Commentary: Liquid Biopsy for Average-Risk Colorectal Cancer Screening

Introduction

Despite well-known benefits and clear evidence of programmatic effectiveness in the average-risk population, colorectal cancer (CRC) screening in the US remains underutilized. Only 60–65% of age-eligible individuals are up to date with screening.¹ Emerging blood tests (liquid biopsies) detecting circulating nucleotides such as cell-free DNA and/or metabolic products associated with CRC and its precursors could help expand screening. Introduction of new non-invasive screening tests could mitigate disparities², although high-cost could widen disparities. ³

To help define the properties of a successful CRC screening LB and help guide key stakeholders across the healthcare continuum, the AGA convened a workshop comprising an international panel of CRC experts to summarize the clinical status of, and scientific findings on, emerging blood tests. The aim was to define properties of successful CRC screening blood tests and compare possible outcomes with the established strategies of annual fecal immunochemical tests (FIT), triennial multitarget stool DNA (MT-sDNA) tests, and every-10-year colonoscopies.

We employed previously validated decision models to estimate effects of a new LB on an average-risk population aged 45–75, assuming the test met minimal CMS criteria for CRC sensitivity (74%) and specificity (90%). Key endpoints were CRC incidence, CRC mortality, and quality-adjusted life years gained (QALYG). The panel also considered the possible impact of a new test on recognized health disparities in CRC.

This commentary summarizes workshop presentations and the deliberations of a multi-disciplinary panel of experts reviewing them. Our discussion focused on key LB characteristics that may produce favorable outcomes, including improved adherence rates, sensitivity for CRC and advanced adenomas (AA, i.e., adenomas ≥ 1cm, or with villous histology or high-grade dysplasia), rates of colonoscopy for abnormal test results, and costs. The resulting consensus statements are intended to be a decision-making companion to recently released CMS coverage guidelines.

Current Landscape of CRC screening

Colorectal cancer (CRC) remains the second-leading cause of cancer death in the US. Despite increasing CRC incidence in younger persons, overall US CRC incidence and mortality have declined in the past 20 years.

Despite this progress, glaring disparities remain in participation based on race. ^{4, 5 6} Beyond race, participation in CRC screening is suboptimal among individuals with low income, low educational attainment, lack of insurance, and immigrant status.^{4, 7, 8} Completion of screening in these groups is associated with lower CRC incidence and mortality inequities.⁹

Colonoscopy and fecal immunochemical test (FIT) are the two most common forms of screening in the US. ¹⁰ Colonoscopy is more invasive, inconvenient, and costly than FIT or other non-invasive options, factors that may reduce adherence. ¹¹ Colonoscopy’s success as a screening or follow-up test depends on the quality of AA detection: endoscopists with low adenoma detection rates have higher rates of post-colonoscopy CRC.¹²

In contrast to colonoscopy, successful non-invasive screening involves a two-step process because abnormal results in the initial test require a subsequent colonoscopy. Completion of timely (<6 months) colonoscopy after a positive FIT is just 50–60% in many US health settings, and many patients with positive results fail to have any subsequent colonoscopy.¹³ Individuals not receiving a colonoscopy after a positive FIT have a two-fold higher risk of death from CRC than those receiving colonoscopies.¹⁴ Depending on the setting, only 20–60% of US patients screened with FIT may undergo repeat screening after an initial normal test.¹⁵

Despite these challenges, high adherence using non-invasive screening tests have been demonstrated in organized systems.⁹ Providing universal access to such screening tests can also reduce or eliminate disparities in key outcomes such as CRC incidence and mortality.¹⁶ These experiences suggest that non-invasive test programs require ensuring 1) access and adherence for intended-use population; 2) navigation to a high-quality colonoscopy if the initial test is abnormal; and 3) appropriate interval follow-up testing if the initial test is negative.

Emerging Blood-Based Tests

Current non-invasive CRC screening primarily involves stool-based tests., but with advances in biomarker discovery a new class of blood tests is emerging.¹⁷ These tests are based on detecting cell-free DNA, circulating tumor DNA, proteomics, and immunological and glycoprotein signatures from CRC, combined with artificial intelligence and machine learning methods to construct assays for CRC and possibly precursor lesions, such as AAs.¹⁸ Currently, two LBs (from Guardant and Freenome) are at or beyond prospective population-level studies.

