Limburg, Paul J.

The Centers for Medicare & Medicaid Services (CMS) recommend covering blood-based tests meeting proposed minimum performance thresholds for colorectal cancer (CRC) screening. Outcomes were compared between currently available stool-based screening tests and a hypothetical blood-based test meeting CMS minimum thresholds. Using the Colorectal Cancer and Adenoma Incidence and Mortality Microsimulation Model (CRC-AIM), outcomes were simulated for average-risk individuals screened between ages 45 and 75 years with triennial multitarget stool DNA (mt-sDNA), annual fecal immunochemical test (FIT), and annual fecal occult blood test (FOBT). Per CMS guidance, blood-based CRC screening was modeled triennially, with 74% CRC sensitivity and 90% specificity. Although not specified by CMS, adenoma sensitivity was set between 10% and 20%. Published adenoma and CRC sensitivity and specificity were used for stool-based tests. Adherence was set at (1) 100%, (2) 30%–70%, in 10% increments, and (3) real-world rates for stool-based tests (mt-sDNA = 65.6%; FIT = 42.6%; FOBT = 34.4%). Assuming perfect adherence, a blood-based test produced ≥19 lower life-years gained (LYG) than stool-based strategies. At the best-case scenario for blood-based tests (100% adherence and 20% adenoma sensitivity), mt-sDNA at real-world adherence achieved more LYG (287.2 vs. 297.1, respectively) with 14% fewer colonoscopies. At 100% blood-based test adherence and real-world mt-sDNA and FIT adherence, the blood-based test would require advanced adenoma sensitivity of 30% to reach the LYG of mt-sDNA (297.1) and ∼15% sensitivity to reach the LYG of FIT (258.9). This model suggests that blood-based tests with CMS minimally acceptable CRC sensitivity and low advanced adenoma sensitivity will frequently yield inferior outcomes to stool-based testing across a wide range of adherence assumptions.

Objectives Machine learning (ML)–based emulators improve the calibration of decision-analytical models, but their performance in complex microsimulation models is yet to be determined. Methods We demonstrated the use of an ML-based emulator with the Colorectal Cancer (CRC)-Adenoma Incidence and Mortality (CRC-AIM) model, which includes 23 unknown natural history input parameters to replicate the CRC epidemiology in the United States. We first generated 15,000 input combinations and ran the CRC-AIM model to evaluate CRC incidence, adenoma size distribution, and the percentage of small adenoma detected by colonoscopy. We then used this data set to train several ML algorithms, including deep neural network (DNN), random forest, and several gradient boosting variants (i.e., XGBoost, LightGBM, CatBoost) and compared their performance. We evaluated 10 million potential input combinations using the selected emulator and examined input combinations that best estimated observed calibration targets. Furthermore, we cross-validated outcomes generated by the CRC-AIM model with those made by CISNET models. The calibrated CRC-AIM model was externally validated using the United Kingdom Flexible Sigmoidoscopy Screening Trial (UKFSST). Results The DNN with proper preprocessing outperformed other tested ML algorithms and successfully predicted all 8 outcomes for different input combinations. It took 473 s for the trained DNN to predict outcomes for 10 million inputs, which would have required 190 CPU-years without our DNN. The overall calibration process took 104 CPU-days, which included building the data set, training, selecting, and hyperparameter tuning of the ML algorithms. While 7 input combinations had acceptable fit to the targets, a combination that best fits all outcomes was selected as the best vector. Almost all of the predictions made by the best vector laid within those from the CISNET models, demonstrating CRC-AIM’s cross-model validity. Similarly, CRC-AIM accurately predicted the hazard ratios of CRC incidence and mortality as reported by UKFSST, demonstrating its external validity. Examination of the impact of calibration targets suggested that the selection of the calibration target had a substantial impact on model outcomes in terms of life-year gains with screening. Conclusions Emulators such as a DNN that is meticulously selected and trained can substantially reduce the computational burden of calibrating complex microsimulation models. Highlights Calibrating a microsimulation model, a process to find unobservable parameters so that the model fits observed data, is computationally complex. We used a deep neural network model, a popular machine learning algorithm, to calibrate the Colorectal Cancer Adenoma Incidence and Mortality (CRC-AIM) model. We demonstrated that our approach provides an efficient and accurate method to significantly speed up calibration in microsimulation models. The calibration process successfully provided cross-model validation of CRC-AIM against 3 established CISNET models and also externally validated against a randomized controlled trial.

