Circulating Tumor DNA for Molecular Residual Disease Detection and Multicancer Early Detection: A Systematic Review

1. Introduction

Liquid biopsy has moved in recent years from a largely experimental concept into the domain of clinically meaningful tools in personalized oncology [1-12]. One of its most extensively studied components is the analysis of circulating tumor DNA, which may reflect the presence of a tumor clone, its molecular features, and disease dynamics [1-2,5- 8,10,13]. ctDNA is being evaluated across several clinical uses, including detection of molecular residual disease after curative-intent treatment, early detection of molecular relapse, monitoring of treatment response, and development of blood-based approaches for multicancer early detection [1-24]. The clinical meaning of ctDNA cannot be reduced to the ability to detect tumor-derived DNA fragments in plasma. The more important question is whether test results lead to changes in medical management that improve outcomes or safely reduce unnecessary treatment [14-15,25-28]. This distinction is fundamental in oncology. An analytically sensitive assay may detect a molecular signal accurately, yet still lack demonstrated clinical utility [3-4,9,16-17]. For a ctDNA-guided strategy to be clinically credible, it must show more than an association between a positive result and recurrence risk [1-2,5-8,13,18-22,29]. It must also show an effective downstream action, such as initiation, omission, intensification, or de-escalation of treatment, modification of surveillance, or early therapeutic intervention [14-15,25-28]. Molecular residual disease detection and multicancer early detection represent different clinical scenarios and should not be judged by identical standards. MRD testing is performed in patients with a known cancer diagnosis after treatment given with curative intent [1,5-7,14-15,18-22,26- 29]. In this setting, pre-test recurrence risk remains clinically important, and ctDNA detection may serve as a marker of persistent microscopic tumor burden [1-2,5-8,13,18-22,29]. Its potential value lies in improved selection for adjuvant therapy, reduction of treatment intensity in patients at lower risk, escalation of therapy in patients at high molecular risk, or earlier intervention when molecular relapse is detected [14-15,25-28]. Multicancer early detection requires a different evidentiary logic. Unlike MRD testing, MCED is intended for asymptomatic people or screening-eligible populations [4,9,11,16,23]. Cancer prevalence is substantially lower in this setting. Even high specificity may therefore produce a clinically meaningful number of false-positive results [11,23- 24]. Such results may initiate additional diagnostic cascades, including repeat laboratory testing, imaging, invasive procedures, emotional distress, and potential overdiagnosis [11,23-24]. For MCED, it is insufficient to demonstrate detection of a cancer signal or prediction of the tissue of origin [4,9,16-17]. The test must identify clinically meaningful cancers at earlier stages without creating excessive harm, and should ultimately reduce mortality or produce a substantial and durable shift in population-stage distribution [30-31]. This systematic review therefore evaluates ctDNA-based strategies separately for two key domains. The first domain is molecular residual disease detection after potentially curative treatment [1,5-7,14-15,18-22,25-29]. The second is multicancer early detection [4,9,11,16,23-24,30-32]. This separation avoids methodological mixing of distinct clinical tasks and allows a more precise assessment of where ctDNA can already be considered a tool for clinical decision-making and where further testing is required in randomized trials, diagnostic pathway studies, and implementation studies that assess accuracy, outcomes, safety, and health-system feasibility [14-15,27,30-31].

2. Methods

2.1. Protocol and eligibility criteria

This systematic review was conducted in accordance with PRISMA 2020 and PRISMA-S reporting principles [33-34]. The protocol was not prospectively registered. Eligibility criteria were structured separately for MRD after curativeintent treatment, post-treatment recurrence monitoring, and MCED in asymptomatic or screening-eligible populations. Eligible MRD studies were randomized trials, prospective interventional studies, and longitudinal prognostic cohorts in solid tumors. For intervention studies, the ctDNA result had to be linked to a prespecified management action, including administration, omission, escalation, or de-escalation of systemic therapy, modification of surveillance, or initiation of treatment at molecular relapse [14-15,21,25-28]. Primary outcomes were recurrence-free survival, disease-free survival, overall survival, recurrence, treatment exposure, and toxicity. For prognostic MRD studies, ctDNA had to be measured after completion of curative-intent treatment or during serial follow-up and related to subsequent clinical or radiological recurrence [1-2,5-7,13,18-22,29]. Cross-sectional analytical studies without longitudinal clinical outcomes were used only as technological context and were not graded as evidence of clinical effectiveness. Eligible MCED studies included prospective screening or diagnostic-pathway studies in asymptomatic or screeningeligible participants with reported diagnostic resolution after a positive result [11,23-24,30,32]. Case-control classifierdevelopment studies and biobank-only validation studies were retained as supporting evidence for analytical or diagnostic validity but were not treated as independent evidence of population screening benefit [3-4,8-10,12,16- 17]. Exclusion criteria were hematologic malignancies, single case reports, technical reports without clinically interpretable outcomes, commercial or news materials, reviews without original data, and publications without transparent methods. Ongoing programs without any effectiveness or diagnosticperformance results were described separately and were excluded from risk-of-bias and GRADE assessments. Conference abstracts reporting outcome data were appraised provisionally using the information available at the review cutoff, with limitations related to incomplete reporting incorporated into risk-of-bias and certainty judgments

2.2. Information sources and search strategy

The main search was conducted from database inception to July 9, 2026 in MEDLINE via PubMed, Embase via Elsevier, Scopus, Web of Science Core Collection, and the Cochrane Central Register of Controlled Trials. Trial searches covered ClinicalTrials.gov, WHO ICTRP, ISRCTN, and jRCT. A targeted update search on July 15, 2026 was undertaken to identify newly published reports and major 2026 conference abstracts, including ALTAIR, PATHFINDER 2, and NHS-Galleri. Complete source-specific strategies are provided in Supplementary Table S1. Search concepts combined circulating tumor DNA, cellfree DNA, liquid biopsy, molecular or minimal residual disease, recurrence or surveillance, multicancer early detection, solid tumors, and cancer. No date restriction was applied in the main search. English-language and humanstudy limits were applied where supported by the platform. Conference abstracts were retained only when they represented the most recent report of a major ongoing program and were clearly identified as non-peer-reviewed evidence.

2.3. Study selection and data management

Records were managed in a reference-management workspace. Deduplication used DOI, PMID, normalized title, first author, publication year, and trial registration identifier. Registry entries were linked to publications but retained as separate records when they supplied current recruitment or reporting status. The complete executable search strategies are reported in Supplementary Table S1 rather than repeated in the main text. After deduplication, 54 unique records were available for screening, comprising 42 bibliographic records and 12 trialregistry records. Screening was performed independently by two reviewers, with disagreements resolved by discussion and, when necessary, adjudication by a third reviewer.

2.4. Data extraction

Two reviewers independently extracted bibliographic details, tumor type and stage, study design, sample size, assay type, tumor-informed or tumor-agnostic approach, sampling time, clinical decision algorithm, comparator, endpoint definitions, numerical effect estimates with confidence intervals, follow-up, publication status, funding, and conflicts of interest. The completed study-level extraction table is provided in Supplementary Table S3, and the reusable extraction template is provided in Supplementary Table S7.

