Secondary Logo

Journal Logo

ARTICLE: COLON

Randomized Controlled Trial of Personalized Colorectal Cancer Risk Assessment vs Education to Promote Screening Uptake

Yen, Timothy MD1; Qin, FeiFei MPH2; Sundaram, Vandana MPH2; Asiimwe, Edgar MD, MSc3; Storage, Tina MD1; Ladabaum, Uri MD, MS4

Author Information
The American Journal of Gastroenterology: February 2021 - Volume 116 - Issue 2 - p 391-400
doi: 10.14309/ajg.0000000000000963

Abstract

INTRODUCTION

Colorectal cancer (CRC) is the third most commonly diagnosed cancer among men and women in the United States and the third leading cause of cancer death. It is estimated that in 2019, CRC was diagnosed in 145,600 people, leading to an estimated 51,020 deaths (1). The US rates of CRC incidence and death have decreased since the 1980s, with a more rapid decline in both over the past decade, which has largely been attributed to CRC screening (2,3). Although CRC carries significant morbidity and mortality, the 5-year survival is 90% when CRC is detected at a localized stage. However, CRC screening remains underutilized, and fewer than half of CRCs are diagnosed at a localized stage (4). Current guidelines identify only a relatively small proportion of adults as eligible for higher-intensity CRC screening and then provide uniform recommendations for the remaining average-risk population consisting of a variety of test options, each with its own interval, starting at age 45 or 50 years.

According to data from the National Health Interview Survey, only 57.6% of American men and women aged 50–75 years have undergone CRC screening using one of the most commonly recommended methods: colonoscopy, sigmoidoscopy, fecal occult blood testing (FOBT), or fecal immunochemical test (FIT) (5). Numerous factors have been associated with suboptimal screening rates. These factors can be categorized into nonmodifiable factors (age, ethnicity, and sex), factors that are not easily modified (educational level, health insurance, and socioeconomic status), and potentially modifiable factors (knowledge about CRC and screening, attitudes toward screening, perception of risk for developing CRC, and access) (6–10). Modifiable factors are of particular interest given their susceptibility to change. Risk perception, or perceived susceptibility as defined by the Health Belief Model, refers to an individual's subjective assessment of risk of developing a given health problem (11). Studies have shown that high-risk perception of developing CRC correlates with higher screening uptake (6,12–15). Similarly, underestimation of personal CRC risk is associated with lower screening uptake (13,16).

In this study, we sought to determine whether providing an individual with a personalized CRC risk assessment would impact CRC screening intent and behavior. We used the National Cancer Institute's CRC Risk Assessment Tool (CCRAT) to estimate an individual's lifetime risk of developing CRC by taking into account various risk factors, including demographics, diet, physical activity, medical history, smoking history, and family history (17,18). To assess behavior change, we used the transtheoretical model (TTM), which proposes that individuals pass through various stages in the process of changing a health behavior. This model has been applied to CRC screening behavior (19). The stages include precontemplation (no intention to take action in the near future), contemplation (intention to take action in the near future), preparation (initiating small steps toward making a change), action (completion of change with intention to continue changed behavior), and maintenance (sustained behavior change) (20). Risk assessment could affect overall participation by moving patients forward in stages of the TTM toward screening completion, but also has the potential to decrease participation in screening. We hypothesized that personalized CRC assessment risk would result in net progression through the stages of the TTM, culminating in an overall increase in screening completion.

Our primary objective was to determine whether knowledge of one's CRC risk, as calculated by the National Cancer Institute's CCRAT, increases CRC screening participation in previously unscreened adults compared with general education in the context of no organized outreach efforts. Our secondary objectives were to determine whether knowledge of one's CRC risk influences screening intent and attitudes in previously unscreened adults.

METHODS

General study design and setting

This was a 2-arm, open-label, parallel randomized control trial design, performed in a 1:1 ratio. The study was approved by the Stanford University Committee on Human Research Institutional Review Board, and it is registered in ClinicalTrials.gov, NCT03819920. A REDCap database was used to operationalize study arm allocation and for data entry and management. The Stanford REDCap platform is developed and operated by the Stanford Medicine Research Information Technology team. The REDCap platform services at Stanford are subsidized by (i) the Stanford School of Medicine Research Office and (ii) the National Center for Research Resources and the National Center for Advancing Translational Sciences, National Institutes of Health, through grant UL1 TR001085. All authors had access to the study data and reviewed and approved the final manuscript.

The study was performed in the context of our institution's usual care. Primary care patients who are potentially eligible for screening are tracked (see Eligibility, below), but screening is left to the discretion of individual primary care physicians and staff. There is no uniform organized outreach program such as mailed FIT. At the time of study design, the fraction of primary care patients up to date with CRC screening was an estimated 50%. A recent administrative review suggests that approximately 60% of referrals to the Stanford Digestive Health Center originate from outside Stanford or by patient self-referral.

