Clinical Validation of a Multiplex Urine Biomarker Assay for Surveillance of Recurrent Bladder Cancer

Abstract

Purpose:

More than 50% of patients with non–muscle-invasive bladder cancer experience recurrence, requiring lifelong surveillance with repeated cystoscopy. Given the invasive nature and cost of cystoscopy, accurate noninvasive tools are needed to support risk-adapted monitoring. We evaluated the ability of Oncuria-Monitor, a multiplex urine biomarker assay, to detect recurrent bladder cancer during surveillance.

Patients and Methods:

Between February 2017 and August 2020, six medical centers in the United States and Japan prospectively enrolled 300 patients with a history of bladder cancer, generating 1,248 serial urine samples. Participants were divided into training and validation cohorts. At each surveillance visit over 2 years, urine samples were analyzed in a blinded manner using Oncuria-Monitor and BladderChek, alongside urine cytology. Test performance was compared with cystoscopy and histopathology-confirmed recurrence.

Results:

Recurrent bladder cancer was identified in 31% (93/300) of participants, with 143 total recurrences during follow-up, including 90 tumors in the validation cohort. In the validation cohort, Oncuria-Monitor achieved a sensitivity of 85.6% [95% confidence interval (CI), 78.1%–92.2%] and a negative predictive value (NPV) of 93% (95% CI, 89.2%–96.4%). In comparison, BladderChek demonstrated a sensitivity of 20.0% and an NPV of 88%, whereas urine cytology showed a sensitivity of 36.9% and an NPV of 91.6%. The number needed to evaluate to detect one recurrence was 3 for both cystoscopy and Oncuria-Monitor, compared with 15 for BladderChek and 8 for cytology.

Conclusions:

In this large prospective longitudinal study, Oncuria-Monitor demonstrated clinically actionable performance, enabling a rule-out strategy that could safely reduce cystoscopy in approximately 25% of surveillance visits. These findings support a paradigm shift toward biomarker-guided, risk-adapted surveillance in bladder cancer that reduces unnecessary invasive procedures while maintaining oncologic safety.

Translational Relevance.

Non–muscle-invasive bladder cancer remains a major clinical challenge due to its high recurrence rate and reliance on repeated cystoscopy (an invasive and resource-intensive procedure) for surveillance. There is a critical need for noninvasive biomarkers that enable early detection while supporting risk-adapted monitoring strategies. In this large prospective, longitudinal study of 300 patients with more than 1,200 serial samples collected across 2 years, we demonstrate that a multiplex urine-based immunoassay (Oncuria-Monitor) achieves substantially higher sensitivity and negative predictive value compared with standard noninvasive tests. Importantly, these performance characteristics support a rule-out strategy that could safely reduce the frequency of cystoscopy in approximately 25% of surveillance visits. These findings provide clinical evidence for a paradigm shift from uniform, invasive surveillance toward biomarker-guided, risk-adapted monitoring. More broadly, this work highlights the potential for multiplex biomarker platforms to reduce procedural burden while maintaining oncologic safety, an approach that may be generalizable to surveillance strategies across other malignancies.

Introduction

Approximately 80% of newly diagnosed bladder cancers are non–muscle-invasive bladder cancer (NMIBC; ref. 1). Although NMIBC is rarely fatal initially, there is a high rate of both disease recurrence (50%–70%) and disease progression to muscle-invasive bladder cancer (MIBC) over time (10%–30%; refs. 1, 2). An estimated 700,000 Americans are living with bladder cancer, making it one of the most prevalent cancers in the United States, particularly among men (3). Due to high progression and recurrence rates, guidelines recommend patient monitoring after initial diagnosis and treatment using a risk-based regimen that prescribes surveillance as often as every 3 to 6 months for 4 years and at least annually thereafter (1, 4). Current surveillance relies on the gold standard method of cystoscopy to visually evaluate the bladder lining at each follow-up visit.

This surveillance paradigm represents one of the most intensive and costly follow-up regimens in oncology, with bladder cancer carrying the highest lifetime per-patient management cost of any malignancy. The invasive nature of cystoscopy contributes to patient discomfort, procedural risk, and reduced adherence to recommended surveillance schedules (5, 6). Furthermore, a combination of methods (cystoscopy and urinary cytology) is frequently used to monitor patients at risk of recurrent bladder cancer because no single procedure is 100% sensitive. Cytologic evaluation of urine has low sensitivity (35%–50%) for detecting bladder cancer (7, 8). Consequently, there remains a critical unmet clinical need for highly sensitive urine-based assays capable of reliably excluding recurrent disease and enabling risk-adapted surveillance strategies. Such tests could reduce dependence on invasive cystoscopy while maintaining early detection of clinically significant recurrence. Recurrent bladder cancer is generally more difficult to detect than de novo tumors, largely because they tend to be smaller and flatter and shed fewer detectable biomarkers, often arising within a treatment-altered or chronically inflamed urothelium.

