Patient Preferences for a Blood-Based Colorectal Cancer Screening Test: Insights From a Conjoint Analysis Survey

Abstract

INTRODUCTION:

Over one-third of people are not up-to-date with colorectal cancer (CRC) screening, and blood-based tests offer a promising alternative to existing options. We used conjoint analysis to quantify the proportion of people who would prefer a hypothetical blood test over current methods (e.g., fecal immunochemical test, multitarget stool DNA test, colonoscopy).

METHODS:

We conducted a conjoint analysis survey in a US nationally representative sample of average risk individuals aged 40–75 years who were not up-to-date with CRC screening. We performed latent class analysis to identify groups with similar decision-making profiles and estimated the proportion who would prefer a blood test every 3 years over existing methods.

RESULTS:

Overall, 1,009 participants completed the survey. Using latent class analysis, we identified 2 distinct groups: (i) prioritized how the test is performed—39.4%, and (ii) prioritized the accuracy of detecting CRC and advanced adenomas—60.6%. Through simulations using the conjoint data, most individuals in the first group preferred a blood test every 3 years (65.1%), whereas 53.0% of the second group also favored the blood test. In additional simulations that incorporated test accuracy for CRC and advanced adenoma detection, these performance characteristics emerged as important drivers of screening preferences across the different testing options.

DISCUSSION:

Among individuals not up-to-date with CRC screening, our findings suggest that many would generally prefer a blood-based screening test over other options, but preference may depend on test accuracy. Offering a blood test option may improve CRC screening uptake, particularly among individuals who are unscreened or overdue for screening.

INTRODUCTION

While the US Preventive Services Task Force recommends that all adults at average risk of colorectal cancer (CRC) start screening at the age of 45 years (1), only 59% of people are up-to-date (2). Current methods for screening include stool, imaging, and endoscopy tests (1,3), each presenting unique barriers to completion. For example, patients frequently cite fear, aversion to the bowel preparation, and lack of time as barriers to undergoing colonoscopy (4). As for stool tests, patients report discomfort with handling stool, forgetting to mail the test to the laboratory, and concerns about test accuracy, among others, as barriers to their use (5). The emergence of blood-based tests offers a promising addition to existing screening options (6,7), as they are less invasive, require no preparation, are more accessible, and can be completed during a routine clinic visit.

While blood tests have advantages over current approaches, it is unclear whether, and to what extent, people would prefer blood tests over traditional options. To address this gap, we used conjoint analysis—a method that assesses how people make complex decisions—among a nationwide sample of 1,009 people in the United States who were not up-to-date with CRC screening to evaluate their preferences for a hypothetical blood test over existing approaches. In addition, at varying CRC and advanced adenoma (AA) accuracies, we determined the proportion of respondents who preferred to complete a blood test, fecal immunochemical test (FIT), multitarget stool DNA (mt-sDNA) test, or colonoscopy for their screening.

METHODS

Study design and participant recruitment

We performed a cross-sectional, self-administered, online survey of a nationwide sample of US adults. Our objective was to understand the trade-offs they make when choosing among currently available CRC screening modalities along with a hypothetical blood-based test. The study was approved by the Cedars-Sinai Institutional Review Board (STUDY3048).

We collaborated with a survey research firm (Cint; Stockholm, Sweden) to recruit respondents. Cint partners with research panels across the United States and drew from a sample of over 20 million panelists who agreed to receive survey invitations; see the Supplementary Methods, http://links.lww.com/CTG/B413 in our previous publication (8) for details about Cint and its participant remuneration policies.

For the overall cohort, we recruited a representative sample by sex and region based on latest US Census data. From February 16 to 28, 2024, Cint sent invitation emails to panelists, and those who clicked the survey link were informed that the study goal was “to learn how people make certain medical decisions.”

Study population

All respondents who accessed the survey were first presented with eligibility questions. We included individuals aged 40–75 years and excluded those who were up-to-date with their CRC screening; been diagnosed with colon polyps, Crohn's disease, or ulcerative colitis; or had a first-degree relative with CRC. Of note, while individuals aged 40–44 years are not yet eligible for CRC screening, they were included in the study because they may become eligible in the future.

