Distinct longitudinal patterns of urine tumor DNA in patients undergoing surveillance for bladder cancer

Abstract

Cystoscopy is the gold standard for surveillance of non‐muscle invasive bladder cancer (NMIBC), but the procedure is invasive and has suboptimal accuracy. The aim of this study was to investigate the potential of analyzing urine samples for the presence of urine tumor DNA (utDNA) to replace cystoscopy for surveillance of bladder cancer recurrence. In this longitudinal, prospective, and observational study, 47 patients were followed for recurrence for 2 years, involving analysis of utDNA using the BladMetrix DNA methylation biomarker test at each cystoscopy. In total, utDNA was detected in 21/23 recurrences (91% sensitivity), including 5/5 T1, T2, and carcinoma in situ (CIS) tumors (100%) and 10/12 Ta tumors (83%), with < 1% false‐negative test results. Importantly, utDNA analysis showed the potential to reduce the number of cystoscopies by 55%, benefitting 79% of the patients. Eleven of 23 recurrences (48%) were detected earlier with utDNA than with cystoscopy, and distinct patterns of residual utDNA post‐surgery indicated minimal residual disease (MRD) or field effect in 6% and 15% of the patients, respectively. In conclusion, utDNA analysis shows high sensitivity to detect tumor recurrence, potential to reduce the number of cystoscopies, and promise to guide patient‐specific surveillance regimens.

In this longitudinal study, we followed 47 bladder cancer patients over 2 years for recurrence, using a urine tumor DNA test (BladMetrix) in parallel to cystoscopy. The urine test detected 91% of the recurrences with < 1% false negative test results. Importantly, the urine test showed potential to reduce cystoscopies by 55%, and to indicate minimal residual disease and field effect.

Abbreviations

CIS

carcinoma in situ

ddPCR

droplet digital PCR

dMIQE

Minimum Information for Publication of Quantitative Digital PCR Experiments

EAU

European Association of Urology

EMA

European Medicines Agency

FDA

Food and Drug Administration

MIBC

muscle invasive bladder cancer

MRD

minimal residual disease

NBI

narrow‐band imaging

NMIBC

non‐muscle invasive bladder cancer

PDD

photodynamic diagnosis

STARD

STAndards for the Reporting of Diagnostic accuracy studies

TURB

transurethral resection of the bladder

utDNA

urine tumor DNA

1. Introduction

With ~ 570 000 new cases each year, bladder cancer is the 10th most commonly diagnosed cancer worldwide, and a leading cause of cancer‐related mortality, causing more than 200 000 deaths in 2020 [1]. Approximately 75% of the patients present with non‐muscle invasive bladder cancer (NMIBC), generally carrying a good prognosis, with a 5‐year cancer‐specific survival close to 90% [2]. Despite a favorable prognosis, there is a risk of recurrence and progression to muscle‐invasive bladder cancer (MIBC), where survival drops to 35% [3]. In high‐risk patients, the probability of recurrence and progression at 5 years can reach 78% and 45%, respectively [4]. The high recurrence and progression rates necessitate frequent patient surveillance, and NMIBC consequently represents a long‐lasting disease burden for the patients and the healthcare system. Cystoscopy is the most commonly used method for surveillance of bladder cancer recurrence but is invasive, uncomfortable, and costly [5, 6]. Also, the sensitivity is suboptimal, and particularly carcinoma in situ (CIS) is easily overlooked [7, 8].

A variety of urine‐based biomarker tests for surveillance of recurrence among patients with NMIBC have been suggested as alternatives to cystoscopy, and a handful have been approved by the Food and Drug Administration (FDA) and the European Medicines Agency (EMA) [9, 10]. Still, no molecular biomarker test is routinely used in clinical practice [11, 12], mainly due to varying accuracy across studies, poor sensitivity for low‐grade tumors, and uncertain clinical benefit. In addition, the vast majority of biomarker studies are limited by a single‐visit design, where the biomarker test has been evaluated at only one surveillance point per patient. Altogether, there is an apparent need for a noninvasive test for the accurate detection of bladder cancer recurrences, with the potential to reduce the number of cystoscopies, and with the ability to guide and improve patient management.

