Screening for Colorectal Cancer: Updated Evidence Report and Systematic Review for the U.S. Preventive Services Task Force

Abstract

Importance:

Colorectal cancer (CRC) remains a significant cause of morbidity and mortality in the United States.

Objective:

Systematically review the effectiveness, diagnostic accuracy, and harms of screening for CRC.

Data Sources:

Searches of MEDLINE, PubMed, and the Cochrane Central Register of Controlled Trials for relevant studies published from January 1, 2008, through December 31, 2014, with surveillance through February 23, 2016.

Study Selection:

English-language studies conducted in asymptomatic populations at general risk of CRC.

Data Extraction and Synthesis:

Two reviewers independently appraised the articles and extracted relevant study data from fair- or good-quality studies. Random-effects meta-analyses were conducted using the profile likelihood and restricted likelihood estimation methods.Main Outcomes and Measures: CRC incidence and mortality, test accuracy to detect CRC or adenomas, serious adverse events.

Results:

Large pragmatic randomized clinical trials evaluating one- or two-time flexible sigmoidoscopy (N=458,002) was associated with decreased CRC-specific mortality compared to no screening (incidence rate ratio, 0.73 [95% CI, 0.66–0.82]). RCTs with multiple rounds of biennial screening with Hemoccult II (N=419,966) showed reduced CRC-specific mortality (relative risk [RR], 0.91; 95% CI, 0.84–0.98 at 19.5 years to RR, 0.78; 95% CI, 0.65–0.93 at 30 years). Studies of the diagnostic accuracy of computed tomographic colonography (CTC) with bowel preparation demonstrated that per-person sensitivity and specificity to detect adenomas ≥6 mm, was comparable to colonoscopy (sensitivity ranged from 73% [95% CI, 58%–84%] to 98.0% [95% CI, 91%–100%], specificity ranged from 88.8% [95% CI 84%–93%] to 91% [95% CI, 88%–93%]). Variability and imprecision regarding CTC test accuracy may be due to differences in study designs or CTC protocols. Four of the CTC studies allowed for the estimation of sensitivity of colonoscopy, and found the sensitivity to detect adenomas ≥6 mm ranged from 75% (95%CI, 63%–84%) to 93% (95%CI, 88%–96%). Fecal immunochemical tests (FITs) did not represent a homogeneous group of tests. On the basis of a single stool specimen, the most commonly evaluated families of FIT tests demonstrated good sensitivity (range, 73%–88%) and specificity (range, 90%–96%). One study (N=9989) found that FIT plus stool DNA test had better sensitivity in detecting CRC than FIT alone (92%) but lower specificity (84%). Serious adverse events from colonoscopy in asymptomatic persons included perforations (4/10,000 procedures, 95% CI 2–5/10,000) and major bleeds (8/10,000 procedures, 95% CI 5–14/10,000). CT colonography may have harms resulting from low-dose ionizing radiation exposure or identification of extra-colonic findings.

Conclusions and Relevance:

Colonoscopy, flexible sigmoidoscopy, CTC, and stool tests have differing levels of evidence to support their use, ability to detect cancer and precursor lesions, and risk of serious adverse events in average risk adults. Although CRC screening has a large body of supporting evidence, additional research is still needed.

Although colorectal cancer (CRC) incidence has been declining over the past 20 years in the United States, it still causes significant morbidity and mortality.¹ Despite increases in screening rates over the past 30 years, in 2012, an estimated 28% of eligible US adults had never been screened for CRC.² A variety of tests are available for screening, including stool-based tests (e.g., guaiac-based fecal occult blood testing [gFOBT], immunochemical-based fecal occult blood testing [FIT], stool DNA [sDNA] testing), endoscopy (e.g., flexible sigmoidoscopy [FS], colonoscopy), and imaging (e.g., double-contrast barium enema, computed tomographic colonography [CTC]).

Currently, most US guideline organizations, including the U.S. Preventive Services Task Force (USPSTF), recommend that options for CRC screening include colonoscopy every 10 years, an annual high-sensitivity gFOBT or FIT, and FS every 5 years with high-sensitivity gFOBT or FIT.^3,4 In 2008, the USPSTF recommended CRC screening using fecal occult blood testing, sigmoidoscopy, or colonoscopy beginning at age 50 years and continuing until age 75 years (A recommendation); selectively offering screening in adults age 76–85 years (C recommendation); and against screening for colorectal cancer in adults older than age 85 years (D recommendation). At that time, the USPSTF had insufficient evidence to assess the benefits and harms of CTC and sDNA testing as screening modalities. A systematic review was conducted to update relevant evidence since 2008 and to help inform a separate modeling exercise, which together were used by the USPSTF in its process of updating the 2008 CRC screening recommendations.

METHODS

Scope of Review

This review addressed three key questions as shown in Figure 1. Additional methodological details regarding search strategies, detailed study inclusion criteria, quality assessment, excluded studies, and description of data analyses are publically available in the full evidence report at http://www.uspreventiveservicestaskforce.org.

Figure 1.

Analytic Framework

CTC, computed tomographic colonography; DNA, deoxyribonucleic acid; FIT, fecal immunochemical test; FS, flexible sigmoidoscopy; gFOBT, guaiac fecal occult blood test; KQ, key question; mSEPT9, circulating methylated septin 9 gene deoxyribonucleic acid.

Data Sources and Searches

MEDLINE, PubMed, and the Cochrane Central Register of Controlled Trials were searched to locate primary studies informing the key questions (eMethods) that were published from the end of the previous review⁵ (January 1, 2008) through December 31, 2014. The database searches were supplemented with expert suggestions and by reviewing reference lists from all other relevant systematic reviews, including the 2008 USPSTF evidence report. The search also included selected grey literature sources, including ClinicalTrials.gov and WHO International Clinical Trials Registry Platform (ICTRP), for ongoing trials. Since December 2014, we continued to conduct ongoing surveillance through article alerts and targeted searches of high-impact journals to identify major studies published in the interim that may affect the conclusions or understanding of the evidence and therefore the related USPSTF recommendation. The last surveillance was conducted on February 23, 2016. Although we identified several potentially relevant new studies, none of these studies would substantively change our review’s interpretation of findings or conclusions.^6–8

Study Selection

Two investigators independently reviewed 8492 titles and abstracts and 696 articles against the specified inclusion criteria (Figure 2). Discrepancies were resolved through consensus and consultation with a third investigator. Inclusion criteria were fair- and good-quality English-language studies of asymptomatic screening populations of individuals who were ≥40 years old, either at average risk for CRC or not selected for inclusion based on CRC risk factors. Studies were included that evaluated the following screening tests: colonoscopy, FS, CTC, gFOBT, FIT, FIT plus sDNA, or a blood test for methylated SEPT9 DNA (mSEPT9).