Eligibility

We determined potential subject eligibility based on CRC screening status for all patients across 5 Stanford Health Care primary care clinics between ages 50 and 75 years from 2015 to 2018 using a population health management list. A dedicated team including primary care nurses generates this list by tracking patients' health maintenance tasks using an automated system embedded in the Epic (Epic Systems Corporation, Verona, WI) electronic medical record (EMR). We used this list to identify patients who might not be up to date with CRC screening. Because this automated system may not capture testing performed outside of Stanford Health Care, we performed further manual electronic chart review to exclude patients with screening documented in notes or media attachments. Screening status was reassessed verbally during the initial telephone call, when ultimate eligibility was determined.

We included patients whether or not they regularly attended clinic visits. Patient were initially contacted using an official letter to the patient's listed home address regarding study details, Health Insurance Portability and Accountability Act authorization, and an option to opt out of follow-up contact for study recruitment (sample letters in Supplemental Digital Content 1, http://links.lww.com/AJG/B690). Subjects who did not opt out were contacted by telephone at least 2 weeks after letters were sent to allow adequate time for response. During this telephone contact, patient screening status and eligibility were verbally reassessed before proceeding to consent, randomization, and intervention (see Supplemental Digital Content 2, Initial Telephone Contact, Section 1, http://links.lww.com/AJG/B691). All contacts with subjects were through telephone, with no face-to-face visits. Exclusion criteria included prior CRC screening (identified through manual chart review or verbally), break-the-glass charts designated as confidential by patient request and/or because of psychiatric care, non-English speakers, personal history of CRC, history of inflammatory bowel disease (ulcerative colitis or Crohn disease), hearing impairment (inability to hear questions on telephone contact), gross lack of decision-making capacity (based on the report by a family member on telephone contact), and history of Lynch syndrome or familial adenomatous polyposis.

Randomization

Subjects were randomized 1:1 using block randomization stratified by sex (female or male) and age group (younger than 60 years or 60 years and older to avoid gross imbalance in age distributions) to either the CCRAT arm or control arm. The randomization scheme was generated using the blockrand (v 1.3; Snow, 2013) package in R 3.5.2 (R Core Team, 2018). The blockrand function randomized subjects to specified arms within sequential blocks; block sizes were randomly chosen from the set of 2, 4, 6, 8. The randomization scheme was incorporated into REDCap to assign participants into the treatment arms and was concealed from study personnel involved in recruitment.

Interventions

On telephone contact, all participants provided verbal informed consent after hearing a standardized script. Then, participants completed a verbally administered preintervention survey assessing demographic information and knowledge and attitudes about CRC screening. These included perceptions of absolute CRC risk, relative CRC risk, knowledge of available CRC screening tests, fear of finding CRC on screening, and screening intent defined by TTM stage (see Supplemental Digital Content 2, Initial Telephone Contact, Sections 2 and 3, http://links.lww.com/AJG/B691). Randomization was then performed, and study interventions were administered immediately during the same telephone encounter.

The CCRAT arm was guided through the CCRAT's series of questions to estimate subjects' personalized risk of CRC. The personalized lifetime risk score was recorded, and participants were provided with scripted feedback regarding their absolute numerical and relative lifetime CRC risk as detailed in the online National Cancer Institute's CCRAT (https://ccrisktool.cancer.gov). Relative lifetime CRC risk was presented as higher or lower or same (within 0.1%) as average risk for the patient's age, sex, and ethnicity. We provided no counseling or decision-making support. The general education control arm received a short generic script on CRC and CRC screening with no numerical risk statistics or counseling (see Supplemental Digital Content 2, Initial Telephone Contact, Sections 4a and 4b, http://links.lww.com/AJG/B691).

Following these interventions, both arms immediately repeated assessment of CRC screening intent (see Supplemental Digital Content 2, Initial Telephone Contact, Section 5, http://links.lww.com/AJG/B691).

At 6 and 12 months after initial intervention, participants' EMRs were individually reviewed for CRC screening completion, defined as completion of any available screening test. For those without recorded screening completion, telephone contact was performed to assess screening status (see Supplemental Digital Content 2, Telephone Contact at 6 and/or 12 months postintervention, http://links.lww.com/AJG/B691). If screening was not completed at 6 months by review of EMR or telephone contact, screening intent was assessed again at 12 months.

There was no research-driven intervention regarding screening modality. All subsequent clinical care, including screening, was usual care.

Outcomes

The primary outcome was time to screening at 12-month follow-up, with the date of initial intervention considered day 0 and the date of screening as recorded in the EMR considered the date of event. The date of follow-up phone call was considered the date of event for patients who self-reported screening but had no screening result recorded in the chart to minimize recall bias.

