FS and Colonoscopy Benefits

Four large population-based RCTs evaluating screening FS showed that intention to screen with one-time FS (or, in the PLCO trial, two rounds of FS) was consistently associated with a decrease in CRC incidence (IRR 0.78 (95% CI, 0.74 to 0.83) and CRC-specific mortality (IRR 0.74 (95% CI, 0.68 to 0.80) compared with no screening at 11 to 17 years of followup. Despite this robust evidence, recent utilization data in the United States suggest that FS (with or without stool testing) is very uncommon (<1%).³⁶⁰ Public and clinician perceptions of accuracy of colonoscopy versus FS, given the reach of endoscopy, also play an important role in the low utilization of FS compared with colonoscopy.³⁶¹ Although from included studies, FS is associated with a reduction in CRC incidence and mortality for both proximal and distal cancers, albeit greater reductions for distal cancers. We found no studies estimating the test accuracy of FS compared with a colonoscopy reference standard. Estimates of FS sensitivity and specificity are based on a limited number of relatively small studies with suboptimal study designs (e.g., tandem FS studies, simulated studies using colonoscopy and assumed FS reach to splenic flexure).⁹⁰ Sensitivity of FS to detect CRC calculated from the PLCO trial is 69.6 percent³⁶²; however, the test accuracy of FS to detect CRC may depend on the referral criteria, as criteria resulting in greater followup colonoscopy may detect a greater number of cancers—particularly proximal cancers. For example, the PLCO trial used nonbiopsy referral-based criteria for followup colonoscopy and had the highest referral rate to colonoscopy (about 33%) of all the trials.

Only one prospective cohort study has evaluated the association of the receipt of screening colonoscopy and CRC mortality in average-risk adults.²¹ However, this study is part of a larger evidence base of population-based case control studies and retrospective cohort studies demonstrating an association of screening colonoscopy and reduction in CRC incidence and/or mortality.^{19, 20, 156, 363–367} This included study using data from the Nurses’ Health Study and the Health Professionals Follow-Up Study found that CRC mortality was lower in people with at least one screening colonoscopy versus those who never had a screening endoscopy (adjusted HR, 0.32 [95% CI, 0.24 to 0.45]) at 24 years of followup. Another included study conducted among Medicare beneficiaries found that receipt of screening colonoscopy was associated with a lower incidence of CRC after 8 years as compared with no screening colonoscopy in people ages 70 to 74 years; this study did not report CRC mortality outcomes.¹²⁸ This magnitude of association from observational studies should not be compared with the magnitude of effect in CRC mortality in intention to treat analyses from RCTs of screening FS. Currently, one large screening RCT in average risk adults, NordICC, evaluating the impact of screening colonoscopy to usual care on CRC incidence and mortality in Norway, Sweden, Poland, and the Netherlands, is underway.³⁶⁸

We included only four studies for which we could derive community-based relevant estimates of test accuracy, evaluating screening colonoscopy against a criterion standard. However, none of them were designed to estimate the test performance for CRC. Based on three studies, the per-person sensitivity for colonoscopy to detect adenomas 10 mm or larger ranged from 89.1 to 94.7 percent and the per-person sensitivity to detect adenomas 6 mm or larger ranged from 74.6 to 92.8 percent. Colonoscopies in these studies were conducted by experienced endoscopists; test performance will vary in clinical practice based on the adequacy of bowel preparation and colonoscopist performance/experience. A separate body of evidence addressing adenoma miss rates from tandem colonoscopy studies, not included in our review, confirms that colonoscopies can miss adenomas. A 2019 systematic review of 43 studies of over 15,000 tandem colonoscopies demonstrated that miss rates for adenomas and AA are higher than previously appreciated.⁹⁵ Both the effectiveness and test accuracy of colonoscopy may vary depending on a number of factors including the examiner quality. The American Society for Gastrointestinal Endoscopy, American College of Gastroenterology, and U.S. Multi-Society Task Force have issued guidance and recommendations for the technical performance and quality improvement targets for colonoscopy.^{369, 370} In addition, there is a growing body of evidence, not included in this review, that evaluates whether technological advancements in colonoscopy to improve adenoma detection, namely chromoendoscopy or digital/virtual chromoendoscopy (e.g., narrow band imaging, flexible spectral imaging color enhancement), endoscopic technologies to increase mucosal surface area inspection (e.g., wide-angle lens or full-spectrum endoscopy, through-the-scope retrograde viewing device), and computer aided detection using artificial intelligence can improve detection, but data are limited to support widespread adoption in screening or average-risk populations.^371–375