Figure 2.

Literature Flow Diagram

FS, flexible sigmoidoscopy; KQ, key question; MA, meta-analysis; SER, systematic evidence review; USPSTF, U.S. Preventive Services Task Force

^a Reasons for Exclusion:

Relevance: Study aim not relevant

Design: Study did not use an included design

Setting: Study was not conducted in a country relevant to U.S. practice

Population: Study was not conducted in an average-risk population

Outcomes: Study did not have relevant outcomes or had incomplete outcomes

Intervention: Study used an excluded intervention/screening approach

Quality: Study did not meet criteria for fair or good quality

Simulated FS: Study used the distal colon results from a colonoscopy to simulate a FS

SER MA Outdated: Study was an existing systematic evidence review with an out of date meta-analysis

For key question 1, randomized clinical trials (RCTs) or otherwise controlled trials of CRC screening versus no screening, as well as trials comparing screening tests, that included outcomes of cancer incidence, CRC-specific mortality, or all-cause mortality. For tests without trial-level evidence (i.e., colonoscopy, FIT), well-conducted prospective cohort or population-based nested case-control studies were examined.

For key question 2, diagnostic accuracy studies that used colonoscopy as a reference standard were included. Studies whose design was subject to a high risk of bias were generally excluded, including studies that did not apply colonoscopy to at least a random subset of screen-negative persons (verification bias),⁹ and studies without an adequate representation of a full spectrum of patients (spectrum bias), such as in case-control studies.^9–13 Selected well-conducted FIT diagnostic accuracy studies that used robust registry follow-up for screen negative participants were included.

For key question 3, all trials and observational studies that reported serious adverse events requiring unexpected or unwanted medical attention or resulting in death were included. These events included, but were not limited to, perforation, major bleeding, severe abdominal symptoms, and cardiovascular events. Studies designed to assess for extra-colonic findings (i.e., incidental findings on CTC) and the resultant diagnostic yield and harms of work-up were also included. Studies reporting extra-colonic findings generally used the CT Colonography Reporting and Data System (C-RADS). Under C-RADS, extra-colonic findings are categorized as E0 (limited exam), E1 (normal exam or normal variant), E2 (clinically unimportant finding in which no work-up is required), E3 (likely unimportant or incompletely characterized in which work-up may be required), or E4 (potentially important finding requiring follow-up).¹⁴

Data Extraction and Quality Assessment

Two reviewers each critically appraised all articles that met inclusion criteria using the USPSTF design-specific quality criteria¹⁵ supplemented by the National Institute for Health and Clinical Excellence methodology checklists,¹⁶ AMSTAR for systematic reviews,¹⁷ Newcastle Ottawa Scales for cohort and case-control studies,¹⁸ and QUADAS and QUADAS II for studies of diagnostic accuracy (eTable 1).^19,20 Poor-quality studies and those with a single fatal flaw or multiple important limitations that could invalidate results were excluded from this review. Disagreements about critical appraisal were resolved by consensus and, if needed, consultation with a third independent reviewer. One reviewer extracted key data from included studies; a second reviewer checked the data for accuracy.

Data Synthesis and Analysis

For each key question, the number and design of included studies, overall results, consistency or precision of results, reporting bias, study quality, limitations of the body of evidence, and applicability of findings were summarized. The results were synthesized by key question, type of screening test, and study design. Studies from the 2008 review that met the updated inclusion criteria were incorporated. The analyses for test performance focused primarily on per-person (i.e., by individual patient rather than by lesion) test sensitivity and specificity to detect adenomas (by size, where reported, <6 mm, ≥6 mm, ≥10 mm), advanced adenomas (as defined by the study), and CRC. The studies used several kinds of FITs, which were grouped as qualitative (fixed cut-off) or quantitative (adjustable cut-off), as well as into families (tests produced by the same manufacturer, using the same components and method, or compatible with different automated analyzers). Tests were compared using similar cutoff values expressed in μg Hb/g feces.

Due to the limited number of studies and the clinical heterogeneity of studies, the analyses were largely descriptive. Random-effects meta-analyses were conducted using the profile likelihood method²¹ to estimate the effect of FS based on the pooled incidence rate ratio (events/person-year) for CRC incidence and mortality across the 4 major FS trials. Random-effects models were also conducted using the restricted maximum likelihood estimation method to estimate rates of serious adverse events for colonoscopy and FS. The presence and magnitude of statistical heterogeneity were assessed among pooled studies using the I² statistic. All tests were 2-sided with a p-value less than 0.05 indicating statistical significance. Meta-analyses were performed using R version 3.0.2 (The R Project for Statistical Computing, http://www.r-project.org/).^22,23

RESULTS

Effectiveness of screening

KQ1. What is the effectiveness of screening programs based on the prespecified screening tests (alone or in combination) in reducing incidence of and mortality from colorectal cancer?

Twenty-five unique fair- or good-quality studies^24–48 (published in 47 articles^24–70) were found that assessed the effectiveness or comparative effectiveness of screening tests on CRC incidence and mortality. These studies included 1 cohort study of screening colonoscopy,³⁵ 4 RCTs of FS (in 7 articles),^{24,38,40,49,59,65,70} and 6 trials (in 11 articles) of Hemoccult II gFOBT.^{28,32–34,39,43,58,61–63,66} In addition, 15 comparative effectiveness studies (in 22 articles) were found that were primarily designed to assess the relative uptake and CRC yield between different screening modalities.^{25–27,29–31,36,37,41,42,44–48,53–57,64,68} Due to limitations in study designs, the observational colonoscopy study and comparative effectiveness studies are not discussed further in this article. Summarized below are the results for CRC-specific mortality, as results for CRC incidence were consistent with CRC mortality findings.

Flexible sigmoidoscopy.

Four large fair-quality pragmatic RCTs (N=458,002) evaluated the effectiveness of 1 or 2 rounds of FS in average-risk adults aged 50–74 years (Table 1).^{24,38,40,49,59,65,70} Adherence to FS in these trials ranged from 58% to 84%, and rates of diagnostic colonoscopy ranged from 5% to 33% due to differences in referral criteria. Based on pooled intention-to-treat analyses, FS was associated with lower CRC-specific mortality compared with no screening at 11–12 years of follow-up (incidence rate ratio, 0.73 [95% CI, 0.66–0.82], I²=0%) (Figure 3); however, the association with mortality benefit was limited to distal CRC (incidence rate ratio, 0.63 [95% CI, 0.49–0.84], I²=44%) (eFigure 1). In one trial, conducted in Norway, half of the participants randomized to FS also received a single FIT test; the FS plus FIT arm had lower CRC mortality than the FS-only arm did (hazard ratio, 0.62 [95% CI, 0.42, 0.90]).⁵⁹

Table 1.