MRD outcomes included recurrence-free survival, disease-free survival, overall survival, molecular clearance, treatment exposure, and toxicity. MCED outcomes included cancer signal detection rate, sensitivity, specificity, positive predictive value, cancer signal origin accuracy, diagnostic resolution, false-positive investigations, invasive procedures, psychosocial outcomes, and stage distribution.

2.5. Risk of bias and certainty of evidence

Randomized trials were assessed with RoB 2 and classified as low risk, some concerns, or high risk [35]. Prognostic factor studies were assessed with QUIPS [36]. Non-randomized studies estimating effects of interventions were assessed with ROBINS-I [37]. Diagnostic accuracy and diagnostic-pathway studies were assessed with QUADAS-2 [38]. Domain-level judgments are reported in Supplementary Table S4. Certainty of evidence was evaluated by outcome using GRADE domains: risk of bias, inconsistency, indirectness, imprecision, and publication bias [39]. Because several questions were informed by a single randomized trial or by heterogeneous observational cohorts, certainty ratings were not inferred directly from study design. The evidence profile is reported in Supplementary Table S5.

2.6. Data synthesis

A structured narrative synthesis was performed. MRDguided interventional studies, prognostic and observational MRD evidence, MCED diagnostic pathway studies, and studies of harm, economic consequences, and implementation were evaluated separately. Interpretation prioritized not only analytical or diagnostic accuracy, but also the presence of a complete clinical chain. This chain required detection of a molecular signal, change in clinical decisionmaking, delivery of a defined intervention, and impact on outcomes or potential harm. The completed PRISMA 2020 checklist is provided in Supplementary Table S6.

3. Results

3.1. Study selection

A total of 54 unique records entered title, abstract, or registry screening. Twelve records were excluded at this stage, leaving 42 reports that were sought and retrieved for full-text assessment. Ten reports were excluded after full-text review. Twenty-nine completed reports representing 25 study programs were included in the narrative synthesis, while three ongoing or incompletely published programs were described separately. The study-selection process is shown in Figure 1. Full-text exclusion categories are provided in Supplementary Table S2.

Figure 1. PRISMA 2020 flow diagram

3.2. Characteristics of included studies

The completed evidence was divided into MRD-guided intervention studies, longitudinal prognostic MRD studies, MCED diagnostic-pathway studies, and analytical or classifier-validation studies used as supporting context. Key characteristics are summarized in Table 1, and full extraction is provided in Supplementary Table S3. The most clinically actionable evidence came from DYNAMIC in stage II colon cancer and IMvigor011 in ctDNApositive muscle-invasive bladder cancer [14-15,27]. DYNAMIC demonstrated a reduction in chemotherapy exposure without an apparent loss of recurrence-free survival, whereas IMvigor011 demonstrated improved disease-free and overall survival with biomarker-selected adjuvant atezolizumab. DYNAMIC-III and ALTAIR showed that detection of molecular residual disease does not guarantee benefit from escalation or treatment at molecular recurrence [26,28].

Table 1. Selected key characteristics of completed studies included in the clinical synthesis

Study Population/sample Design and assay Endpoint Interpretation Garcia-Murillas 2015 [1] Early breast cancer; n=55 Prospective tumor-informed cohort Molecular relapse ctDNA anticipated clinical relapse and established prognostic validity.

Study Population/sample Design and assay Endpoint Interpretation TRACERx 2017/2023 [2,13] Early NSCLC; prospective cohort Tumor-informed phylogenetic ctDNA Relapse and dissemination Strong biological and prognostic evidence; treatment actionability not tested. Reinert 2019 [5] Stage I-III CRC; n=130 Prospective ultradeep sequencing Postsurgical recurrence Postoperative ctDNA strongly stratified recurrence risk. Christensen 2019 [6] Bladder cancer; n=68 Longitudinal tumor-informed cfDNA Metastatic relapse Molecular relapse preceded clinical progression in many patients. Coombes 2019 [7] Early breast cancer; n=49 Personalized serial ctDNA Metastatic recurrence ctDNA frequently preceded metastatic recurrence. IMvigor010 biomarker analysis 2021 [18] Resected MIBC; randomizedtrial biospecimens Exploratory ctDNA subgroup analysis Adjuvant atezolizumab interaction Hypothesis-generating predictive signal; prospective confirmation required. Loupakis 2021 [19] CRC after metastasectomy; n=112 Personalized ctDNA cohort MRD and recurrence Post-resection ctDNA associated with high recurrence risk. Magbanua 2021 [20] Neoadjuvant-treated breast cancer; n=84 Prospective serial ctDNA Response and survival ctDNA dynamics reflected response and prognosis. DYNAMIC 2022/2025 [14-15] Stage II colon cancer; n=455 Randomized ctDNA-guided strategy Chemotherapy use and RFS Reduced chemotherapy exposure with comparable 2and 5-year RFS. GALAXY/CIRCULATE-Japan 2023/2024 [21,29] Resectable CRC; n=1,039 and expanded cohort Prospective observational platform MRD, DFS, OS, ACT interaction Strong prognostic evidence; treatment interaction remains observational. c-TRAK TN 2023 [25] High-risk early TNBC; n=208 Prospective surveillance/intervention ctDNA-triggered pembrolizumab Illustrated limitations of late molecular detection and weak intervention uptake. DYNAMIC-III 2025 [26] Stage III colon cancer; n=968 evaluable Randomized strategy trial Risk-adjusted adjuvant therapy Prognostic separation was strong; escalation benefit was not established. IMvigor011 2025 [27] ctDNA-positive MIBC; n=250 randomized Phase III double-blind RCT DFS and OS Atezolizumab improved DFS and OS versus placebo. ALTAIR 2026 [28] ctDNA-positive resected CRC; n=243 Phase III double-blind RCT DFS after molecular recurrence Primary DFS endpoint not met; toxicity substantial. Martin-Arana 2025 [22] Localized stage II-III colon cancer Tumor-agnostic plasma WES MRD sensitivity and relapse biology Promising detection method; external clinical validation required. DETECT-A 2020 [11] Asymptomatic women; n=10,006 Prospective blood test plus PET-CT pathway Feasibility and diagnostic resolution Feasible pathway without mortality endpoint. PATHFINDER 2023/2025 [23-24] Screening-eligible adults; n=6,662 Prospective MCED return-of-results cohort Diagnostic resolution and psychosocial outcomes Feasible pathway; falsepositive work-up and anxiety remain relevant. CCGA validation 2020/2021 [9,16] Multiple cancers and controls Targeted methylation classifier validation Sensitivity, specificity, tissue of origin Diagnostic validation, not population outcomes evidence. MCED evidence included DETECT-A, PATHFINDER, PATHFINDER 2, NHS-Galleri, CCGA validation studies, and related diagnostic-pathway or psychosocial analyses [9,11,16,23-24,30,32]. These studies support feasibility, high specificity, and targeted diagnostic routing, but they differ in design and endpoint. PATHFINDER 2 and NHS-Galleri were available as 2026 conference abstracts and were therefore treated as preliminary evidence pending full peer-reviewed reports [30,32]. Key numerical findings and their clinical interpretation are summarized in Table 2.