Single-analyte protein biomarker assays (e.g., BladderChek and BTAstat) are limited because not all bladder tumors, or even those in one category of lesion (e.g., NMIBC or low-grade), will harbor the single molecular change the assay detects (9–12), and these tests can produce false-positive results arising from common benign conditions (13, 14). Accordingly, the identification of a single molecular biomarker to aid bladder cancer clinical evaluation has proven elusive. Investigators have begun to identify diagnostic biomarker panels or signatures to aid in bladder cancer clinical evaluation. A rigorous methodologic approach was deployed to derive and validate a unique bladder cancer–associated diagnostic signature and to create and validate a multiplex immunoassay (Oncuria) to detect this bladder cancer signature and calculate a cancer risk score (15–29).

Herein, we report the prospective clinical validation of Oncuria-Monitor, a multiplex urine-based immunoassay designed to detect a bladder cancer–associated protein signature in patients undergoing surveillance. This study, the largest to date in this space, evaluates whether multiplex biomarker testing can improve the detection of recurrent disease and inform more efficient surveillance strategies, representing a critical step toward noninvasive precision monitoring of bladder cancer.

Patients and Methods

Study design

From February 2017 to August 2020, we enrolled participants in this prospective longitudinal study at six academic and private practice settings (five in the United States and one in Japan; Fig. 1; Supplementary Table S1), Clinicaltrials.gov NCT03193515. All relevant Institutional Review Boards and independent ethics committees approved the study, which was performed in accordance with the International Conference on Harmonisation Good Clinical Practice Guideline and the principles of the Declaration of Helsinki (30). Participants provided written informed consent before participation. The trial followed the STARD 2015 reporting guidelines (31).

Figure 1.

Flow diagram of study enrollment.

Study population

Participants were adults (≥18 years old) with an NMIBC diagnosis within the past 24 months and were willing to adhere to the standard American Urological Association (AUA) risk-based surveillance regimen (e.g., low risk seen every 6–12 months for 2 years and intermediate/high risk seen every 3 months for 2 years; ref. 4). Participants may have been treated with adjuvant intravesical therapy at the investigator’s discretion. Exclusion criteria were as follows: (i) radical cystectomy; (ii) previous cancer (excluding bladder, basal, and squamous cell skin cancer) within 3 years; (iii) known active (i.e., current or ongoing) urinary tract infection or urinary retention; (iv) active stone disease (renal or bladder) or renal insufficiency (creatinine ≥2 mg/dL); (v) ureteral stents, nephrostomy tubes, or bowel interposition; (vi) recent genitourinary instrumentation (within 10 days of scheduled cystoscopy); and (vii) unable/unwilling to provide written informed consent.

Clinical procedures

Participants were required to provide a voided urine sample and undergo cystoscopy (baseline) within 30 days after providing informed consent, with subsequent urine samples collected over the ensuing 2-year follow-up period. All voided urine samples were collected before cystoscopy and were refrigerated within 30 minutes. No dietary or fluid restrictions were required. A portion of the urine was sent to the local clinical laboratory for urinalysis, whereas an aliquot was available for Oncuria-Monitor and BladderChek testing. In addition, cytologic evaluation (Paris Classification) of voided urine samples or bladder barbotage specimens was performed and documented according to the standard clinical practice of each participating site. Cytology interpretation was conducted either locally at the participating institution or through affiliated/reference cytopathology laboratories routinely utilized by the site. Urinary cytology results were available for 1,228 samples, with 16 deemed indeterminate due to insufficient cellularity. A positive cytologic result was defined as one in which malignant (n = 46) or suspicious (n = 25) cells were present, whereas a negative cytologic result was defined as one in which no malignant (n = 1,004) or atypical (n = 137) cells were present. Physicians were required to describe the presence, size, and multiplicity of any tumors or abnormalities noted on white-light cystoscopy. Physicians did not have access to Oncuria-Monitor, BladderChek, or cytology results prior to cystoscopy.

Participants were considered positive for malignancy if one or more tumors or suspicious areas were observed during cystoscopy and, if removed, were defined as malignant on pathologic examination. Tumors or suspicious areas seen endoscopically but not removed were considered positive for malignancy and designated as cancer stage Tx if cytology was positive/suspicious or if extensive fulguration led to complete resolution on subsequent follow-up. Participants were considered negative for cancer if no tumor or suspicious areas were seen endoscopically, if tissue underwent biopsy and was defined as nonmalignant on histopathologic examination, or if cytology was anything other than positive/suspicious and subsequent follow-up did not note a tumor or suspicious areas. Pathologic examination of biopsied tissue was performed within each institution or at a reference laboratory per each facility’s standard practice. Staging criteria were as per the American Joint Committee on Cancer (32).

Primary and secondary outcomes

The primary outcome was the ability of the Oncuria-Monitor to noninvasively detect recurrent bladder cancer. The secondary outcome was comparing the Oncuria-Monitor cancer detection rates with those of BladderChek and urinary cytology.

Laboratory procedures

The BladderChek Test (Abbott), an FDA-cleared point-of-care assay for bladder cancer detection, was performed at each participating site as per the manufacturer’s protocol. Of the tests conducted, 1,247 yielded evaluable results and were included in the analysis. One test remained inconclusive despite repeat testing and was excluded.