Effectiveness of Screening to Reduce Colorectal Cancer Mortality: Flexible Sigmoidoscopy and Hemoccult II Randomized Clinical Trials (Key Question 1)

Screening Tool and Reference	Qualitya	Country	Patient Age Range, y	No. of participants	No. of Screening Rounds	Screening Interval, y	Follow-up Period,c y	Positive Screening Results,d %	CRC, %b	No. of CRC deaths/ 100,000 p-y	CRC Mortality RR (95% CI)
Flexible Sigmoidoscopy
NORCCAP, 201459	Fair	Norway	50–64	Intervention: 20,572 Control: 78,220	1	NA	11.0	20.4	1.4	Intervention: 31 Control: 43	0.80 (0.62–1.04)d,e
PLCO, 201238,70	Fair	United States	55–74	Intervention: 77,445 Control: 77,455	2	3–5	12.1	32.9	1.5	Intervention: 29 Control:39	0.74 (0.63–0.87)
SCORE, 201140,65	Fair	Italy	55–64	Intervention: 17,136 Control: 17,136	1	NA	11.4	8.6	1.6	Intervention: 35 Control Group:44	0.78 (0.56–1.08)
UKFSST, 201024,49	Fair	United Kingdom	55–64	Intervention: 57,099 Control: 112,939	1	NA	11.2	5.2	1.5	Intervention: 30 Control:44	0.69 (0.59–0.80)b
Hemoccult IIf
Minnesota Colon Cancer Control Study, 2013 43,62,63,66	Good	United States	50–80	Intervention (Biennial): 15,587 Control: 15,394	6	2	30	NRg	2.9	Intervention: 50 Control: 63	0.78 (0.65–0.93)
				Intervention (Annual): 15,570 Control: 15,394	11	1	30	NRg		Intervention: 42 Control: 63	0.68 (0.56–0.82)
Nottingham, 201239,58	Good	United Kingdom	45–74	Intervention: 76,056 Control: 75,919	3–5	2	28	2.1	3.0	Intervention: 91 Control: 100	0.91 (0.84–0.98)
Göteborg, 200833	Fair	Sweden	60–64	Intervention: 34,144 Control: 34,164	2–3	1–9	19	3.8g	2.2	Intervention: 53 Control: 64	0.84 (0.71–0.99)
Burgundy, 200428	Fair	France	45–74	Intervention: 45,642 Control: 45,557	6	2	11	2.1	1.5	Intervention: 53 Control:64	0.84 (0.71–0.99)
Funen, 200432	Good	Denmark	45–75	Intervention: 30,967 Control: 30,966	9	2	17	1.0	2.8	Intervention: 84 Control: 100	0.84 (0.73–0.96)

Note: The comparator for each of these randomized clinical trials was a control group that was not offered any colorectal cancer screening.

^aQuality assessed using criteria from the U.S. Preventive Services Task Force ¹⁵

^bThe percentage of CRC cases that occurred during the follow-up period among those included in the study at baseline.

^cMedian follow-up time for flexible sigmoidoscopy, longest follow-up time for Hemoccult II

^dFor FS, this refers to the percentage of patients who were referred to colonoscopy out of those who received their FS. For Hemoccult II, it refers to the percentage of patients who tested positive out of those who took the test in round 1 only.

^eCalculated RR (not study reported)

^eNORCCAP reported a statistically significant decrease in CRC mortality for the screened group versus the control (Hazard Ratio: 0.73; 95% CI, 0.56–0.94; p=0.02). In order to present consistent results across studies, we show unadjusted results here.

^fOne trial in Finland has not reported CRC mortality^34,61

^gStudy included rehydrated tests: Göteborg – 91.7% of all tests were rehydrated; Minnesota Colon Cancer Control Study – 82.5% of all tests were rehydrated

CI, confidence interval; NA, not applicable; NR, not reported; p-y, person-years; y, years.

Figure 3.

Forest plot of randomized controlled trials of FS screening on colorectal cancer mortality (Key Question 1)

CI, confidence interval; Control, no colorectal cancer screening; IRR, incidence rate ratio; NORCCAP, Norwegian Colorectal Cancer Prevention; PL, Profile Likelihood; p-y, person-years; PLCO, Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial; SCORE, Screening for COlon Rectum; UKFSST, UK Flexible Sigmoidoscopy Screening Trial.

gFOBT.

Five older, large, pragmatic RCTs (n=419,966) with 11–30 years of follow-up evaluated the effectiveness of annual or biennial screening programs with Hemoccult II (Table 1).^{28,32,33,39,43,58,62,63,66} Based on intention-to-treat analyses, compared to no screening, biennial screening with Hemoccult II resulted in a reduction in CRC-specific mortality after 2–9 rounds of screening (relative risk [RR], 0.91; 95% CI, 0.84–0.98 at 19.5 years to RR, 0.78; 95% CI, 0.65–0.93 at 30 years). Based on one trial, conducted in the United States, annual screening with Hemoccult II after 11 rounds of screening resulted in greater reductions (RR, 0.68; 95% CI 0.56–0.82) than biennial screening at 30 years did (RR, 0.78; 95% CI 0.65–0.93).⁴³

Diagnostic accuracy of screening

KQ2: What are the test performance characteristics of the prespecified screening tests (alone or in combination) for detecting colorectal cancer, advanced adenomas, and/or adenomatous polyps based on size?

Thirty-three unique diagnostic accuracy studies^71–103 (published in 44 articles^71–114) were found that evaluated the 1-time test performance of a screening test compared to an adequate reference standard, including 9 studies of screening CTC (in 10 articles),^{80,81,84–86,88,92,98,100,113} 3 studies of gFOBT Hemoccult Sensa,^71,72,89 20 studies of various FITs^{71–77,79,81–83,87,89,90,93–97,99,101–103} (1 of which evaluated a FIT plus sDNA test⁸²), and 1 study of a blood test to detect circulating mSEPT9.⁷⁸ The study of mSEPT9 (not approved by the US Food and Drug Administration [FDA] for screening) and studies evaluating Hemoccult Sensa and FITs that only applied the colonoscopy reference standard to positive stool tests are not discussed further in this article.

Direct visualization tests.