Survey instrument

Supplementary Digital Content (see Supplementary File 1, http://links.lww.com/CTG/B413) includes the full survey instrument. For this study, we updated our existing conjoint instrument (8)—originally developed in a National Cancer Institute-funded study examining patient preferences for US Multi-Society Task Force-recommended CRC screening tests—to include a hypothetical blood-based test.

Conjoint analysis exercises to assess CRC screening test preferences

Conjoint analysis quantifies how people make trade-offs when considering competing factors. We describe conjoint analysis and development of the instrument elsewhere (8,9). In brief, this approach is based on the principle that any product or service, such as a cancer screening test, can be characterized by its attributes and is valued based on the levels of these attributes. Participants are shown a series of side-by-side profiles (e.g., different tests) and are asked to select their preferred profile based on the stated objective (Figure 1).

Figure 1.

Sample conjoint exercise where participants consider 3 hypothetical CRC screening tests and decide which one, if any, they would be most likely to do. The attribute levels for each of the 3 tests varied independently of one another and corresponded to the ranges presented in Table 1. Participants were shown a total of 10 vignettes.

Table 1 presents the attributes and levels that were tested in the survey. Of note, while the conjoint analysis survey included the range of CRC and AA sensitivities for the various test modalities reported in the literature (10–12), this information was presented to respondents as “accuracy” as it is a more commonly understood term in everyday language. Moreover, we focused on modeling sensitivity, rather than also incorporating specificity or false positive and negative rates, as these metrics are less frequently emphasized in public discussions. Including these additional test characteristics also would have significantly increased the survey length. Similarly, AAs were described to participants as “precancerous colon polyps” to enhance comprehension.

Table 1.

Attributes and levels in the conjoint analysis

CRC screening test attribute	Attribute levels
Way to look for colon cancer	Blood test
	Colonoscopy
	Colon CT scan
	Colon video capsule
	Stool testa
Accuracy of the test at finding colon cancer	75% accurate at finding colon cancer
	80% accurate at finding colon cancer
	85% accurate at finding colon cancer
	90% accurate at finding colon cancer
	95% accurate at finding colon cancer
Accuracy of the test at finding precancerous colon polyps	Not able to find precancerous colon polypsb
	35% accurate at finding precancerous colon polyps
	55% accurate at finding precancerous colon polyps
	75% accurate at finding precancerous colon polyps
	95% accurate at finding precancerous colon polyps
How often you perform the test	Repeat every 1 yr
	Repeat every 3 yr
	Repeat every 5 yr
	Repeat every 10 yr
Diet the day before the testc	Low fiber diet
	Clear liquid diet
Bowel prep before the testc	Drink 1 L of bowel prep both the night before and the morning of the test
	Drink 1.5 L of bowel prep both the night before and the morning of the test
	Drink 2 L of bowel prep both the night before and the morning of the test
Chance of a complication that requires you to go to the hospitalc	0.1% (1 in 1,000) chance of a complication
	0.3% (3 in 1,000) chance of a complication
	0.5% (5 in 1,000) chance of a complication
	0.7% (7 in 1,000) chance of a complication
	0.9% (9 in 1,000) chance of a complication

CRC, colorectal cancer; CT, computed tomography; FIT, fecal immunochemical test; mt-sDNA, multitarget stool DNA test.

^aFor stool tests, the survey did not specifically distinguish between FIT and mt-sDNA because the test completion steps for patients are similar for both.

^bThis level was presented to participants in the survey as “Not able to find precancerous colon polyps” and was mapped to a 15% accuracy rate in the simulation analyses.

^cThe choice-based conjoint used an alternative specific design and levels for this attribute were only shown for colonoscopy, colon CT scan, and colon video capsule.

The attributes and levels were inputted into a conjoint analysis program (Lighthouse Studio 9.15.4; Sawtooth Software, North Orem, UT). We used a choice-based conjoint with an alternative-specific design, and participants were shown a random set of 10 side-by-side profiles (Figure 1) drawn from 300 potential sets generated through a balanced overlap design. Before completing the conjoint exercises, respondents were provided information on the steps involved with each testing modality (e.g., blood test, colonoscopy, stool test) and their attributes (e.g., CRC accuracy, AA accuracy, test frequency). Refer to Supplementary File 1, http://links.lww.com/CTG/B413 to see how these were described to participants. To help reduce order bias, the presentation order of the test modality information was randomized. Participants were also informed that those with a positive noninvasive test (e.g., blood test, stool test) would need a follow-up colonoscopy to confirm the results.

