Blood-based tests used in asymptomatic people offer a seemingly simple and potentially transformative solution for cancer screening. Further, some have suggested these tests have the potential to reduce racial disparities in cancer diagnosis and outcomes by increasing access to screening and reducing variability in quality and accuracy of existing screening tests received in minoritized communities.¹ Commercial interest in such tests is increasing dramatically and in 2021, the U.S. House of Representatives introduced the Medicare Multi-Cancer Early Detection Screening Coverage Act,² which seeks to mandate Centers for Medicare & Medicaid Services (CMS) coverage for multicancer detection tests approved by the U.S. Food and Drug Administration (FDA). However, the simplicity of a single blood draw to screen for many cancers, an attractive feature to both patients and clinicians, belies the extraordinary complexity underlying these tests. Besides the obvious technical intricacies with the assay itself, which may involve next-generation nucleic acid sequencing, methylation analysis, and complex artificial intelligence (AI) algorithms to analyze numerous and varied types of analytes (e.g., cell-free DNA, proteins, metabolites), and the meandering diagnostic pathways that might follow a positive signal on these tests, the enormous policy and downstream cost implications that could occur from widespread adoption of such tests demand a robust synthesis of their clinical benefits, harms, and costs.

Clinicians, patients, and payers need robust evidence about the clinical utility of blood-based tests used in screening for multiple cancers. Unlike treatment, where all who engage have the potential to benefit from quality-of-life improvements, relief of symptoms, prevention of disease progression, or prolongation of life, the evaluation of screening interventions requires precise attention to harms. Many people typically have to be screened for a few to benefit, yet all who engage in screening have the potential to be harmed because of false-positive tests requiring expensive and/or invasive diagnostic evaluations, false-negative tests that offer unwarranted reassurance that could foster risky health behaviors or avoidance of standard of care screening, and overdiagnosis (the identification of indolent cancers that would not otherwise become symptomatic, which can lead to anxiety and unnecessary treatments and all of the side effects, costs, and harms associated with treatment). Clinicians and patients need information about the clinical net benefit of such multiple cancer screening tests (MCSTs) to have conversations about screening with these tests and payers need information about the clinical net benefit to make evidence-informed coverage policies. Should MCSTs have a clinical net benefit, secondary questions include those related to cost-effectiveness, screening intervals, diagnostic pathways after a positive test, and equitable access related to screening, followup testing, and treatment.

Cancer biomarker technology and the analytes evaluated are continually evolving. Tests undergo refinement based on advances in the analysis of tumor- and non-tumor-derived nucleic acids (cell-free or from intact cells), identification of protein biomarkers, and the use of AI in a pan-omics approach.³ Although many tests in development or currently available have evidence on analytic validity, evidence on clinical validity (i.e., accuracy outcomes such as specificity, sensitivity, positive and negative predictive value) varies, and many tests may never progress to be evaluated in rigorously designed trials to establish a clinical net benefit. Because the technology and analytes used as part of MCSTs will continue to evolve and it will not be practical to evaluate each new generation of a test in a large trial to assess clinical net benefit, an evidentiary framework for extrapolating clinical validity data for new generations of tests will be needed to regularly reevaluate the impact of these tests on clinical net benefit.

Objectives of the Review

This systematic review will assess the accuracy, effectiveness, and harms of screening for multiple cancers with blood-based biomarkers. The intended audience includes clinicians, professional organizations such as guideline developers, and payers who determine coverage related to laboratory testing.

A preliminary analytic framework, key questions (KQs), and scope (population, intervention, comparators, outcomes, timing, study design and setting; PICOTS) were posted for public comment in February 2024. In addition, between April and May 2024, we interviewed four Key Informants who represented the following perspectives: patient/caregiver advocacy, primary care clinician, Medicaid payers, and cancer research funders. Key changes from the preliminary scope include the addition of non-cell-free DNA-based tests to the scope and the addition of KQs, outcomes, and study designs related to screening test accuracy. Further, blood-based tests developed for single-cancer screening were removed from the scope. Although questions related to equitable implementation and access to MCSTs were identified as important considerations, it is critical to first establish clinical utility for these tests, and we believe questions related to implementation are premature for this topic area. This review will include 5 KQs:

Key Questions

KQ 1: What is the effectiveness of screening with blood-based multicancer screening tests (MCST) on cancer-specific mortality and all-cause mortality?
KQ 2a: What is the effectiveness of screening with MCSTs on the cumulative detection of cancer overall and by cancer type?
KQ 2b: What is the effectiveness of screening with MCSTs on the cumulative detection of late-stage cancer (i.e., stage shift) overall and by cancer type?
KQ 3: What is the accuracy of MCSTs for detection of cancer and does accuracy vary by cancer type or stage?
KQ 4: What are the harms of screening with MCSTs?
KQ 5: What are the harms of the evaluation and additional testing following a positive MCST or with surveillance following a negative evaluation after a positive MCST?

For the above KQs, the following PICOTS inclusion/exclusion criteria apply; detailed criteria organized by KQ and PICOTS are in Table 1.

