This study is an integrative review conducted according to the methodological framework proposed by Whittemore and Knafl [5], informed by PRISMA 2020 reporting conventions [6] adapted to the integrative-review design, and conducted throughout all stages by two reviewers (A.C., A.P.B.G.). The integrative review design was selected for its capacity to synthesize diverse types of evidence—including experimental, observational, and theoretical studies—thereby enabling a comprehensive appraisal of the multi-paradigmatic literature on ctDNA in GU oncology. This design permits inclusion of narrative and integrative reviews alongside primary studies, a scope explicitly excluded from formal systematic reviews, and is reported as such throughout this manuscript; no PROSPERO registration was sought, consistent with standard practice for integrative reviews, which by definition lack a single a priori quantitative protocol amenable to prospective registration.

A systematic search was performed across four primary bibliographic databases: PubMed/MEDLINE, Cochrane Central Register of Controlled Trials (CENTRAL), Embase, and Web of Science. The search period spanned January 2020 to April 2026, a time frame selected to capture the contemporary era of ctDNA clinical translation while excluding a preparatory literature base. Search strategies were constructed using combinations of Medical Subject Headings (MeSH) and free-text terms, organized around three conceptual domains: 1) the analyte of interest (“circulating tumor DNA,” “ctDNA,” “liquid biopsy,” “cell-free DNA”); 2) the tumor types of interest (“prostate cancer,” “urothelial carcinoma,” “bladder cancer,” “renal cell carcinoma,” “Genitourinary cancer”); and 3) the clinical outcomes of interest (“treatment selection,” “predictive biomarker,” “resistance monitoring,” “treatment response,” “minimal residual disease”). Boolean operators AND/OR were applied to combine term clusters. The integrity and contextual classification of retrieved references were subsequently verified using SCITE.ai, employed exclusively as a bibliometric validation tool and not as a primary literature source. The four databases searched (PubMed/MEDLINE, CENTRAL, Embase, Web of Science) are indexed, non-predatory sources; no additional exclusion filter for predatory journals was required beyond restricting retrieval to these curated indices. The January 2020 start date was selected to capture the period following the first clinically oriented ctDNA companion-diagnostic validation studies in GU oncology, avoiding an earlier exploratory-technology literature base of limited contemporary clinical relevance; the full per-database search strings are provided in Supplementary material. All classifications and inclusion/exclusion judgments derived from SCITE.ai outputs were reviewed and confirmed manually by both reviewers before incorporation; the tool was used solely as a secondary verification aid and did not replace reviewer judgment at any stage, in accordance with the journal’s policy on artificial intelligence (AI)-assisted tools.

Studies were eligible for inclusion if they met the following criteria: 1) published in English in peer-reviewed journals or as preprints of sufficient methodological quality between January 2020 and April 2026; 2) reported original or synthesized data on ctDNA as a biomarker in PC, UC (bladder or upper tract), or RCC; 3) addressed at least one of the primary outcomes—treatment selection, treatment response prediction, or resistance monitoring; and 4) provided sufficient methodological or quantitative data to allow extraction of a meaningful synthesis. Studies were excluded if they addressed exclusively non-GU cancer types, focused on circulating tumor cells (CTCs) or other liquid biopsy analytes without ctDNA-specific data, or were conference abstracts without associated full-text publication.

The selection process followed a two-stage approach conducted independently and in duplicate by both reviewers (A.C., A.P.B.G.). In the first stage, titles and abstracts of all retrieved records were screened independently by each reviewer against the eligibility criteria; discrepancies were resolved by consensus discussion. In the second stage, full texts of candidate studies were retrieved and independently assessed for final inclusion by both reviewers, with disagreements again resolved by consensus; no third-party arbitration was required. Data extraction was performed independently by both reviewers using a standardized extraction form covering study design, cancer type, patient population characteristics, ctDNA detection methodology, primary and secondary outcomes, and key quantitative findings; extracted data were cross-checked between reviewers, and discrepancies were resolved by re-evaluation against the source publication and the eligibility criteria. Full documentation of the screening and extraction process, including reviewer-level decisions at each stage, is maintained and available upon request.