Residual cancer cells after surgery, often referred to as minimal residual disease (MRD), represent a clinical challenge in many cancer types, including NMIBC. MRD can eventually lead to recurrence, and is typically difficult to detect with existing diagnostic tools [13]. Also, field effect or field cancerization may lead to later tumor development in individuals with a former bladder cancer history [14], with an increased risk of high‐grade recurrences [15]. Field effect implies that an area of the normal‐appearing bladder mucosa at cystoscopy acquires molecular changes that predispose the tissue to undergo later malignant transformation [16]. Clinical tools for the detection of biological phenomena such as MRD and field effect are warranted and would be useful in order to tailor successful patient‐specific surveillance and treatment regimens [14].

In the present longitudinal and prospective study, we used BladMetrix, our previously reported DNA methylation test [17], to investigate the performance of urine tumor DNA (utDNA) as an alternative to cystoscopy for patients undergoing surveillance for recurrent bladder cancer. Importantly, our longitudinal study design allowed following changes in utDNA levels over time for individual patients, resulting in observations of interesting biological phenomena including MRD and field effect.

2. Materials and methods

2.1. Patient inclusion, clinical procedure, and urine sampling

The present study was a prospective, longitudinal, observational cohort study. Patients with a suspected bladder cancer scheduled for transurethral resection of the bladder (TURB) at Oslo University Hospital, Aker, Oslo, Norway from January 2017 to September 2018 were evaluated for inclusion. Tumor staging was performed according to the 2017 Tumor‐Node‐Metastasis (TNM) system [11, 18] and grading according to the WHO 2004/2016 classification system [19, 20]. The inclusion and exclusion criteria are summarized in Table S1, and an overview of the patient enrollment process is shown in Fig. S1. All patients considered for inclusion provided two initial urine samples for utDNA analysis, and underwent a subsequent TURB with histological evaluation. Two urine samples, from two independent urinates, were demanded from each patient in order to increase the sensitivity for detection [21]. One sample was typically collected at home in the morning and the other at the hospital prior to the clinical intervention. Two parallel samples were likewise requested before each surveillance cystoscopy. Parallel urine samples were achieved in 93% of the cases. In total, 581 urine samples were included in the analyses. Patients suited for inclusion and with detectable utDNA, i.e., a positive BladMetrix test, prior to TURB qualified for surveillance with utDNA. These patients went through surveillance cystoscopies according to national guidelines (Table S2) [22], and provided urine samples before each cystoscopy control for a 2‐year period. If a cystoscopy was scored as indicative or suspicious of recurrence, histology was performed in most cases. Histological diagnosis was considered the standard reference, but in cases lacking histology (i.e., a suspected tumor was burned away), the cystoscopy result was used as reference.

2.2. BCG treatment regimen

In general, for patients receiving BCG treatment, the following standard regimen was followed [22]: First, an induction regimen is administered, consisting of weekly intravesical installations of BCG for 6 weeks. This is followed by a maintenance regimen, including weekly BCG instillations for 3 weeks, at three, six and 12 months after initiation of the induction regimen. Some patients with high‐risk tumors (T1 high‐grade combined with CIS) may be further treated with weekly installations for 3 weeks every 6^th month, for up to two more years, if side effects are well tolerated.