• Population(s):
  • With the exception of KQ 3, individuals 18 years of age or older without active cancer or a history of cancer.
• Intervention:
  • Blood-based cancer tests designed to detect at least two different types of cancer; analytes include but are not limited to cell-free nucleic acids (DNA, RNA), proteins, small molecules, and metabolites. Tests designed for single-cancer detection or those based on tissue, urine, or other fluids are excluded.
• Comparator:
  • Varies by KQ; no screening or usual cancer screening for KQs 1, 2, and 4; no comparator required for KQ 5; adequate reference standard test for KQ 3 (see Table 1).
• Outcomes:
  • Varies by KQ; cancer-specific mortality, overall and late-stage cancer detection, psychosocial distress, overdiagnosis, radiation exposure, adverse events from invasive diagnostic procedures, and patient costs; accuracy (for KQ 3).
• Timing:
  • At least 5 years for KQ 1; no restrictions on timing for the other KQs.
• Setting:
  • Outpatient settings; countries categorized as high or very high in the 2024 United Nations Human Development Report.
• Study Designs:
  • For KQs 1, 2, 4, and 5, randomized controlled trials (RCTs), controlled trials, and non-randomized studies of interventions (with some restrictions for KQs 1 and 2); test accuracy studies (KQ 3); case series, case reports, modeling studies, and reviews are excluded.

In addition to the KQs that will be systematically reviewed, we have specified several contextual questions (CQs). The purpose of these questions is to provide end users of the review with additional information to put the findings of the KQs into a larger context.

Contextual Questions

CQ 1: What is the relationship between reduction in late-stage cancer detection and reduction in cancer-specific mortality and does this relationship vary by cancer type?
CQ 2: What is the impact on healthcare resource utilization of screening with MCSTs?
CQ 3: What is the cost-effectiveness of screening with MCSTs in U.S. settings from societal and payer perspectives? 
CQ4: What are the out-of-pocket costs incurred by individuals who are screened for cancer with MCSTs?

An evidence synthesis on this topic within the coming year may not definitively answer the KQs needed to determine the clinical net benefit because the most relevant and rigorous studies will not be completed, but it will establish a current baseline of evidence and can focus on the adaptation of existing methods for evaluating cancer screening, which were developed based on one test-one cancer approaches, to an evidentiary framework that can accommodate one test-many cancers.⁴ The analytic framework guiding this systematic review is depicted in Figure 1.

Figure 1. Analytic framework for blood-based tests for multiple cancer screening.

KQ = key question; MCST = multiple cancer screening test.

We will follow the guidance from the Agency for Healthcare Research and Quality (AHRQ) Methods Guide for Effectiveness and Comparative Effectiveness Reviews⁵ and Methods Guide for Medical Test Reviews.⁶ We will supplement this with guidance from the Cochrane Collaboration and the GRADE Working Group, particularly with respect to methods for evaluating test accuracy.^7-9 We will adhere to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guideline and the PRISMA extension for Diagnostic Test Accuracy.¹⁰

Criteria for Inclusion/Exclusion of Studies in the Review: The criteria for inclusion and exclusion of studies for this systematic review are based on the KQs. These criteria were briefly described in the previous section and are detailed more fully in Table 1. We expect most of the eligible evidence to come from peer-reviewed published literature. We will include unpublished evidence only if enough details are available to make a judgment about eligibility based on our review's eligibility criteria and if enough methods are described to assess the study's risk of bias.

Table 1. Detailed inclusion and exclusion criteria for systematic review on blood-based tests for multiple cancer screening

^aKQ 3 prediagnostic accuracy performance studies that use disease-free longitudinal followup as a reference standard should have a minimum of 1 year followup.^12 
bRefers to study registration in ClinicalTrials.gov database, or another study registry such as those included in the World Health Organization International Clinical Trials Registry Platform.
KQ = key question; MCST = multiple cancer screening test; NRSI = non-randomized study of interventions.

Literature Search Strategies to Identify Relevant Studies to Answer the Key Questions

Search Dates: 2013 to the present.

Electronic Bibliographic Databases: MEDLINE via PubMed; Cochrane Library.

Search Terms: We will use terms associated with screening, cancer, blood tests, biomarkers, accuracy outcomes, and costs/cost-effectiveness to identify relevant studies. We will use custom and validated PubMed filters to limit the searches to studies in human and adult populations, as well as to limit the search to eligible study designs (KQs only) and geographic location (CQ 2, 3, 4 only). We will divide the search into two strategies: the first search will cover KQs 1–5 and CQ 1, and the second search will cover costs/cost-effectiveness (CQs 2, 3, 4). We will limit the first search to articles published between 2013 and 2024, the period when tests for multicancer screening and detection were first developed. We will limit the second search for costs/cost-effectiveness to studies published from 2019–the present and conducted in the United States using the rationale that recent (last 5 years) healthcare utilization and cost studies from the United States will be most relevant to decision makers in the United States. The proposed search strategies for MEDLINE via PubMed are provided in Appendix B.

