Projected Impact and Cost-Effectiveness of Novel Molecular Blood-Based or Stool-Based Screening Tests for Colorectal Cancer

Abstract

Background:

Cell-free DNA blood tests (cf-bDNA) and next-generation stool tests could change colorectal cancer (CRC) screening.

Objective:

To estimate the clinical and economic impacts of novel CRC screening tests.

Design:

Cost-effectiveness analysis using MOSAIC (Model of Screening and Surveillance for Colorectal Cancer).

Data sources:

Published data.

Target Population:

Average-risk persons.

Time Horizon:

Ages 45 to 100 years.

Perspective:

Health sector.

Intervention:

Novel versus established CRC screening tests.

Outcome Measures:

Incidence and mortality of CRC, quality-adjusted life-years (QALYs), costs.

Results of Base-Case Analysis:

For colonoscopy every 10 years, annual fecal immunochemical test (FIT), and triennial next-generation multitarget stool DNA, FIT-RNA, cf-bDNA (Guardant Shield), or cf-bDNA (Freenome), the relative rates (RRs) and 95% uncertainty intervals (UIs) versus no screening for CRC incidence were 0.21 (0.19 to 0.22), 0.29 (0.27 to 0.31), 0.33 (0.32 to 0.36), 0.32 (0.30 to 0.34), 0.58 (0.55 to 0.61) and 0.58 (0.55 to 0.60), respectively; the RRs for CRC death were 0.19 (0.17 to 0.20), 0.25 (0.23 to 0.27), 0.28 (0.27 to 0.30), 0.28 (0.26 to 0.30), 0.44 (0.42 to 0.47), and 0.46 (0.44 to 0.49), respectively. The cf-bDNA test (Shield; list price $1495) cost $89 600 ($74 800 to $102 300) per QALY gained versus no screening; alternatives were less costly and more effective.

Results of Sensitivity Analysis:

Incremental costs exceeded incremental benefits when novel test intervals were shortened to 2 or 1 years. The cf-bDNA test matched FIT’s impact on CRC mortality at 1.35 (1.30 to 1.40)-fold FIT’s uptake rate, assuming equal colonoscopy follow-up. If persons who accept colonoscopy or stool tests shifted to cf-bDNA, CRC deaths increased. This adverse effect was overcome if every 3 such substitutions were counterbalanced by cf-bDNA uptake by 2 or more persons refusing alternatives, assuming equal colonoscopy follow-up.

Limitation:

Longitudinal test-specific participation patterns are unknown.

Conclusion:

First-generation cf-bDNA tests may deliver net benefit or harm, depending on the balance between achieving screening in persons who decline alternatives versus substituting cf-bDNA for more effective alternatives.

Primary Funding Source:

The Gorrindo Family Fund.

Screening substantially decreases colorectal cancer (CRC) incidence and mortality through detection of CRC precursors and early-stage CRC (1, 2). The U.S. Preventive Services Task Force (USPSTF) provides a grade A recommendation for CRC screening for ages 50 to 75 years and a grade B recommendation for ages 45 to 49 years (3). Despite the effectiveness of CRC screening, approximately 40% of eligible persons are not up to date with CRC screening in the United States. (4). Access to screening and patient preferences, including the acceptability of available tests, such as the fecal immunochemical test (FIT), the multitarget stool DNA test (MT-sDNA), or colonoscopy, are powerful determinants of screening participation (5, 6).

Emerging blood-based and stool-based tests could alter the CRC screening landscape (7). In 2021, the Centers for Medicare & Medicaid Services (CMS) announced they would cover CRC screening every 3 years with U.S. Food and Drug Administration (FDA)-approved blood-based tests with sensitivity of 74% or more and specificity of 90% or more for CRC (8). The FDA approved a novel cell-free DNA blood test (cf-bDNA; Guardant Shield) for CRC screening on 29 July 2024 (9). This test, which has high sensitivity for CRC stages II to IV but not stage I (65%) or CRC precursors (10), became commercially available at a list price of $1495 on 1 August 2024 (11). In this changing landscape, tradeoffs emerge between tests’ abilities to detect CRC and CRC precursors, their specificity, their ease of use and uptake (12, 13), and their cost. Randomized trials comparing the rapidly expanding alternatives are unlikely to be performed.