The refrigerated voided urine samples were processed per standard operating procedures within 1.5 hours and placed in the freezer (−20°C or −80°C) at the local site. Regularly, frozen urine samples were transported to a central biorepository on dry ice with a temperature data logger, which is standard for clinical trials. Central laboratory testing was performed without knowledge of cytology results or clinical findings (i.e., ProBe design; ref. 33). Urine samples were subsequently sent in batches to BioAgilytix for testing. Oncuria-Monitor, which detects 10 proteins (A1AT, APOE, ANG, CA9, IL8, MMP9, MMP10, PAI1, SDC1, and VEGF) within a single voided urine sample, was performed as per the manufacturer’s instructions (Nonagen Bioscience Corp.).

Sample size

The true sensitivity of Oncuria-Monitor was estimated to be approximately 75%. Based on a type I error rate (α level) of 0.05, with 99% power to find a significant difference of 20% in the sensitivity rate among Oncuria, BladderChek, and cytology, an estimated sample size of 300 participants with at least 300 urine samples was required, allowing for the replacement of any participants removed from analysis and assuming that up to 30% of participants would have a positive cystoscopic evaluation. Partitioning the cohort into training and validation datasets reduced the effective power available for model development from approximately 99% to 80%.

Statistical analysis

Descriptive statistics included the calculation of mean values, standard deviations, percentages, and 95% confidence intervals (95% CI). Comparisons between groups were made using the Kruskal–Wallis test (continuous data) or Fisher exact test (categorical data).

We utilized a previously reported machine learning pipeline to evaluate various predictive models for detecting recurrent bladder cancer (29). Specifically, we explored multiple classification approaches, including logistic regression (glm), elastic net (glmnet), random forest (ranger), XGBoost (xgboost), and decision trees (rpart). Our dataset incorporated participants’ demographics (age, sex, and race), tumor grade and stage, and the 10 protein biomarkers.

The entire cohort of 300 participants consisted of 1,248 voided urine samples, which were split into training (n = 150 participants with 624 urine samples) and validation (n = 150 participants with 624 urine samples) groups stratified by participants, time point, disease status, and institutes (Table 1; Supplementary Tables S2 and S3). Model testing was conducted through cross-validation with stratified folds to ensure robustness. The training set was evaluated using internal cross-validation with 1,000 bootstrap iterations to optimize model performance and assess stability. Each model underwent a comprehensive preprocessing workflow, including data imputation, upsampling, normalization, and categorical variable encoding. We used another 10-fold cross-validation to explore different hyperparameter settings—adjustable factors that affect model performance—and identified the optimal configuration.

Table 1.

Study participant characteristics.

Characteristic	Overall N = 300a	Training			Validation
		Urothelial cancer n = 40b	No urothelial cancer n = 110b	P valueb	Urothelial cancer n = 53b	No urothelial cancer n = 97b	P valueb
Median age, years (IQR)	72 (66–79)	75 (70–80)	73 (66–80)	0.2	72 (65–81)	73 (66–80)	>0.9
Age group	​	​	​	0.3	​	​	0.9
<55	17 (5.7%)	0 (0%)	6 (5.5%)	​	5 (9.4%)	6 (6.2%)	​
55–64	43 (14%)	4 (10%)	18 (16%)	​	8 (15%)	13 (13%)	​
65–74	101 (34%)	13 (33%)	39 (35%)	​	17 (32%)	32 (33%)	​
≥75	139 (46%)	23 (58%)	47 (43%)	​	23 (43%)	46 (47%)	​
Sex	​	​	​	0.6	​	​	0.10
Male	234 (78%)	29 (73%)	84 (76%)	​	39 (74%)	82 (85%)	​
Female	66 (22%)	11 (28%)	26 (24%)	​	14 (26%)	15 (15%)	​
Race	​	​	​	0.068	​	​	0.068
Caucasian	212 (71%)	33 (83%)	74 (67%)	​	42 (79%)	63 (65%)	​
Other	88 (29%)	7 (18%)	36 (33%)	​	11 (21%)	34 (35%)	​
Risk group at enrollment	​	​	​	0.5	​	​	0.8
Low	44 (14.7%)	6 (15%)	16 (15%)	​	8 (15%)	14 (14%)	​
Intermediate	79 (26.3%)	12 (31%)	24 (22%)	​	17 (32%)	26 (27%)	​
High	176 (59.0%)	21 (54%)	70 (64%)	​	28 (53%)	57 (59%)	​
Unknown	1 (<1%)	1 (2.5%)	0 (0%)	​	​	​	​
Smoking status	​	​	​	0.5	​	​	>0.9
Never smoker	93 (31%)	12 (31%)	32 (29%)	​	17 (33%)	32 (33%)	​
Former smoker	172 (58%)	24 (62%)	61 (56%)	​	31 (60%)	56 (58%)	​
Current smoker	31 (10%)	3 (7.7%)	34 (31%)	​	4 (7.7%)	8 (8.3%)	​
Unknown	4 (1%)	1 (2.5%)	1 (2.5%)	​	1 (2%)	1 (1%)	​
Any household smoker	63 (21%)	3 (7.7%)	34 (31%)	0.004	7 (13%)	19 (20%)	0.3
Unknown	2 (<1%)	1 (2.5%)	0 (0%)	​	1 (2%)	1 (1%)	​

NOTE: Bolded values indicate P < 0.05.

^aMedian (IQR Q1–Q3) or n (%), as appropriate.