FS and Colonoscopy Harms

Serious adverse events from screening colonoscopy or colonoscopy in asymptomatic persons are estimated at 14.6 serious bleeds (95% CI, 9.4 to 19.9) and 3.1 perforations (95% CI, 2.3 to 4.0) per 10,000 procedures. This estimate of serious bleeds is higher than appreciated in the prior review to support the 2016 USPSTF recommendation (8.2 serious bleeds per 10,000 procedures, 95% CI, 5.0 to 13.5).¹ Overall, it appears the risk of major bleeding and perforation is higher with increasing age. Other serious harms (e.g., infections, other GI events, cardiopulmonary events) were not consistently reported, and four studies evaluating harms in people who received colonoscopy versus those who did not found no increased risk of serious harms (including MI, CVA, or other cardiovascular events) as a result of colonoscopy. Serious adverse events from screening FS are rare (0.5 [95% CI, 0 to 1.3] serious bleeds and 0.2 perforations [95% CI, 0.1 to 0.4] per 10,000 procedures); however, screening FS may require followup colonoscopy. Serious harms of colonoscopy following screening FS are estimated at 20.7 serious bleeds (95% CI, 8.2 to 33.2) and 12.0 perforations (95% CI, 7.5 to 16.5) per 10,000 colonoscopies.

Case reports of fatal or near-fatal outcomes in average-risk people undergoing routine colonoscopy include splenic rupture, retroperitoneal or intra-abdominal hemorrhage, retroperitoneal gas gangrene, bowel infarction or ischemic colitis, small bowel perforation, colonic gas explosion with electrocautery, and appendicitis or appendiceal abscess.⁸² In addition, there have been case reports of transmission of communicable diseases (i.e., hepatitis C virus, human papillomavirus) using unsanitized colonoscopes and chemical colitis from glutaraldehyde, which is used to disinfect endoscopes.⁸²

We found no studies directly assessing the harms of cancer overdiagnosis (i.e., cancer detected through screening that would have not otherwise clinically manifested during a person’s lifetime). One Markov modeling study using data from over 4 million screening colonoscopies from Germany’s national screening colonoscopy registry, found that the risk of overdiagnosis was very low in people ages 55 to 79 years and 28 percent of the overdiagnoses occurred in people older than age 75 years.³⁷⁶ Another potential harm is the overdetection of adenomas (i.e., adenomas detected through screening that would not develop into cancer and/or otherwise clinically manifested during a person’s lifetime) leading to unnecessary procedures or more intensive colonoscopy surveillance.

CTC Benefits

While we found no studies examining the impact of screening CTC on cancer incidence or mortality, there is a robust evidence base evaluating the test performance of screening CTC in average-risk adults. However, none of these studies were designed to estimate test performance to detect cancer. Based on seven studies of CTC with bowel preparation, the per-person sensitivity and specificity to detect adenomas 10 mm or larger ranged from 66.7 to 93.5 percent and 86.0 to 97.9 percent, respectively; and to detect adenomas 6 mm or larger ranged from 72.7 to 98.0 percent and 79.6 to 93.1 percent, respectively. It is unclear whether the variation in test performance is due to differences in study design or populations studied or differences in bowel preparation, CTC imaging, reading protocols, and radiologist experience. In the included studies and current practice there is variation in bowel preparation (e.g., full, partial, none) and CTC technical enhancements (e.g., increasing detectors, fecal tagging, electronic cleansing, computer aided detection, insufflation techniques). Because some variation in accuracy is likely due to CTC protocol and/or radiologist ability, both the American College of Radiology and the International Collaboration for CT Colonography Standards have recommended practice guidelines and quality metrics, as well as specifications for training and certification.^377–379 In practice, the standard appears to be dry preparation (sodium phosphate, magnesium citrate, bisacodyl) rather than wet preparation (PEG) because of patient preferences and because PEG can leave liquid in the colon that can potentially obscure lesions.³⁸⁰ Fecal tagging now appears to be routinely employed (oral ingestion of high-density oral contrast agent so that residual colonic contents can be differentiated from polyps) and appears to decrease the need for cathartic preparation. Additionally, there are different contrast agents, either barium- or iodine-based (ionic and nonionic), and the selection of which to use is largely based on local experience. Current practice centers on multidetector row CT scanners, which uses much thinner slices with faster scan times, resulting in better imaging and decreased radiation dose. Finally, there are differences in reading software. Commonly used reading software allows for both two- and three-dimensional display. The selection of the primary method appears to depend on radiologist preference. Other practice variations that influence the impact and implementation of screening CTC includes colonoscopy referral or surveillance criteria, as well as coordination with colonoscopy resources. Currently, there is consensus that large lesions (≥10 mm) should be referred to colonoscopy for polypectomy. There is variation in practice for smaller lesions, such that 6- to 9-mm lesions may be referred to colonoscopy for polypectomy or be monitored with CTC surveillance (with a followup CTC in 3 years), and the smallest lesions (≤5 mm) may be ignored or monitored. The American College of Radiology states that people with lesions of 6–9 mm should be offered colonoscopy and lesions smaller than 5 mm need not be reported.^{205, 378, 381, 382} Preference for CTC over colonoscopy may be, in part, due to difference in bowel preparation. Ideally, while same-day colonoscopy could avoid duplicate preparation, it may result in suboptimal colonoscopy if limited bowel preparation is used for CTC and would require close coordination between radiology and gastroenterology departments/services.