We used our validated model MOSAIC (Model of Screening and Surveillance for Colorectal Cancer) (13) to project reductions in CRC incidence and mortality, and cost-effectiveness, with emerging blood-based and stool-based CRC screening tests versus established alternatives (14). In probabilistic analyses that incorporate uncertainty around test performance characteristics, we compared the recently approved cf-bDNA (Shield) (10), a second cf-bDNA whose preliminary registration data were recently announced (Freenome) (15), MT-sDNA (Exact Sciences Cologuard) (16), a more specific next-generation MT-sDNA (ngMT-sDNA) (17), a novel FIT-RNA (Geneoscopy ColoSense) (18), and FIT and colonoscopy, which are the most commonly used CRC screening tests (19, 20). We address in detail various dimensions of patient participation, including relative uptake, completion of colonoscopy after positive noninvasive screen, and the potential tradeoffs at the population level of capturing previously unscreened persons (an addition effect) versus shifting from current strategies to blood-based testing (a substitution effect) (21, 22).

Methods

Study Design and Perspective

We performed a probabilistic decision and cost-effectiveness analysis from a health sector perspective (23) using our validated model of CRC screening (13, 24–32), MOSAIC v2023.1 (Supplement, available at Annals.org); MOSAIC is constructed in TreeAge Pro 2023 (TreeAge Software) with outputs analyzed in Excel (Microsoft) and SAS 9.4 (SAS Institute).

All results are presented as means (95% uncertainty intervals [UIs]), reflecting 10 000 independent model iterations. For each iteration, values were drawn randomly for each model input from distributions based on published literature. For incremental comparisons (for example, additional CRCs prevented, cost per quality-adjusted life-year [QALY] gained), the incremental outcome was first calculated for each iteration, and the distribution of the incremental outcome across all 10 000 iterations is reported.

MOSAIC, Model Inputs, and Data Sources

Previous publications and the Supplement describe MOSAIC in detail (13, 24–30) (Supplement Figure 1, available at Annals.org), including validations (33) against randomized controlled trials of fecal occult blood testing (34, 35) and sigmoidoscopy (36–38), concordance with other established models (31, 39), and recalibrations to contemporary data on screening and colorectal neoplasia (13, 40–54) with validations (Supplement Tables 1 to 3 and Supplement Figures 2 to 15, available at Annals.org). In brief, the Natural History module reproduces the natural history and age-specific incidence and prevalence of nonadvanced colorectal adenomas, advanced precancerous lesions (APLs), and CRC by stage in the United States without screening (24, 26, 28). Persons transition between health states of normal; nonadvanced polyp; APL; localized, regional, or disseminated CRC; and death, in 1-year cycles. Screening and postpolypectomy surveillance are superimposed on the Natural History module from age 45 years through age 75 years for screening, and postpolypectomy surveillance through age 80 years, accounting for test performance characteristics, complications, and costs. In sensitivity analyses, we approximate the potential impact of contrasting test sensitivities for adenomas versus sessile serrated lesions (SSLs) depending on the fraction of CRCs that arise from SSLs. We account for age-specific non-CRC mortality. Persons are followed until age 100 years or death. A strategy’s impact is determined by test performance characteristics, screening interval, screening ages, cost, and participation rates. The Supplement details all model inputs (Supplement Tables 4 and 5, and Supplement Figure 16, available at Annals.org), including the derivation of costs in 2023 dollars, and all data sources.

Screening Strategies and Test Performance Characteristics

The established comparators were screening colonoscopy every 10 years, annual FIT (20 mcg Hb/g feces threshold), and MT-sDNA every 3 years, as endorsed by the USPSTF (3). The novel strategies consisted of ngMT-sDNA and FIT-RNA every 3 years (same interval as MT-sDNA), and cf-bDNA (Shield or Freenome) every 3 years, based on the CMS coverage determination. The distributions for test sensitivities for colorectal neoplastic lesions and specificity were drawn from large prospective studies (10, 15–18), including age-dependent APL sensitivity and/or test specificity for cf-bDNA (Shield) (10), ngMT-sDNA (17), and FIT-RNA (18).

Clinical and Economic Outcomes

The principal model outputs were CRC cases and deaths, QALYs, costs, and number of colonoscopies in hypothetical cohorts of 100 000 persons starting at age 45 years (23). Future QALYs and costs were discounted by 3% annually (55). The CRC stage-specific health-state utilities were applied for 5 years after diagnosis. We summarized the clinical impact of each strategy as the relative risks (RRs) for CRC cases and CRC deaths versus no screening. We calculated costs per QALYs gained between strategies and considered $100 000 to 150 000 per QALY gained as cost-effective (56, 57).

Base Case

The base case reflects idealized screening implementation, including 100% uptake, per-round adherence, and colonoscopy follow-up after an abnormal noninvasive screen. Given our focus on novel tests, we based FIT inputs on the 2024 direct comparison to ngMT-sDNA, including CRC sensitivity of 67.3% (17). We also modeled FIT based on the 2014 direct comparison to MT-sDNA, including CRC sensitivity of 73.8% (16), which is consistent with meta-analysis findings (58). Given the expectation that the more specific ngMT-sDNA will replace MT-sDNA, subsequent analyses compared cf-bDNA with ngMT-sDNA.