Table 2. Key numerical results and clinical conclusions

Study/program Sample size Primary endpoint Main numerical result Clinical conclusion DYNAMIC [14-15] 455 randomized Chemotherapy use; RFS Chemotherapy 15% vs 28%; 2y RFS 93.5% vs 92.4%; 5-y RFS 88% vs 87% Supports de-escalation in stage II colon cancer. DYNAMIC-III [26] 968 evaluable 3-y RFS in ctDNA-negative; 2-y RFS in ctDNA-positive 72.5% were ctDNA-negative; strong prognostic separation; escalation superiority not established Not a basis for universal escalation/de-escalation. GALAXY [21,29] 1,039 initial; expanded cohort DFS/OS and ACT interaction Postoperative ctDNA strongly associated with recurrence; treatment interaction observational High prognostic value; causal treatment inference limited. c-TRAK TN [25] 208 enrolled ctDNA detection and pembrolizumab response ctDNA detected in 27.3%; many had radiographic metastases; no ctDNA clearance in treated subgroup Cautionary actionability evidence. IMvigor011 [27] 761 monitored; 250 randomized DFS and OS DFS 9.9 vs 4.8 mo, HR 0.64 (0.47-0.87); OS 32.8 vs 21.1 mo, HR 0.59 (0.39-0.90) Direct evidence of MRD-guided treatment benefit. ALTAIR [28] 243 randomized DFS Median DFS 9.30 vs 5.55 mo; HR 0.79 (0.60-1.05); P=0.107 Primary endpoint not met. DETECT-A [11] 10,006 Cancer detection pathway Prospective blood test plus PET-CT pathway; mortality not assessed Feasibility only. PATHFINDER [23] 6,662 Diagnostic resolution Median resolution 79 d; truepositive 57 d; false-positive 162 d Feasibility with substantial downstream testing. PATHFINDER 2 [32] 35,878 enrolled; 32,007 performance set PPV, specificity, sensitivity, resolution PPV 60.3%; specificity 99.6%; all-cancer sensitivity 39.3%; median resolution 48 d Conference-level performance evidence; no mortality endpoint. NHS-Galleri [30] 142,924 randomized Stage III-IV cancer incidence Primary endpoint not met: 706 vs 688 events; IRR 1.03; stage IV reduction secondary No definitive population implementation conclusion. Vanguard [31] Active feasibility trial Recruitment and diagnostic logistics No effectiveness result available Designed to inform a future definitive RCT.

3.3. Evidence for MRD-guided interventions

In DYNAMIC, 455 patients with stage II colon cancer were randomized. Chemotherapy was administered to 15% of the ctDNA-guided group and 28% of the standard-management group. Two-year recurrence-free survival was 93.5% versus 92.4%, and at a median follow-up of 59.7 months five-year recurrence-free survival was 88% versus 87% [14-15]. The clinical contribution of the strategy was therefore reduced treatment exposure without an apparent survival penalty, rather than improved survival. In IMvigor011, 761 patients entered serial ctDNA surveillance and 250 ctDNA-positive patients were randomized to atezolizumab or placebo. Median diseasefree survival was 9.9 versus 4.8 months (hazard ratio 0.64, 95% confidence interval 0.47-0.87), and median overall survival was 32.8 versus 21.1 months (hazard ratio 0.59, 95% confidence interval 0.39-0.90) [27]. This provides direct evidence that an MRD result can select a population in which a defined intervention improves outcomes. DYNAMIC-III enrolled 968 evaluable patients with stage III colon cancer and confirmed marked prognostic separation by postoperative ctDNA status, but the ctDNA-guided escalation component did not establish superior recurrencefree survival [26]. In ALTAIR, 243 ctDNA-positive patients were randomized after curative-intent treatment; trifluridinetipiracil did not significantly improve the primary disease-free survival endpoint compared with placebo (median 9.30 versus 5.55 months; hazard ratio 0.79, 95% confidence interval 0.60-1.05) and produced substantial grade 3 or higher toxicity [28].

3.4. Prognostic and observational MRD evidence

Across colorectal cancer, early-stage non-small cell lung cancer, breast cancer, and bladder cancer, postoperative ctDNA positivity was consistently associated with increased recurrence risk and often preceded clinical or radiological progression [1-2,5-7,13,18-22,29]. The magnitude of association varied by assay, sampling time, tumor shedding, treatment exposure, and endpoint definition. Prognostic discrimination is not equivalent to clinical benefit. c-TRAK TN illustrated this distinction: ctDNA was detected in 27.3% of 208 enrolled patients, but many already had radiologically detectable metastatic disease at molecular detection and pembrolizumab did not produce ctDNA clearance in the small treated group [25]. Negative ctDNA also cannot fully exclude recurrence, particularly in lowshedding tumors or sanctuary sites.

3.5. Evidence on MCED

DETECT-A enrolled 10,006 women and demonstrated that a multi-analyte blood test followed by PET-CT could be incorporated into a prospective diagnostic pathway [11]. PATHFINDER enrolled 6,662 participants; among participants with a cancer signal detected result, median time to diagnostic resolution was 79 days, and false-positive participants underwent substantial laboratory and imaging evaluation [23]. These studies established feasibility but were not designed to demonstrate mortality reduction. PATHFINDER 2 enrolled 35,878 participants; the 2026 ASCO abstract reported 287 positive tests among 32,007 participants with 12-month cancer assessment, 173 confirmed cancers, positive predictive value of 60.3%, specificity of 99.6%, and median diagnostic resolution of 48 days [32]. NHS-Galleri randomized 142,924 participants. Its primary endpoint - a statistically significant reduction in stage III-IV cancers across 12 prespecified cancers - was not met (706 versus 688 events; incidence rate ratio 1.03), although reductions in stage IV disease and increases in stage I-II diagnoses were reported as secondary findings [30]. The NCI Vanguard Study remains an active feasibility study for a future definitive randomized trial [31].

3.6. Clinical applicability of ctDNA and cfDNA

Clinical applicability requires a complete chain from molecular detection to a valid decision, an effective intervention, and improved patient-important outcomes. DYNAMIC and IMvigor011 satisfy this chain in defined settings [14-15,27]. DYNAMIC-III, c-TRAK TN, and ALTAIR show why the result cannot be generalized across tumors or interventions [25-26,28]. For MCED, current evidence supports pathway feasibility and high specificity but not mortality reduction. Conferencelevel results from PATHFINDER 2 improve estimates of diagnostic performance, while NHS-Galleri provides randomized evidence that the prespecified late-stage primary endpoint was not met [30,32]. Implementation decisions must therefore consider false-positive investigations, overdiagnosis, costs, capacity for confirmatory diagnostics, and equity of access.