e22514 Background: Several colorectal cancer (CRC) screening tests, shown to reduce CRC incidence and mortality, are recommended for average-risk adults aged 45-75 years. CRC screening is underutilized in the US with access to facility-based options, such as colonoscopy, representing a barrier to screening. Stool-based tests are guideline-endorsed and offer patients a convenient and more accessible CRC screening option that can be completed at home. Provider recommendation is consistently identified as the primary factor driving patient adherence. Understanding how aspects of patient-provider communication influence CRC screening can inform interventions to improve adherence. We developed a survey, informed by behavioral theory, to identify communication factors associated with completion of mt-sDNA screening. Methods: Adults aged 45-75 years, with a US postal address, who received a valid order for mt-sDNA and whose kit was shipped between 5/21-9/21 were eligible for sampling. The survey was administered by RTI International between 3/22-6/22. Participants could complete electronic or paper surveys. We fit a multivariable logistic regression to identify patient-provider communication factors associated with mt-sDNA test completion. Results: 2,973 respondents completed the survey (response rate = 21.7%). The majority (77.3%) completed the mt-sDNA test and said they would be ‘very likely’ (76.8%) to complete Cologuard in the future. The proportion of respondents who had conversations with their healthcare provider about Cologuard, and the specific content of those conversations, was significantly different between completers and non-completers and between respondents who said they were ‘very likely’ to complete Cologuard in the future compared to respondents who were not. Results from a binary logistic regression suggested that discussions with a healthcare provider related to step-by-step instructions for completing Cologuard (OR = 1.62) and needing to have a colonoscopy follow-up (OR = 1.93), as well as feeling included ‘a great deal’ in the decision-process with respondents’ healthcare provider (compared to ‘somewhat’ or ‘not at all’) increase the odds of kit completion. Additionally, healthcare provider discussions relating to test cost (OR = 1.47) and performance (OR = 1.72), as well as feeling included ‘a great deal’ in the decision-process was positively associated with respondents indicating they were ‘very likely’ to use Cologuard in the future. Provider discussions about the cost of a follow-up colonoscopy was negatively associated with a ‘very likely’ future user (OR = 0.52). Conclusions: Healthcare providers play a key role in patient adherence to CRC screening and must be appropriately prepared and resourced to educate and to engage patients in shared decision-making about CRC screening.

10580 Background: Blood-based tests are an emerging option for non-invasive colorectal cancer (CRC) screening for which the Centers for Medicare & Medicaid Services (CMS) has proposed coverage every 3 years for average-risk individuals. Analyses with the CRC-AIM microsimulation model demonstrated that triennial screening with blood-based tests with CMS minimum performance often yields inferior outcomes to stool-based CRC screening (fecal immunochemical test [FIT] and multi-target stool DNA [mt-sDNA]). CRC-AIM was used to explore the impact of blood-based screening intervals, using reported performance characteristics from the ECLIPSE study (Guardant Health, Inc.)1, on estimated clinical outcomes vs FIT and mt-sDNA screening. Methods: Outcomes were simulated for average-risk individuals screened between ages 45-75 years. Intervals were set at 1, 2, or 3 years for blood-based, 3 years for mt-sDNA and annually for FIT screening. The blood-based test was modeled at 83% CRC sensitivity, 13% advanced adenoma sensitivity, and 90% specificity. Perfect adherence (100%) was modeled for blood-based tests and real-world adherence rates used for CRC initial screening tests (mt-sDNA=65.6%; FIT=42.6%) and follow-up colonoscopy (mt-sDNA=72.1%; FIT=46.0%). Primary outcomes were life-years gained (LYG) and colonoscopies (COL) per 1000 individuals for blood-based screening compared to mt-sDNA and FIT. As a secondary analysis, estimates were generated with perfect (100%) adherence assumed for initial screening and follow-up colonoscopy for mt-sDNA and FIT as well. Results: Blood-based testing at a 3-year interval with 100% adherence led to 11% fewer LYG than mt-sDNA at RWE adherence (240 vs 270, respectively), and 9% more COL (1402/1000 vs 1288/1000). Meanwhile, blood-based testing at a 1-year interval and perfect adherence (best case scenario) resulted in the lowest LYG per COL ratio across all scenarios (16% more LYG and 87% more COL compared to mt-sDNA at RWE adherence). Blood-based test results showed similar trends when compared to FIT. Conclusions: Outcomes for blood-based screening every 3 years based on reported test performance characteristics and unrealistic perfect adherence demonstrate fewer LYG as compared to either mt-sDNA or FIT. When compared to real-world adherence for mt-sDNA and FIT, blood-based tests with 1- and 2-year intervals and perfect adherence had higher LYG with the tradeoff of a greater number of diagnostic colonoscopies. [Table: see text]