2.3. Urine processing, DNA extraction, and bisulfite conversion

Urine samples were collected from January 2017 to October 2020. All samples were sent by car to the Norwegian Radium Hospital (~ 7 km) for processing. The majority of samples were processed within 3 h after sampling (median = 1 h 47 min, interquartile range; IQR = 55 min – 2 h 40 min) using a centrifugation protocol. Samples were centrifuged at 4 °C for 20 min (3200 g ), washed with PBS, and re‐centrifuged. Pellets were finally suspended in 1 mL PBS, centrifuged at 4 °C for 5 min (6000 g ), and stored (up to 12 months) at −80 °C until DNA extraction. The median urine volume was 65 mL (IQR = 37–100 mL). Urine pellets were chosen over urine supernatant, mainly due to increased yield and an easier processing workflow [23, 24]. DNA was extracted from urine pellets using the QIAamp DNA Mini Kit (Qiagen, Hilden, Germany) according to the manufactures' protocol (“DNA Purification from Blood or Body Fluids ‐ Spin Protocol”). DNA was eluted in two steps in order to increase yields; first in 60 μL, and subsequently in 20 μL AE‐buffer (QIAamp DNA Mini Kit; Qiagen). DNA concentrations were measured using the NanoDrop 1000 Spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA), with a median of 30.2 ng·μL⁻¹ (IQR = 13.6‐92 ng·μL⁻¹). The EpiTect Bisulfite Kit (Qiagen) was used for bisulfite conversion of up to 430 ng DNA according to the manufacturers' protocol (“Sodium Bisulfite Conversion of Unmethylated Cytosines in DNA Isolated from FFPE Tissue Samples”). Bisulfite‐converted samples were automatically purified by the QIAcube System (Qiagen) and eluted in 40 μL EB‐buffer (Qiagen) to a final concentration of 10.75 ng·μL⁻¹.

2.4. Droplet digital PCR and utDNA analysis

The QX200™ Droplet Digital™ PCR System (BioRad, Hercules, CA, USA) was used to analyze the eight DNA methylation biomarkers composing the BladMatrix test in bisulfite‐converted urine DNA as previously described [17]. The droplet digital PCR (ddPCR) reaction consisted of 1x ddPCR Supermix for Probes (BioRad), 818 nm of each primer, 182 nm of each probe, and 3 μL bisulfite converted DNA template in a final volume of 22 μL. Droplet generation of 20 μL of the pre‐reaction volume was performed using the Automated Droplet Generator (BioRad). The PCR was performed either in a T100 Thermal Cycler (BioRad) or in a Tetrad 2 Thermal Cycler (BioRad). See Table S3 for PCR cycling conditions. For all analyses, universal methylated human DNA standard (Zymo Research, Irvine, CA, USA) was used as methylation‐positive control, normal blood of healthy donors and/or human WGA non‐methylated DNA (Zymo Research) as methylation‐negative controls, and RNase‐free water (Sigma Aldrich, St. Louis, MO, USA) as non‐template control (NTC). See Fig. S2 for a representative example of amplification of positive and negative controls.

Positive droplet calling was performed in r (version 4.1.0) using the PoDCall shiny interface (https://bioconductor.org/packages/PoDCall/). The 4Plex was used as an internal control for normalization [17, 25, 26]. For reactions with < 3 positive droplets for the methylated target (considered background), methylation concentrations were set to zero. A 4Plex lower limit of 60 copies per μL was applied, and samples below this threshold were excluded from further analyses. The ddPCR methylation concentrations (cop·μL⁻¹) were calculated by dividing the concentration of the target (i.e., the biomarker) by the concentration of the 4Plex, and multiplying by a constant of 400. The biomarkers were scored methylated or unmethylated based on previously defined methylation concentration thresholds [17]. The two parallel urine samples, i.e., from two independent urinations by the same patient on the day of the hospital visit, were analyzed separately, and the highest methylation concentration per biomarker among the two samples was used in downstream analyses. In 67% of the cases, the first of the two parallel samples was used in the analysis. Of note, in cases with equal biomarker concentration for two parallel samples, the first sample was used by default. Diverging results for parallel samples (positive utDNA test for one urine sample and negative utDNA test for the other) were found for 13% (3/23) of the recurrences, underscoring the added value of parallel urine sampling.

The utDNA analysis was defined as positive if ≥ 2/8 biomarkers were methylated (i.e., a positive BladMetrix test), and negative if < 2/8 biomarkers were methylated (i.e., a negative BladMetrix test), as previously elaborated in Pharo et al. [17]. All ddPCR primer and probe sequences are listed in Table S4. Analyses were performed according to the “Minimum Information for Publication of Quantitative Digital PCR Experiments” (dMIQE) guidelines [27] (Table S5) and to the “STAndards for the Reporting of Diagnostic accuracy studies” (STARD) checklist [28] (Table S6).