Consistent with integrative review methodology, no single risk-of-bias instrument [e.g., the Cochrane Risk of Bias tool (RoB 2), the Risk Of Bias In Non-randomised Studies of Interventions tool (ROBINS-I), and A MeaSurement Tool to Assess systematic Reviews (AMSTAR-2)] was applied uniformly across the heterogeneous study designs included in this synthesis, as such tools are designed for single study-design strata and are not intended to appraise a mixed corpus of primary studies, case series, and narrative or systematic reviews within a single scale. Instead, certainty of evidence was assessed narratively, by cancer-site stratum and clinical domain (prognosis/MRD, treatment monitoring, and predictive/therapy-selection use), considering study design, sample size, prospective versus retrospective ascertainment, and assay maturity; quantitative claims were anchored preferentially in primary cohort and trial data rather than in review-derived estimates, and heterogeneous assays or cancer sites were not pooled into a single quantitative effect. This narrative certainty appraisal is presented alongside the stratified synthesis in the Results section.

Database searches across PubMed/MEDLINE, CENTRAL, Embase, and Web of Science for the period January 2020 to April 2026 yielded records that, after deduplication, comprised 118 unique records, screened independently by both reviewers at title and abstract level. Following this stage, 39 records were retrieved and assessed in full text (retrieval complete, 39/39); 25 were excluded, most commonly for wrong population (n = 13; not a GU-dedicated synthesis or population) or wrong intervention/analyte (n = 12; analyte other than ctDNA, e.g., AR-V7 assessed exclusively via CTCs or tissue). The final synthesis comprised 15 studies: 14 identified through the database search and one additional study identified outside the database search (Figure 1).

A total of 15 studies met all inclusion criteria and formed the basis of this integrative synthesis. The corpus encompassed original research studies (n = 8; 53.3%), narrative reviews (n = 5; 33.3%), a systematic review with qualitative synthesis (n = 1; 6.7%), and a case series (n = 1; 6.7%). Studies were stratified by cancer site: prostate [number of studies per stratum (k) = 4], urothelial/bladder (k = 4), renal (k = 4; see in ctDNA in RCC: emerging evidence and current limitations section), upper tract UC (UTUC; k = 1), and mixed GU cohorts (k = 2). Detailed characteristics of all included studies are presented in Table 1.

The most clinically advanced application of ctDNA in GU oncology involves its use as a predictive and monitoring biomarker in metastatic CRPC (mCRPC). PC presents unique challenges for tissue-based genomic profiling: the predominance of bone metastases renders repeat biopsy both technically difficult and clinically hazardous, while tumor heterogeneity between metastatic deposits complicates the representativeness of any single specimen. ctDNA addresses these limitations by providing a composite, systemic view of the evolving genomic landscape [3].

A pivotal contribution to the field was provided by Sweeney et al. [9], who analyzed ctDNA TF data from IMbassador250, a phase III trial evaluating enzalutamide with or without atezolizumab in post-abiraterone mCRPC patients. In 494 evaluable participants, baseline ctDNA TF detection and its dynamics through cycle 3, day 1 (C3D1) were evaluated as biomarkers of clinical outcome. Patients with detectable ctDNA TF at C3D1, irrespective of baseline status, demonstrated significantly shorter radiographic progression-free survival (rPFS) and overall survival (OS) than those without ctDNA detection. Critically, when ctDNA TF and prostate-specific antigen (PSA) response were discordant at C3D1, patients who were ctDNA-undetected but PSA-non-reduced achieved superior OS [median OS (mOS) 22.1 months] relative to those who were ctDNA-detected despite PSA reduction (mOS 16.0 months; P < 0.001). This finding established ctDNA TF as an independent and complementary biomarker to PSA in assessing treatment response, with the capacity to identify a clinically distinct patient subgroup inadequately characterized by conventional PSA kinetics alone.

Circulating markers of the androgen receptor (AR) pathway offer a mechanistically grounded, ctDNA/cfDNA-based approach to treatment selection in CRPC. AR-V7, a constitutively active splice variant lacking the ligand-binding domain of full-length AR, confers primary resistance to the AR signaling inhibitors (ARSIs) enzalutamide and abiraterone; detection of this pathway via cell-free DNA and exosomal RNA—rather than via CTCs or tissue, the more classical detection routes—is the focus of this ctDNA-specific synthesis. Del Re et al. [7] demonstrated in a retrospective cohort of 84 chemotherapy-naive mCRPC patients starting first-line abiraterone or enzalutamide that AR copy-number gain in cfDNA (detected in 14.3% of patients) and AR-V7 detection in circulating exosomal RNA (35.7%) each independently predicted primary ARSI resistance; patients with both aberrations had markedly shorter progression-free survival (PFS; 3.8 vs. 28.0 months) and OS (6.1 months vs. not reached; P < 0.0001 for both) than patients with neither. Complementary cfDNA-based approaches to treatment selection in mCRPC include genome-wide aneuploidy (GWA) scoring: in a prospective cohort, a high cell-free DNA aneuploidy score at baseline and at an early on-treatment timepoint was associated with poorer response to ARSIs, and the early-timepoint score also predicted poorer response to taxanes [8].