Sensitivity Analyses

Lower Test Costs and Shorter Intervals

We modeled cf-bDNA at a cost equal or lower versus MT-sDNAs, and we explored the novel tests at intervals of 2 and 1 years.

Multiple Dimensions of Participation

For noninvasive CRC screening to be effective, colonoscopy must follow abnormal screens. Current U.S. colonoscopy follow-up rates after abnormal stool tests are approximately 60% on average (59). We modeled noninvasive strategies with colonoscopy follow-up rates of 40% to 100%, including the possibility of worse rates after cf-bDNA versus stool testing if persons who took up cf-bDNA were particularly averse to colonoscopy.

Organized programs must decide whether to offer only a single test, considering tradeoffs between test performance characteristics and uptake (21, 22). We determined thresholds for relative participation with cf-bDNA versus FIT to match clinical outcomes with FIT.

The cf-bDNA test could be taken up by persons who consistently decline colonoscopy or stool tests (addition), or it could be substituted for these alternatives by persons open to them (substitution). In a population representative of the current state (40% unscreened; 60% screened [colonoscopy, 40%, FIT, 10%, ngMT-sDNA, 10%]), we explored cf-bDNA’s potential addition effect as a function of the colonoscopy follow-up rate. We then explored the interaction between cf-bDNA addition and substitution effects, assuming a future state with an improved colonoscopy follow-up rate of 80%.

SSLs

The MT-sDNA test has better SSL sensitivity than other noninvasive tests (10, 15–17). We modeled a scenario with approximately 20% of CRCs arising from SSLs (13) (Supplement).

Role of the Funding Source

The funding source had no role in the design, conduct, analysis, or decision to submit the manuscript for publication.

Results

Base Case

Clinical Outcomes

Without screening, 7470 (95% UI, 6606 to 8322) CRC cases and 3624 (UI, 3211 to 4030) CRC deaths occurred in a 100 000–person cohort starting at age 45 years (Table 1). Assuming 100% participation in all steps of screening, colonoscopy and FIT yielded reductions of more than 70% in CRC incidence (RR, 0.21 [UI, 0.19 to 0.22] and RR, 0.29 [UI, 0.27 to 0.31]) and reductions of 75% or more in CRC mortality (RR, 0.19 [UI, 0.17 to 0.20] and RR, 0.25 [UI, 0.23 to 0.27]), respectively, versus no screening (Table 1; Figure 1). The average number of colonoscopies per person starting at age 45 years was approximately half for FIT (2.0 [UI, 1.8 to 2.1]) versus screening colonoscopy (4.0 [UI, 3.9 to 4.1]) (Table 1).

Table 1.

Clinical and Economic Outcomes, and Cost-Effectiveness of Screening Strategies, in a Hypothetical Cohort of 100 000 Persons Starting at Age 45 Years^*

Variable	No Screen	cf-bDNA Every 3 y (Guardant Shield) (10)	cf-bDNA Every 3 y (Freenome) (15)	MT-sDNA Every 3 y (Exact Sciences Cologuard) (16)	ngMT-sDNA Every 3 y (Exact Sciences) (17)	FIT-RNA Every 3 y (Geneoscopy ColoSense) (18)	FIT Every Year (20 mcg Hb/g Feces Threshold) (17)	Colonoscopy Every 10 y
Outcome
CRC cases	7470 (6606–8322)	4365 (3817–4933)	4310 (3785–4835)	2357 (2057–2676)	2498 (2188–2815)	2406 (2104–2722)	2181 (1901–2472)	1543 (1328–1769)
Relative ratevs. no screen	1	0.58 (0.55–0.61)	0.58 (0.55–0.60)	0.32 (0.30–0.34)	0.33 (0.32–0.36)	0.32 (0.30–0.34)	0.29 (0.27–0.31)	0.21 (0.19–0.22)
CRC deaths	3624 (3211–4030)	1604 (1403–1811)	1679 (1479–1881)	970 (851–1097)	1025 (902–1151)	1006 (885–1132)	904 (792–1022)	672 (581–768)
Relative rate vs. no screen	1	0.44 (0.42–0.47)	0.46 (0.44–0.49)	0.27 (0.25–0.29)	0.28 (0.27–0.30)	0.28 (0.26–0.30)	0.25 (0.23–0.27)	0.19 (0.17–0.20)
QALYs/person	21.2864 (21.2586–21.3073)	21.3587 (21.3442–21.3693)	21.3581 (21.3441–21.3685)	21.3791 (21.3713–21.3848)	21.3784 (21.3704–21.3843)	21.3794 (21.3716–21.3851)	21.3817 (21.3749–21.3869)	21.3845 (21.3789–21.3887)
Costs/person,† $	5997 (4538–7626)	12 425 (10 396–14 550)	Cost to be determined	6566 (5447–7751)	6573 (5444–7767)†	Cost to be determined	3575 (2818–4400)	5221 (4296–6211)
Colonoscopies/person (avg)	0.2 (0.2–0.2)	1.4 (1.3–1.5)	1.4 (1.3–1.4)	1.9 (1.8–2.1)	1.6 (1.4–1.7)	1.9 (1.7–2.0)	2.0 (1.8–2.1)	4.0 (3.9–4.1)
Cost/QALY gained versus:
No screen, $		89 600 (74 800–102 300)	Cost to be determined	6300 (1100–11 800)	6500 (1300–12 000)‡	Cost to be determined	FIT dominates	Colonoscopy dominates
cf-bDNA	–	–	–	MT-sDNA dominates	ngMT-sDNA dominates	–	FIT dominates	Colonoscopy dominates
ngMT-sDNA	–	–	–	–	–	–	FIT dominates	Colonoscopy dominates
FIT, $	–	–	–	–	–	–	–	729 500 (345 200–1 550 000)