Secondary outcomes included CRC screening completion rates at 6- and 12-month follow-up; screening intent at baseline; change in screening intent immediately postintervention; changes in screening intent at 6- and 12-month follow-up for those remaining unscreened; and CRC screening rates at 12 months as a function of CCRAT score tertile in the CCRAT arm. The screening completion rate at 6 months and 12 months was determined by review of participants' EMRs. Participant self-report on screening was used for those for whom a screening result was not recorded in the EMR. Intent was coded as a multilevel categorical variable: 0 = precontemplation, 1 = contemplation, and 2 = preparation. Change in intention to screen for CRC was evaluated by calculating the difference in score pre- and postintervention. For instance, if a participant was in the precontemplation stage preintervention and was in the contemplation stage postintervention, this was considered a 1-step positive change.

Sample size

Based on prior internal data, we assumed that the education control arm would have a screening completion rate of 50% by 12 months (event rate for remaining unscreened of 50%) and that the intervention would increase this rate to 65% (event rate for remaining unscreened of 35%, leading to a hazard ratio of 1.5). We assumed uniform accrual over a 6-month period, minimum follow-up of 12 months, and total time of study of 18 months for all subjects. To achieve a power of 80% at a 2-sided alpha level of 0.05, we required 226 patients total. Sample size calculations were performed using a log-rank test in PASS Version 14.0 (PASS 14 Power Analysis and Sample Size Software, 2015; NCSS, LLC, Kaysville, UT, ncss.com/software/pass).

Statistical analysis

This was an intention-to-treat analysis. Descriptive statistics on demographics and preintervention knowledge and attitudes on CRC screening were presented as mean with standard deviation for continuous variables and frequencies with percentages for categorical variables. Multinomial univariable logistic regression was used to analyze the directionality and significance of association between baseline screening intent and demographic characteristics, knowledge, fear, and perception of CRC in all participants (not by study arm). A Cox proportional hazards regression was used to evaluate the primary end point of time to screening by study arm. Screening completion rates at 6 and 12 months were calculated as the proportion of participants who were screened by 6 and 12 months, respectively, and logistic regression was used to assess differences in screening rates between the 2 arms. The χ2 tests were also used to determine the association between screening rates and tertiles of CCRAT risk score among those in the intervention arm. A multinomial logistic regression was used to evaluate the secondary end point of change in screening intent at postintervention. Age group (<60 and 60+ years) and sex were included as the demographic covariates in all models. All statistical tests were 2 sided and evaluated at an alpha of 0.05. SAS Version 9.4 (SAS Institute, Cary, NC) and R Version 3.5.2 (R Core Team, 2018) were used for all data analyses. Data input was performed using REDCap software.

RESULTS

Study participants

From our population management list of 5,618 patients due for CRC screening between October 2015 and January 2018, 1,711 patients had no prior CRC screening documented (Figure 1). Of these, 538 patients did not respond to multiple (up to 3) contact attempts by telephone, 357 declined to consent, and 223 reported prior CRC screening on contact. We excluded 129 patients with age over 75 years at the time of contact, 87 non-English speakers, 67 deceased, 44 with a personal history of CRC, 23 with restricted charts, 9 with inflammatory bowel disease (ulcerative colitis or Crohn's disease), 3 with significant hearing impairment, and 2 with lack of decision-making capacity. No remaining patients had known Lynch syndrome or familial adenomatous polyposis. A total of 230 patients were enrolled, with 114 patients randomized to the intervention (CCRAT) arm and 116 patients randomized to the education control arm, with 5 lost to follow-up in the intervention arm and 2 lost to follow-up in the control arm (Figure 1).

Figure 1.
Figure 1.:
CONSORT flow diagram. Screening completion was assessed at 6- and 12-month follow-up. Screening intent was assessed before randomization, immediately postintervention, and at 6- and 12-month follow-up if screening was not completed. CCRAT, Colorectal Cancer Risk Assessment Tool; CRC, colorectal cancer; Inflamm bowel disease, inflammatory bowel disease.

Baseline characteristics

Most patients were younger than 60 years, female, White or Asian, had some college or higher education, and had an annual household income of $75,000 or more (Table 1). A small minority had a family history of CRC or a personal history of another cancer (Table 1). There were no significant differences between arms in any demographic category (Table 1).

Table 1.
Table 1.:
Baseline characteristics

Most patients had heard of colonoscopy (99.1%) and FIT (74.3%) or FOBT (80.0%), but rarely other tests (Table 2). Over half (56.9%) expressed fear of finding CRC on screenings, but the majority (57.4%) considered it neither likely nor unlikely that they would develop CRC, and 40.0% considered it unlikely or very unlikely. Perceptions of risk relative to the average person were similar to perceptions of absolute risk (Table 2). Only 9.1% had already scheduled screening tests, 47.8.8% contemplated screening within 6 months, and 43.0% were in the precontemplative state (Table 2).