Gray Literature Searches: For identifying unpublished studies from the gray literature, we will search ClinicalTrials.gov, the World Health Organization International Clinical Trials Registry Platform, the Centers for Disease Control and Prevention website, and the CMS website to identify completed and ongoing studies in this topic area. We will also check the websites of test manufacturers with commercially available MCSTs to identify relevant citations.

Quality Control and Supplementary Searching: The search strategy will be peer reviewed by an information specialist from another AHRQ Evidence-based Practice Center (EPC) using the 2015 Peer Review of Electronic Search Strategy Guideline Statement.¹³ We will conduct quality checks of the search by evaluating whether known relevant citations are identified by the search and will revise the search accordingly. We will examine studies cited in recent narrative reviews, editorials, and systematic reviews to identify relevant studies that may have been missed by our search. Additionally, we will employ the Similar articles PubMed feature to identify any missed relevant articles related to key PubMed citations retrieved by the above search approaches.

Update Searches: We will update all electronic searches while the draft report is posted for public comment to capture any new publications. We will use the same search strategies with an updated date.

Supplemental Evidence and Data for Systematic Reviews (SEADS): Supplemental evidence and data will be solicited from the public on the Effective Health Care website for 4 weeks following posting of the final version of this protocol. A notice will be published on the Federal Register to widely disseminate this opportunity to submit SEADS.

Process for Selecting Studies: EndNote (version 21) will be used to manage all retrieved citations. We will use DistillerSR (version 2024.3 release) to screen titles and abstracts and full-text articles identified from our electronic database searches, from supplementary searching, or that were suggested by public comments or peer reviewers.

Two reviewers will independently review titles and abstracts or study registry entries against the criteria specified in Table 1. Titles/abstracts excluded by two reviewers will not be considered further. Title/abstracts without sufficient information to determine exclusion or inclusion will be included for full-text review. Initially, two reviewers will be required to include a citation for full-text review and conflicts within the team will be adjudicated by a third reviewer. Once the team is sufficiently calibrated as evidence by minimal conflicts with respect to include/exclude decisions, citations included by one reviewer will move forward to full-text review without a second reviewer. We will use DistillerSR's AI feature to prioritize citations for screening. Once the highest remaining priority score falls below 0.20, one reviewer will be substituted with DistillerSR's AI function that has been trained based on the abstracts that have already been included and excluded by the team. Any discrepancies between the single reviewer and DistillerSR will be adjudicated by a second human reviewer.

Two reviewers will independently review full-text articles included at title/abstract review against the criteria specified in Table 1. Full-text articles excluded by two reviewers will not be considered further. Reviewers will record one reason for exclusion for documentation in the evidence report. Full-text articles included by two reviewers will be provisionally included. In the event of inclusion/exclusion conflicts between reviewers, we will adjudicate through discussion or by a third reviewer. The principal investigator will review all included studies to ensure they meet eligibility criteria for inclusion prior to data abstraction. During full-text review, we expect to find multiple articles reporting on the same study. We will consider the study our unit of analysis and all relevant companion articles will be identified and linked to an index article for the study for all subsequent steps in the review process.

Ongoing studies that appear to be eligible based on information in the study registration but for which no peer-reviewed publications can be identified will be cataloged in the report as relevant ongoing studies. If the study completion date has passed, we will contact the authors to inquire about approximate dates of publication.

Literature or data identified during the updated search or that is suggested by SEADS submissions, external peer reviewers, or public comment will be assessed using the same methods as the initially identified records. New eligible studies identified from the update search will be included in the final report.

Data Abstraction and Data Management

We will abstract data from included studies using structured forms that we will design in DistillerSR. We will abstract data from companion articles onto the DistillerSR record of the index article. One reviewer will abstract the data into the forms and a second reviewer will check the data for accuracy. We will contact study authors if key data are missing or unclear in the article. Data abstracted will be compiled into detailed evidence tables for the evidence report and will be used for our synthesis of findings.

One data abstraction form will be tailored to abstract data relevant to KQ 1, 2, 4, and 5 studies. A second data abstraction form will be tailored to abstract data relevant to KQ 3 (test accuracy). The structured forms will include data elements relating to study identifiers and sponsors, study population, screening intervention including details related to the test technology, analytes used, and an indication of whether AI is involved in generating the test result; comparator intervention or test; and reported outcomes. Templates for the data elements we plan to abstract are in Appendix C. For trials and non-randomized studies of interventions (NRSIs), we will abstract outcomes for all reported time points for each study. However, our strength of evidence assessments may be limited to a time point that is common across studies.

Assessment of Methodological Risk of Bias of Individual Studies

We will use design-specific instruments for evaluating the risk of bias (ROB) of included studies. The RoB 2 instrument¹⁴ will be used to evaluate RCTs, ROBINS-I¹⁵ will be used to evaluate NRSIs, and QUADAS-2¹⁶ will be used to evaluate test accuracy studies. For test accuracy studies, we will tailor the QUADAS-2 tool for both diagnostic performance and prediagnostic performance study designs.¹² We will report domain-specific ROB ratings and an overall ROB rating for each included study. In some cases, we may report an outcome-specific ROB rating within a study if we assess the outcome as having a different ROB rating than the overall study. Studies will not be excluded from the report based on ROB ratings. Two reviewers will independently assess the ROB for each study; disagreements on domain-level ratings will be adjudicated through discussion or a third reviewer.

