NordICC

Published

Bretthauer M, et al. "Effect of colonoscopy screening on risks of colorectal cancer and related death". The New England Journal of Medicine. 2022. (e-pub 2022-10-09):1-10.
PubMed • Full text • PDF • ClinicalTrials.gov

Clinical Question

In European countries that do not routinely screen for colorectal cancer, does a mailed letter inviting adults 50-64 for a screening colonoscopy reduce risk of incident colorectal cancer and colorectal cancer mortality?

Bottom Line

In European countries that do not routinely screen for colorectal cancer, a mailed letter inviting participants for a screening colonoscopy was associated with lower risk of incident colorectal cancer but not colorectal cancer mortality in intention-to-screen analyses.

Major Points

In the US, colorectal cancer (CRC) is the 3rd leading cause of cancer mortality among men and women. The lifetime risk is ~3% for men and ~2% for women. The widely-accepted colorectal tumorigenesis theory proposes that most CRCs originate from sequential mutations. Central to this theory is the development of resectable precancerous polyps that can be identified on a screening modality and removed, with removal typically completed via colonoscopy, to prevent progression to CRC. There are several screening modalities that may be used, including fecal immunohistochemical testing (FIT), multitarget stool DNA testing (e.g., Cologuard), fecal occult blood testing (FOBT), barium enema, sigmoidoscopies, and colonoscopies. These modalities are used at regular intervals for screening by themselves or in combination, but typically any abnormal screening leads to a colonoscopy for definitive diagnosis and potential preventive intervention (i.e., polypectomy). Prior to 2022, FOBT and sigmoidoscopies were the only screening modalities with high-quality trial evidence supporting their use to reduce CRC mortality. Screening sigmoidoscopy, which visualizes the rectum and sigmoid colon, was shown to reduce mortality in an RCT of >100,000 persons in the UK in the UK Flexible Sigmoidoscopy Trial (2010). This trial randomized individuals who were interested in having CRC to an invitation to screening or no invitation. Invitation to screening (and, an implied downstream sigmoidoscopy) was associated with a 23% reduction in CRC incidence and 31% reduction in mortality. Importantly, 71% of invited individuals attended their screening. Of these, 5% were later referred for colonoscopy because of detection of high-risk polyps.

Sigmoidoscopies, which like colonoscopies have burdensome preps, do not visualize more proximal portions of the colon. A seminal 2000 publication from the VA described that colonoscopies in a cohort of 3121 asymptomatic veterans aged 50-75 years, 40% of screenees had a pre-cancerous lesion detected. Most patients with proximal precancerous lesions also had distal precancerous lesions, which would have presumably led to a colonoscopy. However, among the 61% with normal rectum and sigmoid colon, 71 (4%) had proximal precancerous lesions that would have been missed on sigmoidoscopy that were identified on colonoscopy. Another paper published simultaneous with the VA analysis described a cohort of 1,994 patients at one medical center who underwent screening colonoscopies. Of the 50 precancerous or cancer lesions that were found, 23 (46%) were among individuals with normal distal colons that would not have led to colonoscopy. Another seminal paper describe a decrease in left-sided CRC and no change in right-sided CRC between 1978-1998. The authors hypothesized that the reduction in left-sided but not right-sided CRC may relate to use of sigmoidoscopies in that timeframe, leading to lack of screening of the right colon. And not all colonoscopies are equal. A 2014 study from Kaiser Permanente compared risk of CRC mortality by quality of gastroenterologist, as determined by the proportion of adenoma detection rate for each gastroenterologist (# of colonoscopies performed that detected an adenoma/# of colonoscopies performed). In this analysis, quintile 1 had a detection rate of 7 to 19%, while quintile 5 had a detection rate from 34-52%. The risk of incident cancer was reduced by about half, and about 2/3rds for CRC mortality, in the 5th relative to 1st quintile. From this, a defined Medicare quality metric is proportion of screening colonoscopies with a detected adenoma, noting that guidelines and clinical evidence support an adenoma detection rate of 25% among adults (20% among women and 30% among men).

In 2001, Medicare began to cover colonoscopies for cancer screenings, which drove the use of colonoscopies as the predominate screening modality for CRC in the US. This move has been criticized for several reasons: (1) There was no clinical trial evidence supporting the use of colonoscopies to reduce incident CRC or CRC mortality in 2001, and (2) Colonoscopies have higher reimbursement than other modalities, meaning they are more costly than other options. To shore up evidence, two large trials were designed. One is COLONPREV, which randomized 53,000 Spanish adults aged 50-69 years to FIT q2y or a one-time colonoscopy between 2009-2011. As of 2022, ten-year follow-up results have not been published. The other trial is Nordic-European Initiative on Colorectal Cancer (NordICC) trial, published in 2022.

