Bladder EpiCheck clinical utility to predict BCG response in non‐muscle‐invasive bladder cancer

Abstract

Objective

To evaluate the performance of Bladder Epicheck® (BE; Nucleix Ltd., Rehovot, Israel) in predicting tumour recurrence and bacillus Calmette–Guérin (BCG) failure during the first year after induction treatment.

Patients and methods

Prospective study including 65 patients with non‐muscle‐invasive bladder cancer treated with BCG between 2018 and 2021. Urine samples analysed with BE were collected before and after BCG induction. Logistic binary regression was used to assess the association between clinical and pathological variables and BE results with tumour recurrence and BCG failure during the first year after induction treatment.

Results

During follow‐up, 16 (24.6%) patients experienced a bladder cancer event, 11 (68.8%) of which were BCG failure (high‐grade recurrence) and five (31.2%) were low‐grade recurrences. The median (range) time to overall recurrence was 7.3 (3.8–17.4) months. A significant association was found between the risk of tumour recurrence/BCG failure and post‐BCG cystoscopy (odds ratio [OR] 10.0; P < 0.001 and OR 13.1; P < 0.001, respectively), post‐BCG BE result (OR 16.9; P < 0.001 and OR 33.1; P < 0.001, respectively) and pre/post‐BCG EpiScore value variation (OR 14.4; P = 0.001 and OR 7.1; P = 0.018, respectively). A nomogram including these three variables outperformed the Club Urológico Español de Tratamiento Oncológico (CUETO) risk tables to predict any bladder cancer event after BCG induction (area under the curve 95.1% vs 67.1%). Result validation in a larger and independent series is needed.

Conclusions

The BE post‐BCG status and variations in EpiScore values can help us identify patients at higher risk of any bladder cancer event and BCG failure promptly. These data can have an impact on disease management.

Abbreviations

AUC

area under the curve

BE

Bladder Epicheck

BCG

Bacillus Calmetter-Guérin

CIS

carcinoma in situ

CUETO

Club Urológico Español de Tratamiento Oncológico

EAU

European Association of Urology

HG

high grade

LG

low grade

(N)MIBC

(non‐)muscle‐invasive bladder cancer

OR

odds ratio

TURBT

transurethral resection of the bladder tumour

Introduction

The optimal treatment for non‐muscle‐invasive bladder cancer (NMIBC) at higher risk of disease progression is transurethral resection of the bladder tumour (TURBT) followed by adjuvant intravesical Calmette–Guérin (BCG) instillations. BCG treatment significantly reduces the odds of tumour recurrence after TURBT, especially when maintenance is completed [1]. However, up to 20–40% of these patients will experience BCG failure, resulting in tumour recurrence and eventually progression to MIBC [2, 3]. Furthermore, the cancer‐specific survival rate of progressive patients failing intravesical therapy is significantly lower than de novo MIBC [1, 4, 5]. This underlines the limited time window for an appropriate management approach, an early cystectomy if the patient is compliant, or being enrolled in a clinical trial.

Clinicopathological parameters can help identify patient risk groups [6]; however, they cannot accurately predict response to BCG treatment on an individual basis [3]. Urine markers have been described as having prediction qualities [7, 8]; however, the inflammatory response produced by BCG can impact the performance of the tests and, therefore, might not be ideal to assess BCG response.

Bladder Epicheck® (BE; Nucleix Ltd., Rehovot, Israel) is a urine assay that analyses the degree of methylation of a group of genes implicated in bladder cancer, using quantitative real‐time PCR. The test has achieved a sensitivity of >90% in high‐grade (HG) NMIBC and similar specificity compared to cytology in a high‐risk population. Moreover, its performance is not influenced by BCG treatment [9]. Consequently, it has been suggested that BE could be a useful tool for assessing BCG response in patients with NMIBC. Predicting BCG outcomes is of clinical importance to tailor disease management. This study aimed to investigate the potential role of BE in predicting the BCG response in this group of patients.