Data Synthesis

We will summarize data narratively and in tabular formats organized by KQ and then by test analyte (e.g., cell-free DNA, protein biomarkers). If we have at least two similar studies with minimal clinical and methodologic heterogeneity that report the same outcome at reasonably similar time points, we will assess the feasibility of conducting quantitative synthesis.¹⁷ Quantitative synthesis will be conducted with Stata (version 17.0). A priori subgroups of interest for KQs 1, 2, 4, and 5 include (1) residence in a medically underserved community including rural areas, (2) public, private, or no health insurance coverage, and (3) those defined by race or ethnicity. These subgroups are of interest because of differential access to followup diagnostic evaluation or treatment experienced by such persons, particularly in the United States. Other subgroups of interest include those defined by age (to examine whether the benefits and harms of MCSTs vary across the adult lifespan) and sex (to examine whether the benefits and harms of MCSTs vary in relationship to sex-specific cancers that may be detected). Finally, we will stratify findings for populations defined by average cancer risk versus higher cancer risk as defined by study authors. Populations defined as higher risk may include those with a strong family history of cancer, those with genetic mutations that predispose them to cancer, those with a history of childhood cancer, those with comorbidities associated with a higher cancer risk, and those with a history of precancerous lesions.

For trials and NRSIs (KQs 1, 2, 4, 5), we will synthesize dichotomous outcomes (e.g., mortality, cancer cumulative detection) using relative (e.g., relative risk, hazard ratios, odds ratios) and absolute (e.g., absolute risk difference) effect measures. If quantitative synthesis is not possible, we will report the range of observed relative and absolute effects across studies. If quantitative synthesis is possible, we will report pooled relative effects and transform pooled relative effects into absolute effects (e.g., number of deaths averted per 1,000 people screened) using established methods.¹⁸ Wherever possible, we will report cumulative cancer detection and mortality outcomes overall and by specific cancer type. For continuous outcomes, we will synthesize outcomes as mean differences or standardized mean differences. If outcomes are quantitatively synthesized, we will conduct a sensitivity analysis excluding studies we rate as high ROB to determine impact on pooled estimates. For outcomes that have studies reporting zero or rare events (<1% incidence), we will use appropriate methods for quantitative synthesis of rare events such as Mantel-Haenszel fixed-effect models or Peto odds ratio. We will assess statistical heterogeneity for any quantitative synthesis with the I² statistic and will use Cochrane methods guidance to interpret this statistic.^{18, 19} If at least 10 studies are available for quantitative synthesis, we will consider assessing for reporting bias using Egger’s test.²⁰

For test accuracy studies (KQ 3), we will synthesize data on accuracy for each cancer type (e.g., breast, lung, pancreas) and stage. Further, we will stratify findings based on study design (diagnostic performance vs. prediagnostic performance). For quantitative syntheses, we will conduct sensitivity analyses excluding studies evaluated as high ROB to assess impact on pooled estimates. Because diagnostic performance studies use a case-control study design, any quantitative synthesis of sensitivity and false-negative outcomes will be conducted separately from synthesis of specificity and false-positive outcomes because these outcomes are not correlated as they are derived from different source populations.

Grading the Strength of Evidence (SOE) for Major Comparisons and Outcomes: We will use AHRQ Effective Health Care Program Guidance^{5, 6, 21} to assess the SOE, supplemented by selected guidance from the GRADE working group.²² We will grade outcomes from RCT bodies of evidence separately from outcomes from NRSI bodies of evidence. Based on input from our review’s Key Informants and Technical Expert Panel, we will grade SOE for the outcomes detailed in Table 2 as they are critical or important to decision making around the use of MCSTs.

Table 2. Outcomes for strength of evidence assessment

KQ = key question.

We consider cancer-specific mortality outcomes as the most critical for decision making. This outcome is critical to understanding which cancers may have a mortality benefit from MCSTs and may offer information related to the impact of these tests on cancers with standards of care screening tests. We consider all-cause mortality an important outcome. All-cause mortality is less biased than cancer-specific mortality because it is not subject to misclassification bias based on how deaths are attributed. Further, it is easier to interpret than numerous different cancer mortality outcomes that may not be consistent with each other with respect to magnitude or direction of effect. However, we do not consider all-cause mortality as critical because the proportion of cancer deaths among all deaths is already low and reductions in cancer-specific mortality from screening are typically small resulting in minimal impact on all-cause mortality (estimated at 1% to 3%). Although this impact is minimal, across a population it may still have public health importance.^{23, 24}