4. Risk of bias

RoB 2 judgments were low risk or some concerns for the principal randomized trials [35]. DYNAMIC and IMvigor011 had low overall risk for their primary endpoints [14,27]. DYNAMIC-III and ALTAIR were judged as some concerns because interpretation depended on complex strategy algorithms, incomplete maturity for some endpoints, or reliance on exploratory subgroup and sensitivity analyses [26,28]. QUIPS, ROBINS-I, and QUADAS-2 assessments identified moderate or serious limitations in many nonrandomized studies, particularly confounding, heterogeneous sampling schedules, selective analysis, spectrum effects, verification bias, and incomplete assessment of false-negative participants [36-38]. Table 3 summarizes overall judgments; domain-level assessments are provided in Supplementary Table S4.

Table 3. Risk of bias

Study/evidence group Design and tool Overall judgment Main concern DYNAMIC [14] RCT; RoB 2 Low risk Open-label strategy but objective outcomes and prespecified analysis. DYNAMIC 5-y follow-up [15] Extended RCT follow-up; RoB 2 Some concerns Some post hoc ctDNA-clearance analyses. DYNAMIC-III [26] Randomized strategy trial; RoB 2 Some concerns Complex prespecified escalation/deescalation algorithms and subgroup interpretation. ALTAIR [28] Phase III RCT; RoB 2 Some concerns Primary endpoint negative; central review and subgroup findings exploratory. IMvigor011 [27] Phase III RCT; RoB 2 Low risk Double-blind design and objective primary endpoint. GALAXY [21,29] Prospective observational; QUIPS/ROBINS-I Moderate to serious Confounding in estimates of ACT benefit. IMvigor010 ctDNA analysis [18] Exploratory biomarker analysis; ROBINS-I Serious Post hoc subgroup and treatment interaction. c-TRAK TN [25] Prospective intervention cohort; ROBINS-I Serious Small treated subgroup and overt disease at ctDNA detection. TRACERx [2,13] Prospective prognostic cohort; QUIPS Moderate Selection and assay-timing limitations; no intervention test. Breast MRD cohorts [1,7,20] Prognostic cohorts; QUIPS Moderate Small samples and heterogeneous timing. DETECT-A [11] Diagnostic pathway; QUADAS-2 Moderate Verification and pathway effects; no randomized outcome endpoint. PATHFINDER [23-24] Diagnostic pathway; QUADAS-2 Moderate No randomized control and incomplete long-term verification of negatives. PATHFINDER 2 [32] Conference abstract; QUADAS-2 framework Some concerns Incomplete peer-reviewed reporting at cutoff. NHS-Galleri [30] Randomized screening trial; RoB 2 preliminary Some concerns Conference-level reporting and immature mortality follow-up. CCGA/Klein [9,16] Classifier validation; QUADAS-2 Moderate Spectrum and case-control applicability concerns.

Study/evidence group Design and tool Overall judgment Main concern Kim 2023 [17] Classifier study; QUADAS-2 High risk External intended-use population validation required.

5. Certainty of evidence

Certainty was moderate for ctDNA-guided de-escalation in stage II colon cancer because the evidence is based primarily on one randomized strategy trial with a relatively wide non-inferiority margin, despite consistent five-year follow-up [14-15]. Certainty was high for the relative diseasefree and overall survival effects of atezolizumab among ctDNA-positive patients in IMvigor011, but applicability remains limited to the studied disease, assay, and treatment context [27]. For other MRD settings, certainty was low to moderate. Prognostic associations were consistent, but direct evidence that a ctDNA-triggered intervention improves patientimportant outcomes was limited or negative [1-2,5-7,13,18- 22,25-26,28-29]. For MCED, certainty was moderate for diagnosticpathway feasibility and low or very low for reduction in latestage disease or mortality. PATHFINDER 2 and NHS-Galleri were conference reports at the review cutoff, and the NHS- Galleri primary endpoint was not met [30,32]. The complete GRADE profile is provided in Supplementary Table S5. Outcome-level certainty ratings are summarized in Table 4.

Table 4. Certainty of evidence

Finding Evidence base Certainty GRADE rationale Conclusion ctDNA-guided de-escalation in stage II colon cancer [14-15] RCT plus 5-y follow-up Moderate No serious inconsistency; some indirectness and imprecision from single strategy trial and non-inferiority margin Reduced chemotherapy without apparent RFS/OS loss. ctDNA-guided atezolizumab in ctDNA-positive MIBC [27] Phase III RCT High Low risk of bias; precise DFS and OS effects; limited settingspecific applicability Improved DFS and OS. ctDNA-guided strategy in stage III colon cancer [26] Randomized strategy trial Low to moderate Complex intervention, phase 2 escalation component, imprecision Prognostic value high; strategy benefit not established. Treatment at molecular recurrence [25,28] Prospective intervention study plus phase III RCT Low Inconsistent interventions, imprecision, negative primary endpoint Detection alone does not ensure benefit. Post-treatment MRD prognosis across solid tumors [1-2,5- 7,13,18-22,29] Multiple prospective cohorts Moderate Consistent direction but heterogeneity and confounding Strong prognostic value. MCED diagnostic-pathway feasibility [11,23-24,32] Prospective pathway studies; one conference report Moderate No mortality endpoint; indirectness for screening benefit Feasibility and high specificity supported. MCED reduction in late-stage cancer or mortality [30] One large randomized trial reported in conference abstract form Very low to low Primary endpoint not met; mortality immature; incomplete reporting Broad implementation remains premature.

6. Discussion

This review separates analytical validity, prognostic validity, and clinical utility. ctDNA is clinically actionable only when the test identifies a population for whom a defined management strategy produces net benefit. DYNAMIC and IMvigor011 are the clearest examples, but they support setting-specific decisions rather than universal ctDNAguided oncology [14-15,27]. MCED evidence has advanced from retrospective classifier validation to prospective pathways and a large randomized trial [11,23,30,32]. Nevertheless, the evidentiary standard for population screening is not satisfied by specificity, positive predictive value, or stage distribution alone. The failure of NHS-Galleri to meet its primary stage III-IV endpoint strengthens the need for mature mortality follow-up, transparent assessment of harms, and replication independent of test manufacturers [30]. The practical implication is that ctDNA and cfDNA tests provide clinical information rather than treatment. Their value depends on timing, assay sensitivity, tumor biology, availability of effective interventions, and the capacity of the health system to resolve positive findings without disproportionate harm.

7. Limitations

First, the review was based on aggregate published data and did not have access to individual participant data. Second, clinical and methodological heterogeneity precluded a defensible meta-analysis. Third, several 2026 findings were available only as conference abstracts and should be updated when full peer-reviewed reports appear. Fourth, certainty judgments remain sensitive to the choice of outcome and to the distinction between prognostic performance and treatment effect.

8. Conclusions ctDNA has demonstrated clinical utility in selected solid- tumor MRD settings. The strongest evidence supports

reduced chemotherapy exposure in stage II colon cancer without an apparent loss of recurrence-free survival and improved disease-free and overall survival with atezolizumab in ctDNA-positive muscle-invasive bladder cancer [14-15,27]. In most other settings, ctDNA remains a powerful prognostic biomarker whose optimal treatment response has not been established. MCED should not be described as an established population screening method. Prospective studies support feasibility and high specificity, but NHS-Galleri did not meet its prespecified primary late-stage endpoint, and mortality benefit remains unknown [11,23,30-32]. Future implementation requires randomized evidence of net benefit, mature mortality outcomes, acceptable false-positive and overdiagnosis burdens, and feasible diagnostic pathways.