Patients and Methods

Patients and Treatment Regimens

All patients who were scheduled to undergo intravesical BCG immunotherapy in our centre from September 2018 to February 2021 were offered to participate in this prospective study. Patients were eligible if they had pathologically confirmed primary or recurrent intermediate, high‐risk, or very high‐risk NMIBC, according to the European Association of Urology (EAU) risk group classification [6]. Exclusion criteria were any concomitant neoplasm, pelvic radiation, or non‐urothelial carcinomas. Patients included with T1 HG underwent re‐TURBT between 1 and 2 months after the initial diagnosis to evaluate for invasive disease and had normal upper urinary tract imaging. All patients included received a complete BCG induction, including five or six weekly treatments. Maintenance consisted of three weekly treatments at 3, 6, and 12 months after induction. Complete maintenance was not mandatory. Overall, 23% (15/65) of patients did not receive any maintenance because of intolerance, tumour recurrence, or BCG shortage. All patients with Ta or T1 had cystoscopy and cytology after BCG induction. Patients with carcinoma in situ (CIS) alone or associated underwent random bladder biopsies after induction. Patients with persistent CIS after the first BCG induction received a second complete BCG induction treatment. After BCG induction, a standard cystoscopy surveillance schedule was used in accordance with the EAU guidelines [6].

The Club Urológico Español de Tratamiento Oncológico (CUETO) group risk tables (https://www.aeu.es/Cueto.html), a clinical pathological scoring model for NMIBC treated with BCG instillations [2], were used to calculate the recurrence risk in each patient during the first year.

All methods were conducted according to relevant guidelines and regulations, with written informed consent obtained from all participants (HCB/2013/8753). The study adhered to the Declaration of Helsinki standards and received approval from the Clinical Research Ethics Committee of the Hospital Clinic of Barcelona (HCB/2018/0746).

The BE Assay

Voided urine samples were collected immediately before the first intravesical BCG instillation (pre‐BCG sample) and after completion of BCG induction at the time of the first follow‐up cystoscopy (post‐BCG sample). The BE test (Nucleix) was performed and analysed following the manufacturer's instructions [9]. Briefly, samples were centrifuged twice at 1000  g for 10 min. The DNA extracted using the Bladder EpiCheck DNA extraction kit was digested using a methylation‐sensitive restriction enzyme. Quantitative PCR results were analysed using proprietary BE software that provides an EpiScore (a number between 0 and 100) [9]. An EpiScore of ≥60 indicates a positive result, while a score of <60 is considered negative. The physicians and patients were ‘blinded’ to the BE results that were not used in clinical practice.

Endpoints

The BCG response was evaluated according to two primary endpoints: tumour recurrence and BCG failure. All patients were followed up until the end of the maintenance treatment or until 1 year after BCG induction for those without maintenance. Recurrence was defined as any bladder cancer event found during the follow‐up period, regardless of grade or stage. BCG failure was defined as the presence of any HG tumour during the established follow‐up period [6].

Statistical Analysis

Study variables included patient and tumour characteristics (demographics, tumour stage, grade, primary/recurrent, size, and multifocality, and the presence of concomitant CIS) and BE assay results (pre‐BCG status, post‐BCG status, and EpiScore value pre/post‐BCG variation). Tumour grade was scored according to the world health organization (WHO) 2004/2016 grading system [10]. Descriptive statistics were used to summarise the study population characteristics. Binary logistic regression was used to investigate potential predictive factors of BCG response. Selection of predictors was performed using the forward stepwise approach. The Kaplan–Meier method and the log‐rank test were used to estimate and compare tumour recurrence between groups. The area under the  curve (AUC) was used to compare the performance of the models (Youden's index). Statistical analysis was performed using the Statistical Package for the Social Sciences (SPSS) version 29.0 (IBM Corp., Armonk, NY, USA) and Stata/BE 18.0 (Stata Corp., College Station, TX, USA). Statistical significance was set at α = 0.05.