10567 Background: Average-risk screening for colorectal cancer (CRC) is effective and recommended. The U.S. Preventive Services Task Force issued grade A and B recommendations for screening ages 50-75 and 45-49, respectively, however the recommendation for the older population (ages 76-85) is grade C (moderate certainty that the net benefit is small). With emerging blood-based CRC screening, we evaluated the benefits of screening the older population with continued multi-target stool DNA (mt-sDNA) or FIT strategies and/or with a modality shift to blood-based tests after age 75. Methods: Using the validated CRC-AIM microsimulation model, triennial mt-sDNA and annual FIT screening (ages 45-75) followed by four strategies for ages 76-85 (no screening, continued mt-sDNA/FIT, or a modality shift to blood-based screening using either CMS criteria [CMS] or recently reported blood test data from the ECLIPSE study1 [BT; Guardant Health, Inc.] were considered. For CMS criteria and BT, CRC sensitivity were set at 74% and 83%; and adenoma sensitivity was set to 20%, and 13%; and CRC specificity was set to 90% for both tests. Real-world (RWE) adherence for initial screening was set to 80% (CMS, BT), 65.6% (mt-sDNA), and 42.6% (FIT). Adherence for follow-up colonoscopy was set to 72.1% (mt-sDNA, CMS, BT) and 46.0% (FIT). Results: Screening among the population aged 45-85 resulted in 281 life years gained (LYG) with continued mt-sDNA as compared to 193 with continued FIT. Furthermore, mt-sDNA reduced CRC mortality by 70% as compared to 49% by FIT (Table). When screening modality was shifted to CMS (76-85 year olds), mt-sDNA/CMS resulted in 279 LYG as compared to 196 for FIT/CMS. CRC incidence was reduced by 59% (mt-sDNA/CMS) as compared to 38% (FIT/CMS), and mortality declined by 69% (mt-sDNA/CMS) as compared to 51% (FIT/CMS). Reductions in disease burden were lower when modality was switched to BT after age 75 due to reduced test performance. Conclusions: Extending CRC screening to the older population with continued mt-sDNA demonstrated the best health outcomes compared to other modalities. A modality shift after age 75 to blood-based testing resulted in greater LYG as compared to no screening, with performance at CMS criteria yielding better results than BT. Therefore, despite inferior performance, blood-based tests may offer a CRC screening option for older patients who are unable to complete stool testing. [Table: see text]

e18911 Background: Stool-based colorectal cancer (CRC) screening with the multi-target stool DNA (mt-sDNA) test is a guideline endorsed strategy for average-risk individuals age ≥45 years and provides a reference for value-based pricing (VBP); maximum price at which a willingness-to-pay threshold is reached. Blood-based tests (BT) are an emerging screening option. Using recent reported BT performance data from the ECLIPSE study (Guardant Health, Inc.)1, the potential VBP of BT compared with multi-target stool DNA (mt-sDNA) was analyzed. Methods: Using the validated CRC-AIM model, we simulated a cohort of 2 million average-risk individuals who received screening with BT at 1, 2, or 3-year intervals or triennial mt-sDNA between the ages of 45-75 years. For BT, CRC sensitivity, advanced adenoma sensitivity, and specificity were based on reported clinical trial data (83%, 13%, and 90%, respectively). Adherence assumptions for BT and mt-sDNA screening were modeled at perfect (100%) or real-world mt-sDNA adherence (RWE). Key outcomes were life-years gained (LYG) per 1000 individuals vs no screening, and VBP of BT vs mt-sDNA ($895 vs $508) was assessed at a willingness-to-pay (WTP) threshold of $50,000 per quality-adjusted life-year. Results: The triennial mt-sDNA was more effective and less costly vs BT at all intervals with both perfect and RWE adherence at the current price. Compared with mt-sDNA screening, LYG of screening with BT was lower (320 vs. 241-314) and the VBP should be at most $49 and $11 and no-cost at the 1-, 2-, and 3-year intervals, respectively (Table), when perfect adherence is assumed. With RWE adherence, the LYG with mt-sDNA was 306 compared to 229-284 with BT across all intervals. To be comparable to RWE triennial mt-sDNA screening, the maximum VBP for BT would need to be $13 at the 1-year interval and no cost for 2- and 3-year interval. Conclusions: Compared to annual, biannual, and triennial BT, triennial mt-sDNA was more effective at all intervals and 44% to 102% less costly. The maximum price of BT should not exceed $49 to be considered a cost-effective option compared to mt-sDNA when WTP is 50,000. [Table: see text]