Nine fair- or good-quality studies (N=6497) evaluated the diagnostic accuracy of multi-detector CTC in average-risk screening populations (Table 2).^{80,81,84–86,88,92,98,100,113} The 2 largest and best-quality studies were multicenter trials conducted in the United States evaluating CTC with bowel preparation (prep) and fecal tagging.^84,98 Overall, the studies were not powered to estimate test performance to detect CRC. Based on 7 studies of CTC with bowel prep (N=5328), the per-person sensitivity to detect adenomas ≥10 mm and ranged from 67% (95% CI, 45%–84%) to 94% (95% CI, 84%–98%), and specificity ranged from 98% (95% CI, 96%–99%) to 96% (95% CI, 95%–97%). The per-person sensitivity to detect adenomas ≥6 mm ranged from 73% (95% CI, 58%–84%) to 98% (95% CI, 91%–100%), and specificity ranged from 89% (95% CI 84%–93%) to 91% (95% CI, 88%–93%). Two studies (N=1169) evaluated CTC without bowel prep.^80,100 Although the data were limited, the sensitivity of CTC without bowel prep to detect adenomas ≥6 mm appeared to be lower than the sensitivity of CTC protocols including bowel prep.

Table 2.

Prospective Diagnostic Accuracy Studies of Screening Computed Tomographic (CT) Colonography (Key Question 2)

Study	Qualitya	Study Site	Cohort Size	Mean Patient Age, y	No. of Readers, Trainingc	Reading Strategyd		Reference Standard		Adenoma ≥6 mm, % (95% CI)		Adenoma ≥10 mm, % (95% CI)
									Sensitivity	Specificity	Sensitivity	Specificity
With bowel preparation
Lefere et al., 201388	Fair	Portugal	496	60	Yes	1, >5000 exams	3D (with 2D)	Repeat colonoscopy if indicated	98 (91–100)	91 (88–93)	NR	NR
Graser et al., 200981	Good	Germany	307	60	No	3, >300 exams	3D (with 2D)	Colonoscopy, segmental unblindinge	91 (80–97)	93 (90–96)	92 (76–98)	98 (96–99)
Johnson et al., 200884 ACRINf	Good	United States	2531	58	Yes	15, >500 exams	3D (with 2D)	Repeat colonoscopy if indicated	78 (72–83)	90 (88–91)	90 (83–95)	86 (85–87)
Kim et al., 200886	Fair	South Korea	241	58	No	2, >100 exams	2D (with 3D)	Single colonoscopy	68g (55–80)	89g (84–93)	87h (64–97)	97h (95–99)
Johnson et al., 200785	Fair	United States	452	65	No	3, >1000 exams	3D (with 2D)i	Single colonoscopy	NR	NR	67 (45–84)	98 (96–99)
Macari et al., 200492	Fair	United States	68	55	No	1, 5 y	NR	Single colonoscopy	NR	NR	100j (46–100)	98j (93–100)
Pickhardt et al., 200398	Good	United States	1233	58	Yes	6, >25 exams	3D (with 2D)	Colonoscopy, segmental unblinding	89 (83–93)	80 (77–82)	94 (84–98)	96 (95–97)
Without bowel preparation
Fletcher et al., 201380	Good	United States	564	NR	Yes	2, >150 exams	2D and 3D	Single colonoscopy	75 (59–87)	92 (90–94)	67 (42–86)	97 (96–98)
Zalis et al., 2012100	Good	United States	605	60	Yes	3, >200 exams	2D and 3D	Colonoscopy, segmental unblinding	58 (46–69)	88 (85–91)	90 (70–98)	85 (82–88)

^aQuality assessed using criteria from Quality Assessment of Diagnostic Accuracy Studies (QUADAS) I¹⁹ and II²⁰ instrument

^boral ingestion of high-density oral contrast agent so that residual colonic contents can be differentiated from soft tissue density polyps

^cnumber of exams or years training required by each reader/radiologist

^dreader/radiologist procedure for using 2- and 3-dimensional images

^eCT colonography enhanced colonoscopy, in which endoscopist was shown respective segment of colon on CT colonography after examination of segment by colonoscopy

^fNational CT Colonography Trial.

^gAny histology ≥6 mm; sensitivity for adenomas ≥6 mm 72.7% (95% CI, 58.4–84.1), specificity not reported.

^hAny histology ≥10 mm; sensitivity for adenomas ≥10 mm 90.0% (95% CI, 61.9–99.0), specificity not reported.

ⁱStudy evaluated different reading strategies; data shown reflect primary 3D strategy.

^jAny histology ≥10 mm.

NR, not reported.

Four (N=4821) of the 9 CTC studies allowed for the estimation of sensitivity of colonoscopy generalizable to community practice.^84,85,98,100 Compared to CTC or colonoscopy plus CTC (e.g., segmental unblinding), the sensitivity for colonoscopy to detect adenomas ≥10 mm ranged from 89% (95% CI, 78%–96%) to 98% (95% CI, 74%–100%) and for adenomas ≥6 mm ranged from 75% (95% CI, 63%–84%) to 93% (95% CI, 88%–96%) (see full report). Therefore, CTC with bowel prep had sensitivity to detect adenomas ≥6 mm comparable with colonoscopy, albeit with wider variability in estimated performance. It is unclear whether the observed variation in CTC performance was due to differences in study design, populations, bowel prep, CTC technologies, or differences in reader experience or reading protocols.

Stool tests.

Fourteen fair- or good-quality studies (N=59,425) that used colonoscopy reference standard in all participants reported sensitivity and specificity for 19 different types of qualitative or quantitative FITs, including one FIT plus sDNA test (Table 3).^{73,76,77,79,81,82,90,93,94,96,97,99,102,103,106,107,114} Overall, the sensitivity for CRC and advanced adenomas varied widely, including a discontinued test with very low sensitivity.⁹⁹ Given the heterogeneity among FITs and their test performance, focus was placed on the performance characteristics of currently available tests evaluated in more than 1 study. Two families of FDA-cleared tests, OC-Light (qualitative, no. of studies=3, N=25,924) and OC FIT-CHEK (e.g. OC-Sensor Diana, OC-Micro, OC-Auto) (quantitative, no. of studies=5, N=12,794), had relatively high sensitivity and specificity. With a single stool specimen, the lowest sensitivity demonstrated for CRC was 73% (95% CI, 48%–90%) and specificity was 96% (95% CI, 95%–96%). Similarly, the highest sensitivity with paired specificity for CRC was 88% (95% CI, 55%–99%) and 91% (95% CI, 89%–92%), respectively. In the largest studies, sensitivity ranged from 74% (95% CI, 62%–83%) for quantitative test categories (n=9989) to 79% (95% CI 61%–91%) for qualitative test categories (n=18,296). In a small study (N=770) that tested 3 stool specimens, sensitivity was 92% (95% CI, 69%–99%) but specificity was 87% (95% CI, 85%–89%). OC-Light or OC-FIT-CHEK test sensitivity and specificity for advanced adenomas ranged from 22% (95% CI, 17%–28%) to 40% (95% CI, 30%–51%) and specificity ranged from 97% (95% CI, 97%–98%) to 91%, (95% CI, 91%–92%). While higher sensitivities to detect advanced adenomas were obtained for certain other FITs or by using 3 stool specimens, the corresponding specificities were lower.