Cancer-specific mortality from randomized trials that minimize lead- and length-time bias is the most rigorous endpoint to assess benefit, but such trials require exceptionally large sample sizes and long durations. Late-stage detection 3 to 4 years after randomization is the primary endpoint of the NHS-Galleri Trial.²⁵ Late-stage detection is often touted as a proxy for cancer-specific mortality measured at later followup time points. This might be an acceptable proxy for ovarian cancer, where a 20% reduction in late-stage disease translates to a 13% reduction in mortality, but reduction in late-stage disease does not result in a similar magnitude of reduction in mortality for some cancers.²⁶ A recent metanalysis of 41 trials found that the detection of late-stage cancer was correlated to cancer-specific mortality for some cancers, but not for others.²⁷ The reasons why late-stage detection may not be a suitable proxy for cancer-specific mortality for some cancers is not entirely clear.²⁸ Whether cases shifted from late to early detection through screening have similar prognosis as cases detected early without screening is unknown; evidence from some cancers suggest these are biologically different tumors.²⁶ Earlier diagnosis of tumors with certain molecular signatures may have worse prognosis than site- and stage-matched tumors without such molecular signals. Thus, detecting such tumors early may not alter the clinical trajectory or reduce mortality. Such tumors might require far more aggressive therapy or new therapies for there to be a benefit from detection. In summary, extrapolation from late-stage detection to cancer-specific mortality may not be valid for some cancers.^{27, 28}

Other outcomes that we have identified for SOE assessment within KQs 1, 2, 4, and 5 are important to decision making as they will allow an evaluation of the clinical net benefit of screening with MCSTs. We consider the accuracy outcomes associated with KQ 3 as less important to decision making as they do not directly inform an assessment of clinical net benefit. However, they may be useful for describing the landscape of current tests and determining which tests may be most promising for studying in future screening trials.

If no studies are identified that report an outcome selected for SOE grading, we will grade the outcome as insufficient because of no evidence available. Otherwise, we will assign an overall SOE grade for each comparison and outcome as high, moderate, low, or insufficient SOE. Suggested definitions of these grades are listed in Table 3.21 Further, for KQs 1, 2, 4, and 5 we will include the direction of effect (i.e., decreased, increased, no effect) as part of the SOE assessment.

Table 3. Strength of evidence grades and definitions from AHRQ EPC Program.²¹

AHRQ = Agency for Healthcare Research and Quality; EPC = Evidence-based Practice Center.

For each KQ 1, 2, 4, and 5 outcome and comparison, we will summarize the magnitude and direction of effect across the body of evidence. To grade the SOE, we will consider the domains of consistency, precision, directness, study limitations (i.e., ROB), and reporting bias. For NRSI bodies of evidence, we will also consider the additional domains of plausible confounding, strength of association, and dose-response association, which in this review's context might refer to the frequency of screening, timeliness of evaluation of a positive signal, or timeliness of starting treatment. We consider all of the outcomes we have specified for grading SOE as direct outcomes as they represent patient-centered health outcomes; but depending on how data is ascertained, they may be judged as indirect. For judging consistency in KQs 1, 2, 4, and 5, we will consider both the direction and magnitude of relative effects and evaluate whether confidence intervals around point estimates overlap and whether inconsistencies among estimates can be explained. For pooled estimates, we will also consider the I² statistic. For judging precision, we will focus on the confidence interval around the absolute effect using a minimally contextualized approach as described by the GRADE working group.²⁹ For KQ 1, we will consider the absolute effect in relationship to a minimally important difference (MID). For mortality outcomes, we will consider the MID to be 3 deaths averted per 10,000 patients screened. This absolute difference represents the smallest observed mortality benefit observed among cancer screening tests recommended by the U.S. Preventive Services Task Force.^30-33 We will use this threshold only to evaluate the precision of the mortality effect estimates as part of assessing SOE. Decision makers who use the findings from this review may consider using a different threshold for determining a clinical net benefit from screening, including weighing the mortality benefit observed against the harms of screening. For evaluating the precision of outcomes associated with KQs 2, 4, and 5, we will consider the absolute effect in relationship to a null effect because MIDs for cancer detection and harms from screening are not established. For KQs 1 and 2, we will assess reporting bias by comparing protocols to final reported methods and outcomes.

For KQ 3 (accuracy), for each major class of analyte (e.g., cell-free DNA, specific protein biomarker(s) and cancer type), we will summarize the accuracy outcomes with pooled estimates where possible, and when not possible, we will describe the range of estimates. We will not grade SOE for KQ 3 outcomes because we expect to find numerous studies evaluating various unique tests, with few tests evaluated by more than one study. Further, we expect most of the studies to report accuracy outcomes by cancer type, with data reported for between 2 to 10 different types of cancers. Because SOE must be applied across a body of evidence reporting the same intervention (i.e., screening test) and outcome, this would require hundreds of different SOE assessments for an outcome that is the least important for decision making. Although this outcome is less important to clinical and policy decision making, it reflects the early stages of research on MCSTs, and a synthesis of accuracy is useful for decisions about future research investments in these tests.