CTC Harms

Immediate serious adverse events from screening CTC appear to be uncommon. Perforations were the most commonly reported harms (estimated at 1.3 per 10,000 examinations [95% CI, 0 to 2.9]); however, these perforations were detected radiographically (not symptomatic) and sustained by room-air manual insufflation which is no longer used in practice. However, like FS, CTC may require followup or therapeutic colonoscopy, and we did not find sufficient evidence to estimate serious adverse events from colonoscopy followup procedures.

Potential harms from CTC include exposure to radiation, especially if used in a program of screening that requires repeated examinations. Radiation dose in our included studies ranged from 0.8 to 5.3 mSv, consistent with a 2012 survey of academic and nonacademic institutions which found that the median radiation dose per screening CTC examination was 4.4 mSv,^{383 384, 385}and a 2018 narrative review reporting the typical radiation exposure associated with a CTC examination at ≤3 to 6 mSv (which is higher than radiation exposure from digital mammography or CT for lung cancer screening).³⁸⁶

Given that the average amount of radiation exposure from background sources in the United States is about 3.0 mSv per year,³⁸⁷ ionizing radiation from a single CTC examination is low. Even low doses of ionizing radiation, however, may convey a small excess risk of cancer.^{388, 389} We identified no studies directly measuring the risk for stochastic effects (i.e., cancer) caused by radiation exposure from CTC. We can indirectly estimate these adverse effects, however, based on the range of effective radiation dose for CTC reported in the literature and estimates of lifetime attributable risk of malignancy (i.e., all solid cancers and leukemia) from the National Research Council report “Health Risks From Exposure to Low Levels of Ionizing Radiation.”³⁸⁷ Based on this report, the council predicts that approximately one additional individual per 1,000 would develop cancer (solid cancer or leukemia) from an exposure of 10 mSv above background using the linear no-threshold (LNT) model. In comparison, 420 individuals per 1,000 would be expected to develop cancer from other causes over their lifetimes. Because of limitations in the data used to develop risk models, the risk estimates are uncertain, and variation by a factor of 2 or 3 cannot be excluded.³⁸⁷ Multiple organizations support the LNT model to estimate potential harms of radiation exposure of less than 100 mSv, including the Nuclear Regulatory Commission, the International Commission on Radiological Protection, the U.S. National Council on Radiation Protection and Measurements, the United Nations Scientific Committee on the Effects of Atomic Radiation, and the U.K. National Radiological Protection Board. Other organizations, however, believe that the LNT model is an oversimplification and likely overestimates potential harms of low-dose radiation exposure, including the Health Physics Society, the France Academy of Sciences/National Academy of Medicine, and the American Nuclear Society.³⁹⁰ The effective radiation dose in CTC targets the abdomen and would not likely increase the risk of certain prevalent cancers (e.g., cancers of the breast, thyroid, or lung), although the risk for leukemia or abdominal organ cancer may remain. This risk estimate is consistent with other published literature on radiation exposure risk from CT.^{388, 391}

Modeled data based on the National Research Council’s assumptions, and using a mean dose of 8 mSv for women and 7 mSv for men per CTC examination, found that the benefits of CTC screening every 5 years (from ages 50 to 80 years) far outweigh any potential radiation risks, with 15 cases of radiation-related cancers per 10,000 persons screened (95% CI, 8 to 28) versus 358 to 519 CRC cases prevented per 10,000 persons screened.³⁹²