Table 3.

Prospective Diagnostic Accuracy Studies of FIT Tests (With or Without Stool DNA Test) Using Colonoscopy Reference Standard (Key Question 2)

Study	Qualitya	Mean Patient Age, y	CRC, %	Cohort Size	Test Family Name	Cutoff, μg Hb/g feces	No. of Stool Samples	CRC		Advanced Adenomas
								Sensitivity, % (95% CI)	Specificity, % (95% CI)	Sensitivity, % (95% CI)	Specificity, % (95% CI)
Qualitative FIT Tests
Levy et al., 201490b	Fair	56.9	NR	308	Clearview (cassette)	6	NR	NR	NR	NR	NR
			NR	44	Clearview (test strip)	50	NR	NR	NR	NR	NR
			NR	217	OC-Light	10	NR	NR	NR	NR	NR
			NR	52	QuickVue	50	NR	NR	NR	NR	NR
Chiu et al., 201377	Good		0.15	18296	OC-Light	10	1	79 (61–90)	93 (92–93)	28 (25–32)	94 (93–94)
Ng et al., 201377,96	Fair	57.7	0.48	4539	Hemosure	50	NR	54 (32–74)	89 (88–90)	37 (30–44)	91 (90–91)
Brenner et al., 2010106	Good	63c	0.8	1319	Bionexia Hb	NR	NR	NR	NR	52 (44–61)	80 (77–82)
				1328	Bionexia Hb-Hp	NR	NR	NR	NR	72 (63–79)	56 (54–59)
				1330	FOB advanced	NR	NR	NR	NR	27 (20–35)	91 (90–93)
				1319	immoCARE-C	30	NR	NR	NR	25 (18–33)	96 (95–97)
				1330	PreventID CC	NR	NR	NR	NR	49 (41–58)	81 (79–84)
				1330	QuickVue	50	NR	NR	NR	56 (48–64)	68 (65–70)
Cheng et al., 200276	Fair	46.8	0.22	7411	OC-Light	10	NR	88 (66–97)	91 (90–92)	40 (30–51)	91 (91–92)
						~1,000	1	56 (33–76)	97 (96–97)	NR	NR
Nakama et al., 199994	Fair	NR	0.39	4611	Monohaem	~1,000	2	83 (62–95)	95 (95–96)	NR	NR
						~1,000	3	89 (69–98)	93 (92–94)	NR	NR
Quantitative FIT Tests
Hernandez et al., 2014102	Good	57.6	0.6	779	OC FIT-CHEK	10	1	100 (62–100)	92 (90–94)	NR	NR
						20	1	100 (62–100)	94 (92–95)	NR	NR
						10	2	100 (62–100)	88 (85–90)	NR	NR
						20	2	100 (62–100)	90 (88–92)	NR	NR
Imperiale et al., 201482	Fair	64.2	0.65	9989	OC FIT-CHEK	20	1	74 (62–83)	93 (93–94)	24 (21–27)	95 (94–95)
					Cologuard (FIT plus stool DNA test)	NA	1	92 (84–97)	84 (84–85)	42 (39–46)	87 (86–87)
Lee et al., 2014103	Good	58c	NR	NR	Hemo Techt NS-Plus C system	6.3	NR	86 (57–98)	94 (93–95)	NR	NR
Brenner and Tao, 201373	Good	62.7	0.67	2220	OC FIT-CHEK	20	1	73 (48–90)	96 (95–96)	22 (17–28)	97 (97–98)
				2220	RIDASCREEN Hb	2	1	60 (35–81)	95 (94–96)	21 (16–27)	97 (96–98)
				2235	RIDASCREEN Hb-Hp	2	1	53 (29–76)	95 (94–96)	18 (13–24)	97 (96–98)
de Wijkerslooth, 201279	Good	60a	0.64	1256	OC FIT-CHEK	10	1	88 (55–99)	91 (89–92)	34 (26–43)	93 (92–95)
						20	1	75 (41–94)	95 (93–96)	28 (20–37)	97 (96–98)
Park et al., 201097	Fair	59.3	1.7	770	OC FIT-CHEK	10	3	92 (69–99)	87 (85–89)	44 (32–57)	89.8 (87.4–91.9)
				757		20	3	92.3 (69.3–99.2)	90.1 (87.8–92.1)	33.9 (22.8–46.5)	92.1 (89.9–94.0)
Graser, 200981	Good	60.5	0.33	285	FOB Gold	NR	2	100 (14.7–100)	NR	29.2 (14.1–48.9)	85.8 (81.1–89.6)
Morikawa, 200593	Fair	48	0.4	21805	Magstream/ HemeSelect	100–200	1	65.8 (54.9–75.6)	94.6 (94.3–94.9)	NR	NR
Sohn et al., 200599	Fair	48.9	0.3	3794	OC Hemodia	20	1	25.0	NR	6.0	NR

^aQuality assessed using criteria from Quality Assessment of Diagnostic Accuracy Studies (QUADAS) I¹⁹ and II²⁰ instrument

^bResults reported for advanced neoplasia (composite of CRC and advanced adenoma) only

^cMedian.

CRC, colorectal cancer; NA, not applicable; NR, not reported.

Cologuard (Exact Sciences Corp., Madison, WI) is an FDA-approved stool test that combines stool DNA with a proprietary FIT component. One fair-quality diagnostic accuracy study (N=9989) evaluated Cologuard compared to OC FIT-CHEK.⁸² In that study, Cologuard had a statistically significant higher sensitivity to detect CRC and advanced adenoma compared with OC FIT-CHEK. The higher sensitivity for CRC (92%; 95% CI, 84%–97%) and for advanced adenoma (42%; 95% CI, 39%–46%) was accompanied by lower specificity (84%; 95% CI, 84%–85% for CRC, and 87%; 95% CI, 86%–87% for advanced adenoma). In our active surveillance of the literature, we identified one additional diagnostic accuracy study of FIT plus sDNA (n=661) in asymptomatic Alaska Native adults.⁸ This study was not powered to find a difference in detection of CRC, nonetheless, findings were generally consistent with the included study on FIT plus sDNA.⁸²

Harms of screening

KQ3a: What are the adverse effects of the different screening tests (either as single application or in a screening program)?

KQ3b: Do adverse effects vary by important subpopulations (e.g., age)?