NordICC randomized 94,959 adults in Poland, Norway, Sweden, and the Netherlands to participate in a colonoscopy or no invitation to colonoscopy in a 1:2 fashion. The present manuscript reports the 10-year follow up data, excluding the Netherlands because of data sharing issues related to a local law. In the end, 42% of those invited to have a colonoscopy actually underwent screening. Invitation to screen was associated with an 18% reduction in incident CRC (0.98 vs. 1.20% 10-year risk; RR 0.82, 95% CI 0.70 to 0.93). There was no difference in CRC mortality in the intention-to-screen analysis (0.28 vs 0.31% 10-year risk; RR 0.90, 95% CI 0.64 to 1.16), though there was in the per-protocol analysis (0.15 vs. 0.30% 10-year risk; RR 0.50, 95% CI 0.27 to 0.77). There were few complications from the colonoscopy.

NordICC has several limitations, which are detailed in the Criticisms section below. The attendance at colonoscopy was lower at 42% (expected was 50%). Perhaps this might have been from a lack of a prescreening step that occurred in the UK Flexible Sigmoidoscopy Trial that ensured invited persons were at least interested. The UK Flexible Sigmoidoscopy study had 71% attendance among invitees. Other limitations in NordICC are inherent to pragmatic trials in general. "Explanatory" (i.e., more conventional) RCTs have narrow eligibility criteria, close follow-up, and a protocol delivered in a highly controlled environment, that intentionally maximizes internal validity at the cost of external validity. In contrast, pragmatic trials are often implemented within routine clinical care and aim to answer more generalizable questions about how an intervention might affect usual practice. It is reasonable to argue that NordICC, as a pragmatic trial, is not intended to answer questions about how much CRC incidence and mortality is reduced when applied to an adult aged 50-64. Instead, it tests how inviting participants to a screening program with an expected 50% attendance rate affects CRC patterns in several European countries. This manuscript reports the 10 year follow-up for most participants but it was not powered appropriately for this interval analysis (which itself was missing ~10% of participants in the Netherlands), it was instead powered for 15 year follow-up. Ultimately, NordICC provides solid evidence that invitation to colonoscopy -- even with poor attendance at the colonoscopy -- in countries without routine CRC screening is associated with lower incident CRC. That CRC mortality did not differ between arms in the intention-to-screen analysis but did in the per-protocol analysis confuses the interpretation of this trial. The inherent limitations of pragmatic trials means that this trial is not optimized for internal validity (i.e., does colonoscopy reduce CRC and CRC mortality?) and instead maximizes generalizability (i.e., what happens when we mail letters inviting persons to a colonoscopy or not, in a population without routine CRC screening?). There is solid evidence that sigmoidoscopy reduces cancer mortality, and a colonoscopy incorporates all components of a sigmoidoscopy plus proximal colon screening and polypectomy, so it's reasonable to surmise that the lack of evidence of CRC mortality reduction relates to the specific implementation of screening in this trial, or perhaps that it is underpowered to detect an effect at 10 years. The estimated intention-to-screen RR should not be interpreted like an explanatory RCT that estimates how much an intervention reduces an outcome when isolating that signal from noise. Readers should be cautioned about interpretation of the per-protocol analyses as well, since per-protocol analyses introduce bias and do not flip this pragmatic trial to an explanatory trial. Hopefully the 15-year follow up of NordICC and the impending results of COLONPREV will provide additional insight into the implications of the use of invitations for colonoscopy-based CRC screening in populations.

Guidelines

As of October 2022, no guidelines have been published that reflect the results of this trial.

Design

• Multicenter, pragmatic, open-label, randomized trial
• N=84,585 of 94,959 persons invited to participate as individuals from Netherlands were not included in this manuscript
  • Invited to screen (n=28,220)
  • No invitation to screen so received usual care group (n=56,365)
• Setting: Poland, Norway, Sweden, and the Netherlands
• Screening timeframe: 2009-2014
• Median follow-up: 10 years
• Analysis: Intention-to-screen
• Primary outcomes:
  • Incident CRC
  • CRC mortality

Population

Inclusion Criteria

• Aged 55-64 years, living in geographical rage of interest

Exclusion Criteria

• Prior colorectal surgery
• Requiring long-term nursing care
• Treatment for cancer with chemotherapy or radiation
• Severe cardiac or pulmonary disease
• Extended anticoagulation with warfarin
• CVD hospitalization in prior 3 mo
• Cerebrovascular event in prior 3 mo
• Lives abroad
• Unable to reach by mail

Baseline Characteristics

The Table 1 is atypical in that it presents the analytical population by country and not by group assignment; scant details by group assignment appear in Table S3 of the supplementary appendix.' From the screening group

• Women: 50%
• Age group: 55-59 years 51%, 60-64 years 49%
• Race/ethnicity: Not recorded.
• Country: Norway 31%, Poland 64%, Sweden 4%, Netherlands 0% (not included in this analysis because of a Dutch law)

Interventions

• Participants were randomized by household (i.e., if another member of that household was also eligible, they were randomized to the same arm together)
  • Invited to screen - Mailed a letter with an appointment for colonoscopy 6-7 weeks out, and also details about where to get colon prep and how to take it. A reply form allowed participants to decline, and/or also provide health details and telephone number for screening. Participants were contacted 3 days prior to the appointment to confirm their planned attendance.
    • At the colonoscopy, all precancerous lesions were removed.
  • No invitation to screen so received usual care - Not offered screening. In the protocol, the authors seem to imply that no other screening modality is used in any of the participating countries.