These emerging LBs are appealing for numerous reasons: they are minimally invasive, require no bowel preparation or stool handling, and can be coupled with other routine blood draws, allowing screenings to be completed screenings in one primary-care visit. These features hold promise to improve adherence, a key means of improving screening effectiveness. Given the robust literature on benefits of CRC screening, LBs for CRC are probably closer to clinical implementation than those for other cancers (e.g. lung, pancreas, liver, breast), including multi-cancer early detection (MCED) tests.

Careful evaluation of the potential of LBs for CRC is urgently needed. ¹⁹ CMS has determined it will cover a CRC screening LB every three years if it achieves 74% sensitivity for CRC, 90% specificity, and FDA approval.²⁰

The ECLIPSE trial (Guardant) of a new LB evaluated diagnostic performance in approximately 20,000 average-risk individuals undergoing screening in the US.²¹ Full peer-reviewed publication is anticipated; a preliminary abstract reported CRC sensitivity of 83% and specificity of 90% in individuals without advanced neoplasia or with negative colonoscopies. This test also demonstrated 13% sensitivity for AAs. The PREEMPT study (Freenome) has also completed recruitment of approximately 42,000 average-risk individuals and is conducting final analysis.²²

Modeling Results

Workshop participants considered modeling performed by four independent groups, including three constituting the Cancer Intervention and Surveillance Modeling Network (CISNET) Colorectal Cancer consortium and a fourth from Stanford University.^{23, 24} The groups aligned on core questions of interest, but each performed modeling independently. Issues addressed were: 1) expected clinical utility and economic impact of testing a screen-eligible population every three years with a LB test, with performance characteristics modeled with minimum CMS thresholds for sensitivity and specificity; 2) comparison of this LB to annual FIT, MTsDNA every three years, or colonoscopy every ten years assuming perfect participation; and 3) effect of differential participation between tests, particularly vs. FIT, and its impact on relative effectiveness and cost-effectiveness of LBs.

Each modeling group tested different assumptions for adherence, CRC sensitivity, AA sensitivity, intervals and cost. The four models yielded consistent conclusions with respect to relative effectiveness and cost-effectiveness of LBs. All showed that FIT, MTsDNA or colonoscopy had a substantially greater impact on CRC incidence, mortality, and QALYs gained than an LB meeting minimum CMS thresholds. Even at participation rates as high as 80%, LBs matching minimum thresholds performed every three years would yield more CRC cases and deaths and life-years lost to CRC than annual FIT screening at 60% participation. Improved AA sensitivity can enhance relative effectiveness of LBs over FIT. However, even with AA sensitivities up to 50%, LB every three years would not match the effectiveness of annual FIT screening, assuming an independent yield at each testing cycle. Modeling did suggest that LB meeting minimum threshold could approach FIT effectiveness (QALYG) if performed annually.

Assuming the cost is approximately $500, similar to the cost of MTsDNA (Cologuard), LB meeting minimum CMS performance thresholds was more cost-effective than no screening, at the typical $100,000 willingness-to-pay threshold. However, it was not more cost-effective than FIT or colonoscopy screening. Even if optimistic scenarios of 50% sensitivity for AAs and 80% participation with LB were realized, LBs would still need to cost considerably less to be potentially more cost-effective than FIT.

Discussion and Consensus Statements

New LB tests for CRC screening of average-risk individuals could potentially expand screening participation in screen-eligible participants to determine if outcomes would be improved. and could also replace currently recommended screening tests. The panel explored both scenarios

Expansion of screening to individuals who decline recommended screening tests.

The models suggest that blood test meeting minimal CMS criteria would result in reduced CRC incidence and mortality, increased QALYG, and be cost-effective, compared to no screening. The panel discussion, which included patient perspectives, suggested that LB offers a simple process that could encourage more people to participate in screening. Shared decision-making will be a critical step in outlining relative benefits and limitations of LBs. Patients who may have declined colonoscopy, should understand the need for a colonoscopy if findings are abnormal.

LB replacement of a recommended screening test.

