Lowering the colorectal cancer screening age improves predicted outcomes in a microsimulation model

Abstract

Aims

While most guidelines still recommend colorectal cancer (CRC) screening initiation at age 50 years in average-risk individuals, guideline-creating bodies are starting to lower the recommended age of initiation to 45 years to mitigate the trend of increasing CRC rates in younger populations. Using CRC-AIM, we modeled the impact of lowering the CRC screening initiation age, incorporating theoretical and reported adherence rates, for triennial multi-target stool DNA (mt-sDNA) or annual fecal immunochemical test (FIT) screening.

Methods and Materials

Screening strategies were simulated for individuals without CRC at age 40 and screened from ages 50 to 75 or 45 to 75 years. Outcomes included CRC incidence, CRC mortality, and life-years gained (LYG) per 1000 individuals screened (compared with no screening). Models used theoretically perfect (100%) and previously reported (71% mt-sDNA; 43% FIT) adherence rates.

Results

With perfect adherence, mt-sDNA and FIT resulted in 22.2 and 23.4 more predicted LYG, respectively, with screening initiation at age 45 versus 50 years; reported adherence resulted in 23.9 and 24.4 more LYG, respectively. With perfect adherence, screening initiation at age 45 versus 50 years resulted in 26.1 and 28.6 CRC cases, respectively, with mt-sDNA and 22.8 and 25.5 cases with FIT; with reported real-world adherence there were 28.5 and 31.2 cases, respectively, with mt-sDNA and 37.1 and 40.2 cases with FIT. Similar patterns were observed for CRC deaths. With screening initiation at age 45 and reported adherence, mt-sDNA averted 8.6 more CRC cases and 3.3 more deaths per 1000 individuals than FIT.

Conclusions

Estimated CRC screening outcomes improved by lowering the initiation age from 50 to 45 years. Incorporating reported adherence rates yields greater benefits from triennial mt-sDNA versus annual FIT screening.

Keywords:

• Adenoma
• adherence
• age
• colorectal cancer
• Medicare
• screening
• simulation model

Background

Screening for colorectal cancer (CRC) has been shown to reduce CRC incidence and mortality^1,2. Recommended screening options include tests that detect blood in the stool (e.g. fecal immunochemical test [FIT] and high-sensitivity guaiac-based fecal occult blood test), a test that detects DNA markers associated with CRCs and advanced precancerous lesions in cells shed from the colon or rectum into the stool (e.g. multitarget stool DNA [mt-sDNA] test), or tests that directly visualize CRCs and precancerous lesions (e.g. colonoscopy, flexible sigmoidoscopy, and computed tomography colonography)^3,4. Historically, guidelines in the United States have recommended initiation of CRC screening for average-risk individuals at age 50 years^4,5. However, the incidence of CRC over the last few decades has increased in younger patients⁶. Patients ages 45–50 years accounted for 5.1% of CRC deaths in the US between the years 2010 to 2014⁴, and it is estimated that by 2030, 11% of colon and 23% of rectal cancer cases will occur in individuals younger than 50 years⁷. Thus, recent guideline updates have re-assessed the benefits and drawbacks of lowering the CRC screening initiation age to 45 years. In 2018, the American Cancer Society updated their CRC screening guideline to include a qualified recommendation for initiation of average-risk CRC screening at age 45 years⁴, and in late 2020 the U.S. Preventive Services Task Force posted a draft update to their CRC screening recommendations including a “B” recommendation for CRC screening among average-risk adults ages 45–49 years³. In 2021, the American College of Gastroenterology released guidelines that included a strong recommendation to begin CRC screening at age 45 in average-risk individuals⁸.