Outcomes

Presented as invited to screen vs. no invitation to screen so received usual care group. RD is risk difference. RR is risk ratio. The primary analysis is intention-to-screen and is shown except where specified.

Primary Outcomes

Incident CRC

0.98 vs. 1.20% 10-year risk (RD -0.22%, 95% CI -0.37 to -0.07%; RR 0.82, 95% CI 0.70 to 0.93)

Per-protocol: 0.84 vs. 1.22% 10-year risk (RD -0.38%, 95% CI -0.58 to -0.20%; RR 0.69, 95% CI 0.55 to 0.83)

CRC mortality

0.28 vs 0.31% 10-year risk (RD -0.03%, 95% CI -0.11 to 0.05%; RR 0.90, 95% CI 0.64 to 1.16)

Per-protocol: 0.15 vs. 0.30% 10-year risk (RD -0.15, 95% CI -0.24 to -0.07; RR 0.50, 95% CI 0.27 to 0.77)

Secondary Outcomes

All-cause mortality

11.03 vs. 11.04% 10-year risk (RD -0.01, 95% CI -0.47 to 0.44; RR 0.99, 95% CI 0.96 to 1.04)

Additional Details and Outcomes

Participation in screening among those invited to be screened

All: 42%

By sex

Women 41%

Men': 43%

By age

55-59y: 41%

60-64y: 43%

By country

Norway: 61% of 8,815 invited

Poland: 33% of 18,184 invited

Sweden: 40% of 1,221 invited

Colonoscopy details

Good or very good prep quality: 91%

Intubation of cecum: 97%

Detection of adenomas: 31%

Norway: 27%

Poland: 35%

Sweden: 14%

Colonoscopies/screening in the "no invitation to screen so received usual care group"

The authors note that they "did not observe screening of any meaningful extent in the usual-care group." It's not clear how they determined this since they do not comment on if this population was monitored for CRC screening in their analytical database

Diagnosis of CRC at the time of colonoscopy

62 (0.5%)

Stage at CRC diagnosis

A: 0.14% vs. 0.14%

B: 0.24% vs. 0.30%

C: 0.23% vs. 0.31%

D: 0.17% vs. 0.18%

Unknown: 0.13% vs. 0.17%

Subgroup Analysis

By country

In Table S5 of the supplementary appendix.

Norway: 1.20 vs. 1.57% 10-year risk (RD -0.37, 95% CI -0.68 to -0.07; RR 0.76, 95% CI 0.58 to 0.94)

Poland: 0.83 vs. 0.99% 10-year risk (RD -0.16, 95% CI -0.33 to 0.02; RR 0.84, 95% CI 0.67 to 1.01)

Sweden: Not provided

Adverse Events

Among those receiving screening colonoscopy.

Perforation

None

Major bleeding

Defined as bleeding that warranted treatment.

0.13%

Criticisms

• The authors did not appear to systematically collect or report data on CRC screening that might have occurred outside of this protocol, though they explain at great lengths in this manuscript how they think it probably wasn't done in any of the participating countries to a substantial degree.
• Limited reporting of population characteristics, inadequate reporting of randomization scheme, minimal subgroup analyses.
• Data from the Netherlands was not included reportedly because of a local law related to data protection. The authors note in the protocol that they intended to recruit participants from Iceland, though none were included, perhaps because of lack of funding.
• The adenoma detection in Sweden was 14%, which equates with the low-quality screening by Medicare quality standards, which likely attenuated any observed benefit from colonoscopy in this country However, this was only 4% of the invited population.
• Lower than expected colonoscopy completion in Poland with led to underpowering with the initial recruitment target, though the authors note that they increased the enrollment target to account for the lack of participation in Poland. Of note, the authors based their power calculation on a and 50% compliance, a 50% screening efficacy (presumably reduction in CRC incidence, which is what was observed in the per-protocol analysis), and 25% CRC mortality reduction (~16% reduction in per-protocol analysis), gave 64% power at 10 years and 90% power at 15 years. The present 10-year analysis is therefore underpowered.
• No reported per-protocol all-cause mortality estimates.

Funding

Multiple funders, presumably all major funders were public.