avg = average; cf-bDNA = circulating cell-free blood-based DNA test; CRC = colorectal cancer; dominates = more effective (more QALYs/person) and less costly; FIT = fecal immunochemical test; FIT-RNA = FIT/stool RNAtest; MT-sDNA = multitarget stool DNA test; ngMT-sDNA = next-generation multitarget stool DNA test; QALY = quality-adjusted life-year (discounted).

^*Results are shown as means (95% uncertainty intervals) across 10 000 model iterations.

^†One-time test costs at ages younger than 65 years/ages 65 years and older are: FIT, $18/$18; colonoscopy without polypectomy, $1415/$786; MT-sDNA and ngMT-sDNA, $681/$509; cf-bDNA (Guardant Shield), $1495/$1495.

^‡Test cost of ngMT-sDNA assumed to be the same as test cost of current MT-sDNA.

Figure 1. Clinical outcomes with emerging and established CRC screening strategies.

The estimated numbers of CRC cases and CRC deaths in a cohort of 100000 45-year-old persons are shown as means with 95% uncertainty intervals (95% UIs). The number labels above each bar represent the relative rates and 95% UIs for each screening strategy compared with no screening. cf-bDNA = circulating cell-free blood-based DNA test; CRC = colorectal cancer; FIT = fecal immunochemical test; FIT-RNA = FIT/stool RNA test; MT-sDNA = multitarget stool DNA test; ngMT-sDNA = next-generation multitarget stool DNA test.

The CRC incidence and mortality reductions were 68% and 73% with MT-sDNA (RR, 0.32 [UI, 0.30 to 0.34] and RR, 0.27 [UI, 0.25 to 0.29]) versus no screening, with similar results for ngMT-sDNA and FIT-RNA (Table 1; Figure 1). Compared with MT-sDNA, ngMT-sDNA had a slightly less profound impact on CRC cases and deaths, but the average number of colonoscopies per person was lower (1.6 [UI, 1.4 to 1.7] versus 1.9 [UI, 1.8 to 2.1]) (Table 1). These results are attributable to ngMT-sDNA’s improved specificity, but slightly lower age-specific APL sensitivities, compared with MT-sDNA.

The CRC incidence and mortality reductions were 42% and 56% with cf-bDNA (Shield) (RR, 0.58 [UI, 0.55 to 0.61] and RR, 0.44 [UI, 0.42 to 0.47]), with similar results for cf-bDNA (Freenome) (Table 1; Figure 1). Fewer colonoscopies were required with cf-bDNA than with the stool-based strategies (Table 1) due largely to cf-bDNA’s lower sensitivity for APLs, which also affected cf-bDNA’s effectiveness.

The 95% UIs around the RRs were determined primarily by the uncertainty distributions around test performance characteristic inputs (Supplement Tables 6 and 7, available at Annals.org). The improvements in CRC-related outcomes were associated with progressively greater gains in quality-adjusted life expectancy with cf-bDNA, MT-sDNA or FIT-RNA, and FIT and colonoscopy (Table 1).

The FIT with the higher CRC sensitivity of 73.8% reported in 2014 versus MT-sDNA (instead of 67.3% vs. ngMT-sDNA reported in 2024) achieved slightly improved outcomes, making it even more favorable versus alternatives (Supplement Table 8, available at Annals.org).

Cost-Effectiveness

The FIT and colonoscopy were more effective and less costly than no screening (that is, dominant) (Table 1; Figure 2). The MT-sDNA test cost $6300 (UI, $1100 to $11 800) per QALY gained versus no screening, and cf-bDNA (Shield) cost $89 600 (UI, $74 800 to $102 300) per QALY gained versus no screening at the list price of $1495 (Table 1; Figure 2). The costs for the other novel tests are not yet known.