Table 2.
Table 2.:
Baseline knowledge, attitudes toward CRC screening, and screening intent (N = 230)

Baseline screening intent among all participants

Compared with a reference of the precontemplative stage, those of older age were less likely to be in the contemplative stage (odds ratio [OR] 0.95 [95% confidence interval (CI) 0.92–0.99], P = 0.022), whereas those with a family history of CRC (OR 11.87 [95% CI 3.08–45.83], P = 0.0003) or perception of high relative CRC risk (OR 14.00 [2.15–91.11], P = 0.0068) were more likely to be in the preparative stage (see Supplemental Digital Content 3, Table A, http://links.lww.com/AJG/B692).

Primary outcome: time to screening

The median time to follow-up was 360 days (interquartile range 147–363 days) in each group. Time to screening did not differ in the CCRAT vs the education control arm (hazard ratio 0.78 [95% CI 0.52–1.18], P = 0.24) (Figure 2).

Figure 2.
Figure 2.:
Time to screening. A Cox proportional hazards model of proportion remaining unscreened over time showed no significant differences between the Colorectal Cancer Risk Assessment Tool (CCRAT) intervention vs the education control arm.

Secondary outcome: screening completion

At 12 months, screening completion rates were 38.6% in the CCRAT arm vs 44.0% in the education control arm (OR 0.80 [95% CI 0.47–1.37], P = 0.41) (Figure 3). At 6 months, screening completion rates were 25.4% in the CCRAT arm vs 34.5% in the education control arm (OR 0.65 [95% CI 0.36–1.15], P = 0.14) (see Supplemental Digital Content 3, Table B, http://links.lww.com/AJG/B692).

Figure 3.
Figure 3.:
Screening completion at 12 months. A logistic regression model for screening completion in the Colorectal Cancer Risk Assessment Tool (CCRAT) intervention arm vs the education control arm showed no significant difference. In the CCRAT arm, there was a higher screening completion rate in the top risk score tertile compared with the bottom and middle risk score tertiles, but this trend was not statistically significant (P = 0.10).

Among those who completed screening at 12 months, the type of test chosen was similar between arms. In the CCRAT arm, 50.0% chose FIT and 50.0% chose colonoscopy vs 52.9% FIT and 47.1% colonoscopy in the education control arm.

Secondary outcome: screening intent

Most patients were either in the precontemplative or contemplative stages at baseline, and they remained in their respective stages of intent immediately after the interventions of CCRAT assessment or education control (Figure 4). Overall, the proportion of patients in the precontemplative stage shrank, whereas the proportion in the contemplative stage grew in both arms immediately after the interventions (Figure 4). Specifically, an absolute 26% of patients in the CCRAT arm and 19% of patients in the education control arm progressed from the precontemplative to the contemplative stage, but the difference between arms was not statistically significant (OR 1.52 [95% CI 0.81–2.86], P = 0.19). An absolute 4% of patients in the CCRAT arm and 3% of patients in the control arm moved in the opposite direction from the contemplative to precontemplative stage, with no statistically significant difference between arms (OR 1.93 [95% CI 0.45–8.34], P = 0.38]) (Figure 4) (see Supplemental Digital Content 3, Table C, http://links.lww.com/AJG/B692).

Figure 4.
Figure 4.:
Screening intent. A logistic regression model of changes in screening intent from before (pre) to immediately after (post) intervention showed no significant differences between study arms. Bubble sizes reflect the absolute percentage of persons in a given stage in each study arm. CCRAT, Colorectal Cancer Risk Assessment Tool; Edu, education control.

At 6 months, among the 84 patients remaining unscreened in the CCRAT intervention arm, 19 (22.6%) were in the precontemplative stage, 55 (65.5%) were in the contemplative stage, and 10 (11.9%) were in the preparation stage vs 27 (35.5%), 43 (56.6%), and 6 (7.9%), respectively, among the 76 patients remaining unscreened in the education control arm (P = 0.18). At 12 months, the respective proportions in the precontemplative, contemplative, and preparation stages were 24/65 (36.9%), 37/65 (56.9%), and 4/65 (6.2%) in the CCRAT intervention arm vs 24/63 (54.0%), 21/63 (33.3%), and 8/63 (12.7%) in the education control arm (P = 0.021). In both arms, among those patients who went on to complete CRC screening by 6 months or 12 months, the majority had been in the contemplative stage at the previous assessment (see Supplemental Digital Content 3, Table C, http://links.lww.com/AJG/B692).