^bWilcoxon rank-sum test, Fisher exact test, or Pearson χ² test, as appropriate.

Model sensitivity and specificity were determined based on the correct classification of samples (cancer vs. no cancer) as confirmed by cystoscopy and histologic evaluation of excised tissue. The area under the ROC curve (AUROC) was used to evaluate accuracy in detecting bladder cancer and stratifying risk. Based on the aforementioned workflow, a multiclass random forest model incorporating both clinical and molecular features and classifying outcomes as no cancer, low-grade cancer, or high-grade cancer (including high-stage, aggressive disease) demonstrated superior performance across key metrics, including sensitivity, specificity, negative predictive value (NPV), positive predictive value (PPV), AUROC, and overall accuracy. Accordingly, this model was selected (“locked down”) for subsequent validation in an independent cohort of 150 participants comprising 624 urine samples. Classification thresholds were applied to the predicted class probabilities: A predicted high-grade probability ≥0.213 was classified as high-grade (cancer); otherwise, a predicted low-grade probability ≥0.17 was classified as low-grade (cancer); otherwise, the observation is classified as control (no cancer).

Sensitivity and specificity of the Oncuria-Monitor test, BladderChek, and urine cytologic analysis were compared using a McNemar χ² test. All statistical analyses were performed using R statistical software (v4.5.1; R Foundation for Statistical Computing; ref. 34). Tests were two-sided, and P < 0.05 was considered statistically significant.

Results

Study population

A total of 331 participants were enrolled at six sites, of which 300 met enrollment criteria. Demographic and baseline characteristics are summarized in Table 1. Most participants (78%; n = 234) were male, and 71% (n = 212) self-identified as Caucasian; the median age was 72 years. Of the 300 participants, 14.7% (n = 44) had a history of low-risk NMIBC, 26.3% (n = 79) were intermediate-risk NMIBC, and 59.0% (n = 176) were high-risk NMIBC, which included Ta + Tis (n = 2) and T1 + Tis (n = 1). All 300 participants had baseline visits with cystoscopy, BladderChek, urine cytology, and Oncuria-Monitor testing, with a median number of follow-up visits of 4 (IQR 2–6).

Of the 300 eligible participants, 31% (n = 93) had recurrent tumors while on study (including three with upper tract urothelial carcinoma), with 32 patients having multiple recurrences while on study. The median number of recurrences was 4 (IQR 2–6) accounting for a total tumor recurrence in the study of 143. The median age at recurrence was 72 years; 29% of men and 37% of women experienced recurrence. Of 143 recurrences, 90.2% (n = 129) were resected, and 9.8% (n = 14) did not undergo surgery and therefore were unstaged (Tx). Of 129 resected tumors, 95.3% (n = 123) were NMIBC, of which 66.2% (n = 82/123) were stage Ta, 18.7% (n = 23/123) were Tis [which included Ta + Tis (n = 1) and T1 + Tis (n = 2)], and 14.6% (n = 18/123) were T1. The six remaining tumors (4.7%) were MIBC. Of 129 resected recurrent tumors, 39.5% (n = 51) were low-grade, and 60.5% (n = 78) were high-grade.

Oncuria-Monitor test characteristics

Mean and median biomarker concentrations of the training and validation sets (624 urine samples each) are reported in Supplementary Table S2. Clinical and molecular features of the logistic regression diagnostic algorithm, using the aforementioned cutoff, in the training cohort noted a sensitivity of 98.3% (95% CI, 94.4%–100%), specificity of 39.6% (95% CI, 35.6%–43.8%), NPV of 99.6% (95% CI, 98.6%–100%), and PPV of 14.5% (95% CI, 11.3%–18.1%), allowing the lockdown of the algorithm and deployment in the validation cohort. All subsequent data are related only to the validation cohort, which included 150 participants (n = 624 urine samples).

Oncuria-Monitor resulted in a sensitivity of 85.6% (95% CI, 78.1%–92.2%), detecting 77 of 90 cancers at a specificity of 32.6% (95% CI, 28.7%–36.2%; Table 2). Oncuria-Monitor was positive in 93.8% (95% CI, 84.8%–100%; n = 30) of the 32 low-grade tumors, 88.2% (95% CI, 80%–95.6%; n = 60) of the 68 NMIBC, 82.9% (95% CI, 70.4%–93.6%; n = 34) of the 41 high-grade tumors, and 80% (95% CI, 33.3%–100%; n = 4) of the 5 MIBC (Table 2).

Table 2.

Sensitivity and specificity of Oncuria-Monitor, BladderChek, and cytology for urothelial cancer detection.