Figure 2. Discounted mean cost per person and QALYs per person with emerging and established colorectal cancer screening strategies.

The results of probabilistic analyses with 10000 iterations are shown for each strategy. cf-bDNA (Shield) = circulating cell-free blood-based DNA test (Shield); FIT = fecal immunochemical test; MT-sDNA = multitarget stool DNA test; ngMT-sDNA=next-generation multitarget stool DNA test; QALY = quality-adjusted life-year.

In incremental comparisons between strategies, FIT and colonoscopy were more effective and less costly than cf-bDNA (Shield) and MT-sDNA (that is, FIT and colonoscopy were dominant over cf-bDNA and MT-sDNA), MT-sDNA was more effective and less costly than cf-bDNA (that is, MT-sDNA was dominant over cf-bDNA), and colonoscopy cost $729 500 (UI, $345 200 to $1 550 000) per QALY gained versus FIT (Table 1), assuming comparable participation across strategies.

Cost-Effectiveness: Impact of Test Cost

At a test cost matching that of MT-sDNA, cf-bDNA (Shield) would cost $24 900 (UI, $18 900 to $31 100) per QALY gained versus no screening, but it would still be less effective and more costly than ngMT-sDNA, FIT, or colonoscopy (that is, cf-bDNA remained dominated by the alternatives), assuming comparable participation across strategies (Figure 2; Supplement Table 9, available at Annals.org). At a test cost of $300, cf-bDNA (Shield) could be cost-saving versus no screening (cost-saving [UI, cost-saving to $2441/QALY gained]); it would still be dominated by FIT and colonoscopy, and ngMT-sDNA would cost $39 100 (UI, $23 400 to $58 400) per QALY gained versus cf-bDNA (Shield).

Shorter Screening Intervals

As the cf-bDNA (Shield) screening interval decreased from 3 to 2 to 1 years, the RRs for CRC incidence decreased from 0.58 [UI, 0.55 to 0.61], to 0.50 [UI, 0.47 to 0.53], to 0.35 [UI, 0.33 to 0.38], and the RRs for CRC mortality decreased from 0.44 [UI, 0.42 to 0.47], to 0.35 [UI, 0.33 to 0.38], to 0.26 [UI, 0.24 to 0.28], respectively, versus no screening, assuming 100% colonoscopy follow-up (Supplement Table 8). However, the incremental costs of cf-bDNA (Shield) every 2 versus 3 years, or every 1 versus 2 years, were very high (Supplement Table 8).

As the ngMT-sDNA screening interval decreased from 3 to 2 to 1 years, the incremental benefits in CRC outcomes were smaller than when shortening the interval for cf-bDNA (Shield) (Supplement Table 8)—the more sensitive a test is for CRC and APL, the less there is to gain by shortening the testing interval. The incremental costs of ngMT-sDNA every 2 versus 3 years, or every 1 versus 2 years, were very high, assuming the same test cost as the current MT-sDNA (Supplement Table 8).

Impact of Colonoscopy Follow-up Rate After Abnormal Noninvasive Screen

As the colonoscopy follow-up rate after abnormal cf-bDNA (Shield) decreased from 100% to 80%, 60%, or 40%, the effectiveness of screening eroded substantially (CRC incidence RRs of 0.58 [UI, 0.55 to 0.61], 0.65 [UI, 0.62 to 0.67], 0.72 [UI, 0.70 to 0.74], and 0.80 [UI, 0.78 to 0.82]; CRC mortality RRs of 0.44 [UI, 0.42 to 0.47], 0.52 [UI, 0.50 to 0.55], 0.62 [UI, 0.59 to 0.64], and 0.73 [UI, 0.71 to 0.74], respectively) (Supplement Table 10, available at Annals.org). Similar results were seen for ngMT-sDNA and FIT (Supplement Table 10).

Failure to follow-up with colonoscopy also eroded cost-effectiveness, as payments were made for noninvasive tests that did not lead to polypectomy or early CRC treatment (Supplement Table 10).

Thresholds for Improved Participation With cf-bDNA Versus FIT to Match Clinical Outcomes With FIT

Assuming comparable colonoscopy follow-up rates, the uptake of cf-bDNA (Shield) every 3 years would have to be 1.70-fold (UI, 1.60- to 1.82-fold) that of yearly FIT to match FIT’s impact on CRC incidence, 1.35-fold (UI, 1.30- to 1.40-fold) that of yearly FIT to match FIT’s impact on CRC mortality, and 1.32-fold (UI, 1.25- to 1.41-fold) that of yearly FIT to match FIT’s impact on quality-adjusted life expectancy (Table 2).