2.5. Scoring of the utDNA result according to the standard reference

The utDNA analysis was considered (1) true positive if a positive utDNA result was confirmed by either (i) histology or (ii) cystoscopy with “not available” (NA) histology; (2) false negative if the utDNA result was negative but with positive cystoscopy that was either (i) confirmed by histology or (ii) NA histology; (iii) true negative if the utDNA result was negative with either (i) negative cystoscopy with negative or lacking histology or (ii) positive cystoscopy with negative histology; (iii) false positive if the utDNA analysis was positive but with negative cystoscopy, and with no later confirmed recurrences. In addition, we defined early detection as a positive utDNA result in a patient with later confirmed recurrence but with (i) negative cystoscopy and NA or negative histology for the current sample or (ii) positive cystoscopy but with negative histology of the current sample.

2.6. Statistics

Statistical analyses were performed using imb spss statistics 25 (Chicago, IL, USA) and graphpad prism 9.1.2 (San Diego, CA, USA). Sensitivity, false negative test rate and number of spared cystoscopies were calculated and used to evaluate the performance of utDNA for detecting bladder cancer recurrence.

2.7. Ethical approval of studies and informed consent

This study was carried out according to the Helsinki Declaration, and the included research biobanks have been registered according to national legislation. The study was approved by the Regional Committee for Medical and Health Research Ethics (REC number 28461), and informed written consent was obtained from all patients enrolled in the study.

3. Results

3.1. The patient cohort

Of the 59 patients that fulfilled the inclusion criteria, 47 (80%) had positive utDNA analysis at inclusion and qualified for surveillance with utDNA (Fig. S1). Clinicopathological patient characteristics are shown in Table 1. The 47 patients were followed with utDNA analysis using the BladMetrix test for 2 years, in parallel to all scheduled cystoscopy controls, planned according to the national guidelines (Table S2) [22]. Each patient provided urine at seven time points on average (range 2–9; IQR = 6–7), and in total 261 surveillance points were included for downstream analysis. Altogether 7/268 (2.6%) of the surveillance points were excluded from downstream analysis. The reason for exclusion was either no urine sample provided (n = 4), canceled cystoscopy control (n = 1), or too low DNA concentration for scoring (4Plex control < 60 cop·μL⁻¹; n = 2).

Table 1.

Clinicopathological characteristics of the 47 patients qualifying for surveillance with utDNA. CIS, carcinoma in situ; No, number.

	No. (%)
Total	47
Median age (range)	71 (54–88)
Treatment
BCG	16 (34)
Mitomycin	2 (4)
Antibiotics	2 (4)
Adjuvant chemotherapy	1 (2)
None	26 (55)
Sex
Male	37 (79)
Female	10 (21)
Stage
Ta	30 (64)
Ta + CIS	7 (15)
T1	4 (8.5)
T1 + CIS	2 (4)
CIS	4 (8.5)
Grade
High	27 (57)
Low	20 (43)

3.2. utDNA accuracy and spared cystoscopies

Over the study period, 23 recurrences were observed, of which 17 were confirmed by histology and six by cystoscopy alone. The utDNA analysis was positive in 21 of the 23 recurrences, resulting in 91% sensitivity. A contingency table of the accuracy of utDNA to detect recurrence is shown in Table 2, and the same data is illustrated patient‐wise in Fig. 1. All T1 (1/1), T2 (1/1) and CIS (3/3) recurrent tumors were detected (100%), as well as 10/12 of the Ta tumors (83%; Table 3). Moreover, urine samples from 7/8 (88%) high‐grade recurrences and 8/9 (89%) low‐grade recurrences showed detectable levels of utDNA. In a subgroup analysis, the patients were stratified based on the grade of their tumor at inclusion: (i) high‐grade tumors (n = 27), and (ii) low‐grade tumors (n = 20). The sensitivity for detecting cancer recurrences using utDNA was high in both groups; 100% and 85% in the high‐grade and low‐grade groups, respectively (Table 4). Strikingly, only 0.8% (2/261) of all utDNA analyses were false negatives. Considering a total of 143 true negative test results (Table 2), pre‐screening with utDNA prior to cystoscopy could have reduced the number of cystoscopies by 55% (143/261), sparing 79% (37/47) of the patients one or more unnecessary cystoscopies. Figure S3 illustrates a proposed clinical use of utDNA for patients undergoing surveillance for bladder cancer recurrence.

