD-CRCs after 3 completed the distal colon (71%, 63%, 62%, respectively), whereas
screening rounds), reflecting a FIT sensitivity for detecting colonoscopy interval cancers were more often located in the
CRC of 77%. FIT levels of participants with SD-CRCs and FIT proximal colon (78%; P ¼ .063).
interval cancers are shown in Supplementary Table 2; 31% of The stage distribution differed significantly among the 4
those with a SD-CRC had a positive FIT between the range of groups, with more favorable stages in participants with SD-
10 to 46 mg Hb/g feces. Twelve participants (44%) with a FIT CRC (P < .001). Stage distribution was similar for partici-
interval cancer had undetectable hemoglobin concentrations pants with FIT interval CRC and for nonparticipants with CRC
at the time of the last screening. For FIT interval cancers, (P ¼ .39). Participants with FIT interval cancers had more
Supplementary Table 3 shows the quantitative FIT level per advanced stages than SD-CRC (P ¼ .019). Tumor location and
case, the total number of rounds in which they participated, stage distribution in nonparticipants with CRC were compa-
and after which round the FIT interval cancer was diagnosed. rable to those not offered screening (see Table 3).
Non–SD-CRC
CLINICAL AT
Male 73 (63) 16 (59) 4 (44) 69 (63) 42,486 (59)
Female 43 (37) 11 (41) 5 (56) 40 (37) 30,126 (42)
SES score, n (%) .763
Low 13 (11) 2 (7) 2 (22) Unknown Unknown
Average 82 (71) 21 (78) 6 (67) Unknown Unknown
High 21 (18) 4 (15) 1 (11) Unknown Unknown
a
Four cases with 2 incidents, age at diagnosis first incident selected.
months (IQR 18.1 to 72.1 months). There was a significant gFOBT-based screening. Over 4 rounds of screening, fewer
difference in survival among SD-CRC, FIT interval cancers, than 1 in 4 cancer cases comprised FIT interval cancers. We
and colonoscopy interval cancers (P < .001; Figure 3). observed a more favorable survival for participants with FIT
Those with colonoscopy interval cancers had the lowest interval cancers than for the population that was not offered
survival: survival at 24 months was 67% (95% CI, 35%– screening (ie, general population). Participants with a colo-
88%) compared with 85% (95% CI, 68%–94%) for FIT noscopy interval cancer had the worst outcome, which am-
interval cancers and 93% (95% CI, 87%–96%) for SD-CRCs. plifies the importance of colonoscopy quality control.
Survival at 36 months was 56% (95% CI, 27%–81%), 85% We are the first to describe interval cancers in a FIT-
(95% CI, 68%–94%) and 91% (95% CI, 85%–95%), based screening program with long-term follow-up. Our
respectively. Table 4 lists the number of deaths in the cohort is sampled from an average-risk population,
respective groups. The number of deaths among persons comprising all age ranges commonly invited for CRC
with FIT interval cancers was lower than among non- screening programs worldwide, which makes our results
participants with CRC and those not offered CRC screening applicable for many countries considering CRC screening.
(19% vs 40% and 35%, respectively). This population was screen-naive when first approached,
without the presence of any other CRC screening initiatives
in the population. All data were prospectively collected and
Discussion all invitees were linked to the Netherlands Cancer Registry
This study confirms that the proportion of interval can- by the Netherlands Comprehensive Cancer Organization to
cers in FIT-based CRC screening seems to be lower than in identify all non–SD-CRCs.
Non–SD-CRC
a
Proximal (¼ splenic flexure and proximal).
444 van der Vlugt et al Gastroenterology Vol. 153, No. 2
CLINICAL AT
The success of a FIT-based CRC screening program not before being invited for screening. In such case, refraining
only depends on FIT test characteristics (sensitivity and from participation would have been a justified decision.