Secondary outcome: screening completion by risk score tertile

In the CCRAT arm, the predicted lifetime CRC risk ranges were 7.1%–11.0% in the top risk score tertile, 5.0%–6.9% in the middle risk score tertile, and 3.3%–4.9% in the bottom risk score tertile. There was a higher screening completion rate at 12 months in the top risk score tertile (52.6%) compared with the bottom (32.4%) and middle (31.6%) risk score tertiles, but differences in completion rates between tertiles were not statistically significant (P = 0.10) (Figure 3) (see Supplemental Digital Content 3, Table D, http://links.lww.com/AJG/B692).

DISCUSSION

In this prospective randomized controlled trial of previously unscreened adults, CRC risk assessment with the CCRAT did not result in an overall increase in screening participation or screening intent compared with a general nontailored education control. In the CCRAT arm, we observed a higher screening completion rate in the top risk score tertile compared with the other 2 tertiles, but this trend was not statistically significant, possibly due to the tertiles' sample size.

In prior studies of patients due for CRC screening, several institution-driven interventions have improved screening completion rates, including mailed FOBT (OR 1.9) (21,22), tailored education (OR 2.2) (23), telehealth patient navigators (OR 1.82–2.11) (24,25), in-person digital navigators (OR 2.5) (26), and the combination of tailored education and telehealth patient navigators (OR 2.69) (27). However, in a previous study, CRC risk stratification had no significant impact on what CRC screening test was ordered by primary care physicians (28). Our study focused on the impact of CRC risk assessment on patient-driven decisions and behavior related to CRC screening. To study this end point, we intentionally omitted counseling and patient navigators as study interventions, and we excluded patients with prior CRC screening.

In our study, we observed no overall differences in screening completion rates between patients receiving personalized risk assessment vs patients receiving a general education control. Previous studies suggest that a primary motivator for screening completion may be provider-driven outreach, rather than specific knowledge about CRC risk (29). Therefore, it may be preferable to focus on achieving screening completion by addressing logistical barriers for all screen-eligible patients, until CRC risk can be predicted accurately and systems are established to deliver screening tailored to risk. Navigation programs have been shown to be cost-effective in various patient populations (30,31). It remains to be determined whether this will remain the case if younger patients undergo CRC screening (32).

Although the CCRAT has been shown to be only modestly predictive of future CRC occurrence (18) and of advanced adenoma prevalence at the time of screening (33,34), our trial studied the impact of risk stratification on screening intent and behavior rather than the predictive accuracy of the tool itself. Although there are simpler CRC risk prediction tools, as well as more complex tools that incorporate biomarkers (35,36), we chose the CCRAT because it is relatively practical to administer and because it has been externally and prospectively validated for the outcome of advanced neoplasia at colonoscopy (18,33,34). More complex prediction tools, including those that incorporate biomarkers, have not been proven to have substantially better discriminatory power (37). Our observations may be informative regarding patient behavior in the face of personalized risk stratification in general, regardless of the tool used. However, we provided no explicit information to patients regarding the accuracy of the CCRAT risk estimate. Despite high educational levels (Table 1), our study population demonstrated poor comprehension of absolute vs relative CRC risk and tended to overestimate true CRC risk (Table 2). It is conceivable that risk stratification with a highly accurate tool, combined with information to patients about the tool's high accuracy, could lead to different outcomes than those observed in this study. Such a tool could also identify very-low-risk persons for whom the benefits of screening are uncertain, allocate higher-cost colonoscopy screening tests to higher-risk individuals while allocating lower-cost less invasive FIT tests to lower-risk individuals, and aid in design of specific screening programs (screening interval, start date, and stop date) tailored to an individual's risk. The logistical challenges of implementing any risk prediction tool in clinical practice must be acknowledged. Although guiding study participants through the CCRAT by phone was relatively straightforward during this study, participants had self-selected, and it did require dedicated effort and time. Such time may not be available during a routine health maintenance office visit, and completion of risk assessment outside of an office visit would require the appropriate infrastructure and resources.

Our study was not powered to detect differences in screening completion rates by risk score tertile. Future studies may aim to achieve the power needed to answer this question, given that our results suggest that personalized risk assessment might motivate persons classified as higher risk for CRC to complete screening, but that it might have the unintended consequence of discouraging screening completion among persons not identified as higher risk (38). This has been suggested by a prior study of online CCRAT users showing a decrease in screening intent among those with high baseline screening intent but a subsequent CCRAT lifetime CRC risk less than 5% compared with those with risk greater than 7% (39), which matches well with our bottom and top risk tertiles, respectively. Unlike our current study, that previous study did not examine screening completion, which is the most clinically relevant near-term outcome.