Table 2.

The accuracy of utDNA, using the BladMetrix test, to detect recurrences. ED, early detection; FN, false negative; FP, false positive; NR, not relevant; TN, true negative; TP, true positive.

	Recurrence (#)	No recurrence (#)	Early detections
Positive utDNA analysis (#)	21 (TP)	71 (FP)	24 (ED)
Negative utDNA analysis (#)	2 (FN)	143 (TN)	NR

Fig. 1.

The overall performance of urine tumor DNA (utDNA) for surveillance of bladder cancer patients. Each line represents one patient (y‐axis), and each dot represents a longitudinal surveillance event (in months; x‐axis). The color of the dot indicates the performance of the utDNA analysis compared to the standard reference (i.e., histology if available, or else cystoscopy). Of the 261 surveillance controls utDNA correctly identified 21 recurrences (true positives; green dots) and misclassified 71 non‐recurrences as recurrences (false positives; orange dots). The utDNA analysis was correctly negative at 143 surveillance events with no recurrence (true negatives; blue dots), while two recurrences were missed by utDNA (false negatives; red dots). In 24 cases, utDNA analysis revealed early recurrence detections (green triangles). If treatment was administered to the patient during the study course, this is indicated to the left on the figure by a star (BCG), cross (mitomycin), or a diamond (adjuvant chemotherapy).

Table 3.

Stage and grade of the recurrences. CIS, carcinoma in situ; NA, not available; No, number.

	No. (%)	No. detected by utDNA (%)
Total	23	21 (91)
Stage
Ta	12 (52)	10 (83)
T1	1 (4)	1 (100)
T1 + CIS	1 (4)	1 (100)
T2	1 (4)	1 (100)
Ta + CIS	2 (9)	2 (100)
NA	6 (26)	6 (100)
Grade
High	8 (35)	7 (88)
Low	9 (39)	8 (89)
NA	6 (26)	6 (100)

Table 4.

Recurrences stratified by grade of the tumor at inclusion. HG, high grade; LG, low grade; NA, not available; No, number.

Grade at inclusion	No. of recurrences	No. detected by utDNA (%)	Stage at recurrence	Grade at recurrence
Total	23	21 (91)
HG patients
1	1	1	Ta	HG
2	1	1	Ta	LG
3	1	1	Ta + CIS	HG
4	2	2	Ta	LG
5	2	2	1 T1 + 1 T2	HG
6	1	1	T1 + CIS	HG
7	1	1	Ta + CIS	HG
8	1	1	Ta	HG
Total HG	10	10 (100)
LG patients
1	1	1	NA
2	3	3	2 Ta + 1 NA	2 LG +1 NA
3	2	2	Ta	LG
4	1	1	NA
5	3	3	1 Ta + 2 NA	1 LG + 2 NA
6	1	0	Ta	LG
7	1	0	Ta	HG
8	1	1	Ta	LG
Total LG	13	11 (85)

3.3. Distinct longitudinal patterns of utDNA

The longitudinal design of the present study allowed following patient‐specific changes in utDNA levels over time. Altogether 16 of the patients (34%) experienced in total 23 recurrences, while the remaining 31 patients (66%) had no clinically confirmed recurrence over the two‐year study period. Representative examples of changes in utDNA levels in patients with and without recurrences are shown in Fig. 2 (A and B, respectively). Early detection of tumor recurrence by utDNA was observed in almost half of the cases (11/23; 48%) and in altogether eight patients (17%; Fig. 2C). The number of utDNA positive surveillance points preceding a clinically confirmed recurrence varied from one to four across patients, translating to an early detection range of 3–18 months, with a median of 6.5 months (IQR = 3.9–9.7 months). In total, utDNA indicated early recurrence at 24 surveillance points (Table 2). For 20 of these, the cystoscopy performed in parallel to the utDNA analysis was negative, but later TURB with histology or repeat cystoscopy with tumor destruction confirmed tumor recurrence. In the remaining four cases, both the utDNA and the cystoscopy results were positive, but the histology was negative. However, a recurrence was clinically confirmed at a later surveillance control.