Table 2.

Overall All-or-None^* Participation Rates Required for Screening Every 3 Years With cf-bDNA (Guardant Shield) to Match the Clinical Outcomes of Annual FIT (17) at Varying Levels of All-or-None Participation Rates With FIT^†

Rates to Match Specific Outcomes	Overall FIT Participation Rate With Blood-Based Test (cf-bDNA; Guardant Shield) Every 3 y That Yields Equal Outcomes to Annual FIT (17) (All or None With Every Round Over Time)
	10%	20%	30%	40%	50%	60%	70%	80%
Participation rate (%) of cf-bDNA to match CRC cases prevented with FIT	17.0 (16.0–18.2)	34.1 (32.0–36.4)	51.1 (48.0–54.5)	68.2 (64.1–72.7)	85.2 (80.1–90.9)	cf-bDNA cannot match FIT	cf-bDNA cannot match FIT	cf-bDNA cannot match FIT
Participation rate (%) of cf-bDNA to match CRC deaths prevented with FIT	13.5 (13.0–14.0)	26.9 (25.9–28.1)	40.4 (38.9–42.1)	53.9 (51.9–56.2)	67.4 (64.8–70.2)	80.8 (77.8–84.2)	94.3 (90.8–98.3)	cf-bDNA cannot match FIT
Participation rate (%) of cf-bDNA to match QALYs gained versus no screening with FIT	13.2 (12.5–14.1)	26.4 (24.9–28.2)	39.6 (37.4–42.4)	52.8 (49.9–56.5)	66.0 (62.3–70.6)	79.2 (74.8–84.7)	92.3 (87.3–98.8)	cf-bDNA cannot match FIT

cf-bDNA = circulating cell-free blood-based DNA test; CRC = colorectal cancer; FIT = fecal immunochemical testing; QALY = quality-adjusted life-year (discounted).

^*For illustrative purposes, scenarios reflect perfect participation with every screening round over time in a given fraction of the population with a given test (defined as "participation rate"), and no screening at all in the remainder.

^†Results are shown as overall participation rates (percentage of population) with cf-bDNA in the population (95% uncertainty intervals).

Population-Level Impact of Adding New Screenees With Blood-Based Testingon Top of Current Screening Participation

In a cohort representative of the current population with 40% of persons unscreened and 60% screened (colonoscopy, 40%; FIT, 10%; ngMT-sDNA, 10%; follow-up colonoscopy rate 60% after abnormal screen), there were 2007 (UI, 1775 to 2237) CRC deaths per 100 000 persons. This is approximately 45% fewer CRC deaths than the 3624 (UI, 3211 to 4030) CRC deaths per 100 000 persons with no screening at all (Table 1).

Figure 3 and Supplement Table 11 (available at Annals.org) show the incremental reductions in CRC deaths per 100000 persons when cf-bDNA (Shield) was added in illustrative fractions of currently unscreened persons, across a range of colonoscopy follow-up rates after abnormal cf-bDNA (Shield). For example, with a colonoscopy follow-up rate of 60%, when an absolute 5% of the population shifted from no screening to cf-bDNA (Shield), the additional number of CRC deaths prevented was 70 (UI, 61 to 79) per 100 000 persons, and when an absolute 25% of the population shifted from no screening to cf-bDNA (Shield), the additional number of CRC deaths prevented was 348 (UI, 303 to 394) per 100 000 persons.

Figure 3. Incremental CRC deaths prevented with uptake of cf-bDNA (Shield) by persons who consistently decline screening colonoscopy or stool tests.

The incremental number of CRC deaths prevented depends on the absolute fraction of the population moving from no screening to screening with cf-bDNA (Shield), and the colonoscopy follow-up rate after an abnormal cf-bDNA (Shield) result. cf-bDNA (Shield) = circulating cell-free blood-based DNA test.

Balance Between Improved Outcomes Attributed to Capture of New Screenees Versus Worsened Outcomes Attributed to Diversion From More Effective Strategies

In a population with 40% of persons unscreened and 60% screened (colonoscopy, 40%; FIT, 10%; ngMT-sDNA, 10%), and assuming future improvement to 80% colonoscopy follow-up after abnormal screen, there were 1949 (UI, 1723 to 2172) CRC deaths per 100 000 persons. As cf-bDNA (Shield) was taken up by increasing numbers of currently unscreened persons (addition effect), without any substitution in currently screened persons, CRC deaths decreased progressively (for example, decrease of 22% [UI, 21% to 24%] if 25% of the population shifted from no screening to cf-bDNA) (Table 3). In contrast, if cf-bDNA (Shield) only substituted for current strategies, CRC deaths increased (for example, increase of 13.6% [UI, 12.6% to 14.6%] if 25% of the population shifted from current screening strategies to cf-bDNA) (Table 3).