Ninety-eight fair- or good-quality studies^{26,28,76,77,80,81,84–86,115–154,36–38,44,47,63,65,88,91,92,96,98,100,155–190} (in 113 articles^{26,28,49,52,76,77,80,81,84,85,113,115–149,191–194,33,36–38,43,44,63,65,86,88,91,92,96,98,150–180,189,195–198,47,69,100,181–188,190,199}) were included that evaluated the harms of CRC screening. These studies included 14 studies of screening programs using stool testing or FS, 55 studies of colonoscopy in asymptomatic adults, ^{36,44,76,77,84,96,100,115,116,118,120–123,127–130,132,135,139,141,143,146,147,149,150,152–154,157,158,160–162,169–172,174–177,179–182,186,187,119,125,155,173,188,189} 18 studies of screening FS,^{26,37,38,42,47,49,65,142,145,150,156,161,175,182,184,185,190–193,199} and 15 studies of screening CTC in asymptomatic adults.^{44,80,81,84,86,88,100,117,134,144,149,161,168,178} Twelve CTC studies provided estimates of radiation exposure per exam,^{80,81,84–86,88,92,98,100,117,134,161} and another 21 CTC studies reported information on extra-colonic findings.^{44,84,98,100,113,124,126,133,136–138,140,149,151,159,163,165–167,183,194,197}

Endoscopy harms.

Approximately half of colonoscopy harms studies (29 of 55 studies) were in explicitly screening or asymptomatic populations (eTable 2). By pooling 26 studies (N=3,414,108) in screening populations or generally asymptomatic persons,^{36,44,76,77,84,96,100,119,120,123,125,129,130,135,149,155,162,169,173,175,179–181,187–189} it was estimated that the risk of perforations from colonoscopy was 4/10,000 procedures (95% CI, 2–5/10,000; I²=86%) (Figure 4). On the basis of 22 of those studies (N=3,347,101),^{36,44,76,84,96,100,119,120,123,125,129,130,155,162,169,173,179–181,187–189} it was estimated that the risk of major bleeding from colonoscopy was 8/10,000 procedures (95% CI, 5–14/10,000; I²=97%) (Figure 5). Only eight studies (n=204,614) explicitly reported if perforation or major bleeding was related to polypectomy or biopsy.^{44,84,119,135,157,172,176,177} Based on this limited subset of studies reporting adequate information, 36% (15/42) of perforations and 96% (49/51) of major bleeding events were from polypectomy.

Figure 4.

Forest plot of perforations from colonoscopy, asymptomatic population (Key Question 3)

CI, confidence interval; n, number; RE, restricted maximum likelihood.

Note: One trial has been excluded from the meta-analysis due to very small n (n=63).158 There were no episodes of serious bleeding or perforation in the study.

Figure 5.

Forest plot of major bleeding from colonoscopy, asymptomatic population (Key Question 3)

RE, restricted maximum likelihood; CI, confidence interval; n, number.

Note: One trial has been excluded from the meta-analysis due to very small n (n=63). ¹⁶¹ There were no episodes of serious bleeding or perforation in the study.

All 18 FS harms studies were conducted in general-risk screening populations (eTable 3). Based on the results of 16 studies (N=329,698),^{37,38,42,47,49,65,142,145,150,156,175,182,184,185,190,191} perforations from FS in average-risk screening populations were relatively uncommon: the pooled point estimate was 1/10,000 procedures (95% CI, 0.4–1.4/10,000; I²=18.4%). In 10 studies (N=137,987),^{26,37,47,49,65,142,145,156,184,185} major bleeding episodes from FS were also relatively uncommon, with a pooled point estimate of 2/10,000 procedures (95% CI, 0.7–4/10,000; I²=52.5%) (Figures 6 and 7). FS, however, may require follow-up diagnostic or therapeutic colonoscopy. From five FS screening trials, the pooled estimate was 14 perforations per 10,000 (95% CI, 9–26/10,000), and 34 major bleeds per 10,000 (95% CI, 5–63/10,000) for follow-up colonoscopy for positive screening FS from four trials.

Figure 6:

Forest plot of perforations from flexible sigmoidoscopy (Key Question 3)

CI, confidence interval; n, number; RE, restricted maximum likelihood.

Note: One trial has been excluded from the meta-analysis due to very small n (n=52).¹⁶¹ There were no episodes of serious bleeding or perforation in the study.

Figure 7.

Forest plot of major bleeding from flexible sigmoidoscopy (Key Question 3)

CI, confidence interval; n, number; RE, restricted maximum likelihood;

Note: One trial has been excluded from the meta-analysis due to very small n (n=52).¹⁶¹ There were no episodes of serious bleeding or perforation in the study.

Other serious harms from endoscopy were not routinely reported or consistently defined. Only 2 studies compared harms other than perforation and bleeding in persons who had a colonoscopy versus those who had not.^179,186 Both of these studies found no statistically significant higher risks of serious harms (including myocardial infarction, cerebrovascular accident, other cardiovascular events, and mortality) attributable to colonoscopy. Due to reporting bias around serious harms other than perforation and bleeding as well as the lack of evidence for other serious harms attributable to colonoscopy in the few studies with control groups, these data were not quantitatively pooled.

Nineteen studies examined differential harms of colonoscopy by age group.^{115,118,121,122,127,128,130,135,139,153,155,158,160,169,171,173,186,188,189} These studies generally found increasing rates of serious adverse events with increasing age, including perforation and bleeding.

CTC harms.

Fifteen fair- or good-quality studies addressed serious adverse effects of screening CTC (eTable 4).^{44,80,81,84,86,88,100,117,134,144,149,161,168,178} Evidence suggested little to no risk of serious adverse events, including perforation, from CTC based on 11 prospective studies (N=10,272) performed in screening populations.^{44,80,81,84,86,88,100,117,134,144,149,161,168,178}

Many of the CTC studies in this review did not report actual radiation exposure or provide sufficient information to calculate it. Based on 4 included diagnostic accuracy studies of CTC,^80,81,84,100 the estimated radiation dose for 1 full-screening CTC exam (dual positioning supine and prone) was about 4.5–7 mSv. In 3 additional recent CTC screening studies^117,134,161 (2004–2008), the estimated radiation dose decreased to a range of 1 to <5 mSv.

CTC extra-colonic findings.

Incidental extra-colonic findings detected on CTC can be beneficial or harmful depending on the finding. Twenty-one studies (N=38,293)^{44,84,98,100,124,126,133,136–138,140,149,151,159,163,165–167,183,194,197} in 22 articles^{44,84,98,100,113,124,126,133,136–138,140,149,151,159,163,165–167,183,194,197} (7 studies with overlapping populations reported different types extra-colonic findings) reported on extra-colonic findings in asymptomatic persons (eTable 5). In general, these studies varied greatly in their ability to accurately assess follow-up and the duration of follow-up.