Fig. 2.

Longitudinal urine tumor DNA (utDNA) analysis shown for five bladder cancer patients representing different clinical scenarios. (A) patient with recurrences, (B) patient without recurrence, (C) patient with early recurrence detection, (D) patient with MRD, (E) patient with field effect. The eight DNA methylation biomarkers (i.e., the BladMetrix test) used for the utDNA analysis are shown vertically in different colors with individual y‐axis values. The dotted lines indicate the biomarker‐specific methylation scoring threshold. A biomarker is scored positive if the DNA methylation level is above the threshold (closed circles) and negative if the DNA methylation level is below the threshold (open circles). The results of the utDNA analysis, cystoscopy, and histology for each surveillance control are shown below the chart. The numbers of methylated biomarkers are indicated inside the squares. The utDNA analysis is scored positive if ≥ 2/8 biomarkers are positive and negative if < 2/8 biomarkers are positive. MRD, minimal residual disease; NA, not available; R, recurrence; utDNA, urine tumor DNA.

In three of the eight patients with early detection, utDNA was observed at all surveillance points from TURB to histologically confirmed recurrence, and was subsequently absent after clinical removal of the recurrent tumor, suggesting the presence of MRD (Fig. 2D). In three additional patients with early detection a possible methylation field effect was observed, characterized by utDNA being present at all surveillance points, even after surgically removal of the recurrent tumor. A potential field effect was observed also among four patients without clinically detected tumor recurrence during the study period (Fig. 2E). Figure 3 schematically illustrates the result of the utDNA test (positive/negative) during the clinical course of a representative patient with ‘regular’ recurrence (A), early detection (B), MRD (C) and field effect (D).

Fig. 3.

Urine tumor DNA (utDNA) identifies various clinical scenarios. Illustrative examples of utDNA results (positive: green plus sign, negative: red minus sign) at diagnosis, post‐surgery, during surveillance and recurrence in the setting of (A) regular recurrence detection, (B) early detection, (C) MRD, and (D) field effect. (A) utDNA correctly detects a recurrence, and the utDNA level drops below threshold post‐surgery. (B) A recurrence is detected with utDNA before it is detected by cystoscopy in the clinic. Post‐surgery, the level of utDNA drops below threshold. (C) Residual tumor cells are left after surgery, and the utDNA test is positive at all surveillance points until the recurrence is surgically removed. After surgery, with the successful removal of the entire recurrent tumor, the utDNA analysis is negative. (D) The tumor at diagnosis imposes cancer‐specific methylation changes to the surrounding normal‐appearing bladder tissue (indicated by pink tissue color), predisposing that tissue to undergo later malignant transformation. Consequently, the utDNA analysis is positive at all surveillance points, also after surgical removal of the recurrent tumor. MRD, minimal residual disease; utDNA, urine tumor DNA. Created with BioRender.

4. Discussion

There is an apparent need for a noninvasive test that can reduce the numerous cystoscopies performed among patients undergoing surveillance for bladder cancer recurrence. In the present longitudinal study, analysis of utDNA using the BladMetrix test detected cancer recurrences with high‐sensitivity across all stages and grades. In almost half of the cases, the recurrence was detected earlier with utDNA than with cystoscopy. We also observed distinct longitudinal patterns of utDNA, suggesting MRD and an underlying field effect in the bladder mucosa.