During the conjoint exercises, participants were instructed to “choose which (screening test), if any, you would be most likely to do (for CRC screening)” and to “assume that medical insurance will cover each one and that you will not have any out-of-pocket costs.” Respondents were also informed to “please remember that all the tests are hypothetical; they may or may not resemble ones that are available in the real world.” In line with this instruction—while the CRC and AA sensitivities for each test have been established in the literature (10–12)—the conjoint exercises varied these sensitivity levels above and below their real-world values. This approach allowed us to examine how changes in test accuracy influenced participants' decision making across the different CRC screening options.

CRC screening knowledge, attitudes, and beliefs

We asked participants whether they planned to get screened for CRC. We also adapted the CRC knowledge, perceptions, and screening survey to collect information on respondents' self-perceived susceptibility to CRC and their perceived benefits of screening (13).

Demographics and comorbidities

The survey collected data on self-reported sociodemographics, health status (14), and comorbidities (15). Respondents also completed the Big Five Inventory neuroticism subscale (16).

Sample size and statistical analysis

Informed by conjoint analysis sample size precedents (17), our goal was to recruit at least 300 individuals. However, to enhance explanatory power, we aimed to recruit 1,000 individuals. After completing data collection, we used hierarchical Bayes regression to estimate individual-level importance scores and part-worth utilities for each tested attribute and level, respectively; attributes and levels with higher scores were more highly valued in the decision-making process.

We also performed a latent class analysis to identify distinct segments of respondents with similar choice patterns (18). This is a model-based statistical technique that probabilistically groups respondents into unobserved (latent) classes based on their response patterns, estimating the likelihood that each individual belongs to each class. This approach complements conjoint analysis by using individual-level part-worth utilities to classify respondents into groups that differentially prioritize certain screening test attributes.

Statistical analyses were performed using SAS version 9.4 (Cary, NC). A 2-tailed P value of < 0.05 was considered statistically significant. We performed a multivariable logistic regression to identify demographics, comorbidities, and CRC beliefs associated with latent class assignment. The results were reported as adjusted odds ratios with 95% confidence intervals.

We also conducted simulations using the individual-level part-worth utilities for test modality and frequency to estimate the proportion of individuals who would prefer each of the following tests (3): blood test every 3 years, FIT every year, mt-sDNA every 3 years, and colonoscopy every 10 years. Although the conjoint instrument also included colon CT scan and colon video capsule as screening options, our analysis focused on the most used tests in the United States

Simulations were also performed incorporating part-worth utilities for CRC and AA sensitivities (described to participants as “accuracy” since it is a more commonly used term) in addition to test modality and frequency. For the hypothetical blood test, CRC sensitivities ranging from 80% to 90%, AA sensitivities ranging from 15% to 95%, and test frequencies of every 1, 2, and 3 years were evaluated. As for the other screening options, CRC and AA sensitivities for FIT (CRC sensitivity = 80%; AA sensitivity = 25%), mt-sDNA (CRC sensitivity = 95%; AA sensitivity = 45%), and colonoscopy (CRC sensitivity = 95%; AA sensitivity = 95%) were informed by the literature (10–12).

RESULTS

Study population

Invitations were sent to 6,386 individuals, and 6,177 (96.7%) accessed the survey. We excluded those who met an ineligibility criterion (n = 3,385, 54.8%), did not finish the survey (n = 1,739, 28.2%), or had implausible responses (19) (n = 44, 0.7%). The final analytic set included 1,009 respondents, and Table 2 presents their demographics.

Table 2.