Extracolonic Findings

CTC also detects extracolonic findings, which could be a benefit (e.g., detection of intervenable extracolonic cancer, abdominal aortic aneurysm) or harm (e.g., overdiagnosis, procedural harms from subsequent testing). Extracolonic findings are very common and increase with age. Approximately 1.3 to 11.4 percent of CTC exams have extracolonic findings that necessitate actual diagnostic followup. Only a small proportion of CTC exams have findings that ultimately require any type of definitive treatment (≤3%). Therefore, judicious handling of the reporting and diagnostic workup of extracolonic findings is crucial to minimize the burden of testing (and associated cost and harms of testing), as many findings ultimately prove to be of no clinical consequence. Additional reading software may allow for repurposing CTC examinations to obtain bone mineral density from the lumbar spine to screen for osteoporosis if desired/indicated.^{393, 394} It remains unclear whether detection of extracolonic findings represents a true overall benefit or harm based on empirical evidence.

Age

Because of the higher incidence of CRC in adults under age 50 years over time, in 2018 the ACS issued a qualified recommendation to start screening at age 45 years. Earlier age to initiate screening is primarily based on the epidemiology of disease and modeling studies accounting for the incidence of CRC by age. To date, we have little to no empiric evidence evaluating potential differences in the effectiveness of screening, test performance of screening tests, and the harms of screening in younger age groups (i.e., <50 years vs. older than 50 years). While a few studies of effectiveness (KQ1) recruited adults less than 50 years, none of these studies report stratified analyses by younger age subgroups. Any age differences in older gFOBT and FS screening trials were not statistically significant. Any differences in the effectiveness of screening in younger ages would be attributable to varying the underlying risk/incidence of CRC and/or natural history of disease, as well as differences in test accuracy by age. Limited studies demonstrate no difference in test performance (KQ2) of stool testing or harms of colonoscopy in people younger than 50 years. Although we do not hypothesize that colonoscopy or CTC are more harmful in younger adults than older adults, starting screening at younger ages will accrue more procedural harms and ECF, which should be weighed against any incremental benefit of earlier start to screening.

It is yet unclear whether the spectrum of sporadic CRC in younger adults mimics that seen in a traditionally screened age group, as there is evidence to suggest that a large proportion of the increase in CRC in those under age 50 is rectal versus colon cancer, and those with earlier onset CRC tend to have distinctive clinical features, have a more advanced stage at diagnosis, and poorer overall survival rates, which may be due to a difference in screen- versus symptom-detected disease and/or a more aggressive natural history.⁴³⁸

Current recommendations also differ on the age to stop screening; they range from ages 74 to 85 years. Few studies include older adults age 75 years and older to conduct robust subgroup analyses for the effectiveness, test accuracy and harms of screening. Limited empiric evidence suggests that screening colonoscopy may not result in the same benefit in reduction of CRC incidence in adults ages 75 to 79 years compared with those ages 70 to 74 years.¹²⁸ In addition, limited evidence suggests that CTC has lower sensitivity in older adults³⁵⁰ and the specificity of sDNA-FIT decreases with advancing age³⁵¹ (higher false positive screening). And more robust evidence consistently demonstrates increasing serious harms from colonoscopy (as well as ECF on CTC exams) with advancing age.

Race/Ethnicity

Due to the higher incidence of CRC in Black people compared with white people (and other races/ethnicities), the USMSTF in 2017 recommended screening African Americans at age 45 years, and others at age 50 years. To date, we have little to no empiric evidence evaluating potential differences in the effectiveness of screening, test performance of screening tests, and the harms of screening by race/ethnicity (i.e., Black versus white). While effectiveness studies (KQ1) include nonwhite adults, none report stratified analyses by racial/ethnic subgroups. Again, any differences in the effectiveness of screening would be attributable to varying underlying risk/incidence of CRC and/or natural history of disease. We do not hypothesize that there are any differences in test performance or harms of screening tests by race/ethnicity; and as expected there are limited studies demonstrate no difference or inconsistent findings in test performance (KQ2) of stool testing or harms of colonoscopy by race/ethnicity.