Assessing Applicability: For KQs 1, 2, 4, and 5, we will consider factors such as characteristics of the patient population, for example, how patients were recruited, their underlying cancer risk or concern for cancer including other preventive health behaviors such as smoking, socioeconomic status, insurance status, and other healthcare system factors in determining applicability. For KQ 3, we will consider the pace of evolution and enhancements of the tests evaluated, whether tests are replicable or available outside of the studies where they have been evaluated, secular trends in diagnostic evaluations used for reference standards, and comparability of laboratory processes across in international settings.³⁴

Assessing Conflict of Interest: We will use the Tool for Addressing Conflicts of Interest in Trials (TACIT) to systematically evaluate conflicts of interest (COI) in the study sponsors and investigators of included studies.³⁵ We will report the findings from our COI assessments using this tool in a report appendix. All studies will be included in the review regardless of any notable concerns for COI. However, we may conduct sensitivity analyses based on the presence of notable COI concerns.

Use of Artificial Intelligence and/or Machine Learning: We will use the AI prioritization feature available in the DistillerSR platform to prioritize titles and abstracts and full-text articles for screening. For title and abstract review, we will monitor DistillerSR’s priority rankings and when the highest remaining citation rank score is less than 0.20, we will transition to a single human reviewer and use the AI feature as the second reviewer to screen the remaining titles and abstracts. We will use one or more large language models (LLMs) such as ChatGPT or Claude to extract data required for assessing COI with TACIT. We may use one or more LLMs as an editorial assistant to help improve the clarity or succinctness of text written by team members in the draft evidence report. All such uses will be reviewed by a team member to ensure that accuracy and intent is maintained.