Results

Patient Characteristics

A total of 68 patients with NMIBC who underwent TURBT followed by BCG instillations were enrolled in this study. After excluding three patients (4.6%) with failed BE tests (2.2% BE technical failures), the analysis was conducted on a cohort of 65 patients (Table 1). During the follow‐up period, 16 patients (24.6%) experienced a tumour event, with a median (range) recurrence time of 7.3 (3.7–17) months. Of these patients, 11 (68.8%) were considered BCG failure: seven patients (43.8%) had a HG recurrence and four (25%) presented disease progression, two of whom progressed to MIBC and the other two presented metastatic disease without evidence of local recurrence. Five low‐grade (LG) tumour recurrences were also found in the series. Most patients (77%) had BCG maintenance treatment.

Table 1.

Clinicopathological characteristics of the cohort.

Characteristic	Value
Number of patients	65
Age, years, median (range)	72 (42–96)
Sex (male/female), n (%)	56 (86.2)/9 (13.8)
Primary/recurrent disease, n (%)	37 (56.9)/28 (43.1)
Stage before BCG induction, n (%)
Ta	31 (47.7)
T1	25 (38.5)
CIS alone/concomitant	9 (13.8)/14 (21.5)
Grade before BCG induction, n (%)
HG	61 (93.8)
LG	4 (6.2)
BCG maintenance (no/yes)	15 (23.1)/50 (76.9)
NMIBC risk group*, n (%)
Intermediate	23 (35.4)
High	35 (53.8)
Very high	7 (10.8)

^*EAU guidelines 2024 [6].

Clinical–Pathological Predictors

Potential predictive factors of BCG response are shown in Table 2. Cystoscopy result post‐BCG proved to be a significant independent predictive factor for overall tumour recurrence (odds ratio [OR] 10; P < 0.001) and for BCG failure (OR 13.3; P < 0.001). In addition, patients with BCG maintenance had a lower probability of tumour recurrence (OR 0.3; P = 0.029). Despite tumour status (de novo or recurrent) initially being a predictive factor of BCG response in the univariate analysis, multivariate logistic regression analysis revealed that it was not an independent predictor of BCG response (P = 0.634).

Table 2.

Binary logistic regression analysis of predictive factors of BCG response.

Factor	Category	BC event		OR (95% CI)	P	BCG failure		OR (95% CI)	P
		Yes	No			Yes	No
Sex	Female	2	7	Ref.	0.858	1	8	Ref.	0.620
	Male	14	42	1.2 (0.2–6.3)		10	46	1.7 (0.2–15.5)
Age	<70 years	6	23	Ref.	0.511	5	24	Ref.	0.951
	>70 years	10	26	1.5 (0.5–4.7)		6	30	1.0 (0.3–3.53)
Status	Primary	5	32	Ref.	0.021	3	34	Ref.	0.039
	Recurrent	11	17	4.1 (1.2–13.9)		8	20	4.5 (1.1–19.1)
Stage	Ta	8	23	Ref.		5	26	Ref
	T1	5	20	1.4 (0.4–4.9)	0.601	3	22	1.4 (0.3–6.6)	0.662
	CIS	3	6	1.4 (0.3–7.1)	0.657	3	6	2.6 (0.5–14.0)	0.266
EAU risk groups	Intermediate	5	18	Ref.		2	21	Ref.
	High risk	9	26	1.2 (0.4–4.3)	0.730	8	27	3.1 (0.6–16.2)	0.186
	Very high risk	2	5	1.4 (0.2–9.8)	0.709	1	6	1.8 (0.1–22.7)	0.669
Concomitant CIS	No	8	32	Ref.	0.278	4	36	Ref.	0.069
	Yes	8	17	1.9 (0.6–5.9)		7	18	3.5 (0.9–13.5)
Size	<3 cm	13	38	Ref.	0.755	8	43	Ref.	0.613
	>3 cm	3	11	0.8 (0.2–3.3)		3	11	1.5 (0.3–6.5)
Multifocality	Single	9	32	Ref.	0.516	6	35	Ref.	0.522
	Multiple	7	17	1.5 (0.5–4.6)		5	19	1.5 (0.4–5.7)
Post‐BCG cystoscopy	−	6	42	Ref.	<0.001	3	45	Ref.	<0.001
	+	10	7	10.0 (2.8–36.3)		8	9	13.3 (3.0–60.2)
BCG maintenance	Yes	9	41	Ref	0.029	7	43	Ref.	0.259
	No	8	7	0.3 (0.1–0.9)		4	11	0.4 (0.1–1.8)
BE pre‐BCG status	−	7	30	Ref.	0.225	2	35	Ref.	0.013
	+	9	19	2.0 (0.6–6.4)		9	19	7.9 (1.5–40.1)
BE post‐BCG status	−	3	39	Ref.	<0.001	1	41	Ref.	<0.001
	+	13	10	16.9 (4.0–71.0)		10	13	33.1 (3.9–283.2)
Pre/post‐BCG EpiScore variation	Decrease	2	33	Ref.	0.001	2	33	Ref.	0.018
	Increase/stable	14	16	14.4 (2.9–71.3)		9	21	7.1 (1.4–36)