By design, we did not administer the CCRAT to the control arm to avoid contamination of the intervention. Thus, we do not know the screening completion rates stratified by the CCRAT score in the control arm. It is possible that some patients in the control arm performed informal self-risk assessment on their own, based on family history for instance, and that there was also a gradient of screening completion as a function of risk in the control arm. Therefore, we cannot be certain whether personalized risk assessment truly might have encouraged those in the higher-risk tertile to be screened, while at the same time discouraging those in the other risk tiers, resulting on balance in no overall impact on the screening completion rate vs the education control. Even without an absolute improvement in screening participation overall, if risk stratification resulted in a shift in screening uptake from lower- to higher-risk persons, this might be desirable at the population level because a given set of screening resources would be applied preferentially to persons predicted to have higher CRC risk. This would be expected to yield greater impact on CRC incidence and mortality than equal apportionment of screening resources irrespective of CRC risk. However, for this improvement in long-term outcomes to be achieved, the risk stratification tool would need to be accurate enough to discriminate between persons truly at higher vs lower CRC risk (40).

Our study has limitations. First, this was a single-center study with a fairly small sample size, but one that was based on a priori sample size calculations, and with a low rate of loss to follow-up that can probably be attributed to the substantial time and effort that was invested in telephone contact during evenings, weekends, and holidays. Although the contact at 6 months may have influenced subsequent screening behavior, it is unlikely that this produced a differential effect between study arms. Second, we did not assess CRC risk in the education control arm, but this was by design. We intentionally omitted ascertainment of CRC risk and components of the CCRAT such as diet, physical activity, and aspirin use in the control arm to minimize any undue influence on risk perception and screening behavior. Third, we did not assess comprehension of absolute and relative CRC risk in the CCRAT patients. This was also done by intent to minimize the inclusion of counseling and navigation. Fourth, our study population may not be representative of all adults eligible for CRC screening, given its relatively high educational, high income levels and absence of prior screening, with most patients in the 50–64 year-old age group. We also excluded those with significant cognitive impairment, hearing impairment, and non-English speakers due to logistical considerations for this telephone call–based study. The screening rates observed in our study may seem low compared with the reported national screening rates of 56.4% in the 50–64 years age group and 71.7% in the 65–75 years age group (41), but we enrolled only patients with no previous screening. It remains speculative whether personalized risk assessment could have a different impact in populations with different socioeconomic or demographic profiles or with specific history of screening offers and deferral—variables that we did not assess. Fifth, we did not assess insurance coverage and type. Although this may have influenced the specific screening test our patients subsequently chose, our primary outcome focused on completion of any screening test regardless of type, with randomization likely balancing confounders between arms. Sixth, regarding baseline knowledge, few patients were familiar with the stool DNA test, but it is unknown whether more would have been familiar with the trademark term Cologuard or the term FIT/Stool DNA. In addition, similar proportions were familiar with FOBT and FIT, but we did not explore whether patients understood the differences between these. Finally, we did not examine recommending a particular screening program (e.g., a specific test, interval and starting and stopping age) based on risk assessment. It can be debated whether risk assessment should be applied primarily in shared decision making with the goal of achieving some form of screening participation at all or whether it should underlie a more direct approach in which a specific, personalized screening recommendation including screening modality and interval is given to a patient, including the possibility of not recommending screening below a certain level of predicted risk (42).

In conclusion, personalized CRC risk assessment had no overall impact on screening behavior compared with general education, but it may have discouraged screening in those categorized in the bottom and middle risk score tertiles compared with those categorized in the top risk score tertile. Because even patients at average or below-average CRC risk are still likely to benefit from risk-adjusted screening (43), it will need to be determined how risk stratification should be incorporated into current evidence-based approaches to increase screening participation and whether specific combinations of outreach, counseling, and messaging strategies (44–46) might mitigate any unintended consequences of risk stratification.

CONFLICTS OF INTEREST

Guarantor of article: Uri Ladabaum, MD, MS. The lead author affirms that this article is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned (and, if relevant, registered) have been explained.

Specific author contributions: T.Y.: study concept and design; acquisition of data; analysis and interpretation of data; drafting of the manuscript; critical revision of the manuscript for important intellectual content; statistical analysis; and administrative, technical, or material support. F.F.Q.: study concept and design; analysis and interpretation of data; critical revision of the manuscript for important intellectual content; and statistical analysis. V.S.: analysis and interpretation of data; critical revision of the manuscript for important intellectual content; and statistical analysis. E.A.: acquisition of data. T.S.: study concept and design; acquisition of data; and critical revision of the manuscript for important intellectual content. U.L.: study concept and design; analysis and interpretation of data; drafting of the manuscript; critical revision of the manuscript for important intellectual content; statistical analysis; administrative, technical, or material support; and study supervision.

Financial support: No grant support or financial assistance. REDCap utilization supported by UL1 RR025744 from NIH/NCRR.

Potential competing interests: U. L. is advisor (UniversalDx, Lean) and consultant (Covidien, Motus GI, Quorum, Clinical Genomics).