6642 Background: Although follow up colonoscopy (COL) after a positive stool-based colorectal cancer (CRC) screening test is necessary to complete the CRC screening process, follow-up COL rates remain suboptimal, especially among underserved populations. As of January 2023, federal policies require commercial insurers and Medicare to eliminate out of pocket costs for follow-up COL. To estimate the implications of this policy change, we compared the clinical and fiscal outcomes for a simulated cohort receiving a follow-up COL after a non-invasive CRC test ($0 cost sharing), compared to the same cohort not completing follow-up COL (status quo). Methods: Using the validated CRC-AIM model, we simulated a cohort of 6 million average-risk individuals who received mt-sDNA (commercial patients at age 45 and Medicare patients at age 65) and further analyzed only the population with positive mt-sDNA. Life-years gained, CRC incidence, CRC-related mortality and total costs of care (inclusive of CRC screening and treatment) were calculated for a cohort that underwent follow-up COL and compared these outcomes to that same cohort that did not complete follow-up COL. The cost of follow-up COL was assumed to be $1,576 for Medicare and $3,086 for commercially insured patients ($0 out of pocket costs for all). Results: Patients undergoing follow-up COL after a positive mt-sDNA experienced more life-years (83 per 1,000 commercially insured patients and 80 per 1,000 Medicare beneficiaries) compared to those who did not undergo follow-up COL. The follow-up COL resulted in reductions in CRC incidence and CRC mortality by 14% and 16% for commercial and 19% and 23% for Medicare beneficiaries, respectively. Compared to those patients with a mt-sDNA positive test result and did not complete follow-up COL, the total cost per patient for those receiving follow-up COL was $3,339 less for commercially insured ($15,690 – no follow up vs. $12,351 – follow up) and $4,382 less for Medicare beneficiaries ($19,337 – no follow up vs. $14,955 – follow up). Total spending for patients receiving follow-up COL was cost-saving when compared to no follow-up as long as the COL costs were less than $6,269 for commercial plans and $5,753 for Medicare. Conclusions: Completing a follow-up COL after a positive mt-sDNA screening saves lives, enhances equity, and lowers total expenditures. Thus, health care stakeholders should invest in efforts to identify and reduce barriers that ultimately lead to an increase uptake of this potentially lifesaving and cost-reducing intervention.

Abstract Background and Aims The FDA-approved multitarget stool-DNA [mt-sDNA] test is a successful colorectal cancer [CRC] screening tool in average-risk individuals but is not indicated for patients with inflammatory bowel disease [IBD]. We determined the performance of the mt-sDNA assay without the haemoglobin component [mt-sDNAHgb-] in patients with IBD, while measuring sensitivity for colorectal cancer and advanced colorectal neoplasia [ACRN]. Methods This was a multi-centre, proof-of-concept investigation in persons aged 18–84 years with a diagnosis of IBD, or primary sclerosing cholangitis [PSC] with IBD. Enrolment occurred between March 2013 and May 2016. Stool was tested with the mt-sDNA molecular markers only, minus the immunochemical haemoglobin component. Results The analysis set contained 355 samples. The median age was 52 [range 39–62] years, 45.6% were female and 93% were White. Two-thirds [63%] had ulcerative colitis [UC] and 10.1% had PSC/IBD. Colonoscopy revealed cancer in 8.5% [N = 30], advanced precancerous lesions [APLs] in 9.3% [N = 33] and non-advanced precancerous lesions in 7.6% [N = 27], and three-quarters [74.7%, N = 265] had negative findings. mt-sDNAHgb- sensitivity was 73.3% for any stage cancers, and 76.2% for ACRN. Sensitivity was highest for IBD-associated high-grade dysplasia at 100% and 84.6% for IBD-associated low-grade dysplasia ≥1 cm. The test showed higher sensitivity and lower specificity in UC than in Crohn’s disease. Increasing inflammation score was associated with a significant decrease in mt-sDNAHgb- test score [ = 0.028] amongst neoplasia-negative individuals, but not in patients with ACRN. Conclusions These data highlight the potential of multitarget stool-DNA marker testing as an important addition to colorectal cancer surveillance by complementing colonoscopic evaluations in IBD patients.