BC, bladder cancer.

Bold values statistically significant at P < 0.05.

Pre‐ and Post‐BCG BE Status

The proportion of bladder cancer events in patients with a positive pre‐BCG BE test (32.1%) was higher than in patients with a negative test (18.9%). However, no association was found between pre‐BCG BE status and tumour recurrence (P = 0.225). On the contrary, a significant association was identified between pre‐BCG BE status and BCG failure; 32.5% of patients with a positive pre‐BCG BE test had BCG failure vs 5.4% of those with a negative test (P = 0.013).

On the other hand, patients who had a positive post‐BCG BE test had a higher rate of tumour recurrence (56.6%) than patients with a negative result (7.1%; P < 0.001). Furthermore, patients who had a positive post‐BCG BE test (43.5%) had a higher rate of BCG failure than patients with a negative test (2.4%; P < 0.001). A strong association was found between post‐BCG BE status with any bladder cancer event (P < 0.001) and BCG failure (P < 0.001). In particular, patients with a positive post‐BCG BE status had a 16.9‐ and 33.3‐fold greater risk of a bladder cancer event and BCG failure, respectively, compared to patients with a negative post‐BCG BE status. Of note, the BE post‐BCG status was unable to predict tumour recurrence (false negatives) in three patients (two LG and one HG [intra‐diverticular CIS] recurrences).

Changes between Pre/Post‐BCG BE Results

When analysing the changes in the BE status before and after BCG therapy, no bladder cancer event was found in those patients with a positive pre‐BCG BE shifting to a negative post‐BCG BE status; patients who remained BE positive at post‐BCG were at higher risk of any bladder cancer event and BCG failure (P = 0.004; Table 3). On the other hand, patients with a negative pre‐BCG BE status shifting to a positive post‐BCG BE status had a higher rate of tumour recurrence (mostly due to the LG recurrences; P = 0.007), but not BCG failure (P = 0.302), than those who remained negative. The Kaplan–Meier curves (Fig. 1A) show that those patients with a negative post‐BCG BE status performed better than those with a positive status (P < 0.001). The increase or decrease in EpiScore values pre/post‐BCG was also a stronger predictor of BCG response (Fig. 1B). Patients with an increased or stable EpiScore pre/post‐BCG had a 14‐ and seven‐fold greater risk of a bladder cancer event and BCG failure, respectively, compared with patients with a decreased score (Table 2).

Table 3.

Recurrence and BCG failure rates according to pre‐ and post‐BCG BE status.

BE status		Total patient number	BC event rate, %	P	BCG failure rate, %	P
Pre‐BCG	Post‐BCG
Positive	Negative	11	0	0.004	0	0.004
	Positive	17	52.9		52.9
Negative	Negative	31	9.7	0.007	3.2	0.302
	Positive	6	66.7		16.7

BC, bladder cancer.

Fig. 1.