There is little evidence from randomized controlled trials on the impact of CRC screening in individuals under age 50 years, although the potential benefits of a lower CRC screening age have been demonstrated in CRC modeling analyses^9–11. The previous Cancer Intervention and Surveillance Modeling Network (CISNET) CRC models used to inform CRC screening guidelines assumed perfect (theoretical) adherence with all recommended testing⁹, which is inconsistent with achievable and reported experience in clinical practice^9,12–16. Analyses conducted using the Colorectal Cancer and Adenoma Incidence and Mortality Microsimulation Model (CRC-AIM) found that, compared with assuming perfect adherence, applying imperfect (real-world) adherence rates changed the benefits and burdens of non-invasive stool-based screening strategies (e.g. mt-sDNA test and FIT) and shifted model-recommended strategies¹⁷. A Markov model that investigated the benefits of reducing CRC screening initiation to age 45 found that assuming an 80% participation rate at a screening start age of 50, as opposed to participation rates of 43–69% depending on age, averted 2.6-fold more CRC cases and 2.9-fold more deaths than reducing the screening age to 45¹¹. Thus, the importance of adherence to the benefits of CRC screening cannot be overemphasized and real-world adherence rates should be considered in CRC screening modeling analyses. Using CRC-AIM, the objective of the current study was to model the impact of lowering the CRC screening initiation age, incorporating theoretical and reported adherence rates, for mt-sDNA or FIT screening at 3-year and 1-year intervals, respectively.

Methods

CRC microsimulation models have a natural history component and screening component. The natural history component incorporates assumptions about the progression of adenomas to CRC in unscreened patients, such as adenoma growth rates, risk by age and sex, and probability of adenoma transition to preclinical CRC. The screening component incorporates assumptions about the screening process, such as the screening test used, screening frequency, test sensitivity and specificity, test complications, and adherence to the screening strategy. The details of the CRC-AIM natural history components, CRC screening components, and validation against the CISNET models have been previously published^17,18. In the current analysis, all CRC screening test performance assumptions (sensitivity, specificity, and complications) were identical to those used in the CISNET modeling analyses⁹. Adherence rates were assumed to be perfect (100%) as assumed in previous CRC models⁹, or as previously reported (71% for mt-sDNA¹⁹ vs. 43% for FIT^20,21).

Outcomes of life-years gained (LYG), incident CRC, and fatal CRC were simulated for 4 million individuals born in 1975 who were without diagnosed colon or rectal cancer at age 40 and screened between ages 50–75 or 45–75 years. Outcomes were also assessed for pre-Medicare eligible and Medicare eligible years (ages 65–75). Predicted outcomes with screening compared with no screening were calculated per 1000 individuals. Statistical differences between CRC cases and deaths for ages 45 and 50 were assessed by McNemar’s test.

Results

Lowering the screening start age from age 50 to 45 years resulted in more predicted LYG per 1000 individuals for both mt-sDNA and FIT and regardless of adherence assumptions (Figure 1). With perfect adherence, the LYG with screening initiation at age 45 versus 50 years was 319.5 and 297.3, respectively, with triennial mt-sDNA and was 339.0 and 315.6 with annual FIT; with reported real-world adherence, LYG was 307.4 and 283.5, respectively, with mt-sDNA and 269.0 and 244.6 with FIT (Table 1). Therefore, with perfect adherence rates, triennial mt-sDNA and annual FIT resulted in 22.2 and 23.4 more predicted LYG, respectively with screening initiation at age 45 versus 50 years, and with reported real-world adherence resulted in 23.9 and 24.4 more predicted LYG, respectively.

Figure 1. Life-years gained (LYG) with triennial mt-sDNA and annual FIT assuming reported real-world or theoretical perfect adherence and screening start ages of 45 or 50 years.

Table 1. Screening outcomes per 1000 individuals by adherence rate for triennial mt-sDNA and annual FIT in individuals free of diagnosed colorectal cancer at age 40 and screened between ages 50–75 years or 45–75 years.