Study population characteristics

Variable	Overall cohort (N = 1,009)
Sociodemographics
Age
40–44 yr	197 (19.5%)
45–49 yr	147 (14.6%)
50–75 yr	665 (65.9%)
Gender
Male	554 (54.9%)
Female	450 (44.6%)
Nonbinary or other	4 (0.4%)
Prefer not to say	1 (0.1%)
Race/ethnicity
Non-Hispanic White only	759 (75.2%)
Non-Hispanic Black only	81 (8.0%)
Hispanic or Latino/a/x	71 (7.0%)
Non-Hispanic Asian only	56 (5.6%)
Other or multiracial	30 (3.0%)
Unknown	12 (1.2%)
Educational attainment
High school degree or less	236 (23.4%)
Some college education	254 (25.2%)
College degree	374 (37.1%)
Graduate degree	145 (14.4%)
Marital status
Married or living with a partner	499 (49.5%)
Single, widowed, divorced, or separated	510 (50.5%)
Total household income
<$50,000	480 (47.6%)
$50,000–$100,000	317 (31.4%)
≥$100,001	182 (18.0%)
Prefer not to say	30 (3.0%)
Employment status
Unemployed, on disability, on leave from work, retired, or homemaker	482 (47.8%)
Employed or student	527 (52.2%)
Has health insurance	901 (89.3%)
Has usual source of care	796 (78.9%)
US region
Northeast	191 (18.9%)
South	340 (33.7%)
Midwest	222 (22.0%)
West	256 (25.4%)
Health indicators
Self-reported health status
Excellent	80 (7.9%)
Very good	301 (29.8%)
Good	446 (44.2%)
Fair/poor	182 (18.0%)
No. of medical comorbiditiesa
0	329 (32.6%)
1	224 (22.2%)
≥2	456 (45.2%)
Neuroticism level (1–5 scale; higher = more neuroticism)	2.6 (2.0–3.3)
CRC screening history and perceptions on CRC and CRC screening
Previously been screened for CRC but not up-to-date	153 (15.2%)
Fecal immunochemical test (>1 yr)	86 (56.2%)
Colonoscopy (>10 yr)	45 (29.4%)
Multitarget stool DNA test (>3 yr)	19 (12.4%)
Colon CT scan (>5 yr)	3 (2.0%)
Plans to get screened (again) for CRC	538 (53.3%)
Self-perceived CRC susceptibility (1–5 scale; higher = more susceptible)	2.6 (1.8–3.0)
Self-perceived benefits of CRC screening (1–5 scale; higher = more beneficial)	4.0 (3.6–4.4)

Data are presented as n (% of column) or median (interquartile range).

CRC, colorectal cancer.

^aIncludes anemia or other blood disease, back pain, cancer, depression, diabetes, heart disease, high blood pressure, kidney disease, liver disease, lung disease, osteoarthritis or degenerative arthritis, rheumatoid arthritis, ulcer or stomach disease, or other medical problems.

CRC screening history, knowledge, attitudes, and beliefs

Table 2 presents data on participants' CRC screening history and perceptions on CRC screening. Most people (84.8%) had not been previously screened. Among the 15.2% of respondents who had been screened but were not up-to-date, FIT was the most performed test, followed by colonoscopy, mt-sDNA, and colon CT scan. About half of individuals stated that they plan to undergo CRC screening (again).

Conjoint analysis to assess CRC screening test preferences

The mean test importance scores among the respondents were the following: test modality—47.6% (SD, 16.8%), accuracy at detecting AA—19.3% (11.0%), accuracy at detecting CRC—14.2% (8.0%), test frequency—6.2% (3.1%), chances of a serious complication—6.2% (3.4%), bowel prep before the test—4.4% (3.0%), and diet changes before the test—2.2% (2.0%).

Latent class analysis to identify groups with similar decision-making profiles

By using latent class analysis among the 1,009 respondents, we identified 2 distinct groups of respondents with similar choice patterns. Group membership was mainly predicated on 3 test attributes: test modality, CRC accuracy, and AA accuracy. The importance scores for the remaining 4 attributes were similar between groups.

We found that 398 people (39.4%) strongly prioritized the way the test is performed in their decision making: test modality—61.6% (SD, 8.9%), AA accuracy—14.2% (6.5%), and CRC accuracy—9.3% (4.3%). Conversely, for 611 respondents (60.6%), while test modality was still an important consideration, the test's ability to accurately detect CRC and AA played a more prominent role in their selection process: test modality—38.5% (14.3%), AA accuracy—22.6% (12.0%), and CRC accuracy—17.4% (8.3%).