Most of the evidence to date suggests that differences in risk of CRC and CRC mortality between Black and white adults is primarily driven by differences in receipt of good quality screening and subsequent care of cancer rather than inherent biological differences.⁴³⁹ Furthermore, race is a social construct reflecting much more than heritable disease risk, and therefore confounded by may other factors increasing risk for developing cancer and progression of cancer, not limited to behavioral and environmental risk factors.⁴⁴⁰ However, there remains uncertainty on the differences in carcinogenic mechanisms contributing to observed disparities in outcomes between Black and white adults.⁴⁴¹ While there is some evidence for a difference in the distribution of adenomas in the proximal versus distal colon, and in tumor markers in Black versus white people, the clinical significance of this difference on CRC incidence and mortality is unclear.

Sex

Although no recommendations tailor screening by sex, there is evidence to suggest differences in the effectiveness of screening, test performance of screening tests, and harms of screening in men versus women. Screening FS and selected gFOBT trials suggest a greater benefit in CRC mortality reduction in men than women. These results may be explained by the differences in sex-specific CRC incidence and mortality, as well as differences in the distribution of CRC in the colon (i.e., distal versus proximal) between men and women.^{442, 329} Results were somewhat inconsistent for the FIT test accuracy, with some evidence to suggest that sensitivity may be higher (with lower specificity) to detect CRC in men compared with women. A 2019 systematic review evaluating the effect of sex (and age and positive threshold) on FIT test accuracy found that the 95% CI intervals overlapped between men and women.⁴⁴³ Likewise, results were inconsistent for serious harms from colonoscopy, with some, albeit limited, evidence to suggest slightly higher rates of complications in men compared with women from screening colonoscopy.

Family History

Family history of CRC represents an approximation of genetic risk and is typically characterized in terms of the number of affected relatives, the degree of relatedness, and their age at CRC diagnosis. Individuals at the highest risk are those from families with known genetic syndromes, multiple affected relatives, and/or relatives with early age cancer diagnosis, particularly before age 50 years. At more moderate risk levels are people with one or more FDR or second degree relative (SDR) with later onset cancer. A systematic review of eight large population-based cohorts found that the prevalence of family history of one FDR with early-onset cancer was approximately 0.3 percent, while the prevalence of a single FDR with history of late-onset (after age 60) CRC was more than 3 percent.⁶⁸ Because our review focuses on the evidence to support screening in generally average risk adults, our discussion about the evidence for screening focuses on those at “moderate risk” as opposed to those with the highest hereditary risk for whom most U.S. guidelines recommend early and more frequent colonoscopy (i.e., colonoscopy is typically recommended at age 40 or 10 years before the relative’s age at diagnosis and repeated at 5–10 year intervals).^{63, 444} (Appendix H Table 1) The evidence on initiation of earlier screening in people with moderate familial risk for CRC is summarized below and detailed in Appendix H.

A large body of observational evidence spanning multiple countries and populations suggests that CRC risk increases as intensity of family history of CRC increases (more relatives, closer in relation, younger age at diagnosis), providing a plausible hypothesis for a screening benefit at earlier ages in these groups. Pooled risk estimates for a single FDR with CRC over age 60 are elevated compared to people with no family history (1.83, 95% CI, 1.47–2.25).⁶³ A systematic review of risk for CRC associated with family history found that the risk for CRC increased from 1.8 percent for a 50 year old with no family history to 3.4 percent with at least one affected relative and to 6.9 percent with two or more affected relatives.⁴⁴⁵ A review of reviews conducted for the Canadian guidelines found similar increased levels of risk across nearly all types of studies and populations.⁴⁴⁴

There is limited empiric evidence on the effectiveness of screening, test performance of screening tests and harms in people at moderately increased risk of CRC due to family history and no evidence in this group under age 50 years. Although some studies do include people with a family history of CRC, most do not report results stratified by familial risk. One included observational colonoscopy study in health professionals found that in people with a FDR family history of CRC, the association with CRC mortality was no longer statistically significant after 5 years (multivariate HR 0.91; 95% CI, 0.55 to 1.52) compared with a sustained association beyond 5 years in people without a family history (multivariate HR 0.43; 95% CI, 0.32 to 0.58) (p=0.04 for interaction).²¹ One excluded population-based case-control study found that previous colonoscopy was associated with decreased CRC risk in people with all levels of family history. Regardless of family history status, colonoscopy was associated with a lower CRC risk (OR 0.25 [95% CI, 0.22 to 0.28] for people without family history and OR 0.45 [95% CI, 0.36 to 0.56] for people with family history).⁴⁴⁶ Neither of these studies report results for adults under age 50 years. No included studies reported variation of test accuracy or harms by family history.