AR-V7 status can also be assessed via CTC-based platforms, most notably the Epic Sciences enumeration assay, independently validated as a prognostic biomarker of OS in men with progressing mCRPC starting an ARSI, across a discovery cohort (Memorial Sloan Kettering, n = 171) and the multicenter PROPHECY validation trial (n = 107) [20]. CTC- and cfDNA/exosome-based approaches to AR-pathway assessment are complementary rather than competing technologies: CTC platforms offer single-cell resolution and established clinical validation in this specific application, whereas cfDNA/exosome-based assays offer higher throughput, do not depend on CTC capture efficiency, and can be co-analyzed alongside other ctDNA-based biomarkers (TF, DNA damage repair alterations) from a single blood draw. This CTC-based literature is not itself synthesized in this review, whose scope is specifically ctDNA/cfDNA, but is noted here as the relevant comparator technology.

In MIBC, ctDNA has been validated across the perioperative and metastatic continuum as a biomarker of MRD, treatment response, and relapse prediction. The biological plausibility is well-established: ctDNA concentrations correlate with tumor burden, and the short plasma half-life of these fragments allows rapid reflection of changes induced by surgical or systemic interventions [2, 12].

Lindskrog et al. [10] reported landmark longitudinal ctDNA data from two independent MIBC cohorts with median follow-up exceeding five years. In 68 neoadjuvant chemotherapy (NAC)-treated patients, post-cystectomy ctDNA assessment identified metastatic relapse with 94% sensitivity and 98% specificity. ctDNA dynamics during NAC were independently associated with patient outcomes when adjusted for pathological downstaging [hazard ratio (HR) 4.7; P = 0.029]. In a separate NAC-naïve cohort of 102 patients, preoperative ctDNA positivity predicted significantly inferior recurrence-free survival (RFS; HR 3.4; P = 0.0005) and was an even stronger post-cystectomy prognostic factor (HR 17.8; P = 0.0002). Biologically, baseline ctDNA positivity was associated with the basal/squamous molecular subtype and enrichment of epithelial-to-mesenchymal transition gene sets. These findings collectively establish ctDNA as a superior prognostic tool to conventional pathological staging in MIBC.

The predictive utility of ctDNA in guiding adjuvant immunotherapy decisions in MIBC was substantiated by data from the IMvigor010 trial [21]. A post-hoc ctDNA analysis demonstrated that patients with detectable ctDNA after radical cystectomy derived meaningful disease-free survival benefit from adjuvant atezolizumab, whereas ctDNA-negative patients did not show equivalent advantage [22]. Moreover, ctDNA clearance during treatment occurred more frequently with atezolizumab than with observation (18% vs. 4%), consistent with an on-treatment effect on molecular residual disease [22]. These data position ctDNA as a predictive, not merely prognostic, instrument for adjuvant therapy selection in MIBC. This effect was formally confirmed in the trial’s ctDNA-stratified OS analysis [23], which reported an OS HR of 0.59 (95% CI 0.42–0.83) favoring atezolizumab specifically among ctDNA-positive patients, establishing IMvigor010 ctDNA status as a predictive, not merely prognostic, biomarker for adjuvant immunotherapy benefit in this population.

Quantitative claims regarding the ABACUS and NABUCCO trials are not included in this synthesis because they could not be traced to a verifiable primary source (see Discussion, limitations).

Nawaf et al. [11] reviewed ctDNA-based MRD detection in the context of adjuvant therapy decision-making for MIBC. TOMBOLA (NCT04138628) and IMvigor011 are ongoing prospective trials designed to test whether withholding or administering adjuvant atezolizumab according to postoperative ctDNA status improves outcomes relative to standard management; mature efficacy results were not yet available in the sources captured by this review’s search (through April 2026) and are therefore not quoted here pending peer-reviewed publication. Crocetto et al. [13] provided a complementary systematic synthesis of liquid biopsy applications in bladder cancer, contextualizing ctDNA within a broader multi-analyte framework that includes CTCs and urinary exosomes; because this synthesis spans multiple analytes beyond ctDNA, its contribution to the present review is limited to the ctDNA-specific findings it reports.