specificity) but also on adherence to screening with FIT and Because FIT is more sensitive in detecting advanced
to follow-up colonoscopy in persons who are FIT-positive. neoplasia than gFOBT, one would expect lower interval
Within our pilot study, high participation rates and high rates for FIT. A previous Scottish FIT-based study, however,
colonoscopy adherence rates were achieved, resulting in a reported a FIT interval cancer proportion of 51% using a
substantial number of persons diagnosed with SD-CRC. To cutoff of 80 mg Hb/g feces.22 Italian researchers reported a
fully appreciate the reported data of this research study, FIT interval cancer proportion of 31% using a cutoff of 20
these adherence rates should be confirmed in the setting of mg Hb/g feces after 2 rounds of FIT screening.23 Our lower
an organized population-based screening program. In 2014, proportion of FIT interval cancers could probably be
in the Netherlands, a national FIT-based CRC screening explained by the substantially lower cutoff (10 mg Hb/g
program was implemented, resulting in even higher feces), thereby detecting CRCs already at low Hb concen-
participation rates: 73% in 2015. Excluding those for whom tration. In our study, 31% of all SD-CRCs had low Hb con-
colonoscopy was not recommended (n ¼ 2637) during centration of <47 mg Hb/g feces. The Scottish study
intake, uptake of colonoscopy was 83.1%.19 This confirms investigated the effect of different cutoff levels of fecal Hb
that the high performance indicators achieved within the on interval cancer proportions and colonoscopy demand. It
pilot study could possibly also be achieved in the setting of reported an interval cancer proportion of 51% at a cutoff
the national organized FIT-based screening program. of 80 mg Hb/g feces vs a proportion of 38% using a cutoff of
To fully appreciate our findings, some limitations also 10 mg Hb/g feces with a significant increase in number of
need to be addressed. Little information was available about required colonoscopies for the latter cutoff.22 Hence,
nonparticipants dying of CRC in our cohort, as these data choosing a very low positivity cutoff leads to a higher
were delivered anonymously for privacy reasons. Conse- number of SD-CRCs and lower proportion of FIT interval
quently, their SES could not be assessed. Previous studies cancers, at the cost of increased colonoscopy demand and
have shown that especially socioeconomically deprived increased numbers of false positives.22
persons are less likely to take part in CRC screening, even Our results showed no evidence for gender differences
though they presumably are at a higher risk of developing in interval cancer rates, although previous studies have
CRC due to poor general health.20,21 Last, their date of shown higher rates of interval cancers among women.8,23–25
diagnosis could not be related to the date of invitation. It is As in other studies, CRC was detected less frequently in
possible that a person had already been diagnosed with CRC women. Notably, more women had a colonoscopy interval
August 2017 FIT Interval Cancers in Colorectal Cancer Screening 445
CLINICAL AT
Figure 3. Survival curves.
cancer than men (56% vs 44%), yet this difference did not Comparable early-stage distributions in FIT screening have
reach statistical significance. These results are in line with a been reported.22,26 Most detected tumors were located
Spanish study, reporting SD-CRC and interval cancers dur- distally, except for colonoscopy interval cancers, which were
ing 4 rounds of FOBT screening mainly using gFOBT and FIT mainly located in the proximal colon. These findings are also
in a small proportion of screenees.9 The Scottish researchers in line with previous studies.9,28
also reported higher rates of FIT interval cancers in women, In contrast to previous literature reporting that prog-
without reaching statistical significance.22,24 nosis of nonparticipants with CRC is poorer than that of
Previous gFOBT-based studies have shown that most SD- participants with symptomatically diagnosed CRC, we did
CRC cases were at an early stage, resulting in a better survival not observe this from our data.5,24 The number of deaths
than observed in non–screen-detected cancer cases.4,5,10,26 In was significantly lower in participants with FIT interval
our study, most screen-detected cases were stage I cancers, cancers than in nonparticipants with CRC. This is in contrast
which confirms the findings of previous studies.9,10,24,27 to findings of an English study, which reported similar
In our FIT-based program, a higher percentage of stage I outcomes for gFOBT interval cancers and a control group.8
CRCs were found than in previously reported gFOBT Persons diagnosed with FIT interval cancers showed bet-
screening programs. This could possibly be explained by the ter survival as compared with nonparticipants and the
relatively low FIT cutoff and the higher sensitivity of FIT. general population. This implies that the theoretical risk of
446 van der Vlugt et al Gastroenterology Vol. 153, No. 2
Table 4.Mortality
Non-SD CRC
a
At end of cohort.