Writing assistance: None to report.

Clinical trials identifier: NCT03819920 in ClinicalTrials.gov.

Study Highlights

WHAT IS KNOWN

  • ✓ CRC is a leading cause of cancer.
  • ✓ CRC screening remains underutilized.

WHAT IS NEW HERE

  • ✓ CRC risk assessment did not increase CRC screening participation or intent.
  • ✓ Risk stratification might motivate persons classified as higher CRC risk to complete screening.
  • ✓ Risk stratification might unintentionally discourage screening among persons not identified as higher risk.

REFERENCES

1. Siegel RL, Miller KD, Jemal A. Cancer statistics, 2019. CA Cancer J Clin 2019;69:7–34.
2. Noone AM, Howlader N, Krapcho M, et al. SEER Cancer Statistics Review, 1975–2015. National Cancer Institute: Bethesda, 2018.
3. Edwards BK, Ward E, Kohler BA, et al. Annual report to the nation on the status of cancer, 1975-2006, featuring colorectal cancer trends and impact of interventions (risk factors, screening, and treatment) to reduce future rates. Cancer 2010;116:544–73.
4. Siegel RL, Miller KD, Fedewa SA, et al. Colorectal cancer statistics, 2017. CA Cancer J Clin 2017;67:177–93.
5. Sauer AG, Liu B, Siegel RL, et al. Comparing cancer screening estimates: Behavioral risk factor Surveillance System and National Health Interview Survey. Prev Med 2018;106:94–100.
6. Gimeno Garcia AZ. Factors influencing colorectal cancer screening participation. Gastroenterol Res Pract 2012;2012:483417.
7. Gimeno Garcia AZ, Hernandez Alvarez Buylla N, Nicolas-Perez D, et al. Public awareness of colorectal cancer screening: Knowledge, attitudes, and interventions for increasing screening uptake. ISRN Oncol 2014;2014:425787.
8. Jepson R, Clegg A, Forbes C, et al. The determinants of screening uptake and interventions for increasing uptake: A systematic review. Health Technol Assess 2000;4:i–vii, 1–133.
9. McCaffery K, Wardle J, Waller J. Knowledge, attitudes, and behavioral intentions in relation to the early detection of colorectal cancer in the United Kingdom. Prev Med 2003;36:525–35.
10. Peterson NB, Dwyer KA, Mulvaney SA, et al. The influence of health literacy on colorectal cancer screening knowledge, beliefs and behavior. J Natl Med Assoc 2007;99:1105–12.
11. Rosenstock IM. Historical origins of the health belief model. Health Educ Monogr 1974;2:328–35.
12. Bae N, Park S, Lim S. Factors associated with adherence to fecal occult blood testing for colorectal cancer screening among adults in the Republic of Korea. Eur J Oncol Nurs 2014;18:72–7.
13. Robb KA, Miles A, Wardle J. Demographic and psychosocial factors associated with perceived risk for colorectal cancer. Cancer Epidemiol Biomarkers Prev 2004;13:366–72.
14. Robb KA, Miles A, Wardle J. Perceived risk of colorectal cancer: Sources of risk judgments. Cancer Epidemiol Biomarkers Prev 2007;16:694–702.
15. Wardle J, Sutton S, Williamson S, et al. Psychosocial influences on older adults' interest in participating in bowel cancer screening. Prev Med 2000;31:323–34.
16. Hodge F, Maliski S, Itty T, et al. Colorectal cancer screening: The role of perceived susceptibility, risk and cultural illness beliefs among American Indians. J Cult Divers 2014;21:48–55.
17. Freedman AN, Slattery ML, Ballard-Barbash R, et al. Colorectal cancer risk prediction tool for white men and women without known susceptibility. J Clin Oncol 2009;27:686–93.
18. Park Y, Freedman AN, Gail MH, et al. Validation of a colorectal cancer risk prediction model among white patients age 50 years and older. J Clin Oncol 2009;27:694–8.
19. Trauth JM, Ling BS, Weissfeld JL, et al. Using the transtheoretical model to stage screening behavior for colorectal cancer. Health Educ Behav 2003;30:322–36.
20. Prochaska J, Diclemente C. Trans-theoretical therapy: Toward a more integrative model of change. Psychotherapy 1982;19:276–88.
21. Green BB, Wang CY, Anderson ML, et al. An automated intervention with stepped increases in support to increase uptake of colorectal cancer screening: A randomized trial. Ann Intern Med 2013;158:301–11.
22. Myers RE, Sifri R, Hyslop T, et al. A randomized controlled trial of the impact of targeted and tailored interventions on colorectal cancer screening. Cancer 2007;110:2083–91.
23. Menon U, Belue R, Wahab S, et al. A randomized trial comparing the effect of two phone-based interventions on colorectal cancer screening adherence. Ann Behav Med 2011;42:294–303.