Table 3.

Balance Between an Addition Effect (Unscreened Person Taking Up cf-bDNA) Versus a Substitution Effect (Persons Shifting From Colonoscopy or Stool Testing to cf-bDNA)^*

Absolute Fraction of the Population (Among the 60% Currently Screened) for Whom cf-bDNA Was Substituted for Colonoscopy or Fecal Tests	Absolute Fraction of the Population (Among the 40% Currently Unscreened) for Whom cf-bDNA Was Added, With Results Showing the Percentage Change in the 1949 CRC Deaths per 100 000 Persons* Occurring in the Population Before Introduction of cf-bDNA†
	0%	5%	10%	15%	20%	25%
0%	No change	−4.4 (−4.7 to −4.2)	−8.9 (−9.4 to −8.3)	−13 (−14 to −12)	−18 (−19 to −17)	−22 (−24 to −21)
5%	2.7 (2.5 to 2.9)	−1.7 (−2.1 to −1.3)	−6.2 (−6.8 to −5.5)	−11 (−12 to −9.7)	−15 (−16 to −14)	−19 (−21 to −18)
10%	5.4 (5.1 to 5.8)	1.0 (0.4 to 1.6)	−3.5 (−4.3 to −2.6)	−7.9 (−9.0 to −6.8)	−12 (−14 to −11)	−17 (−18 to −15)
15%	8.2 (7.6 to 8.8)	3.7 (3.0 to 4.5)	−0.7 (−1.7 to 0.3)	−5.2 (−6.4 to −3.9)	−9.6 (−11 to −8.1)	−14 (−16 to −12)
20%	10.9 (10.1 to 11.7)	6.4 (5.5 to 7.4)	2.0 (0.8 to 3.2)	−2.5 (−3.9 to −1.0)	−6.9 (−8.5 to −5.2)	−11 (−13 to −9.4)
25%	13.6 (12.6 to 14.6)	9.2 (8.0 to 10.3)	4.7 (3.4 to 6.1)	0.3 (−1.3 to 1.9)	−4.2 (−6.0 to −2.3)	−8.6 (−11 to −6.5)
30%	16.3 (15.1 to 17.5)	11.9 (10.6 to 13.3)	7.4 (5.9 to 9.0)	3.0 (1.2 to 4.8)	−1.5 (−3.4 to 0.6)	−5.9 (−8.1 to −3.6)

cf-bDNA = circulating cell-free blood-based DNA test; CRC = colorectal cancer.

^*This scenario reflects a current state with 40% of persons unscreened, and 60% of persons screened (40% colonoscopy, 10% fecal immunochemical test [FIT], 10% next-generation multitarget stool DNA test [ngMT-sDNA]), and an illustrative 80% colonoscopy follow-up rate after abnormal screen, in which 1949 CRC deaths per 100 000 persons occur (compared with 3624 CRC deaths per 100 000 persons with no screening at all). Results in this table reflect the impact on CRC mortality after addition and/or substitution of cf-bDNA on top of this current state.

^†Results are shown as percentage changes (95% uncertainty intervals) in the number of CRC deaths in the population. The columns show the impact of various levels of uptake of cf-bDNA in unscreened persons (addition), and the rows show the impact of various levels of substitution of cf-bDNA for colonoscopy, FIT, and ngMT-sDNA (substitution). As there are more new screenees (columns) and fewer substitutions (rows), outcomes improve.

Table 3 and Supplement Table 12 (available at Annals.org) capture the interaction between addition and substitution effects. If cf-bDNA (Shield) substituted for current screening strategies in 15% of the population—for instance—which would increase CRC deaths, then a shift from no screening to cf-bDNA (Shield) in 9% to 10% or more of the population would be needed to counterbalance the substitution effect and avert a population-level increase in CRC deaths.

Differential Detection of SSLs

When 20% of CRCs were assumed to arise from SSLs, there were negligible changes in the incremental results between strategies (Supplement Table 13, available at Annals.org).

Discussion

Our analyses can inform the implementation of the next generation of CRC screening tests in clinical practice. In probabilistic analyses that account for uncertainty in test performance characteristics, ngMT-sDNA and FIT-RNA achieved benefits approaching those of FIT. In comparison, the first-generation cf-bDNA tests were substantially less effective. These results mandate that implementation of cf-bDNA in practice be guided by consideration of relevant tradeoffs, including effectiveness versus ease of use, and cost.

At the recently announced list price of $1495, which is substantially higher than the cost of MT-sDNA or FIT and the Medicare reimbursement for colonoscopy, cf-bDNA (Shield) is not directly competitive with the alternatives. However, in persons who will absolutely not take up screening colonoscopy or stool tests, cf-bDNA (Shield) emerges as cost-effective versus no screening, assuming a high colonoscopy follow-up rate. Although shortening cf-bDNA or ngMT-sDNA testing intervals could improve outcomes substantially, this is not cost-effective at current test costs. The costs of various emerging tests remain to be established, and competition might affect test costs.