12. Stegeman I, van Doorn SC, Mundt MW, et al. Participa- 25. Brenner H, Chang-Claude J, Seiler CM, et al. Interval
tion, yield, and interval carcinomas in three rounds of cancers after negative colonoscopy: population-based
biennial FIT-based colorectal cancer screening. Cancer case-control study. Gut 2012;61:15768–15762.
Epidemiol 2015;39:388–393. 26. Pande R, Froggatt P. Baragwanath. Survival outcome
13. Kapidzic A, Grobbee EJ, Hol L, et al. Attendance and of patients with screening versus symptomatically
yield over three rounds of population-based fecal detected colorectal cancers. Colorect Dis 2012;15:
immunochemical test screening. Am J Gastroenterol 74–79.
2014;109:1257–1264. 27. Mansouri D, McMillan DC, McIlveen E, et al.
14. Grobbee EJ, van der Vlugt M, van Vuuren AJ, et al. A A comparison of tumour and host prognostic factors in
randomised comparison of two faecal immunochemical screen-detected versus non screen-detected colorectal
tests in population-based colorectal cancer screening cancer: a contemporaneous study. Colorectal Dis 2016;
[published online ahead of print August 9, 2016]. Gut doi: 18:967–975.
CLINICAL AT
http://dx.doi.org/10.1136/gutjnl-2016-311819. 28. le Clercq CM. Sanduleanu S. Interval colorectal cancers:
15. Rembacken B, Hassan C, Riemann JF, et al. Quality in what and why. Curr Gastroenterol Rep 2014;16:375.
screening colonoscopy: position statement of the Euro- 29. Singh H, Nugent Z, Demers A, et al. Rate and predictors
pean Society of Gastrointestinal Endoscopy (ESGE). of early/missed colorectal cancers after colonoscopy in
Endoscopy 2012;33:957–968. Manitoba: a population-based study. Am J Gastroenterol
16. Nederlandse richtlijn coloscopie surveillance NVMDL nov 2010;105:2588–2596.
2013. Available at: https://www.mdl.nl/sites/www.mdl.nl/ 30. Bressler B, Paszat LF, Chen Z, et al. Rates of new or
files/richlijnen/Richtlijn_Coloscopie_Surveillance_ missed colorectal cancers after colonoscopy and their
definitief_2013.pdf. Accessed June 28, 2017. risk factors: a population-based analysis. Gastroenter-
17. Schlemper RJ, Riddell RH, Kato Y, et al. The Vienna ology 2007;132:96–102.
classification of gastrointestinal epithelial neoplasia. Gut 31. le Clercq CM, Bouwens MW, Rondagh EJ, et al. Post-
2000;47:251–255. colonoscopy colorectal cancers are preventable: a
18. Edge SB, Compton CC. The American Joint Committee population-based study. Gut 2014;63:957–963.
on Cancer: the 7th edition of the AJCC cancer staging 32. Stoffel E, Erichsen R, Froslev T, et al. Clinical and mo-
manual and the future of TNM. Ann Surg Oncol 2010; lecular characteristics of post-colonoscopy colorectal
17:1471–1474. cancer: a population-based study. Gastroenterology
19. Toes-Zoutendijk E, van Leerdam ME, Dekker E, et al. 2016;151:870–878.e3.
Real-time monitoring of results during first year of Dutch
colorectal cancer screening program and optimization by
altering fecal immunochemical test cut-off levels. Received January 15, 2017. Accepted May 1, 2017.
Gastroenterology 2017;152:767–775.e2.
Reprint requests
20. Pornet C, Dejardin O, Morlais F, et al. Socioeconomic Address requests for reprints to: Evelien Dekker, MD, PhD, Department of
determinants for compliance to colorectal cancer Gastroenterology and Hepatology, Academic Medical Center, Meibergdreef
screening. A multilevel analysis. J Epidemiol Community 9, 1105 AZ, Amsterdam, The Netherlands. e-mail: e.dekker@amc.uva.nl;
fax: þ31 20 691 7033.
Health 2010;64:318–324.