Table 3 presents findings from the multivariable logistic regression on latent class group assignment. We found that women, individuals aged 50–75 years, and those with health insurance had higher odds for prioritizing test modality in their decision-making process. Conversely, non-Hispanic Blacks, individuals who plan to get screened for CRC, and those with a usual source of care were less likely to prioritize test modality; in other words, they valued CRC and AA accuracy more highly when selecting a test. Similarly, people who perceived greater benefits from CRC screening and those who considered themselves at higher risk of developing CRC were also more likely to prioritize test accuracy.

Table 3.

Logistic regression analysis on prioritizing the way the CRC screening test is performed (n = 398) over the accuracy of the test at finding CRC and AAs (n = 611)^a

Variable	Strongly prioritizes way CRC screening test is done aOR (95% CI)
Age
40–44 y	Ref
45–49 y	1.19 (0.68–2.08)
50–75 y	2.56 (1.63–4.01)
Gender
Male	Ref
Female	1.57 (1.15–2.14)
Race/ethnicity
Non-Hispanic White only	Ref
Non-Hispanic Black only	0.45 (0.24–0.82)
Hispanic or Latino/a/x	0.79 (0.42–1.48)
Non-Hispanic Asian only	1.38 (0.69–2.76)
Other or multiracial	2.28 (0.93–5.57)
Unknown	2.30 (0.41–13.02)
Educational attainment
High school degree or less	Ref
Some college education	1.21 (0.80–1.83)
College degree	0.90 (0.61–1.34)
Graduate degree	1.20 (0.70–2.06)
Marital status
Married or living with a partner	Ref
Single, widowed, divorced, or separated	0.73 (0.53–1.01)
Total household income
<$50,000	Ref
$50,000–$100,000	0.81 (0.56–1.15)
≥$100,001	0.73 (0.45–1.19)
Prefer not to say	0.52 (0.22–1.27)
Employment status
Unemployed, on disability, on leave from work, retired, or homemaker	Ref
Employed or student	0.85 (0.61–1.17)
Has health insurance	1.77 (1.05–2.98)
Has usual source of care	0.62 (0.41–0.94)
Self-reported health status
Excellent	Ref
Very good	1.38 (0.74–2.57)
Good	1.13 (0.61–2.11)
Fair/Poor	1.25 (0.61–2.56)
No. of medical comorbiditiesb
0	Ref
1	1.39 (0.91–2.13)
≥2	1.29 (0.85–1.94)
Neuroticism level (1–5 scale; higher = more neuroticism)	0.97 (0.80–1.18)
US region
Northeast	Ref
South	0.89 (0.59–1.36)
Midwest	1.09 (0.69–1.71)
West	0.78 (0.49–1.22)
CRC screening history
No prior CRC screening	Ref
Previously been screened for CRC but not up-to-date	1.49 (0.98–2.27)
Plans to get screened (again) for CRC	0.45 (0.32–0.63)
Self-perceived benefits of CRC screening (1–5 scale; higher = more beneficial)	0.52 (0.40–0.66)
Self-perceived CRC susceptibility (1–5 scale; higher = more susceptible)	0.71 (0.59–0.86)

All the variables in the table were included in the multivariable logistic regression model; sex–prefer not to say, nonbinary, or other (n = 5) were not included in the regression model due to the very small sample size, which subsequently would have led to very wide CIs.

AA, advanced adenoma; aOR, adjusted odds ratio; CI, confidence interval; CRC, colorectal cancer.

^aGroups were determined by latent class analysis of the conjoint analysis data.

^bIncludes anemia or other blood disease, back pain, cancer, depression, diabetes, heart disease, high blood pressure, kidney disease, liver disease, lung disease, osteoarthritis or degenerative arthritis, rheumatoid arthritis, ulcer or stomach disease, or other medical problems.