Overall, extra-colonic findings were common, occurring in 27%–69% of exams. Similarly, the studies suggested a very wide range of findings needing additional work-up: 5%–37% had E3 or E4 findings, and 1.7%–12% had E4 findings. Among the studies that also reported medical follow-up of extra-colonic findings, 1.4%–11% went on to diagnostic evaluation, which is similar to the prevalence of E4 category findings. Among studies that adequately reported subsequent treatment, only up to 3% required definitive medical or surgical treatment. Extra-colonic cancers were not common, occurring in 0.5% of persons undergoing CTC exams. In the largest series of exams (N=10,286), which had about 4 years follow-up, 0.35% of exams revealed an extracolonic malignancy, 32 of which received definitive treatment.¹⁶⁶ Abdominal aortic aneurysms were identified in ≤1.4% of persons. In our active surveillance of the literature, we identified one additional study evaluating extra-colonic findings in screening CTC (n=7952).⁶ This study’s population overlapped with several already included studies, and reported that 2.5% of exams had E4 category findings, consistent with findings from included studies.^{149,163,165–167,194,197}

DISCUSSION

CRC screening continues to be a necessary and active field of research. Since the 2008 USPSTF recommendation was published, 95 new studies were identified, including more evidence on 1) the effectiveness of FS for reducing CRC mortality, 2) the test performance of screening CTC and decreasing radiation exposure from CTC, and 3) the test performance of a number of FDA-approved FITs (including one FIT plus sDNA test). Colonoscopy, FS, CTC, and stool testing (gFOBT, FIT, and FIT plus sDNA test) each have differing levels of evidence to support their use, ability to detect cancer and precursor lesions, and risk of serious adverse events in screening average risk adults for CRC (Table 4).

Table 4.

Summary of Evidence by Key Question (KQ) and Screening Test^a

Test Name	No. of Studies	Sample Size (n),	Study Design	Summary of Findings (includes consistency, precision)	Applicabilityb	Limitations (includes reporting bias)	Overall Quality
KQ1: What is the effectiveness of screening programs in adults at average-risk for colorectal cancer, compared to no screening, in reducing the incidence of or mortality from colorectal cancer?
FS	4	458,002	RCT	FS consistently decreased CRC-specific mortality compared with no screening at 11–12 years of follow-up (IRR 0.73; 95% CI 0.66–0.82). Mortality benefit was limited to distal CRC.	Fair to poor. No longer widely used in the United States.	Only one trial evaluated more than a single round of screening. Variation in referral criteria led to differing rates of follow-up colonoscopy.	Fair to good
gFOBT, Hemoccult II	5	442,088	RCT	Biennial screening with Hemoccult II compared to no screening consistently resulted in reduction of CRC-specific mortality (ranging from 9% to 22% after 2–9 rounds of screening with 11–30 years of follow-up).	Poor. No longer widely used.	Variation in number of screening rounds, use of rehydrated samples, definition of “test positive,” and recommended diagnostic follow-up.	Fair to good
KQ 2: In adults at average-risk for colorectal cancer, what are the test performance characteristics (e.g., sensitivity and specificity) of a one-time application of a screening test, compared to an adequate reference standard, for detecting colorectal cancers, advanced adenomas, or adenomas based on size?
Colonoscopy	4	4821	Prospective diagnostic accuracy	Comparing colonoscopy to CTC or CTC plus colonoscopy, the per-person (or per-lesion) sensitivity for adenomas ≥10 mm was 89% to 98% and the per-person sensitivity for adenomas ≥6 mm was 75% to 93%.	Fair. Colonoscopies were conducted or supervised by “experienced” specialists.	Studies were not designed to assess diagnostic accuracy to detect cancers. Limited studies with large number of endoscopists that were applicable to community practice.	Fair to good
CTC	9	6497	Prospective diagnostic accuracy	The per-person sensitivity and specificity of CTC using bowel preparation to detect adenomas ≥10 mm ranged from 67% to 94% and 86% to 98%, respectively. The per-person sensitivity and specificity to detect adenomas ≥6 mm ranged from 73% to 98% and 80% to 93%, respectively. In 2 studies, sensitivity without bowel preparation to detect adenomas was lower than that of CTC protocols using bowel preparation.	Fair. Mostly single-center studies, with ≤3 highly trained radiologists. Current practice may use different technologies and protocols.	Studies were not designed to assess diagnostic accuracy to detect cancers. Unclear if the variation of test performance was due to differences in study design, populations, bowel preparation, CTC technology, or differences in reader experience or reading protocols.	Fair to good
FIT	Qualitative 6	36,808	Prospective diagnostic accuracy	In studies with colonoscopy follow-up for all, FIT sensitivity varied considerably across assays for each outcome. OC-Light had the highest sensitivity and specificity for CRC, from 88% and 91%, respectively, to 79% and 93%, respectively). OC FIT-CHEK had the best sensitivity and specificity for CRC, from 73% and 96%, respectively, to 92% and 87%. A FIT plus sDNA assay (Cologuard) had better sensitivity but lower specificity, 92% (95% CI, 84–97) and 84% (95% CI, 84–85), respectively, compared to OC FIT-CHEK.	Fair to good. There is a wide range in costs for specific tests (OC-Light, OC-FIT CHEK, Cologuard). Quantitative FITs included some that are older and now discontinued.	Variation in test performance resulted from the use of 18 different FITs (FIT families), different numbers of stool samples, and to some extent different assay cutoff values. Sparse data on most individual tests limited comparisons. FIT plus sDNA was limited to a single study with 6% inadequate stool samples.	Fair to good
	Quantitative 7	40,134
	FIT plus sDNA 1	9989
KQ3: What are the serious adverse effects of colorectal cancer screening tests in asymptomatic adults? Do these adverse effects vary by important subpopulations (e.g., age)?
Endoscopy	FS 18	331,181	Prospective and retrospective studies	Harms from screening FS were estimated at 1 perforation/10,000 procedures (95% CI, 0.4–1.4/10,000) (no. of studies=16) and 2 major bleeds/10,000 procedures (95% CI, 0.7–4/10,000) (no. of studies=10). Harms from screening colonoscopy or colonoscopy in asymptomatic persons was estimated at 4 perforations/10,000 procedures (95% CI, 2–5/10,000) (no. of studies=26) and 8 major bleeds/10,000 procedures (95% CI, 5–14/10,000) (no. of studies=22). Risk of perforations, bleeding, and other serious harms from colonoscopy increased with age.	Good. Reflects community practice.	Only 2 studies reported serious adverse events in persons without colonoscopy (no difference in serious harms other than perforation and bleeding). Likely reporting bias of serious harms other than perforation and bleeding.	Fair
	Colonoscopy 55	10,398,876
CTC	Harms 15	75,354	Prospective and retrospective studies	Harms from CTC in asymptomatic persons were uncommon. Risk of perforation for screening CTC was <2/10,000 exams. The range of low-dose ionizing radiation per exam was 1–7 mSv. ECF was estimated to occur in up to 69% of examinations, and 5–37% of exams might necessitate diagnostic follow-up; however, only ≤3% required any type of definitive treatment. Higher prevalence of ECF with increasing age.	Fair to good. Radiation exposure per exam may be decreasing over time.	No studies reported serious adverse events in persons without CTC. Limited evidence in true average-risk screening populations. Likely reporting bias of serious harms other than perforation. No studies reported differential harms by age group. No studies were able to quantify net benefits and harms of ECF findings. Varying levels of follow-up and few studies with final disposition of ECF. Very limited studies comparing ECF by age group.	Fair
	ECF 21	38,193

^aExcerpted from full Summary of Evidence Table in full report, available at http://www.uspreventiveservicestaskforce.org

^bApplicability or external validity to US practice.