Disease classification	TP	FN	FP	TN	Total	Sensitivity	Specificity	PPV	NPV
Oncuria-Monitor
Any	77	13	360	174	624	0.856 (0.781–0.922)	0.326 (0.287–0.362)	0.176 (0.140–0.210)	0.930 (0.892–0.964)
MIBC	4	1	433	186	624	0.800 (0.333–1)	0.300 (0.266–0.333)	0.009 (0.002–0.018)	0.995 (0.982–1)
NMIBC	60	8	377	179	624	0.882 (0.800–0.956)	0.322 (0.284–0.358)	0.137 (0.106–0.169)	0.957 (0.926–0.984)
High grade	34	7	403	180	624	0.829 (0.704–0.936)	0.309 (0.273–0.343)	0.078 (0.054–0.102)	0.963 (0.934–0.989)
Low grade	30	2	407	185	624	0.938 (0.848–1)	0.312 (0.277–0.346)	0.069 (0.045–0.094)	0.989 (0.973–1)
BladderChek
Any	18	72	6	528	624	0.200 (0.121–0.281)	0.989 (0.979–0.996)	0.750 (0.560–0.923)	0.880 (0.852–0.905)
MIBC	2	3	22	597	624	0.400 (0–1)	0.964 (0.950–0.979)	0.083 (0–0.208)	0.995 (0.988–1)
NMIBC	15	53	9	547	624	0.221 (0.133–0.317)	0.984 (0.973–0.995)	0.625 (0.429–0.828)	0.912 (0.887–0.934)
High grade	16	25	8	575	624	0.390 (0.250–0.539)	0.986 (0.976–0.995)	0.667 (0.462–0.846)	0.958 (0.941–0.972)
Low grade	1	31	23	569	624	0.031 (0–0.097)	0.961 (0.946–0.976)	0.042 (0–0.129)	0.948 (0.930–0.967)
Cytology
Any	24	41	22	450	537	0.369 (0.247–0.487)	0.953 (0.934–0.972)	0.522 (0.371–0.659)	0.916 (0.892–0.941)
MIBC	4	0	42	491	537	1 (1–1)	0.921 (0.900–0.944)	0.087 (0.022–0.173)	1 (1–1)
NMIBC	15	38	31	453	537	0.283 (0.160–0.404)	0.936 (0.915–0.957)	0.326 (0.195–0.462)	0.923 (0.899–0.945)
High grade	17	12	29	479	537	0.586 (0.393–0.769)	0.943 (0.924–0.963)	0.370 (0.228–0.500)	0.976 (0.961–0.988)
Low grade	2	26	44	465	537	0.071 (0–0.182)	0.914 (0.888–0.939)	0.043 (0–0.102)	0.947 (0.9927–0.967)

NOTE: Values shown as n or n (95% CI).

Abbreviations: FN, false negative; FP, false positive; TN, true negative; TP, true positive.

Comparison with cytology and BladderChek

BladderChek detected 20% (95% CI, 12.1%–28.1%; n = 18/90) recurrent tumors in the validation sample set at a specificity of 98.9% (95% CI, 97.9%–99.6%; Table 2). BladderChek was positive in 3.1% (95% CI, 0%–99.7%; n = 1/32) of low-grade tumors, 22.1% (95% CI, 13.3%–31.7%; n = 15/68) of NMIBC, 39.0% (95% CI, 25%–53.9%; n = 16/41) of high-grade tumors, and 40% (95% CI, 0%–100%; n = 2/5) of MIBC (Table 2).

Cytology detected 36.9% (95% CI, 24.7%–48.7%; n = 24/65) of cancers at a specificity of 95.3% (95% CI, 93.4%–97.2%). Cytology was positive in 7.1% (95% CI, 0%–18.2%; n = 2/28) of low-grade tumors, 28.3% (29%; 95% CI, 16%–40.4%; n = 15/53) of NMIBC, 58.6% (95% CI, 39.3%–76.9%; n = 17/29) of high-grade tumors, and 100% (95% CI, 100%–100%; n = 4/4) of MIBC.

Oncuria-Monitor, BladderChek, and cytology’s specificities were only slightly affected by inflammation, urolithiasis, other cancers, infection, and hematuria (Supplementary Table S4).

The NPV of Oncuria-Monitor was 93% (95% CI, 89.2%–96.4%) compared with BladderChek at 88% (95% CI, 85.2%–90.5%) and cytology at 91.6% (95% CI, 89.2%–94.1%). The PPV of Oncuria-Monitor was 17.6% (95% CI, 14%–21%) compared with BladderChek at 75% (95% CI, 56%–92.3%) and cytology at 52.2% (95% CI, 37.1%–65.9%; Table 2).

Table 3 shows the number of persons who would need to be evaluated with cystoscopy, Oncuria-Monitor, BladderChek, and cytology to detect one recurrent bladder cancer, that is, n = 3 (95% CI, 3–4) for cystoscopy, n = 3 (95% CI, 3–4) with Oncuria-Monitor, n = 15 (95% CI, 10–23) with BladderChek, and n = 8 (95% CI, 6–11) with cytology. These trends held when assessing high-grade MIBC.

Table 3.

Numbers of persons who would need to be evaluated with cystoscopy, Oncuria-Monitor, BladderChek, and cytology to detect one urothelial cancer.

Disease classification	Prevalence	Cystoscopya,b	Cytologya,b	BladderCheka,b	Oncuria-Monitora,b
Any	0.3500	3 (3–4)	8 (6–11)	15 (10–23)	3 (3–4)
MIBC	0.1225	8 (8–8)	8 (8–8)	22 (8–57)	11 (8–16)
NMIBC	0.2275	5 (4–6)	16 (11–28)	21 (14–33)	5 (5–5)
High grade	0.1225	10 (8–13)	14 (11–21)	22 (15–32)	10 (9–12)
Low grade	0.2275	4 (4–4)	71 (24–152)	112 (40–185)	5 (4–5)

NOTE: Values shown are prevalence fraction and n (95% CI).