1. Ward A, Van Nuys K, Lakdawalla D. Reducing Racial Disparities In Early Cancer Diagnosis With Blood-Based Tests. University of Southern California, Leonard D. Schaeffer Center for Health Policy and Economics; 2021. Accessed on February 9, 2024.
2. 117th Congress (2021-22) Senate Finance Committee. Medicare Multi-Cancer Early Detection Screening Coverage Act of 2021. 2021.
3. Connal S, Cameron JM, Sala A, et al. Liquid biopsies: the future of cancer early detection. J Transl Med. 2023 Feb 11;21(1):118. doi: 10.1186/s12967-023-03960-8. PMID: 36774504.
4. Deverka PA, Douglas MP, Phillips KA. Multicancer screening tests: anticipating and addressing considerations for payer coverage and patient access. Health Aff (Millwood). 2022 Mar;41(3):383-9. doi: 10.1377/hlthaff.2021.01316. PMID: 35254936.
5. Effective Health Care Program. Methods Guide for Effectiveness and Comparative Effectiveness Reviews. Agency for Healthcare Research and Quality, Rockville, MD; 2022. Accessed on May 21, 2024.
6. Effective Health Care Program. Methods guide for medical test reviews. Agency for Healthcare Research and Quality, Rockville, MD; 2012. Accessed on May 22, 2024.
7. Deeks J, Bossuyt P, Leeflang M, Takwoingi Y, eds. Cochrane Handbook for Systematic Reviews of Diagnostic Test Accuracy. Version 2.0: Cochrane; 2023. Accessed on July 2023.
8. Schünemann HJ, Mustafa RA, Brozek J, et al. GRADE guidelines: 21 part 1. Study design, risk of bias, and indirectness in rating the certainty across a body of evidence for test accuracy. J Clin Epidemiol. 2020;122:129-41.
9. Schünemann HJ, Mustafa RA, Brozek J, et al. GRADE guidelines: 22. The GRADE approach for tests and strategies—from test accuracy to patient-important outcomes and recommendations. J Clin Epidemiol. 2019;111:69-82.
10. Salameh J-P, Bossuyt PM, McGrath TA, et al. Preferred reporting items for systematic review and meta-analysis of diagnostic test accuracy studies (PRISMA-DTA): explanation, elaboration, and checklist. BMJ. 2020;370.
11. UNDP (United Nations Development Programme). Human development report 2023-24: breaking the gridlock: reimagining cooperation in a polarized world. New York: 2024.
12. Baker SG, Etzioni R. Prediagnostic evaluation of multicancer detection tests: Design and analysis considerations. J Natl Cancer Inst. 2024 Feb 28. doi: 10.1093/jnci/djae050. PMID: 38419575.
13. McGowan J, Sampson M, Salzwedel DM, et al. PRESS Peer Review of Electronic Search Strategies: 2015 Guideline Statement. J Clin Epidemiol. 2016 2016/07/01/;75:40-6. doi: https://doi.org/10.1016/j.jclinepi.2016.01.021.
14. Sterne JAC, Savović J, Page MJ, et al. RoB 2: a revised tool for assessing risk of bias in randomised trials. BMJ. 2019 Aug 28;366:l4898. doi: 10.1136/bmj.l4898. PMID: 31462531.
15. Sterne JA, Hernán MA, Reeves BC, et al. ROBINS-I: a tool for assessing risk of bias in non-randomised studies of interventions. BMJ. 2016 Oct 12;355:i4919. doi: 10.1136/bmj.i4919. PMID: 27733354.
16. Whiting PF, Rutjes AW, Westwood ME, et al. QUADAS-2: a revised tool for the quality assessment of diagnostic accuracy studies. Ann Intern Med. 2011 Oct 18;155(8):529-36. doi: 10.7326/0003-4819-155-8-201110180-00009. PMID: 22007046.
17. Morton SC, Murad MH, O'Connor E, et al. AHRQ Methods for Effective Health Care Quantitative Synthesis—An Update.  Methods Guide for Effectiveness and Comparative Effectiveness Reviews. Rockville (MD): Agency for Healthcare Research and Quality (US); 2008.
18. Deeks J, Higgins J, Altman D. Chapter 10: Analysing data and undertaking meta-analyses. In Cochrane Handbook for Systematic Reviews of Interventions version 6.4. Cochrane. 2023. 
19. Lin L. Comparison of four heterogeneity measures for meta‐analysis. J Eval Clin Pract. 2020;26(1):376-84.
20. Egger M, Davey Smith G, Schneider M, et al. Bias in meta-analysis detected by a simple, graphical test. BMJ. 1997 Sep 13;315(7109):629-34. doi: 10.1136/bmj.315.7109.629. PMID: 9310563.
21. Berkman ND, Lohr KN, Ansari MT, et al. Grading the strength of a body of evidence when assessing health care interventions: an EPC update. J Clin Epidemiol. 2015 Nov;68(11):1312-24. doi: 10.1016/j.jclinepi.2014.11.023. PMID: 25721570.
22. GRADE Working Group. Welcome to the GRADE working group. 2024. Accessed on May 29, 2024.
23. Bretthauer M, Wieszczy P, Løberg M, et al. Estimated Lifetime Gained With Cancer Screening Tests: A Meta-Analysis of Randomized Clinical Trials. JAMA Intern Med. 2023 Nov 1;183(11):1196-203. doi: 10.1001/jamainternmed.2023.3798. PMID: 37639247.
24. Stang A, Jöckel KH. The Impact of Cancer Screening on All-Cause Mortality. Dtsch Arztebl Int. 2018 Jul 23;115(29-30):481-6. doi: 10.3238/arztebl.2018.0481. PMID: 30135006.
25. Neal RD, Johnson P, Clarke CA, et al. Cell-Free DNA-Based Multi-Cancer Early Detection Test in an Asymptomatic Screening Population (NHS-Galleri): Design of a Pragmatic, Prospective Randomised Controlled Trial. Cancers (Basel). 2022 Oct 1;14(19). doi: 10.3390/cancers14194818. PMID: 36230741.
26. Owens L, Gulati R, Etzioni R. Stage shift as an endpoint in cancer screening trials: implications for evaluating multicancer early detection tests. Cancer Epidemiol Biomarkers Prev. 2022 Jul 1;31(7):1298-304. doi: 10.1158/1055-9965.Epi-22-0024. PMID: 35477176.
27. Feng X, Zahed H, Onwuka J, et al. Cancer Stage Compared With Mortality as End Points in Randomized Clinical Trials of Cancer Screening: A Systematic Review and Meta-Analysis. JAMA. 2024 Apr 7. doi: 10.1001/jama.2024.5814. PMID: 38583868.
28. Bach PB. Late-Stage Cancer End Points to Speed Cancer Screening Clinical Trials-Not So Fast. JAMA. 2024 Apr 7. doi: 10.1001/jama.2024.5821. PMID: 38583869.
29. Zeng L, Brignardello-Petersen R, Hultcrantz M, et al. GRADE Guidance 34: update on rating imprecision using a minimally contextualized approach. J Clin Epidemiol. 2022;150:216-24.
30. Kim JJ, Burger EA, Regan C, et al. U.S. Preventive Services Task Force Evidence Syntheses, formerly Systematic Evidence Reviews.  Screening for cervical cancer in primary care: a decision analysis for the U.S. Preventive Services Task Force. Rockville (MD): Agency for Healthcare Research and Quality (US); 2018.
31. Nelson HD, Cantor A, Humphrey L, et al. U.S. Preventive Services Task Force Evidence Syntheses, formerly Systematic Evidence Reviews.  Screening for Breast Cancer: A Systematic Review to Update the 2009 U.S. Preventive Services Task Force Recommendation. Rockville (MD): Agency for Healthcare Research and Quality (US); 2016.
32. Jonas DE, Reuland DS, Reddy SM, et al. U.S. Preventive Services Task Force Evidence Syntheses, formerly Systematic Evidence Reviews.  Screening for Lung Cancer With Low-Dose Computed Tomography: An Evidence Review for the U.S. Preventive Services Task Force. Rockville (MD): Agency for Healthcare Research and Quality (US); 2021.
33. US Preventive Services Task Force. Colorectal cancer: screening. 2021. Accessed on May 31, 2024.
34. Effective Health Care Program. Methods Guide – Chapter: Chapter 6: Assessing Applicability of Medical Test Studies in Systematic Reviews. Agency for Healthcare Research and Quality, Rockville, MD. 2012. 
35. Cochrane Bias Methods Group. TACIT Tool for Addressing Conflicts of Interest in Trials. 2024. Accessed on June 5, 2024.

None

If there is a need to amend the protocol, we will provide a numbered list of versions with the date of posting, which will be hyperlinked to previous versions; and a table with the date of each amendment, description of the change, and the rationale. Changes will be incorporated into the protocol.