Strategy	Stool test adherence (%)	Stool tests	COLs	Follow-up COLs	Surveillance COLs	COLs for symptoms	Total COLs	Complications	CRC cases	CRC deaths	LY with CRC	LYG	Incidence reduction (%)	Mortality reduction (%)
No screening	N/A	0	0	0	0	80	80	2	80.1	36.4	644.6	0.0	0.0	0.0
mt-sDNA 45-75, 3	100	7214	0	934	1241	16	2191	11	26.1	9.2	287.5	319.5	67.4	74.7
mt-sDNA 50-75, 3	100	6078	0	813	1121	18	1951	11	28.6	10.2	332.9	297.3	64.3	72.1
mt-sDNA 45-75, 3	71	6516	0	858	1171	18	2047	11	28.5	10.3	310.5	307.4	64.5	71.7
mt-sDNA 50-75, 3	71	5433	0	734	1059	20	1813	10	31.2	11.4	351.8	283.5	61.1	68.7
FIT 45-75, 1	100	19,167	0	937	1330	14	2280	11	22.8	7.7	262.3	339.0	71.5	78.8
FIT 50-75, 1	100	15,861	0	808	1208	15	2031	11	25.5	8.7	312.0	315.6	68.1	76.1
FIT 45-75, 1	43	9795	0	533	927	25	1484	9	37.1	13.6	395.8	269.0	53.7	62.7
FIT 50-75, 1	43	8099	0	461	824	27	1312	9	40.2	14.9	436.1	244.6	49.9	59.0

Abbreviations. COL, colonoscopy; CRC, colorectal cancer; FIT, fecal immunochemical test; LY, life-years; LYG, life-years gained; mt-sDNA, multitarget stool DNA test.

Lowering the screening start age from 50 to 45 years resulted in greater reductions in CRC incidence and mortality for both mt-sDNA and FIT, regardless of adherence assumptions (Table 1). The number of CRC cases and deaths were significantly less with a screening start age of 45 versus 50 years (p<.0001; Figure 2). With perfect adherence, screening initiation at age 45 versus 50 years resulted in 26.1 and 28.6 CRC cases per 1000 individuals, respectively, with triennial mt-sDNA and 22.8 and 25.5 cases with annual FIT; with reported real-world adherence there were 28.5 and 31.2 CRC cases, respectively, with mt-sDNA and 37.1 and 40.2 cases with FIT (Table 1). With perfect adherence, screening initiation at age 45 versus 50 years resulted in 9.2 and 10.2 CRC deaths, respectively, with mt-sDNA and 7.7 and 8.7 deaths with FIT; with reported real-world adherence, there were 10.3 and 11.4 CRC deaths, respectively, with mt-sDNA and 13.6 and 14.9 deaths with FIT (Table 1). Therefore, with perfect adherence, starting screening at age 45 averted 2.5 cases and 1.0 deaths per 1000 individuals with the mt-sDNA test and 2.7 cases and 1.0 deaths per 1000 individuals with FIT compared with starting at age 50. With reported real-world adherence, starting screening at age 45 averted 2.7 cases and 1.1 deaths per 1000 individuals with the mt-sDNA test and 3.1 cases and 1.3 deaths per 1000 individuals with FIT compared with starting at age 50. The increases in LYG and reductions in CRC cases and deaths with a screening start age of 45 versus age 50 were observed for both pre-Medicare and Medicare populations.

Figure 2. CRC cases and deaths per 1000 individuals with triennial mt-sDNA and annual FIT assuming reported real-world or theoretical perfect adherence and screening start ages of 45 or 50 years. *p<.0001 vs. age 50.

With reported real-world adherence, the LYG were greater and the total number of CRC cases and deaths were lower with triennial mt-sDNA than annual FIT regardless of screening start age (Figures 1 and 2). At a screening start age of 45 and at reported real-world adherence, triennial mt-sDNA averted 8.6 more CRC cases and 3.3 more deaths per 1000 individuals than annual FIT.

Discussion

Using the CRC-AIM microsimulation model, we found that greater benefits (increased LYG and decreased CRC incidence and mortality) were achieved from CRC screening by initiating mt-sDNA or FIT testing at age 45 versus age 50 years, using either theoretical (perfect) or previously reported real-world adherence rates. Notably, the application of reported adherence rates resulted in a shift in the benefits observed between the two stool-based screening strategies, regardless of screening initiation age. Annual FIT resulted in greater LYG and fewer CRC cases and deaths than triennial mt-sDNA when perfect adherence was applied, whereas triennial mt-sDNA resulted in a greater benefit than annual FIT when reported real-world adherence was applied.