^aSensitivity of cystoscopy: 0.892, Oncuria: 0.856, BladderChek: 0.200, and cytology: 0.369.

^bNumber needed to screen = 1/(Prevalence × Sensitivity).

In an extrapolation of our results to a hypothetical reference population of 10,000 participants with a history of NMIBC on tumor surveillance at average risk for recurrent bladder cancer, cystoscopy, Oncuria-Monitor, BladderChek, and urine cytology would identify, respectively, 3,122, 3,008, 699, and 1,315 persons with recurrent bladder cancer (Table 4). In that hypothetical population, the numbers of persons who would be referred for cystoscopy based on positive test results would be 7,400 with Oncuria-Monitor, 773 with BladderChek, and 1,612 with cytology. Of all positive test results, false positives would be 59.4% (n = 4,392/7,400) with Oncuria-Monitor, 9.6% (n = 74/773) with BladderChek, and 18.4% (n = 297/1,612) with cytology. Of all negative results, false negatives would be 18.9% (n = 492/2,600) for Oncuria-Monitor, 30.4% (n = 2,801/9,227) with BladderChek, and 26.0% (n = 2,185/8,388) with urine cytology. Higher false-negative results were associated with NMIBC for all tests.

Table 4.

Extrapolation of findings to a hypothetical population of 10,000 persons at average risk for urothelial cancer undergoing screening with cystoscopy, Oncuria-Monitor, BladderChek, and cytology.

Disease classification	Prevalence	Total cancer cases	Total noncancer cases	Sensitivity	Specificity	TNa	FPb	TPc	FNd
Oncuria-Monitor
Any	0.35	3,500	6,500	0.859 (0.781–0.927)	0.324 (0.286–0.361)	2,108 (1,858–2,349)	4,392 (4,151–4,642)	3,008 (2,733–3,246)	492 (254–767)
MIBC	0.122	1,225	8,775	0.802 (0.414–1)	0.299 (0.265–0.331)	2,623 (2,321–2,906)	6,152 (5,869–6,454)	983 (507–1,225)	242 (0–718)
NMIBC	0.227	2,275	7,725	0.885 (0.799–0.964)	0.320 (0.284–0.356)	2,473 (2,192–2,754)	5,252 (4,971–5,533)	2,013 (1,819–2,194)	262 (81–456)
High grade	0.122	1,225	8,775	0.833 (0.702–0.933)	0.307 (0.273–0.341)	2,696 (2,397–2,990)	6,079 (5,785–6,378)	1,021 (860–1,142)	204 (83–365)
Low grade	0.227	2,275	7,725	0.938 (0.842–1)	0.311 (0.274–0.345)	2,400 (2,118–2,664)	5,325 (5,061–5,607)	2,135 (1,916–2,275)	140 (0–359)
BladderChek
Any	0.35	3,500	6,500	0.200 (0.122–0.280)	0.989 (0.980–0.996)	6,426 (6,368–6,476)	74 (24–132)	699 (428–981)	2,801 (2,519–3,072)
MIBC	0.122	1,225	8,775	0.403 (0–0.947)	0.965 (0.948–0.979)	8,464 (8,320–8,590)	311 (185–455)	494 (0–1,160)	731 (65–1,225)
NMIBC	0.227	2,275	7,725	0.220 (0.134–0.316)	0.984 (0.974–0.993)	7,598 (7,523–7,669)	127 (56–202)	500 (306–720)	1,775 (1,555–1,969)
High grade	0.122	1,225	8,775	0.390 (0.255–0.529)	0.986 (0.978–0.995)	8,653 (8,578–8,730)	122 (45–197)	477 (313–648)	748 (577–912)
Low grade	0.227	2,275	7,725	0.031 (0–0.100)	0.961 (0.945–0.977)	7,426 (7,302–7,545)	299 (180–423)	70 (0–228)	2,205 (2,047–2,275)
Cytology
Any	0.35	3,500	6,500	0.376 (0.250–0.496)	0.954 (0.936–0.972)	6,203 (6,083–6,319)	297 (181–417)	1,315 (874–1,738)	2,185 (1,762–2,626)
MIBC	0.122	1,225	8,775	1 (1–1)	0.921 (0.900–0.943)	8,085 (7,895–8,278)	690 (497–880)	1,225 (1,225–1,225)	0 (0–0)
NMIBC	0.227	2,275	7,725	0.290 (0.158–0.408)	0.937 (0.918–0.957)	7,236 (7,090–7,395)	489 (330–635)	659 (360–929)	1,616 (1,346–1,915)
High grade	0.122	1,225	8,775	0.600 (0.396–0.764)	0.944 (0.925–0.963)	8,281 (8,113–8,452)	494 (323–662)	735 (485–936)	490 (289–740)
Low grade	0.227	2,275	7,725	0.071 (0–0.182)	0.914 (0.890–0.938)	7,058 (6,879–7,243)	667 (482–846)	162 (0–414)	2,113 (1,861–2,275)
Cystoscopy
Any	0.35	3,500	6,500	0.892 (0.773–1)	0.978 (0.953–1)	6,359 (6,193–6,500)	141 (0–307)	3,122 (2,707–3,500)	378 (0–793)
MIBC	0.122	1,225	8,775	1 (1–1)	0.844 (0.790–0.901)	7,409 (6,935–7,907)	1,366 (868–1,840)	1,225 (1,225–1,225)	0 (0–0)
NMIBC	0.227	2,275	7,725	0.868 (0.729–1)	0.943 (0.904–0.978)	7,283 (6,981–7,558)	442 (167–744)	1,976 (1,658–2,275)	299 (0–617)
High grade	0.122	1,225	8,775	0.817 (0.630–1)	0.897 (0.849–0.947)	7,875 (7,448–8,307)	900 (468–1,327)	1,000 (772–1,225)	225 (0–453)
Low grade	0.227	2,275	7,725	1 (1–1)	0.883 (0.835–0.930)	6,823 (6,450–7,184)	902 (541–1,275)	2,275 (2,275–2,275)	0 (0–0)