Funding

The source manuscript did not include a completed funding declaration.

Conflicts of interest

The source manuscript did not include a completed conflict-of-interest declaration.

Data availability

All study-level data used in the narrative synthesis, search strategies, exclusion categories, risk-of-bias judgments, GRADE assessments, and the completed PRISMA checklist are provided in the manuscript and Supplementary Tables S1-S7.

References

[1] Garcia-Murillas I, Schiavon G, Weigelt B, et al. Mutation tracking in circulating tumor DNA predicts relapse in early breast cancer. Sci Transl Med. 2015;7(302);302ra133. doi 10.1126/scitranslmed.aab0021. [2] Abbosh C, Birkbak NJ, Wilson GA, et al. Phylogenetic ctDNA analysis depicts early-stage lung cancer evolution. Nature. 2017;545;446-451. doi 10.1038/nature22364. [3] Phallen J, Sausen M, Adleff V, et al. Direct detection of early-stage cancers using circulating tumor DNA. Sci Transl Med. 2017;9(403);eaan2415. doi 10.1126/scitranslmed.aan2415. [4] Cohen JD, Li L, Wang Y, et al. Detection and localization of surgically resectable cancers with a multi-analyte blood test. Science. 2018;359(6378);926-930. doi 10.1126/science.aar3247. [5] Reinert T, Henriksen TV, Christensen E, et al. Analysis of plasma cell-free DNA by ultradeep sequencing in patients with stages I to III colorectal cancer. JAMA Oncol. 2019;5(8);1124-1131. doi 10.1001/jamaoncol.2019.0528. [6] Christensen E, Birkenkamp-Demtroder K, Sethi H, et al. Early detection of metastatic relapse and monitoring of therapeutic efficacy by ultra-deep sequencing of plasma cell-free DNA in bladder cancer. J Clin Oncol. 2019;37(18);1547-1557. doi 10.1200/JCO.18.02052. [7] Coombes RC, Page K, Salari R, et al. Personalized detection of circulating tumor DNA antedates breast cancer metastatic recurrence. Clin Cancer Res. 2019;25(14);4255-4263. doi 10.1158/1078-0432.CCR-18-3663. [8] Cristiano S, Leal A, Phallen J, et al. Genome-wide cellfree DNA fragmentation in patients with cancer. Nature. 2019;570(7761);385-389. doi 10.1038/s41586-019-1272-6. [9] Liu MC, Oxnard GR, Klein EA, Swanton C, Seiden MV, CCGA Consortium. Sensitive and specific multi-cancer detection and localization using methylation signatures in cell-free DNA. Ann Oncol. 2020;31(6);745-759. doi 10.1016/j.annonc.2020.02.011. [10] Zviran A, Schulman RC, Shah M, et al. Genome-wide cell-free DNA mutational integration enables ultra-sensitive cancer monitoring. Nat Med. 2020;26(7);1114-1124. doi 10.1038/s41591- 020-0915-3. [11] Lennon AM, Buchanan AH, Kinde I, et al. Feasibility of blood testing combined with PET-CT to screen for cancer and guide intervention. Science. 2020;369(6499);eabb9601. doi 10.1126/science.abb9601. [12] Chen X, Gole J, Gore A, et al. Non-invasive early detection of cancer four years before conventional diagnosis using a blood test. Nat Commun. 2020;11;3475. doi 10.1038/s41467-020- 17316-z. [13] Abbosh C, Frankell AM, Harrison T, et al. Tracking early lung cancer metastatic dissemination in TRACERx using ctDNA. Nature. 2023;616;553-562. doi 10.1038/s41586-023-05776-4. [14] Tie J, Cohen JD, Lahouel K, et al. Circulating tumor DNA analysis guiding adjuvant therapy in stage II colon cancer. N Engl J Med. 2022;386(24);2261-2272. doi 10.1056/NEJMoa2200075. [15] Tie J, Wang Y, Lo SN, et al. Circulating tumor DNA analysis guiding adjuvant therapy in stage II colon cancer. Five-year outcomes of the randomized DYNAMIC trial. Nat Med. 2025;31;1509-1518. doi 10.1038/s41591-025-03579-w. [16] Klein EA, Richards D, Cohn A, et al. Clinical validation of a targeted methylation-based multi-cancer early detection test using an independent validation set. Ann Oncol. 2021;32(9);1167- 1177. doi 10.1016/j.annonc.2021.05.806. [17] Kim SY, Jeong S, Lee W, et al. Cancer signature ensemble integrating cfDNA methylation, copy number, and fragmentation facilitates multi-cancer early detection. Exp Mol Med. 2023;55;2445-2460. doi 10.1038/s12276-023-01119-5. [18] Powles T, Assaf ZJ, Davarpanah N, et al. ctDNA guiding adjuvant immunotherapy in urothelial carcinoma. Nature. 2021;595(7867);432-437. doi 10.1038/s41586-021-03642-9. [19] Loupakis F, Sharma S, Derouazi M, et al. Detection of molecular residual disease using a personalized circulating tumor DNA assay in colorectal cancer patients undergoing resection of metastases. JCO Precis Oncol. 2021;5;1166-1177. doi 10.1200/PO.21.00101. [20] Magbanua MJM, Swigart LB, Wu HT, et al. Circulating tumor DNA in neoadjuvant-treated breast cancer reflects response and survival. Ann Oncol. 2021;32(2);229-239. doi 10.1016/j.annonc.2020.11.007. [21] Kotani D, Oki E, Nakamura Y, et al. Molecular residual disease and efficacy of adjuvant chemotherapy in patients with colorectal cancer. Nat Med. 2023;29;127-134. doi 10.1038/s41591- 022-02115-4. [22] Martin-Arana J, Gimeno-Valiente F, Henriksen TV, et al. Whole-exome tumor-agnostic ctDNA analysis enhances minimal residual disease detection and reveals relapse mechanisms in localized colon cancer. Nat Cancer. 2025;6;1000-1016. doi 10.1038/s43018-025-00960-z. [23] Schrag D, Beer TM, McDonnell CH, et al. Blood-based tests for multicancer early detection. PATHFINDER, a prospective cohort study. Lancet. 2023;402;1251-1260. doi 10.1016/S0140- 6736(23)01700-2. [24] Nadauld L, McDonnell CH, et al. Psychosocial impact associated with a multicancer early detection test in PATHFINDER. A prospective multicentre cohort study. Lancet Oncol. 2025;26(2). doi 10.1016/S1470-2045(24)00645-4. [25] Turner NC, Swift C, Jenkins B, et al. Results of the c- TRAK TN trial. A clinical trial using ctDNA mutation tracking to detect molecular residual disease and trigger intervention in patients with moderate-risk and high-risk early-stage triple-negative breast cancer. Ann Oncol. 2023;34(2);200-211. doi 10.1016/j.annonc.2022.11.005. [26] Tie J, et al. Circulating tumor DNA-guided adjuvant therapy in locally advanced colon cancer. The randomized phase 2 and 3 DYNAMIC-III trial. Nat Med. 2025;31(12);4291-4300. doi 10.1038/s41591-025-04030-w. [27] Powles T, et al. ctDNA-guided adjuvant atezolizumab in muscle-invasive bladder cancer. N Engl J Med. 2025;393(24);2395- 2408. doi 10.1056/NEJMoa2511885. [28] Bando H, et al. Post-adjuvant chemotherapy in ctDNApositive patients with colorectal cancer. The ALTAIR trial. Nat Med. 2026. doi 10.1038/s41591-026-04428-0. [29] Nakamura Y, et al. ctDNA-based molecular residual disease and survival in resectable colorectal cancer. Nat Med. 2024;30;3272-3283. doi 10.1038/s41591-024-03254-6. [30] Swanton C, Johnson P, Round T, et al. NHS-Galleri: primary results from a randomised controlled trial to assess the clinical utility of a multi-cancer early detection test in population screening. J Clin Oncol. 2026;44(17 suppl):LBA100. doi:10.1200/JCO.2026.44.17_suppl.LBA100.