Multivariable Risk Assessment

Although the concept of individualizing CRC screening recommendations has become more compelling as we have learned more about modifiable and non-modifiable risk factors, multivariate risk assessment for CRC risk is not commonly used in clinical practice^{70, 447} and currently there is no commonly used/accepted risk assessment tool to help tailor CRC screening.⁶⁹ In 2019, one international guideline panel, as part of the BMJ Rapid Recommendations series, issued a weak recommendation against screening in asymptomatic adults ages 50 to 79 years with an estimated 15-year CRC risk below 3 percent using a validated multivariate risk assessment tool (QCancer) which includes a number of variables in addition to age, sex, race/ethnicity, and family history.⁸⁰ In theory, multivariate risk assessment could also identify persons at higher risk for CRC and in whom to initiate screening earlier than age 50 years.

While many risk models or scores have been developed to predict the risk of CRC and/or advanced neoplasia, there are no trials evaluating the benefits and harms of implementing risk assessment to guide CRC screening. Two recent systematic reviews summarize the performance (mainly discrimination) of risk prediction models for CRC and/or advanced neoplasia in asymptomatic general risk adults.^{447, 448} A 2016 systematic review identified 52 models described in 40 studies for assessing risk of CRC or advanced neoplasia in average-risk populations; in aggregate these 52 models considered 87 different risk factors obtained through medical records, self-reported questionnaires, and laboratory testing inclusive of genetic biomarkers.⁴⁴⁷ Commonly included factors were age, sex, family history (generally specified as FDR), BMI, and lifestyle factors (e.g., smoking, alcohol, diet, exercise). Overall, the discrimination of the models ranged from an area under the curve (AUC) of 0.65 to 0.70. The authors found that, in general, models including lifestyle behaviors (obtained by questionnaire) and genetic biomarkers did not have better discrimination than models with risk factors that could be routinely obtained through medical records (i.e., age, sex, family history, smoking, +/− alcohol). In external validation studies, 10 of these models showed acceptable discrimination, AUC 0.71 to 0.78. These include two models containing only three variables (age, sex, and BMI or family history).⁴⁴⁷ A 2018 review focused on multivariate risk tools for advanced neoplasia only and identified 17 original risk scores described in 22 unique studies.⁴⁴⁸ Findings from this review were consistent with the 2016 review in the commonly included factors and discrimination (AUC) of the risk tools. This review also demonstrated a substantial variation in discrimination even for the same risk score across different studies. The review conducted meta-analyses of discrimination for each risk score evaluated in more than one study and found that the most evaluated risk scores (4 or more studies) had less optimal discrimination (AUC 0.61 [95% CI, 0.59 to 0.64] to 0.64 [95% CI, 0.60 to 0.68]. The risk tool with the highest discrimination (AUC 0.70 [95% CI, 0.61 to 0.79]) was only evaluated in two studies.

Two publications externally validated a series of risk models identified in the 2016 review in large population-based cohorts in the United Kingdom and Europe.^{449, 450} One study externally validated 14 different risk models to predict CRC in a large (n=373,112) population-based cohort in the UK (UK Biobank).⁴⁴⁹ Another study externally validated 16 different risk models for CRC in two large population-based cohorts, the European Prospective Investigation into Cancer and Nutrition (EPIC) (n=491,992) and United Kingdom Biobank n=475,629).⁴⁵⁰ These two studies externally validated overlapping risk models. Overall these two studies found that the performance of published risk models for CRC varied widely. Both studies concluded that there are several models (including QCancer) with easily identifiable risk factors that possess good calibration and discrimination, and thus are promising for implementation. Both studies call for modeling plus or minus clinical impact studies to further evaluate their promise for clinical practice.

Only four studies examined risk prediction for advanced colorectal neoplasia specifically in adults younger than age 50 years.^451–454 These studies were development and initial validation studies in large generally asymptomatic populations in Korea. The models demonstrated that a combination of risk factors similar to those in other models (e.g., age, sex, BMI, family history, smoking, laboratory tests) can identify people at higher risk for advanced neoplasia (AUC from 0.66 to 0.72). These models do not appear to be externally validated. In general, these studies included populations with lower average BMI (when reported) than U.S. populations, and given the 10-fold difference in CRC incidence internationally, there is a need to validate in broader populations applicable to U.S. populations.