As one of few studies with a longitudinal study design [29], we here followed bladder cancer patients for 2 years with repeated utDNA analysis at each cystoscopy control, averaging to seven longitudinal data points per patient. In contrast, the vast majority of studies have a single‐visit design. While such studies can provide test accuracy at a specific moment in time, they are unable to capture the longitudinal clinical course for individual patients, which could result in unrepresentative accuracy measures. Also, in a single‐visit setting, a negative test cannot exclude that the cancer will recur in the future, and what appears to be a false positive test can actually be a sign of early disease detection. In the present work, utDNA indicated early detection of recurrence at altogether 24 surveillance points, encompassing 17% of the patients and almost 50% of all recurrences. In a future clinical scenario, it might be relevant to offer photodynamic diagnosis (PDD) cystoscopy to patients with detectable utDNA and negative white light cystoscopy, as PDD cystoscopy has been shown to have higher sensitivity than white light cystoscopy [7, 30].

The longitudinal study design also revealed distinct patterns of utDNA that could be of both biological and clinical interest. In a small subgroup of patients (6%) we observed utDNA signals at all surveillance points from the initial TURB until recurrence, followed by absence of utDNA after surgical removal of the recurrent tumor (Fig. 2D). This could indicate the presence of MRD. In another patient subgroup (15%), utDNA was consistently detected at all surveillance points, regardless of recurrence (Fig. 2E), an observation that points toward a methylation field effect. A methylation field effect in bladder cancer has previously been described by others [16, 31, 32], and in line with our results, Reinert et al. [31] estimated that for a substantial proportion of the patients (31%), the apparently false‐positive results were in fact related to a potential field effect. Both MRD and field effect are typically difficult to detect with existing clinical tools, and utDNA analysis thus shows potential to add an important layer of clinical information. It has previously been shown that patients with MRD or field effect have a higher risk of tumor recurrence [14], or a possibly later tumor development [31], respectively. Recently, Strandgaard et al. [15] also reported that high field cancerization is associated with an increased risk of high‐grade recurrences. From that, one could speculate that a more liberal initiation of BCG treatment might be favorable for patients with MRD or field effect, or, regardless, that these patients could be offered a more frequent or prolonged surveillance schedule, involving the more sensitive PDD or narrow‐band imaging (NBI) cystoscopies [7]. Further studies are required to determine how MRD and field effect should be optimally managed in the clinic.

With high sensitivity across all stages and grades, utDNA analysis using the BladMetrix test fulfills the requirements from the European Association of Urology (EAU), which states that for a molecular biomarker test to replace or reduce the number of cystoscopies, high sensitivity is required across all risk groups [33]. Also, a noninvasive biomarker test with sensitivity above 90% is in line with patient preferences [34]. Indeed, utDNA analysis detects both high and low‐grade lesions with almost 90% sensitivity, Ta recurrences with 83% sensitivity and T1‐T2 and CIS recurrences with 100% sensitivity. Of particular notice, the high sensitivity for CIS tumors largely outperforms white light cystoscopy, which has been reported with 50% sensitivity for these lesions [7]. Moreover, among patients presenting with high‐grade tumors at inclusion, the sensitivity for detecting recurrence is 100%, while for patients presenting with low‐grade disease the sensitivity is 85%. In comparison, commercially available urine tests with similar overall sensitivity, including Bladder EpiCheck (Nucleix, Rehovot, Israel) [35, 36], CxBladder Monitor (Pasific Edge, Dunedin, New Zealand) [37, 38] and Uromonitor (U‐Monitor, Porto, Portugal) [39, 40], generally suffer from lower sensitivity for low‐grade recurrences. In addition, in the present study, utDNA analysis returned < 1% false negative test results, and showed the potential to reduce the number of cystoscopies by 55%, benefitting almost 80% of the patients. It should be noted that 20% of the patients in this study had a negative utDNA test at inclusion, and were thus not eligible for surveillance with utDNA. This is in line with what Reinert et al. [31] reported, stating that 11–18% of the patients showed no methylation at inclusion. However, in case of a future recurrence, these patients might qualify for surveillance with utDNA, given a positive urine test at that time point. In the current study, two parallel urine samples were requested from the patients at all cystoscopy controls, as it has been shown by others to increase the sensitivity [21]. Indeed, we did find that two samples enhanced the test sensitivity by detecting 13% (3/23) more recurrences. However, the collection and analysis of two parallel urine samples increase the test costs, and should be evaluated with a cost–benefit analysis in a separate study.