CI, confidence interval; CRC, colorectal cancer; CTC, computer tomographic colonography; ECF, extra-colonic findings; FDA, Food and Drug Administration; FIT, fecal immunochemical test; FS, flexible sigmoidoscopy; FOBT, fecal occult blood test; HR, hazard ratio; IRR, incidence rate ratio; mSv, millisievert; n, number; RCT, randomized controlled trial; RR, relative risk.

To date, no CRC screening modality has been shown to reduce all-cause mortality. Robust data from well-conducted population-based screening RCTs have demonstrated that both Hemoccult II and FS can reduce CRC mortality, although neither of these tests is widely used for screening in the United States. Therefore, the empiric data on the performance of CRC screening programs using modalities used in clinical practice today are limited. Expensive, large population-based RCTs of newer stool tests may not always be necessary, as evidence-based reasoning supports that screening with stool tests with sensitivity and specificity that are both as good as, or better than, Hemoccult II would result in CRC mortality reductions similar to or better than reductions shown in existing trials.²⁰⁰ Based on this review, stool tests that meet those requirements are available, including specific single-stool sample FITs. However, FITs are not homogenous: they use different assays and have different diagnostic performance levels. The FDA-approved OC-Light and OC FIT-CHEK tests have the most evidence to support their use. Stool tests that maximize sensitivity (e.g., FIT plus sDNA test, multiple sample FITs, or quantitative FIT using lower cut-offs) have lower specificity and therefore need new trials or modelling exercises to understand the tradeoff of higher false-positive findings. Additionally, stool tests vary in cost; for example, the CMS reimbursement is $23 per FIT versus $493 per FIT plus sDNA test.²⁰¹

Even though its superiority in a program of screening has not been empirically established, colonoscopy remains the criterion standard for assessing the test performance of other CRC screening tests. Moreover, colonoscopy is significantly more invasive than other available tests and thus carries a greater possibility of procedural complications, as well as harms of overdiagnosis and overtreatment of smaller lesions (i.e., <10 mm). Three large RCTs of screening colonoscopy in average-risk adults are under way and will provide information about the long-term CRC incidence and mortality outcomes: the Northern European Initiative on Colorectal Cancer (NordICC) trial comparing screening colonoscopy to usual care (estimated primary completion date June 2026),²⁰² COLONPREV comparing colonoscopy to biennial FIT in Spain (estimated primary completion date November 2021),^36,203,204 and CONFIRM comparing colonoscopy versus annual FIT in the United States (estimated primary completion date September 2027).²⁰⁵

Evidence continues to accrue that CTC adequately detects CRC and large potential precursor lesions. Although the risk of immediate harms from screening CTC (e.g., bowel perforation from insufflation) is very low, it is unclear what (if any) true harm is posed by cumulative exposure to low-dose radiation or detection of extra-colonic findings. Although the radiation dose appears to be decreasing over time due to technological and protocol advancements, it still ranges as high as 7 mSv/exam (dual positioning). Given that the average amount of radiation one is exposed to from background sources in the United States is about 3 mSv/year,²⁰⁶ ionizing radiation from a single CTC exam is low. However, current expert recommendations are to repeat CTC every 5 years, and even low doses of ionizing radiation could cumulatively convey a small excess risk of cancer.^207,208 From empiric evidence to date, it remains unclear if detection of extra-colonic findings represents a net benefit or harm.

Limitations

This evidence report and systematic review did not address several important issues: screening in high-risk adults (i.e., those with known family history of CRC), risk assessment to tailor screening, test acceptability, availability of or access to screening tests, methods to increase screening adherence, potential harms of overdiagnosis or unnecessary polypectomy, over- or misuse of screening, and surveillance after adenoma detection. This review was commissioned along with a separate set of microsimulation decision models from the Cancer Intervention and Surveillance Modeling Network (CISNET) that addressed other important gaps in evidence, including ages to start and stop screening, screening intervals, and targeted or tailored screening. The review was limited to evidence conducted in countries with the highest applicability to U.S. practice; additionally only articles published in English were considered for inclusion.

Unlike other routinely recommended or conducted cancer screening, there are multiple viable options for CRC screening. These options have various levels of evidence to support their use, aims (e.g., to detect cancers, potential precursor lesions, or both), test acceptability and adherence, intervals of time to repeat screening, need for follow-up testing (including surveillance incurred), associated serious harms, availability in practice, cost, and advocacy for their use. This complexity is compounded by testing whose quality is more operator-dependent (e.g., colonoscopy, CTC), as well as rapid technologic advancements in improving existing tests or developing new tests. Empiric studies, trials, or well-designed cohort studies with average-risk populations are still needed to evaluate programs of screening using colonoscopy, the best-performing stool tests, and effect of CTC on cancer mortality and cancer incidence. We also need studies of diagnostic accuracy to confirm the screening test performance of promising stool tests based on high sensitivity to detect CRC or advanced adenomas with thus far limited reproducibility (i.e., only 1 study). Diagnostic accuracy studies, particularly those evaluating new or more complex technologies, should report percent inadequate or indeterminate results. It is also important to understand the contribution of technological advancements to existing technology (e.g., enhancements to optical colonoscopy or CTC) on test performance in average-risk adults as well as on reducing harms (e.g., decreasing radiation exposure, less aggressive bowel prep). More complete and consistent reporting regarding downstream benefits and harms from initial detection (i.e., subsequent work-up and definitive treatment) of C-RADS E3 and E4 findings need to be published in observational studies or trials with longer-term follow-up. Data are still needed on the differential uptake of and adherence to screening modalities and on continued adherence to repeated rounds of screening and diagnostic follow-up to screening over longer periods.

Conclusions

Colonoscopy, flexible sigmoidoscopy, CTC, and various stool tests have differing levels of evidence to support their use in CRC screening, ability to detect CRC and precursor lesions, and risk of serious adverse events in average risk adults. Although CRC screening has a large body of supporting evidence, additional research is still needed to weigh the relative benefits and harms of each test in within a program of screening.