Respondents' preferred tests based on simulations using the conjoint data

Figure 2 shows respondents' preferred test after conducting simulations using the individual-level part-worth utilities for test modality and frequency. Among participants who prioritized the test modality (n = 398) in their decision making, most preferred a blood test every 3 years, followed by mt-sDNA every 3 years, FIT every 1 year, and colonoscopy every 10 years. For the group that prioritized test accuracy (n = 611), most people also preferred the blood test, followed by colonoscopy, mt-sDNA, and FIT. Supplementary Digital Content (see Supplementary Figures 1 and 2, http://links.lww.com/CTG/B414) present these analyses stratified by sex and race/ethnicity, respectively. No differences in preferred tests were observed between men and women; however, significant differences emerged between non-Hispanic White individuals and those of other racial/ethnic backgrounds.

Figure 2.

Data from simulations using conjoint analysis data assessing participants' preferred CRC screening test using individual-level part-worth utilities for test modality and frequency. The analyses were stratified by latent class assignment. In both groups, most participants preferred to do a blood test every 3 years over traditional approaches. ***P < 0.001. AA, advanced adenoma; CRC, colorectal cancer; FIT, fecal immunochemical test.

Figure 3a (blood test CRC accuracy set to 80%) and 3b (blood test CRC accuracy at 90%) show results from the simulations determining the proportion of respondents who would prefer the hypothetical blood test at varying levels of blood test AA accuracies vs current options. When the blood test is 80% accurate at detecting CRC, it becomes the overall preferred test when its AA accuracy exceeds ∼43%. At 90% CRC accuracy, the blood test is the preferred choice when the AA accuracy exceeds ∼17%. These findings differ by latent class assignment (see Supplementary Figures 3 and 4, http://links.lww.com/CTG/B414).

Figure 3.

Data from simulations using conjoint analysis data assessing the proportion of people who would prefer a hypothetical blood test every 3 years at varying AA accuracy levels over the following existing tests: (i) FIT every year (80% CRC and 25% AA accuracies); (ii) multitarget stool DNA test every 3 years (95% CRC and 45% AA accuracies); and (iii) colonoscopy every 10 years (95% CRC and 95% AA accuracies). Panels A and B present data for when the blood test CRC accuracy is 80% and 90%. AA, advanced adenoma; CRC, colorectal cancer; FIT, fecal immunochemical test.

Supplementary Digital Content (see Supplementary Figure 5A, http://links.lww.com/CTG/B414) (blood test CRC accuracy set to 80%) and 5B (blood test CRC accuracy at 90%) illustrate the proportion of people who would opt for the hypothetical blood test over current options, across various blood test AA accuracies and frequencies. Varying test frequency for the blood test in the simulations did not significantly affect the results.

DISCUSSION

Using conjoint analysis, we assessed how people consider a blood test alongside other approaches and found that many people who are not up-to-date with CRC screening would generally prefer a blood-based test. However, preference may specifically depend on other factors such as test accuracy and frequency. Overall, these findings suggest that the availability of such tests could improve CRC screening uptake, particularly among those who are unscreened or not up-to-date.

Our study has many important findings. First, through latent class analysis, we identified 2 distinct groups of individuals with similar decision-making profiles in selecting a CRC screening test. Nearly two-fifths of participants strongly prioritize the method by which the test is conducted, whereas the remaining individuals place greater importance on the test's accuracy in detecting CRC and AA. In contrast to our previous conjoint analysis study—which only included stool, endoscopy, and imaging tests and found modality to be the most important factor across the entire sample (8)—this study's latent class analysis provides a more nuanced understanding by revealing distinct decision-making profiles.

Second, by using logistic regression, we identified key demographic factors and CRC beliefs associated with latent class assignment. For example, we observed that women are more likely to prioritize the way the test is performed in their decision making; in other conjoint analyses, the included blood-based options did not assess for differences in decision making by sex (20–23). Conversely, non-Hispanic Blacks, individuals with a usual source of care, and those who view CRC screening as beneficial are more likely to prioritize the test's CRC and AA accuracies in their decision making. Importantly, though, while we observed significant associations between latent class assignment and certain variables, the effect sizes were small-to-moderate. This suggests that relying on sociodemographics and comorbidities to predict an individual's preferred CRC screening test based is unlikely to be helpful or accurate. Instead, decision aids are needed to effectively capture patient preferences for the various CRC screening options. Our research group developed an online decision aid called Protect Your Colon (protectyourcolon.org), supported by National Cancer Institute funding, which educates users about CRC screening and also includes a Colon Protector Selector module. This module uses conjoint analysis to determine their preferred test and generates a report that users can share with their clinician. Protect Your Colon is currently undergoing pilot testing in primary care clinics (24).