A challenge inherent to most studies in the field, including the present work, is that biopsies are not taken for patients with negative cystoscopy results. Moreover, in patients with a positive cystoscopy result, the suspected tumor is sometimes burned away without biopsying. In lack of histological data, the cystoscopy‐based observation is typically used as the reference standard. However, since the reported sensitivity and specificity of white light cystoscopy is approximately 70% [7, 30], use of cystoscopy as a reference represents a limitation to the development and evaluation of molecular biomarker tests, and is likely to cause some seemingly, but incorrect, false‐positive and false‐negative results. Also, in Norway cytology has not routinely been performed for all patients, unless the cystoscopy finding is suspicious, or negative in high‐risk patients. Consequently, we were not able to compare the performance of the utDNA test with cytology, as the study was not designed for that. However, pooled analyses have shown that cytology has an overall sensitivity of 44%, with an even poorer sensitivity for lower risk tumors (median sensitivity 27%) [41]. Compared to these results, the utDNA test largely outperforms cytology in terms of sensitivity.

Finally, the present study includes a modest number of patients and should be validated in a larger patient cohort. However, given our longitudinal study design, each included patient provided on average urine at 7 time points, which increases the number of surveillance points from 47 (in the case of a single‐visit design) to almost 300. Also, our utDNA test has previously been analyzed in a large, prospective and blinded series of 273 patients with gross hematuria, resulting in both high sensitivity and specificity (> 90%) for cancer detection [17].

5. Conclusions

In this longitudinal study, utDNA analysis showed high sensitivity (91%) for identifying tumor recurrence across all stages and grades. Moreover, utDNA has the potential to detect early recurrences, to reduce the number of surveillance cystoscopies by 55%, and to indicate MRD and field effect. By sparing patient discomfort, reducing cystoscopy‐related costs for health services and tailoring patient‐specific surveillance regimens, utDNA analysis has potential to provide significant clinical benefits.

Conflict of interest

GEL is named on a patent (WO/2012/052844) proposing VIM as a biomarker forthe detection of bladder cancer, granted in US (US9797016) and EP (2630261). GEL, RW, and HP are named on a pending PCT patent application (WO/2020/099938) covering the biomarkers used in this study. GEL, MJ, and HP are named on two pending US national applications (US/2019/0221286 and WO/2019/106149) covering the 4Plex control and PoDCall algorithm used in the present study for normalization of DNA methylation ddPCR results. The remaining authors have nothing to disclose.

Author contributions

HMV: Methodology, Validation, Formal analysis, Writing – Original Draft, Visualization. HP: Methodology, Validation, Formal analysis, Investigation, Writing – Original Draft, Visualization, Project administration. AKS: Investigation, Resources. SB‐W: Investigation. M‐BF: Investigation. MJ: Methodology, Formal analysis. PG: Conceptualization, Writing – review & editing. RW: Conceptualization, Resources, Writing – review & editing. GEL: Conceptualization, Funding acquisition, Project administration, Supervision, Validation, Writing – review & editing.

Supporting information

Fig. S1. Flow chart of the patient inclusion.

Fig. S2. A representative example of ddPCR amplification of positive and negative controls.

Fig. S3. Proposed clinical use and scoring of the utDNA test for patients with NMIBC.

Table S1. Inclusion and exclusion criteria for patients considered for attendance in the present study.

Table S2. Recommendations for surveillance cystoscopies according to national guidelines.

Table S3. The PCR thermal cycling conditions for the ddPCR experiments.

Table S4. ddPCR primer and probe sequence used in the present study.

Table S5. The ddPCR analyses were performed according to the dMIQE2020 guidelines.

Table S6. The Standards for Reporting of Diagnostic Accuracy Studies (STARD) checklist.

Data accessibility

All data are included in the manuscript and supplementary or available from the corresponding author upon reasonable request. The raw ddPCR data are available at https://zenodo.org/records/10679272.