The models suggest that LBs meeting minimum CMS performance thresholds performed every three years do not match the effectiveness of FIT, MT-sDNA, and colonoscopy. Although LBs would improve outcomes in currently unscreened people, substituting LB for a currently effective test would worsen patient outcomes and increase cost. The models also demonstrate the insufficiency of focusing on sensitivity to detect CRC alone, showing that effective and cost-effective tests must also detect AAs. In fact, modeling shows that increasing AA sensitivity modestly would likely provide more beneficial than improving CRC sensitivity up to 95%. The panel concluded, that based on modeling, LB should not be recommended to replace current existing tests, since it is neither as effective or cost-effective, and would worsen outcomes. (Figure 1).

Figure 1.

Approximate utilization of current CRC screening tests in the United States

mt-sDNA = multi-target stool DNA test

These are approximations post-2021 when the USPSTF screening guidelines were updated to age 45.

Impact of LB on disparities in CRC screening

As we enter this new frontier in CRC screening, efforts should be heightened to include diverse patient populations in developing and evaluating new tests. We should engage in equity-focused efforts to disseminate culturally appropriate patient and provider education in community settings, facilitate broad access to tests, and reduce test costs. ²⁵

Based on these considerations, the panel developed consensus statements to help guide industry, primary care providers, patients and other groups assessing LBs for CRC (Table 1).

Table 1.

Consensus Statements

Consensus Statements	Comments
Key outcomes of CRC screening are reduction in CRC mortality, incidence, and increased QALYs.	Screening test performance endpoints to achieve these outcomes include: • CRC sensitivity, particularly for potentially treatable stages (I-III) CRC. • AA sensitivity, given prior studies demonstrating that detection and removal AAs can reduce CRC incidence. • Specificity, defined as absence of neoplasia.
Completion of colonoscopy after abnormal non-invasive tests is essential for program effectiveness.	• All noninvasive screening programs are a two-step process, with the first step used to identify high-risk individuals, who should receive follow-up colonoscopy. • Participation costs include costs of achieving adherence and navigation to colonoscopy if a non-invasive test is abnormal.
• Any new LB meeting CMS-minimum standards for currently unscreened individuals, or those declining current recommended tests, should be more effective than no screening. • Uptake in this group would expand the current pool of screened individuals and is likely to improve CRC outcomes.	Since unscreened individuals may have declined colonoscopy screening, shared decision making for a new LB should explain that an abnormal test should be followed by colonoscopy.
• The programmatic use of a new LB replacing current effective screening should be as effective or better than current screening (FIT and colonoscopy). • A new LB test which is not as effective as current screening tests, should not be recommended to screen-eligible individuals as a replacement for current screening.	Tests not meeting these criteria may cannibalize less effective tests and harm patient outcomes.
Models suggest that LB with minimum CMS performance thresholds performed every three years is not more effective than current effective screening.	• FIT dominates LB (with minimum thresholds) with superior CRC outcomes (CRC mortality, incidence and QALYG) and lower cost • Models demonstrate that a key driver of LB effectiveness is detection of AAs.
Any new test must not worsen disparities.	LB could exacerbate existing disparities due to differential access, cost, and insufficient navigation.  Lack of trust in the healthcare system, particularly concerns about a test including genetic information, may also increase disparities.

AA=advanced adenoma, CRC=colorectal cancer, LB=liquid biopsy, QALYs=quality-adjusted life years.

Conclusions

LBs for CRC screening offer a new modality for improving overall participation in CRC screening, due to the ease of obtaining a blood test. Current limitations include detection rates for AA, and cost. Despite these limitations, an LB meeting minimal CMS criteria for sensitivity and performed every three years would likely result in better outcomes than no screening. However, because LB is predicted to be less effective and more costly than currently established screening programs, it cannot be recommended to replace established effective screening methods. (Figure 1). Potential benchmarks that industry might use to assess an effective LB going forward would be sensitivity for stage I-III CRC of >90%, with sensitivity for AA of ≥ 40–50%. Adherence to screening, follow-up of abnormal tests, and reduction of disparities, which could be promoted by effective navigation and education tools, will be important quality metrics of LB screening programs.