21. Moss SM, Campbell C, Melia J, et al. Performance Acknowledgments
The authors thank the registration teams of the Netherlands Comprehensive
measures in three rounds of the English bowel cancer Cancer Organisation for the collection of data for the Netherlands Cancer
screening pilot. Gut 2012;61:101–107. Registry and the scientific staff of the Netherlands Comprehensive Cancer
Organisation. We thank the Netherlands Organization for Health Research
22. Digby J, Fraser C, Carey F, et al. Interval cancers using a and Development of the Dutch Ministry of Health (ZonMW) for funding
quantitative faecal immunochemical test (FIT) for hae- (project numbers 120710007, 63000004, 200340001). The authors thank all
moglobin when colonoscopy capacity is limited. J Med involved coworkers of the Foundation of Population Screening Mid-West and
South-West (Bevolkingsonderzoek Midden-West, Bevolkingsonderzoek Zuid-
Screen 2016;23:130–134. West) for their important contributions to the study. We thank the
23. Zorzi M, Fedato C, Grazzini G, et al. High sensitivity of five gastroenterologists who performed all colonoscopies.
colorectal screening programmes with faecal immuno- Conflicts of interest
chemical test in the Veneto Region, Italy. Gut 2011; All authors have nothing to disclose.
60:944–949.
Funding
24. Steele RJC, McClements P, Watling C, et al. Interval The Netherlands Organization for Health Research and Development of the
cancers in a FOBT-based colorectal cancer population Dutch Ministry of Health (ZonMW) funded all 4 screening rounds (project
numbers 120710007, 63000004, 200340001). The organization was not
screening programme: implications for stage, gender and involved in the analysis and interpretation of the data nor in the writing of the
tumour site. Gut 2012;61:576–581. manuscript or decision to submit the paper.
447.e1 van der Vlugt et al Gastroenterology Vol. 153, No. 2
MW SW MW SW MW SW MW SW
0 – 12 (44)
1–5 – 7 (26)
6–9 – 5 (19)
10–20 20 (17) –
21–46 16 (14) –
47 80 (69) –
Missing – 3 (11)
1 MW 1 1 1 0.94 1.8
2 MW 1 1 1 2.34 –
3 SW 1 1 1 0.5 0.4
4 MW 1 1 1 1.9 0
5 MW 1 1 1 1.53 –
6 SW 1 1 1 0.67 4.8
7 MW 1 1 1 0.48 –
8 SW 2 2 2 0.92 9.6
9 MW 2 2 2 0.94 0
10 MW 2 2 2 2.07 5.0
11 SW 2 2 2 0.67 0
12 MW 2 1 1 1.94 6.0
13 MW 2 2 2 1.93 3.8
14 MW 2 2 2 1.52 0
15 MW 2 2 2 1.59 0
16 SW 2 2 2 0.42 0
17 MW 2 2 2 0.42 9.9
18 MW 2 2 2 1.95 0
19 MW 2 2 1 1.53 3.0
20 MW 3 3 3 1.61 7.6
21 SW 3 3 3 2.24 0
22 MW 3 2 2 1.41 0
23 SW 3 3 3 1.58 0
24 SW 3 3 2 1.08 0
25 MW 3 3 3 2.25 5.4
26 MW 3 3 3 2.29 7.6
27 MW 3 1 1 2.34 0
1 SW Nov 2008 Jan 2009 June 2011 2.58 8 adenoma with Excellent Yes 3y
1 HGD sigmoid
2 SW Feb 2010 March 2010 Febr 2012 1.92 HP – Yes No FU
3 MW Jan 2007 March 2007 July 2008 1.51 No findings Good Yes No FU
4 SW Nov 2009 Nov 2009 Nov 2012 3.00 One TA <10 mm – Yes 6 y
5 SW April 2009 June 2009 June 2013 4.08 HP Good Yes No FU
6 SW May 2009 July 2009 Sept 2012 3.33 >10 adenomas, 1 TVA Good Yes 1 y
7 SW Nov 2008 Dec 2008 Nov 2009 1.00 HP – Yes No FU
8 MW Feb 2012 April 2012 Nov 2013 1.77 No findings Excellent Yes No FU
9 SW June 2008 July 2008 Jan 2015 6.50 No findings Good Yes No FU
FU, follow up; HGD, high-grade dysplasia; HP, hyperplastic polyp; MW, Midwest region; SW, Southwest region; TA, tubular
adenoma; TVA, tubulovillous adenoma with low-grade dysplasia.