None

AHRQ posted the preliminary Key Questions on the AHRQ Effective Health Care Website for public comment from February 12, 2024, to March 4, 2024. The EPC refined and finalized them after reviewing of the public comments and seeking input from Key Informants and the Technical Expert Panel (TEP). This input is intended to ensure that the Key Questions are specific and relevant.

Key Informants are the end users of research; they can include patients and caregivers, practicing clinicians, relevant professional and consumer organizations, purchasers of healthcare, and others with experience in making healthcare decisions. Within the EPC program, the Key Informant role is to provide input into the decisional dilemmas and help keep the focus on Key Questions that will inform healthcare decisions. The EPC solicits input from Key Informants when developing questions for the systematic review or when identifying high-priority research gaps and needed new research. Key Informants are not involved in analyzing the evidence or writing the report. They do not review the report, except as given the opportunity to do so through the peer or public review mechanism.

Key Informants must disclose any financial conflicts of interest greater than $5,000 and any other relevant business or professional conflicts of interest. Because of their role as end users, individuals are invited to serve as Key Informants and those who present with potential conflicts may be retained. The AHRQ Task Order Officer (TOO) and the EPC work to balance, manage, or mitigate any potential conflicts of interest identified.

Technical Experts constitute a multidisciplinary group of clinical, content, and methodological experts who provide input in defining populations, interventions, comparisons, or outcomes and identify particular studies or databases to search. The TEP is selected to provide broad expertise and perspectives specific to the topic under development. Divergent and conflicting opinions are common and perceived as healthy scientific discourse that fosters a thoughtful, relevant systematic review. Therefore, study questions, design, and methodological approaches do not necessarily represent the views of individual technical and content experts. Technical Experts provide information to the EPC to identify literature search strategies and suggest approaches to specific issues as requested by the EPC. Technical Experts do not do analysis of any kind; neither do they contribute to the writing of the report. They do not review the report, except as given the opportunity to do so through the peer or public review mechanism.

Members of the TEP must disclose any financial conflicts of interest greater than $5,000 and any other relevant business or professional conflicts of interest. Because of their unique clinical or content expertise, individuals are invited to serve as Technical Experts and those who present with potential conflicts may be retained. The AHRQ TOO and the EPC work to balance, manage, or mitigate any potential conflicts of interest identified.

Peer reviewers are invited to provide written comments on the draft report based on their clinical, content, or methodological expertise. The EPC considers all peer review comments on the draft report in preparing the final report. Peer reviewers do not participate in writing or editing of the final report or other products. The final report does not necessarily represent the views of individual reviewers.

The EPC will complete a disposition of all peer review comments. The disposition of comments for systematic reviews and technical briefs will be published 3 months after publication of the evidence report.

Potential peer reviewers must disclose any financial conflicts of interest greater than $5,000 and any other relevant business or professional conflicts of interest. Invited peer reviewers with any financial conflict of interest greater than $5,000 will be disqualified from peer review. Peer reviewers who disclose potential business or professional conflicts of interest can submit comments on draft reports through the public comment mechanism.

EPC core team members must disclose any financial conflicts of interest greater than $1,000 and any other relevant business or professional conflicts of interest. Direct financial conflicts of interest that cumulatively total more than $1,000 will usually disqualify an EPC core team investigator.

This project was funded under Contract No. 75Q80120D00007 Task Order 75Q80124F32008 from the Agency for Healthcare Research and Quality, U.S. Department of Health and Human Services. The AHRQ Task Order Officer reviewed the EPC response to contract deliverables for adherence to contract requirements and quality. The authors of this report are responsible for its content. Statements in the report should not be construed as endorsement by either the Agency for Healthcare Research and Quality or the U.S. Department of Health and Human Services.

This protocol will be registered in the international prospective register of systematic reviews (PROSPERO).

Our preliminary scope defines eligibility based on studies conducted in countries with a high or very high Human Development Index (HDI) designation, based on the 2024 United Nations Human Development Report.¹¹ Our preliminary evidence scan identified several tests developed and tested in China with FDA breakthrough device designations. Such studies would not be eligible if we limited the scope to very highly developed countries. We expect to identify few to no studies from high HDI countries other than China.

PubMed KQs 1-5 and CQ 1 search, 4/30/2024

PubMed CQs 2, 3, and 4 search for costs and cost-effectiveness, 4/30/2024

PubMed Patch Search 1 for additional MCD+SCD terms, 5/6/2024

PubMed Patch Search 2, additional multicancer screening terms, 5/7/2024

KQ 1, 2, 4, and 5 Studies

AI = artificial intelligence; CI = confidence interval; HR = hazard ratio; KQ = key question; MCST = multiple cancer screening test; N = number; NCT = National Clinical Trial; NGS = next-generation sequencing; OR = odds ratio; PCR = polymerase chain reaction; RR = risk ratio; SD = standard deviation; U.S. = United States.

KQ3 (Accuracy Studies)

For each test, cancer type, and subpopulation:

AI = artificial intelligence; AUC = analytical ultracentrifugation; CI = confidence interval; N = number; NGS = next-generation sequencing; NZ = New Zealand; PCR = polymerase chain reaction; U.S. = United States.