Other CRC screening modeling analyses have also found that lowering the screening age to 45 years decreases CRC incidence and mortality compared with starting screening at age 50^9,10. However, the previous models assumed perfect adherence. The current analysis demonstrates that outcomes remain better with a screening start age of 45 than 50, even when reported real-world adherence is applied. However, particularly for FIT, changing the adherence assumptions had a greater impact on predicted outcomes than changing the screening age. These findings underscore the value of including real-world adherence rates in CRC screening modeling analyses.

The increased cost and additional burden on colonoscopy resources are a potential downside of implementing a lower screening age^11,22. In modeling guideline-concordant non-invasive tests, we provide information to inform population screening strategies that could achieve improved clinical outcomes with a lower impact on colonoscopy resources. The cost-benefit perspective also needs to be considered by payers such as Medicare. Results of the current analysis indicate that CRC screening could reduce CRC cases and death in the pre-Medicare, as well as the Medicare, population. Although costs were not estimated in the current analysis, another CRC screening model analysis found that savings in long-term CRC treatment costs compensated up to 89% of the additional costs of screening a pre-Medicare population²³. The specificity of FIT in individuals ages 45–49 years is comparable to older age groups and is at least 96%²⁴. Similarly, the specificity of mt-sDNA in patients ages 45–49 years is 95.2%²⁵. These high specificities may decrease the risk of undergoing unnecessary diagnostic procedures²⁵.

The reported adherence rates used in the analysis are from cross-sectional, first-round participation rates since real-world longitudinal adherence rates to mt-sDNA and FIT remain incompletely defined. Using adherence rates from cross-sectional data limits the analysis because it assumes that individuals have the same fixed probability to adhere to screening at every screening opportunity, which may overestimate adherence. Also, the model assumed perfect adherence to follow-up colonoscopy after a positive stool-based test, which is inconsistent with real-world practice^26,27 and would result in an overestimation of the outcome benefits. In addition, the analyses are limited in that sessile serrated polyps were not accounted for in the model. Finally, this model focuses on non-invasive stool tests, a feasible approach to lowering the age of initiation for CRC screening, and did not evaluate all screening strategies included in the guidelines.

Non-invasive stool-based screening tests may serve as effective early screening tools, particularly among younger populations. Modeling analyses demonstrated that CRC outcomes were improved by lowering the age of screening initiation from 50 to 45 years with both reported real-world and perfect adherence assumptions. Incorporating reported real-world adherence rates yields greater benefits with triennial mt-sDNA than annual FIT.

Transparency

Declaration of funding

Financial support for this study was provided by a contract with Exact Sciences Corporation. https://www.exactsciences.com/

The funding agreement ensured the authors’ independence in designing the study, interpreting the data, writing, and publishing the report.

Declaration of financial/other relationships

DAF is a consultant for Exact Sciences and Guardant Health. LS, DB and ABO are employees of Exact Sciences Corporation. LJFR is a consultant for Exact Sciences Corporation through a contractual agreement between Mayo Clinic and Exact Sciences Corporation. PJL serves as Chief Medical Officer for Screening at Exact Sciences through a contracted services agreement with Mayo Clinic. PL and the Mayo Clinic have contractual rights to receive royalties through this agreement. Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Author contributions

DAF contributed to conceptualization of the study and to the review and editing of the manuscript. LS contributed to the methodology, data curation, formal analysis, investigation, software, and visualization of the study and to the review and editing of the manuscript. LJFR contributed to conceptualization of the study and to the review and editing of the manuscript. ABO contributed to the conceptualization, methodology, supervision, and visualization of the study and to the review and editing of the manuscript. PJL contributed to the conceptualization and methodology of the study and to the review and editing of the manuscript. DB contributed to the review and editing of the manuscript.

Data availability

CRC-AIM demonstrates the approach by which existing CRC models can be reproduced from publicly available information and provides a ready opportunity for interested researchers to leverage the model for future collaborative projects or further adaptation and testing. To promote transparency and credibility of this new model, we have made available CRC-AIM’s formulas and parameters on a public repository (https://github.com/CRCAIM/CRC-AIM-Public).

Previous presentations

These data were presented at the American College of Gastroenterology Annual Meeting, Virtual Meeting, 2020.

Acknowledgements

Medical writing and editorial assistance were provided by Erin P. Scott, PhD, of Maple Health Group, LLC, funded by Exact Sciences Corporation.