The extent to which persons who consistently decline screening colonoscopy or stool tests take up cf-bDNA versus the extent to which cf-bDNA attracts persons who would accept the more effective alternatives will determine whether cf-bDNA delivers net benefit or harm at the population level. Organized programs may face a discrete choice about which single test to offer to an entire population, considering that outcomes could be better with a less effective test at a higher participation rate than with a more effective test at a lower participation rate. In complex environments that accommodate patient and clinician choices, the balance between addition versus substitution effects for novel tests must be considered.

We explored this balance in detail. Our results lead to an informative rule of thumb: for every 3 persons substituting cf-bDNA for colonoscopy or stool tests, 2 or more persons who consistently decline colonoscopy or stool tests must take up cf-bDNA (with appropriate colonoscopy follow-up) to reduce CRC mortality from its current level. Similar rules of thumb for cf-bDNA versus specific alternatives follow from our base-case results (Supplement Table 14, available at Annals.org).

Apparently small differences in results between tests translate into large absolute numbers of CRC cases and CRC deaths at the national level. However, from an individual patient’s perspective, reducing the risk for CRC death from approximately 3% or 4% to approximately 1% or less with colonoscopy or stool tests may not seem very different from reducing the risk to approximately 1.6% with a blood test. Tradeoffs between effectiveness, ease of use, and cost may be viewed differently from a population versus individual perspective.

Participation has multiple dimensions. Our analyses highlight how the effectiveness of 2-step screening strategies is degraded substantially by failures to perform follow-up colonoscopy. A recent U.S. study of 39 health care organizations reported a disappointing colonoscopy follow-up rate of only 51.4% within 180 days after abnormal stool test (59). There are calls to elevate the colonoscopy follow-up rate to a quality metric (39, 60). If the first-generation cf-bDNA tests find their primary application in persons who will not take up screening colonoscopy or stool tests, it will be critical for these persons to understand the importance of follow-up colonoscopy (61).

Among some persons who take up screening, longitudinal participation over time is excellent, but in others it is only intermittent (62). Intermittent participation patterns can be expected to degrade the idealized benefits of screening, just as imperfect uptake and follow-up colonoscopy rates do. Participation patterns may differ between tests in ways that are not yet known. As CRC screening options proliferate, it will be critical to gather comparative data on longitudinal participation patterns, and the impact on outcomes.

We previously demonstrated the importance of a screening test’s APL sensitivity (13). The current analysis focused on published test performance characteristics of novel real-world tests, including first-generation cf-bDNA tests that have minimal sensitivity for APL beyond their false-positive rate. Cumulative cf-bDNA programmatic detection of some APLs does deliver some reduction in CRC incidence in our model, but colonoscopy and stool tests are expected to deliver substantially greater reductions in CRC incidence. Improved APL sensitivity for noninvasive tests would be highly consequential.

Limitations of our model point to fundamental open questions. We assumed independent test results from round to round, but it is not known how many lesions might be systematically missed by a given test (for example, a nonbleeding phenotype, or absence of specific biomarkers). Although the trials of cf-bDNA were not powered to detect differences in sensitivity by CRC stage, cf-bDNA sensitivity for stage I CRCs seems modest so far (57% to 65%) (10, 15).

The annual costs of initial CRC screening with colonoscopy in the United States were recently estimated at $23.7 billion (63). When reflecting on the benefits versus costs of cancer screening (64), it is important to consider a long-term horizon. Meta-analyses of flexible sigmoidoscopy trials suggest that CRC screening can reduce all-cause mortality (65, 66), and, in our model, colonoscopic CRC screening emerges as cost-saving over the long-term. However, screening colonoscopy is costly compared with FIT, assuming comparable participation. Widespread implementation of noninvasive CRC screening that is even more costly than screening colonoscopy is problematic from a public health perspective.

In conclusion, first-generation novel cf-bDNA tests have the potential to decrease meaningfully the incidence and mortality of CRC compared with no screening, but substantially less profoundly than screening colonoscopy or stool tests. Net population benefit or harm can follow incorporation of first-generation cf-bDNA CRC screening tests into practice, depending on the balance between bringing unscreened persons into screening (addition) versus shifting persons away from the more effective strategies of colonoscopy or stool testing (substitution). In clinical settings with multiple CRC screening options, shared decision making should consider all of the attributes of competing strategies, including test performance characteristics and cost, and the need for colonoscopy follow-up after an abnormal noninvasive screening test must be emphasized.