[31] National Cancer Institute. The Vanguard Study: testing a new way to screen for cancer. Cancer Screening Research Network multi-cancer detection feasibility study. ClinicalTrials.gov identifier NCT06995898. Accessed July 15, 2026. [32] Giridhar K, McDonnell CH, Kurbegov D, et al. Safety and performance results from PATHFINDER 2 (PF2), a registrational study of a multi-cancer early detection test in an intended-use population. J Clin Oncol. 2026;44(17 suppl):LBA10509. doi:10.1200/JCO.2026.44.17_suppl.LBA10509. [33] Page MJ, McKenzie JE, Bossuyt PM, et al. The PRISMA 2020 statement. An updated guideline for reporting systematic reviews. BMJ. 2021;372;n71. doi 10.1136/bmj.n71. [34] Rethlefsen ML, Kirtley S, Waffenschmidt S, et al. PRISMA-S. An extension to the PRISMA Statement for Reporting Literature Searches in Systematic Reviews. Syst Rev. 2021;10;39. doi 10.1186/s13643-020-01542-z. [35] Higgins JPT, Savovic J, Page MJ, Elbers RG, Sterne JAC. Revised Cochrane risk-of-bias tool for randomized trials, RoB

2. Version 2.0. Cochrane. 2019.

[36] Hayden JA, van der Windt DA, Cartwright JL, Cote P, Bombardier C. Assessing bias in studies of prognostic factors. Ann Intern Med. 2013;158(4);280-286. doi 10.7326/0003-4819-158-4- 201302190-00009. [37] Sterne JAC, Hernan MA, Reeves BC, Savovic J, Berkman ND, Viswanathan M, et al. ROBINS-I. A tool for assessing risk of bias in non-randomized studies of interventions. BMJ. 2016;355;i4919. doi 10.1136/bmj.i4919. [38] Whiting PF, Rutjes AWS, Westwood ME, Mallett S, Deeks JJ, Reitsma JB, et al. QUADAS-2. A revised tool for the quality assessment of diagnostic accuracy studies. Ann Intern Med. 2011;155(8);529-536. doi 10.7326/0003-4819-155-8-201110180- 00009. [39] Guyatt GH, Oxman AD, Vist GE, Kunz R, Falck-Ytter Y, Alonso-Coello P, et al. GRADE. An emerging consensus on rating quality of evidence and strength of recommendations. BMJ. 2008;336(7650);924-926. doi 10.1136/bmj.39489.470347.AD.

Supplementary material

Supplementary Table S1. Complete search strategies Source/platform Search date Coverage and limits Executable strategy MEDLINE via PubMed 9 Jul 2026; update 15 Jul 2026 Inception to search date; humans; English at screening (("circulating tumor DNA"[tiab] OR "circulating tumour DNA"[tiab] OR ctDNA[tiab] OR "cell-free DNA"[tiab] OR cfDNA[tiab] OR "liquid biopsy"[tiab]) AND ("molecular residual disease"[tiab] OR "minimal residual disease"[tiab] OR recurrence[tiab] OR relapse[tiab] OR surveillance[tiab] OR "multi-cancer early detection"[tiab] OR multicancer[tiab]) AND (cancer*[tiab] OR tumor*[tiab] OR tumour*[tiab] OR neoplasm*[tiab])) NOT (leukemia[tiab] OR lymphoma[tiab] OR myeloma[tiab]) Embase via Elsevier 9 Jul 2026 Inception to search date; humans; English at screening ('circulating tumor DNA'/exp OR 'cell free DNA'/exp OR 'liquid biopsy'/exp OR ctdna:ti,ab,kw OR cfdna:ti,ab,kw) AND ('molecular residual disease':ti,ab,kw OR 'minimal residual disease'/exp OR recurrence:ti,ab,kw OR relapse:ti,ab,kw OR surveillance:ti,ab,kw OR 'multicancer early detection':ti,ab,kw OR multicancer:ti,ab,kw) AND 'neoplasm'/exp NOT ('leukemia'/exp OR 'lymphoma'/exp OR 'multiple myeloma'/exp) Scopus 9 Jul 2026 Inception to search date; article or conference paper; English TITLE-ABS-KEY((ctdna OR "circulating tumor DNA" OR "circulating tumour DNA" OR cfdna OR "cell-free DNA" OR "liquid biopsy") AND ("molecular residual disease" OR "minimal residual disease" OR recurrence OR relapse OR surveillance OR "multi-cancer early detection" OR multicancer) AND (cancer OR tumor OR tumour OR neoplasm)) AND NOT TITLE-ABS- KEY(leukemia OR lymphoma OR myeloma) Web of Science Core Collection 9 Jul 2026 Inception to search date; English TS=((ctDNA OR "circulating tumor DNA" OR "circulating tumour DNA" OR cfDNA OR "cell-free DNA" OR "liquid biopsy") AND ("molecular residual disease" OR "minimal residual disease" OR recurrence OR relapse OR surveillance OR "multi-cancer early detection" OR multicancer) AND (cancer OR tumor OR tumour OR neoplasm)) NOT TS=(leukemia OR lymphoma OR myeloma) Cochrane CENTRAL 9 Jul 2026 Inception to search date (ctDNA OR "circulating tumor DNA" OR cfDNA OR "cell-free DNA" OR "liquid biopsy") AND ("molecular residual disease" OR "minimal residual disease" OR recurrence OR relapse OR "multi-cancer early detection" OR multicancer) AND (cancer OR tumor OR neoplasm) ClinicalTrials.gov 9 Jul 2026; status update 15 Jul 2026 All study types and statuses ("circulating tumor DNA" OR ctDNA OR "cell-free DNA" OR cfDNA OR "liquid biopsy") AND ("molecular residual disease" OR "minimal residual disease" OR recurrence OR relapse OR "multi-cancer early detection" OR multicancer) WHO ICTRP, ISRCTN, jRCT 9 Jul 2026 All statuses Equivalent combinations of ctDNA/cfDNA/liquid biopsy with MRD/recurrence/MCED terms, adapted to each interface