NOTE: Based on validation cohort performance. Values shown are prevalence fraction, n, or fraction (95% CI).

Abbreviations: FN, false negative; FP, false positive; TN, true negative; TP, true positive.

^aTN = Total Noncancer Cases × Specificity.

^bFP = Total Cancer Cases × (1 – Specificity).

^cTP = Total Cancer Cases × Sensitivity.

^dFN = Total Noncancer Cases × (1 – Sensitivity).

Discussion

Patients with a bladder cancer history require an intense cystoscopic follow-up schedule to evaluate recurrent disease (2), which occurs in approximately 30% to 35% of the surveillance population (35). In the surveillance setting, where patients are already “ruled in” for evaluation, sensitivity and NPV are prioritized over specificity. Therefore, in this study, which is the largest to date in this space, Oncuria-Monitor was noted to harbor the following performance: a sensitivity of 85.6% at a specificity of 32.6% and an NPV of 93.0%. Oncuria-Monitor demonstrated high sensitivity and NPV compared with cytology and BladderChek, supporting its use as a rule-out test in the surveillance setting. Importantly, these performance characteristics enable the deferral of invasive procedures in a subset of patients and support clinical decision-making by identifying individuals at low risk of recurrence. This finding highlights the potential clinical utility of multiplex biomarker testing as a triage tool to guide surveillance intensity. In particular, a negative Oncuria-Monitor result could provide reassurance to defer or extend cystoscopic intervals in appropriately selected patients, reducing exposure to invasive procedures without compromising the detection of clinically significant disease. Reducing cystoscopy frequency in patients with a history of bladder cancer may be supported by a test with such a strong rule-out performance, which could contribute to better patient outcomes (36). Specifically, by enabling earlier detection of recurrence and more efficient surveillance, this approach has the potential to affect disease management, healthcare utilization, and long-term patient outcomes in a high-risk cancer population.

The sensitivity of Oncuria-Monitor for detecting bladder cancer (85.6%) exceeded that of both BladderChek (20%) and urinary cytology (36.9%), representing a notable performance advantage. These values correspond with sensitivity previously documented for urinary cytology (36, 37) and BladderChek (11, 38, 39). Urinary cytology has historically served as the primary noninvasive adjunct to cystoscopy, but its limited sensitivity—particularly for low-grade disease—has constrained its clinical utility. In contrast, the substantially higher sensitivity observed with Oncuria-Monitor across tumor grades suggests that multiplex protein biomarker signatures may provide a more reliable alternative for noninvasive surveillance. This improvement is especially relevant for detecting early or low-volume recurrences that may otherwise escape cytologic detection. Oncuria-Monitor’s performance matches other commercially available tests, for example, Pacific Edge’s Cxbladder Monitor sensitivities of 81% to 91% in multicenter evaluations (40, 41), whereas Nucleix’s Bladder EpiCheck assay demonstrates sensitivities of 73% to 80% in similar settings (42, 43). The UroVysion FISH assay has 65% to 80% sensitivity depending on assay interpretation criteria (44). A legacy urine-based test, BTAstat, has consistently reported lower sensitivities of 57% to 70% (38). Importantly, Oncuria-Monitor demonstrated a sensitivity of >68% across low- and high-grade as well as NMIBC and MIBC diseases, underscoring its capacity to reliably detect clinically significant and indolent recurrences.

The 32.6% specificity of Oncuria-Monitor in our cohort was lower than that of other multiplexed urinary biomarker assays. Cxbladder Monitor has demonstrated specificities of 60% to 68% (40, 41), whereas Bladder EpiCheck has values of 68% to 75% (42, 43). Reported specificities of BladderChek were 40% to 50% (11, 38, 39), and BTAstat was 55% to 65% (38). UroVysion FISH specificity ranges from 65% to 80% depending on positivity thresholds (44). Urine cytology, while maintaining a specificity >85%, is hindered by poor sensitivity, particularly for low-grade recurrences (36, 37). Oncuria-Monitor’s high sensitivity and NPV suggest its utility as a reliable adjunct in surveillance algorithms, particularly in patients at increased risk for recurrence. In the setting of bladder cancer surveillance, patients are inherently “ruled in” for frequent cystoscopic evaluation based on their history of disease. Our data suggest that the use of Oncuria-Monitor enables a complementary “rule-out” strategy, whereby approximately 25% of patients could safely defer invasive cystoscopy based on the assay’s high sensitivity and favorable NPV. Accordingly, the incorporation of this test into routine surveillance paradigms may allow for the deintensification of cystoscopy schedules while maintaining patient safety.