Third, we conducted simulations using the conjoint data to identify the proportion of participants who would prefer a hypothetical blood test over traditional approaches. When accounting for only test modality and frequency, the blood test was the most popular choice by a significant margin across both latent classes. We also performed simulations that included individual with part-worth utilities for CRC and AA accuracies in addition to test modality and frequency. When modeling a blood test with CRC (80%) and AA (15%) accuracies consistent with emerging products (6,7), we observed that 23.7% of people would prefer to do a blood test every 3 years; most people would otherwise prefer to complete a mt-sDNA every 3 years (42.5%) or colonoscopy every 10 years (31.4%). Given that the emerging blood-based tests (6,7) may improve further—coupled with preliminary data from another assay demonstrating sensitivities of 93% for CRC and 54% for AAs (clinical validation results forthcoming) (25)—we conducted additional simulations spanning a range of CRC and AA accuracy levels to determine the thresholds at which blood tests would become the preferred option. For example, a blood test every 3 years with an 80% CRC accuracy becomes the most preferred test when its AA accuracy exceeds ∼43%. On the other hand, when the blood test CRC accuracy is raised to 90%, it needs a lower AA accuracy (17%) for it to become the most preferred option.

Fourth, even though respondents were explicitly informed that they would need to repeat screening at 1-, 3-, 5-, or 10-year intervals, test frequency was not highly valued in the decision-making process, ranking well below test modality and accuracy for detecting CRC and AAs. Even when focusing specifically on blood-based tests, varying their frequency (every 1, 2, or 3 years) did not appreciably change the proportion of respondents who would opt for the hypothetical blood test over current options. These findings suggest that some patients may be willing to undergo certain tests more frequently if doing so increases the likelihood of early cancer detection, although cost and convenience remain important considerations. Nonetheless, regardless of test modality, adherence to repeated testing at recommended intervals is essential to achieving the full preventive and early detection benefits of CRC screening. This remains an important area of research—particularly for noninvasive tests that require more frequent use—given previous evidence of suboptimal adherence over multiple testing rounds (26,27).

While our study found that many people would prefer a blood test for CRC screening over current approaches, it is important to recognize that a positive blood test requires a follow-up colonoscopy. Among patients who initially declined a colonoscopy and FIT, Liang et al conducted a randomized controlled trial comparing reoffering of colonoscopy/FIT only to additionally offering the methylated septin 9 blood test (28). While they found that offering septin 9 as a secondary option increased screening rates by 7.5%, completion of a full screening strategy (i.e., performing a colonoscopy after a positive noninvasive test) was not different between groups (28). Similarly, in a randomized controlled trial by Coronado et al comparing usual care to offer of a CRC blood test, only 50% of individuals with a positive blood test underwent a follow-up colonoscopy (29). Thus, as our study observed that many people would prefer a blood test, it is vital for healthcare systems to prepare for and ensure that patients with a positive blood test have timely access to and understand the importance of completing a confirmatory colonoscopy.

Our study is among very few that examined people's preferences for a blood-based CRC screening test (20–23). For example, Heidenreich et al performed an online discrete choice experiment among people in the United States and similarly modeled their preferences to mt-sDNA, colonoscopy, FIT, and a blood test (characteristics based on septin 9) (22). They found that most would prefer mt-sDNA every 3 years (38.8%), followed by colonoscopy every 10 years (32.5%), FIT every year (19.2%), and a blood test every year (9.4%). Although our data are not directly comparable, we found that 23.7% of people would prefer a blood test with CRC and AA sensitivities corresponding to emerging tests (6,7). We likely observed a higher proportion who would prefer a blood test as the CRC accuracy for the newly emerging blood tests (6,7) are higher than septin 9. Moreover, unlike the study by Heidenreich et al, (22) our base-case assumption was a 3-year blood test interval rather than 1 year.