The pilot study by Jang et al. [19] extended these observations to the metastatic GU setting, prospectively enrolling 20 patients with advanced GU malignancies—15 with RCC, 4 urothelial, and 1 prostate—receiving immune checkpoint inhibitors (ICIs). Using the tumor-informed Signatera multiplex polymerase chain reaction (mPCR)-based next-generation sequencing (NGS) assay, longitudinal ctDNA analysis demonstrated an overall concordance of 83% (15 of 18 evaluable patients) between ctDNA dynamics and radiographic response. The overall objective response rate was 70%. Among three discordant cases, two involved central nervous system metastases, a known blind spot for systemic ctDNA assays, underscoring the importance of integrating ctDNA with anatomic imaging. This study provides proof-of-concept for ctDNA-guided ICI monitoring across GU cancer types.

Beyond MRD and prognosis, ctDNA-based molecular subtyping—including detection of FGFR3 alterations—has been proposed as a route to non-invasive treatment stratification in UC, linking luminal-papillary tumors to FGFR-directed therapy and basal/squamous tumors to immunotherapy or chemotherapy [13]. However, no study meeting this review’s ctDNA-specific eligibility criteria reported a quantitatively verifiable analysis of ctDNA-based FGFR3 detection performance or of erdafitinib response stratified by ctDNA FGFR3 status; this application is therefore noted as an emerging direction rather than as established evidence, and specific response-rate figures are not cited here, pending identification of a verifiable primary source.

Among the three principal GU cancer types, RCC represents the domain where ctDNA evidence is least mature but where clinical need is most acute. Liquid biopsy offers a particularly compelling solution in RCC given the challenges of percutaneous renal mass biopsy in certain anatomical locations. The high prevalence of morphologically heterogeneous tumors and the frequent necessity of distinguishing clear cell from non-clear cell RCC subtypes for systemic therapy selection also contribute to the utility of liquid biopsy [17].

Clear cell RCC is characterized by VHL gene inactivation in up to 90% of cases, often accompanied by concurrent alterations in PBRM1, SETD2, BAP1, and KDM5C. These mutations generate tumor-specific ctDNA fragments detectable in plasma, albeit at levels typically lower than those observed in prostate or UCs, partly because RCC frequently lacks the high tumor mutational burden seen in urothelial malignancies [15, 17]. PBRM1 mutations have been proposed as predictive biomarkers of benefit from ICI therapy in metastatic RCC (mRCC), based on retrospective analyses of tumor tissue; whether ctDNA-based PBRM1 detection achieves equivalent predictive accuracy awaits prospective validation.

Del Re et al. [14] quantified plasma cfDNA amount and tumor protein p53 (TP53) mutational status in 48 patients with metastatic clear cell RCC (mccRCC) receiving first-line immunotherapy (ipilimumab-nivolumab) or VEGFR-tyrosine kinase inhibitors: cfDNA concentration above 0.883 ng/µL was associated with shorter PFS (7.23 vs. not-reached months; P = 0.001) and OS (23.9 vs. not-reached months; P = 0.008), and TP53-mutant cfDNA independently predicted shorter PFS (P = 0.04), with the combination of high cfDNA and TP53 mutation identifying the poorest-prognosis subgroup. In a considerably larger, more recent nationwide genomic profiling study, Kato et al. [15] analyzed serial ctDNA in 124 patients with mRCC enrolled in the SCRUM-Japan MONSTAR-SCREEN project: baseline TF below 1.2% was associated with significantly better first-line PFS (HR 0.467, 95% CI 0.229–0.979), BAP1 mutational status at baseline predicted poorer OS (HR 0.487, 95% CI 0.322–0.736), and 46.8% of patients acquired new ctDNA mutations at disease progression, a finding linked to shorter time to progression. This larger prospective cohort strengthens the primary-study basis for ctDNA prognostic use in RCC, which had previously relied on smaller retrospective cohorts and case series. Alam et al. [16] described a case series of six patients with diverse renal malignancies—including clear cell RCC, papillary RCC, translocation RCC, malignant angiomyolipoma, and squamous cell carcinoma of the renal pelvis—in whom serial ctDNA monitoring informed clinical decisions in adjuvant and metastatic settings in all six cases; the authors explicitly noted lower assay sensitivity and negative predictive value in kidney cancers relative to other tumor types, consistent with the lower tumor mutational burden and lower ctDNA shedding characteristic of RCC noted throughout this synthesis.