Supplementary Table S2. Full-text exclusion categories Reason for exclusion Number of reports Analytical validity or classifier development without longitudinal clinical outcome 3 Review, commentary, or editorial without original eligible data 2 Wrong population, including hematologic malignancy or non-screening population 2 No clinically interpretable outcome or no diagnostic resolution 2 Duplicate or secondary report without unique extractable data 1 A citation-level exclusion log should be retained with the review archive. Supplementary Table S3. Study-level data extraction Report Cancer/ domain Design Sample Assay Endpoint Main result Evidence role Garcia-Murillas 2015 [1] Breast Prospective cohort 55 Tumor-informed mutation tracking Relapse prediction ctDNA anticipated relapse Prognostic Abbosh 2017 [2] NSCLC Prospective translational cohort 100 Tumor-informed phylogenetic assay Relapse dynamics ctDNA tracked tumor evolution Prognostic Phallen 2017 [3] Pan-cancer Validation cohort 200+ Targeted ctDNA sequencing Early detection Technical feasibility Analytical context Cohen 2018 [4] Eight cancers Case-control validation 1,005 CancerSEEK multi-analyte Detection/localization Feasibility; no outcomes Diagnostic context Reinert 2019 [5] CRC Prospective cohort 130 Ultradeep tumor-informed sequencing Postsurgical recurrence Strong prognostic association Prognostic Christensen 2019 [6] Bladder Prospective cohort 68 Tumor-informed cfDNA Metastatic relapse Molecular lead time Prognostic Coombes 2019 [7] Breast Prospective cohort 49 Personalized ctDNA Metastatic recurrence ctDNA antedated recurrence Prognostic Cristiano 2019 [8] Pan-cancer Classifier validation 236 cancers plus controls Fragmentomics Cancer detection Technical validation Analytical context Liu/CCGA 2020 [9] Pan-cancer Validation study 6,689 Targeted methylation Sensitivity/specificity/ CSO High specificity; stagedependent sensitivity Diagnostic context Zviran 2020 [10] Pan-cancer Method validation multiple cohorts Genome-wide cfDNA integration Ultrasensitive monitoring Technical performance Analytical context Lennon/DETECT-A 2020 [11] MCED Prospective interventional 10,006 Multi-analyte test plus PET- CT Diagnostic pathway Feasible; mortality not assessed Pathway Chen 2020 [12] Pan-cancer Nested case-control 605 samples Methylation blood test Prediagnostic detection Signal before diagnosis; design indirect Analytical context Powles/IMvigor010 2021 [18] MIBC Exploratory RCT biomarker analysis 581 biomarkerevaluable Tumor-informed ctDNA Treatment interaction Hypothesisgenerating benefit in ctDNApositive Predictive exploratory Loupakis 2021 [19] Metastatic CRC resection Prospective cohort 112 Personalized ctDNA MRD and recurrence Strong recurrence stratification Prognostic Magbanua 2021 [20] Breast Prospective cohort 84 Tumor-informed ctDNA Response/survival ctDNA dynamics prognostic Prognostic

Report Cancer/ domain Design Sample Assay Endpoint Main result Evidence role Tie/DYNAMIC 2022 [14] Stage II colon Randomized strategy trial 455 Tumor-informed ctDNA Chemotherapy use; RFS 15% vs 28% chemothera py; noninferior 2-y RFS Actionable Tie/DYNAMIC follow-up 2025 [15] Stage II colon Extended RCT follow-up 455 Tumor-informed ctDNA 5-y RFS/OS RFS 88% vs 87%; OS 93.8% vs 93.3% Actionable Kotani/GALAXY 2023 [21] Resectable CRC Prospective observational 1,039 Tumor-informed ctDNA MRD; ACT interaction Strong prognostic association Prognostic/ predictive exploratory Klein 2021 [16] Pan-cancer Independent validation 4,077 Targeted methylation Sensitivity/specificity/ CSO High specificity; no outcome benefit Diagnostic context Schrag/ PATHFINDER 2023 [23] MCED Prospective pathway cohort 6,662 Targeted methylation MCED Diagnostic resolution Median 79 d Pathway Abbosh/TRACERx 2023 [13] NSCLC Prospective translational cohort 421 Tumor-informed ctDNA Dissemination/relapse Strong prognostic evidence Prognostic Kim 2023 [17] Pan-cancer Classifier study multiple cohorts Methylation/CNV/ fragmentation Detection/CSO Improved classifier performance Analytical context Nakamura/GALAXY 2024 [29] Resectable CRC Prospective observational expanded cohort Tumor-informed ctDNA DFS/OS MRD strongly associated with survival Prognostic Turner/c-TRAK TN 2023 [25] TNBC Prospective surveillance/interven tion 208 Tumor-informed ctDNA Detection and pembrolizumab 27.3% positive; limited intervention signal Actionability caution Tie/DYNAMIC-III 2025 [26] Stage III colon Randomized strategy trial 968 evaluable Tumor-informed ctDNA Risk-adjusted ACT Prognostic separation; escalation not superior Actionability uncertain Powles/IMvigor011 2025 [27] MIBC Phase III RCT 250 randomized Tumor-informed ctDNA DFS/OS HR 0.64 and 0.59 Actionable Bando/ALTAIR 2026 [28] Resected CRC Phase III RCT 243 Tumor-informed ctDNA DFS HR 0.79; primary endpoint negative Negative actionability Martin-Arana 2025 [22] Colon Prospective diagnostic/prognosti c localized cohort Tumor-agnostic plasma WES MRD sensitivity Promising; external validation needed Prognostic/ technical Nadauld/ PATHFINDER 2025 [24] MCED Prospective cohort analysis PATHFINDE R cohort Participant-reported outcomes Psychosocial impact Transient anxiety effects Harms

Supplementary Table S4. Domain-level risk-of-bias and applicability assessment Study or evidence group Tool Selection or randomization Deviations or confounding Missing data Outcome or index test Selective reporting Applicability Overall judgment DYNAMIC [14] RoB 2 Low Low Low Low Low Setting-specific ctDNA-guided strategy Low risk DYNAMIC 5-year follow-up [15] RoB 2 Low Low Some concerns Low Some concerns Setting-specific; exploratory ctDNA-clearance analyses Some concerns DYNAMIC-III [26] RoB 2 Low Some concerns Low Low Some concerns Complex escalation and de-escalation strategy Some concerns ALTAIR [28] RoB 2 Low Low Low Low Some concerns Heterogeneous disease stages and ctDNAassessment windows Some concerns IMvigor011 [27] RoB 2 Low Low Low Low Low Limited to biomarker-selected muscle-invasive bladder cancer and the evaluated assay-treatment combination Low risk GALAXY/CIRCULATE-Japan [21,29] QUIPS and ROBINS- I Moderate Serious confounding Moderate Low Moderate Observational estimates of adjuvant-treatment benefit Moderate to serious IMvigor010 ctDNA analysis [18] ROBINS-I Moderate Serious Moderate Low Serious Exploratory biomarker subgroup analysis not prospectively designed to establish treatment effectiveness Serious c-TRAK TN [25] ROBINS-I Moderate Serious Moderate Low Moderate Small treated subgroup and frequent radiologically detectable disease at ctDNA detection Serious TRACERx [2,13] QUIPS Moderate Moderate Moderate Low Moderate Research cohort with assay-timing and treatmentactionability limitations Moderate Breast cancer MRD cohorts [1,7,20] QUIPS Moderate Moderate Moderate Low Moderate Small cohorts, heterogeneous sampling schedules and no randomized intervention test Moderate DETECT-A [11] QUADAS-2 Low Not applicable Moderate Moderate verification concerns Moderate Restricted population and non-randomized diagnostic pathway without mortality assessment Moderate PATHFINDER [23-24] QUADAS-2 Low Not applicable Moderate Moderate verification concerns Moderate No randomized comparator and incomplete longterm verification of participants with negative