There were limitations to our study. First, we conducted our study only in the United States; our findings may not be generalizable to other countries. Second, we used an online survey and our results may not generalize to elderly individuals or those who lack basic computing skills. However, in 2023, 96% and 88% of 50–64 and ≥65-year-old individuals, respectively, used the internet (30). Third, although the survey completion incentive was modest, all respondents were members of paid survey panels which may have introduced selection bias. Fourth, while our survey informed respondents that a positive noninvasive test would necessitate a confirmatory colonoscopy, it did not explicitly highlight colonoscopy's unique preventive advantage—the ability to directly remove precancerous lesions. Inclusion of this information, which is often emphasized in clinical discussions about screening options, might have influenced participants' preferences, potentially increasing selection of colonoscopy. However, our previous research found that instructing people that a positive FIT requires follow-up colonoscopy did not alter their initial choice between FIT and colonoscopy (31). Finally, our conjoint analysis included CRC and AA sensitivities, while excluding other test characteristics such as specificity and false positive and negative rates. This decision was made because these attributes are less well understood by the public, and including them would have substantially increased the survey length for participants. Further research is needed to assess the importance of these characteristics when patients choose between blood-based tests and other options. In addition, it is crucial to develop tools that clearly describe the differences among the tests—including their performance—to help clinicians and patients select the most appropriate and preferred option.

In conclusion, our findings indicate that many people who are not up-to-date with CRC screening generally prefer a blood-based test over traditional methods. We also found that the accuracy of CRC and AA detection significantly influenced decision making. Simulations revealed that nearly one-quarter of individuals favored a blood test with sensitivities comparable with emerging tests (6,7), and this preference increased as test characteristics improved. With only 59% of people aged 45–75 years up-to-date with CRC screening (2), blood tests could significantly enhance screening uptake, especially among those who decline colonoscopy or stool tests.

CONFLICTS OF INTEREST

Guarantor of the article: Christopher V. Almario, MD, MSHPM, FACG.

Specific author contributions: A.C.: study design; acquisition of data; analysis and interpretation of data; drafting of the manuscript; critical revision of the manuscript for important intellectual content. M.L.: statistical analysis; analysis and interpretation of data; critical revision of the manuscript for important intellectual content. N.M.G.: study concept and design; analysis and interpretation of data; critical revision of the manuscript for important intellectual content. L.B.: study concept and design; analysis and interpretation of data; critical revision of the manuscript for important intellectual content. B.M.R.S.: study concept and design; analysis and interpretation of data; critical revision of the manuscript for important intellectual content; technical or material support. C.V.A.: 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; technical or material support; study supervision.

Financial support: This study was supported by Freenome.

Potential competing interests: C.V.A. and B.M.R.S. have served on advisory panels with Exact Sciences. C.V.A. has served on an advisory panel with Universal DX. B.M.R.S. has served on advisory panels with Freenome and Guardant Health. C.V.A. has received grant research support to his institution from Freenome and Guardant Health. L.B. and N.M.G. were employees of Freenome during the time of this research. The remaining authors have no relevant disclosures.

IRB approval statement: The study was approved by the Cedars-Sinai Institutional Review Board (STUDY3048).

Study Highlights.

WHAT IS KNOWN

• ✓ More than one-third of US adults are not up-to-date with colorectal cancer screening.

• ✓ The emergence of blood-based tests offers a promising addition to existing screening options.

WHAT IS NEW HERE

• ✓ Among 1,009 respondents who completed the conjoint survey, 2 distinct decision-making profiles were identified—one group prioritized how the test is performed (39.4%), whereas the other prioritized test accuracy (60.6%).

• ✓ Simulations of the conjoint data showed many prefer a blood-based test, though this preference may depend on test accuracy.

• ✓ Offering a blood test option may help improve colorectal cancer screening uptake, particularly among individuals who are unscreened or overdue for screening.

Supplementary Material

ct9-17-e00945-s001.pdf ^{(3.1MB, pdf)}

ct9-17-e00945-s002.docx ^{(2.4MB, docx)}

ABBREVIATIONS:

AA

advanced adenoma

aOR

adjusted odds ratio

CI

confidence interval

CRC

colorectal cancer

CT

computed tomography

FIT

fecal immunochemical test

mt-sDNA

multitarget stool DNA