The clinical implications of these findings are substantial. Patients with NMIBC often undergo lifelong surveillance, resulting in repeated invasive procedures, significant healthcare expenditures, procedural anxiety, and reduced patient adherence over time. In this setting, a noninvasive assay with high sensitivity and high NPV may provide meaningful clinical utility by identifying patients at low likelihood of recurrence who may be candidates for modified surveillance strategies. Thus, by supporting a rule-out approach, Oncuria-Monitor has the potential to reduce procedural burden, enable deferral of invasive procedures in selected patients, and support biomarker-guided, risk-adapted surveillance paradigms. Importantly, Oncuria-Monitor maintained strong sensitivity across clinically significant disease categories, including high-grade and muscle-invasive recurrences, which are the most critical lesions to detect during surveillance.

Several limitations merit consideration. First, the specificity of Oncuria-Monitor was lower than that of cytology and point-of-care testing. Typically, this may lead to additional diagnostic evaluations or false-positive findings in some patients. However, surveillance populations are already enriched for recurrence risk and routinely undergo cystoscopic evaluation; therefore, in this clinical context, sensitivity and NPV are prioritized to minimize missed recurrences, particularly aggressive disease. Second, although this study was conducted prospectively across multiple centers, surveillance decisions were not guided by biomarker results; therefore, additional utility studies will be necessary to determine whether biomarker-guided surveillance can safely reduce cystoscopy frequency without compromising oncologic outcomes. Third, although the classifier was validated using an independent validation cohort within this multicenter prospective study, additional external validation studies in independent populations will further strengthen generalizability and reduce concerns about overfitting. Finally, formal cost-effectiveness analyses were beyond the scope of this study but represent an important area for future investigation.

Importantly, Oncuria-Monitor is positioned for near-term clinical implementation, with deployment in Clinical Laboratory Improvement Amendments–certified settings underway, supporting the immediate clinical translatability of these findings.

In summary, this prospective multicenter study demonstrates that multiplex urine biomarker testing can detect recurrent bladder cancer with substantially greater sensitivity than conventional noninvasive methods and diagnostic efficiency approaching that of cystoscopy, with similar numbers needed to evaluate to detect recurrence. Integration of such assays into surveillance protocols could enable risk-adapted monitoring, reduce reliance on invasive procedures, and improve patient-centered care. These findings support the role of multiplex biomarker platforms as an emerging paradigm in precision cancer surveillance.

Supplementary Material

Supplementary Table S1

Participating Sites and Patient Enrollment.

Supplementary Table S2

Characteristics of Urine Samples Used for Development and Validation.

Supplementary Table S3

Representativeness of Study Participants.

Supplementary Table S4

Specificities of Oncuria-Monitor, BladderChekTM and Cytology in Benign Urological Conditions.

Data Availability

The datasets generated and analyzed during the current study are not publicly available because they are part of an ongoing regulatory submission supporting FDA review. Following completion of the regulatory process, the data may be made publicly available. Until that time, deidentified original datasets are available from the corresponding author upon reasonable request.

Authors’ Disclosures

Y. Lotan reports grants and personal fees from Nonagen Bioscience during the conduct of the study, as well as grants and personal fees from Pacific Edge, Nucleix, and Convergent Genomics and personal fees from Promis Diagnostics, Vesica Health, and Cleancard outside the submitted work. M. Ahdoot is a consultant for Oncuria. S. Porten reports grants from Cedar Sinai during the conduct of the study, as well as personal fees from Vesica Health, Stryker, J&J, and Pacific Edge and grants from Photocure outside the submitted work. H. Furuya reports other support from Nonagen Bioscience Corp. during the conduct of the study, as well as other support from Nonagen Therapeutics Corp. outside the submitted work. C.J. Rosser reports grants from the NIH during the conduct of the study, as well as nonfinancial support from Nonagen Bioscience outside the submitted work; C.J. Rosser also reports a patent for Nonagen Bioscience issued. No disclosures were reported by the other authors.

Authors’ Contributions

Y. Lotan: Conceptualization, data curation, writing–original draft. M. Luu: Formal analysis, writing–original draft. M. Liu: Data curation. S. Tikhonekov: Data curation. T. Kobayashi: Data curation. M. Ahdoot: Data curation, writing–review and editing. H.L. Kim: Data curation, writing–review and editing. S. Porten: Data curation, writing–review and editing. G. Peers: Data curation, writing–review and editing. Y. Kita: Data curation, writing–review and editing. T. Daskivich: Data curation, writing–review and editing. K. Murakami: Investigation, writing–review and editing. T. Sakatani: Investigation, writing–review and editing. I. Pagano: Data curation, formal analysis. Y. Zheng: Formal analysis, supervision. Z. Zhang: Formal analysis, supervision. H. Furuya: Supervision, methodology, writing–review and editing. C.J. Rosser: Conceptualization, resources, supervision, funding acquisition, project administration, writing–review and editing.