results

Moderate PATHFINDER 2 [32] QUADAS-2 Unclear Not applicable Unclear Low for reported diagnosticperformance measures Unclear Conference abstract with incomplete peer-reviewed methodological reporting Some concerns NHS-Galleri [30] Preliminary RoB 2 Low Low Unclear Low Unclear Conference-level reporting and immature mortality follow-up Some concerns CCGA/Klein validation studies [9,16] QUADAS-2 Moderate Not applicable Low Low Moderate Case-control and spectrum effects limit applicability to an intended-use screening population Moderate Kim 2023 [17] QUADAS-2 High Not applicable Unclear Moderate Moderate Classifier-development evidence without adequate external validation in an intended-use screening population High risk

Supplementary Table S5. GRADE evidence profile Outcome Studies Risk of bias Inconsistency Indirectness Imprecision Publication bias Certainty Reduced ACT without RFS loss in stage II colon [14-15] 1 RCT + follow-up Not serious Not applicable Serious: single setting/strategy Serious: noninferiority margin Undetected Moderate DFS/OS benefit in ctDNA-positive MIBC [27] 1 phase III RCT Not serious Not applicable Not serious for studied setting Not serious Undetected High Stage III colon ctDNA-guided strategy [26] 1 randomized strategy trial Some concerns Not applicable Serious Serious Undetected Low-moderate Treatment at molecular recurrence [25,28] 2 intervention studies Serious Serious Serious Serious Possible Low Post-treatment MRD prognosis [1- 2,5-7,13,18-22,29] Multiple cohorts Moderate Not serious directionally Moderate Variable Possible Moderate MCED diagnostic feasibility [11,23- 24,32] Prospective pathways Moderate Moderate Serious for mortality benefit Not serious for performance Possible industry sponsorship Moderate for feasibility MCED latestage/mortality benefit [30-31] 1 RCT abstract + ongoing trial Some concerns Not applicable Serious Serious Possible Very low-low Supplementary Table S6. Completed PRISMA 2020 checklist Item Topic Requirement addressed Location in manuscript 1 Title Identifies the report as a systematic review Title page 2 Abstract Structured summary with objectives, methods, results, and conclusions Abstract 3 Rationale Rationale in context of existing knowledge Introduction 4 Objectives Explicit review objective and domains Introduction 5 Eligibility criteria Inclusion/exclusion criteria and grouping for synthesis Section 2.1 6 Information sources Databases, registries, and update search with dates Section 2.2; Table S1 7 Search strategy Full strategies for all sources Table S1 8 Selection process Independent screening and conflict resolution Section 2.3 9 Data collection process Independent extraction process Section 2.4 10a Data items - outcomes Outcome definitions Section 2.4 10b Data items - other variables Study and assay variables Section 2.4; Table S3 11 Risk of bias Tools and process by design Section 2.5; Table S4 12 Effect measures HR, RR, proportions, medians, and CIs Tables 2 and S3 13a Synthesis eligibility Separate MRD, prognosis, MCED, and harms groups Section 2.6 13b Data preparation Narrative harmonization and no imputation Section 2.6 13c Presentation methods Structured tables and narrative synthesis Results; Tables 1-4 13d Synthesis methods Meta-analysis not performed because of heterogeneity Section 2.6 13e Heterogeneity exploration Clinical sources discussed narratively Results; Discussion 13f Sensitivity analyses Not performed; not applicable without meta-analysis Section 2.6 14 Reporting bias assessment Considered in GRADE Section 2.5; Table S5 15 Certainty assessment GRADE domains and outcome-level ratings Section 2.5; Table S5 16a Study selection results Flow counts and diagram Section 3.1; Figure 1 16b Excluded studies Category-level reasons; citation log to archive Table S2 17 Study characteristics Characteristics and full extraction Table 1; Table S3 18 Risk of bias results Overall and domain-level judgments Table 3; Table S4

Item Topic Requirement addressed Location in manuscript 19 Individual study results Sample sizes and numerical estimates Table 2; Table S3 20a Synthesis results - contributing studies Studies grouped by clinical domain Sections 3.3-3.6 20b Synthesis results - statistical synthesis Not applicable; no meta-analysis Section 2.6 20c Synthesis results - heterogeneity Narrative explanation Results; Discussion 20d Synthesis results - sensitivity Not applicable Section 2.6 21 Reporting biases Considered qualitatively and in GRADE Section 5; Table S5 22 Certainty of evidence Outcome-level certainty ratings Section 5; Tables 4 and S5 23a Discussion - interpretation Interpretation in relation to evidence Section 6 23b Discussion - evidence limitations Risk of bias, indirectness, heterogeneity Sections 5-7 23c Discussion - review limitations Review-process limitations Section 7 23d Discussion - implications Practice and research implications Sections 6 and 8 24a Registration Not prospectively registered Section 2.1 24b Protocol No public protocol available Section 2.1 24c Amendments No protocol amendments reported Section 2.1 25 Support Funding information was not provided in the source manuscript. Funding 26 Competing interests Conflict-of-interest information was not provided in the source manuscript. Conflicts of interest 27 Availability of data/materials Study-level data and supplementary materials provided Data availability; Tables S1-S7 Supplementary Table S7. Data extraction template Domain Variables Instructions Identification Author, year, journal, DOI, PMID, registry identifier Record exactly as reported Population Cancer type, stage, setting, age, inclusion/exclusion criteria, sample size Separate screened, enrolled, randomized, and analyzed populations Study design RCT, intervention cohort, prognostic cohort, diagnostic pathway, classifier validation Specify prospective/retrospective and single-/multicenter Assay ctDNA/cfDNA target, tumor-informed vs tumor-agnostic, platform, threshold, blinding Record version and timing of assay Sampling Postoperative window, serial schedule, preanalytical handling Record time from treatment to blood draw Clinical algorithm Action after positive/negative test, comparator, confirmatory work-up Required for actionability assessment Outcomes RFS, DFS, OS, recurrence, treatment exposure, toxicity, PPV, specificity, sensitivity, resolution, harms Use study definitions and time horizons Effect estimates Events, proportions, HR/RR/OR, 95% CI, P value Extract adjusted and unadjusted estimates separately Risk of bias Tool, domain judgments, support for judgment RoB 2, QUIPS, ROBINS-I, or QUADAS-2 Certainty and applicability GRADE domains, clinical applicability, funding, conflicts Separate prognostic validity from clinical utility