Jikuya et al. [17] provided a collaborative narrative synthesis of translational RCC research, situating liquid biopsy—including ctDNA, cell-free DNA quantity, and extracellular vesicles—within the broader landscape of RCC biomarker development, and explicitly noting that ctDNA detection rates in RCC remain lower than in other solid tumors under older digital PCR-based assays, with NGS platforms improving sensitivity for low tumor-fraction detection. Gerke et al. [2] similarly identified RCC as the GU cancer type with the greatest unmet need for ctDNA biomarker validation. Taken together, the RCC evidence base—now including one adequately powered primary cohort (Kato et al. [15], n = 124) alongside smaller original studies and case-level data—supports cfDNA quantity, TP53 mutation, TF, and BAP1 mutational status as prognostic and monitoring biomarkers, but does not yet support routine ctDNA-guided treatment selection in RCC outside investigational settings; alternative and complementary biomarkers under active investigation, including DNA methylation-based predictors of systemic-therapy sensitivity, are addressed in the Discussion.

UTUC is addressed here as a distinct stratum. Nakano et al. [18] prospectively studied 50 patients with untreated UTUC using perioperative plasma ctDNA (pan-cancer NGS panel) with matched buffy-coat sequencing for clonal hematopoiesis of indeterminate potential (CHIP) exclusion. ctDNA was detected in 23/50 (46%) of the overall cohort and 17/43 (40%) of localized cases; preoperative ctDNA fraction greater than 2% was the only independent preoperative risk factor for worse RFS among the factors assessed, and early postoperative ctDNA positivity was associated with worse RFS, consistent with MRD. Critically, 19% of detected TP53 mutations were false positives attributable to CHIP, underscoring the magnitude of this analytic confounder in a GU-specific population and reinforcing the buffy-coat-matched sequencing approach as good practice for ctDNA-based UTUC studies. Serial ctDNA also tracked systemic therapy response in metastatic UTUC in this cohort. This single-cohort evidence base is mechanistically informative but requires prospective, multi-institutional validation before ctDNA can inform perioperative treatment decisions in UTUC.

The clinical deployment of ctDNA biomarkers requires robust, standardized, and analytically validated assay platforms. Three principal methodological approaches dominate the contemporary landscape. Digital droplet polymerase chain reaction (ddPCR) offers high sensitivity for the detection of predetermined mutations but is limited in scope to a restricted set of predefined targets. NGS panels, including commercially available platforms such as Guardant360 and FoundationOne Liquid CDx, provide broader genomic coverage and are capable of detecting multiple classes of genomic alterations—point mutations, copy number variations, structural rearrangements, and microsatellite instability—simultaneously. Tumor-informed mPCR-NGS assays, exemplified by Signatera, are designed around patient-specific tumor mutation profiles derived from tissue sequencing and achieve the highest sensitivity for MRD detection in the postoperative setting [10, 19].

A critical technical challenge across all platforms is the confounding effect of CHIP. In this case, age-related somatic mutations in hematopoietic stem cells generate ctDNA-like fragments in plasma that are indistinguishable from tumor-derived signals without matched germline sequencing. Failure to account for CHIP can introduce false-positive findings, particularly in older patients—a demographic that constitutes the majority of GU cancer cases. Pre-analytical standardization, including cell-free DNA collection tube protocols, centrifugation parameters, and storage conditions, also substantially influences ctDNA yield and quality, complicating cross-laboratory comparisons [3]. In UTUC specifically, CHIP accounted for 19% of detected TP53 mutations misclassified as tumor-derived in a dedicated perioperative cohort, illustrating the magnitude of this analytic risk in a GU-specific population when buffy-coat-matched sequencing is not performed [18].

Urine offers an alternative and complementary liquid biopsy matrix in GU oncology, particularly for bladder and urothelial cancers where direct contact between the tumor and the urinary tract generates high concentrations of tumor-derived DNA. TERT promoter mutations, detectable in urine ctDNA with high specificity for bladder cancer, represent a validated molecular marker for surveillance. The complementarity between plasma and urine ctDNA matrices enhances overall biomarker sensitivity, particularly in early-stage and localized disease where systemic ctDNA shedding is low [13].