24. Enard KR, Nevarez L, Hernandez M, et al. Patient navigation to increase colorectal cancer screening among Latino Medicare enrollees: A randomized controlled trial. Cancer Causes Control 2015;26:1351–9.
25. Ritvo PG, Myers RE, Paszat LF, et al. Personal navigation increases colorectal cancer screening uptake. Cancer Epidemiol Biomarkers Prev 2015;24:506–11.
26. Miller DP Jr, Denizard-Thompson N, Weaver KE, et al. Effect of a digital health intervention on receipt of colorectal cancer screening in vulnerable patients: A randomized controlled trial. Ann Intern Med 2018;168:550–7.
27. Champion VL, Christy SM, Rakowski W, et al. A randomized trial to compare a tailored web-based intervention and tailored phone counseling to usual care for increasing colorectal cancer screening. Cancer Epidemiol Biomarkers Prev 2018;27:1433–41.
28. Schroy PC III, Duhovic E, Chen CA, et al. Risk stratification and shared decision making for colorectal cancer screening: A randomized controlled trial. Med Decis Making 2016;36:526–35.
29. Inadomi JM, Issaka RB, Green BB. What multilevel interventions do we need to increase the colorectal cancer screening rate to 80%? Clin Gastroenterol Hepatol 2019. [Epub ahead of print December 27, 2019.]
30. Kim KE, Randal F, Johnson M, et al. Economic assessment of patient navigation to colonoscopy-based colorectal cancer screening in the real-world setting at the University of Chicago Medical Center. Cancer 2018;124:4137–44.
31. Rice K, Sharma K, Li C, et al. Cost-effectiveness of a patient navigation intervention to increase colonoscopy screening among low-income adults in New Hampshire. Cancer 2019;125:601–9.
32. Wolf AMD, Fontham ETH, Church TR, et al. Colorectal cancer screening for average-risk adults: 2018 guideline update from the American Cancer Society. CA Cancer J Clin 2018;68:250–81.
33. Imperiale TF, Yu M, Monahan PO, et al. Risk of advanced neoplasia using the National Cancer Institute's colorectal cancer risk assessment tool. J Natl Cancer Inst 2017;109:djw181.
34. Ladabaum U, Patel A, Mannalithara A, et al. Predicting advanced neoplasia at colonoscopy in a diverse population with the National Cancer Institute colorectal cancer risk-assessment tool. Cancer 2016;122:2663–70.
35. Ma GK, Ladabaum U. Personalizing colorectal cancer screening: A systematic review of models to predict risk of colorectal neoplasia. Clin Gastroenterol Hepatol 2014;12:1624–34.e1.
36. Weinberg DS, Myers RE, Keenan E, et al. Genetic and environmental risk assessment and colorectal cancer screening in an average-risk population: A randomized trial. Ann Intern Med 2014;161:537–45.
37. Usher-Smith JA, Walter FM, Emery JD, et al. Risk prediction models for colorectal cancer: A systematic review. Cancer Prev Res (Phila) 2016;9:13–26.
38. Lieberman D. CRC risk scores not a necessity for GI practice. AGA Perspectives 2016; March 23, 2016.
39. Han PK, Duarte CW, Daggett S, et al. Effects of personalized colorectal cancer risk information on laypersons' interest in colorectal cancer screening: The importance of individual differences. Patient Educ Couns 2015;98:1280–6.
40. Ladabaum U, Mannalithara A, Mitani A, et al. Clinical and economic impact of tailoring screening to predicted colorectal cancer risk: A decision analytic modeling study. Cancer Epidemiol Biomarkers Prev 2020;29:318–28.
41. de Moor JS, Cohen RA, Shapiro JA, et al. Colorectal cancer screening in the United States: Trends from 2008 to 2015 and variation by health insurance coverage. Prev Med 2018;112:199–206.
42. Helsingen LM, Vandvik PO, Jodal HC, et al. Colorectal cancer screening with faecal immunochemical testing, sigmoidoscopy or colonoscopy: A clinical practice guideline. BMJ 2019;367:l5515.
43. Imperiale TF. Personalized medicine: Improving on “average-risk” CRC screening. AGA Perspectives 2016.
44. Christy SM, Rawl SM. Shared decision-making about colorectal cancer screening: A conceptual framework to guide research. Patient Educ Couns 2013;91:310–7.
45. Rimer BK, Briss PA, Zeller PK, et al. Informed decision making: What is its role in cancer screening? Cancer 2004;101:1214–28.
46. Volk RJ, Linder SK, Lopez-Olivo MA, et al. Patient decision aids for colorectal cancer screening: A systematic review and meta-analysis. Am J Prev Med 2016;51:779–91.

Supplemental Digital Content

© 2020 by The American College of Gastroenterology