The BE test results before and after BCG induction as a predictor of BCG response in patients with NMIBC. (A) Kaplan–Meier curve comparing recurrence‐free survival according to the change in BE status pre‐ and post‐BCG induction. (B) Percentage of variation of the EpiScore value pre‐ and post‐BCG induction. Red bars represent patients with any bladder cancer (BC) event and blue lines patients with no events.

Proposed Nomogram and CUETO Tables

We have generated a nomogram including the three independent predictors of BCG response: cystoscopy result post‐BCG, BE post‐BCG status, and variations in EpiScore value pre/post‐BCG. The risk probabilities of presenting any bladder cancer event were estimated based on these predictors. Patients with positive results for the two BE predictors were at higher risk of overall tumour recurrence (a rate of >70% of bladder cancer event) regardless of the cystoscopy. Conversely, when the BE results were both negative, the risk proportion was very low (<4%; Fig. 2). Notably, the proposed nomogram demonstrated statistically significant differences in risk probabilities between patients with and without bladder cancer recurrence (P < 0.001; Fig. S1). Furthermore, these differences were maintained when using only the two BE predictors (P < 0.001). However, the CUETO table probabilities were not able to distinguish the risk of tumour recurrence in our cohort. In alignment with this, the performance of our proposed nomogram (area under the curve [AUC] 0.952; optimal cut‐point 32.95; sensitivity = 88%, specificity = 92%) was more accurate than the association of the two BE predictors (AUC 0.901; optimal cut‐point 52; sensitivity = 69%, specificity = 96%) and CUETO tables (AUC 0.671; optimal cut‐point 18.72; sensitivity = 69%, specificity = 67%; Fig. S2).

Fig. 2.

Proposed nomogram showing recurrence risk probability after BCG induction according to the three independent predictive factors identified through binary logistic regression: post‐BCG BE status, pre/post‐BCG variation in EpiScore value, and cystoscopy result.

Discussion

Bacillus Calmette–Guérin failure in NMIBC represents a crucial issue in the management of patients with bladder cancer. Despite BCG instillations being the standard of care for high‐risk NMIBC, a complete and sustained response during follow‐up is achieved in barely 50% of patients [3]. Furthermore, the prognosis of these NMIBC progressive patients is very poor. Huguet et al. [4] reported a 38% 5‐year cancer‐specific survival for these patients, consistent with other studies [5]. Several factors could account for BCG failure, ranging from the existence of concealed invasive or metastatic disease to an insufficient or diminished immune response [11]. The clinicopathological parameters can help; however, BCG failure cannot be accurately predicted on an individual basis. Identifying these susceptible patients can help us promptly decide whether to opt for a more radical treatment or to enroll them in clinical trials.

The role of tumour biomarkers in this clinical setting is not yet established. Several haematological immune response markers have been described, alongside various urinary cytokines [12]. However, the predictive value of single biomarkers is limited and none of them are currently used in clinical practice. On the other hand, a urinary fluorescence in situ hybridisation‐based assay including various biomarkers, has been shown to be useful in predicting tumour recurrence and progression in this setting, revealing that the earliest positive result after BCG induction conveys a higher risk of disease recurrence [7, 13].

The BE test detects bladder cancer by targeting DNA methylation alterations in specific gene promoters associated with tumour suppressor silencing or oncogene activation. Five individual studies on the BE test, along with two meta‐analyses comparing the currently available commercial urine assays, have provided data on the diagnostic values of the BE test and have consistently shown its high sensitivity in the detection of disease recurrence, especially HG tumours [14, 15, 16, 17, 18, 19]. This has led to its incorporation as part of the NMIBC follow‐up protocol in numerous European hospitals. This is the first study focusing on the BE test's potential in predicting response to intravesical BCG in a prospective series. We found that the results of the BE test and cystoscopy post‐BCG induction significantly correlated with the risk of any tumour event and BCG failure. Notably, not only was the BE status post‐BCG a predictive factor (OR 16.9; P < 0.001) but also the conversion to or persistence of a positive status after BCG induction, compared with patients who became negative from a positive baseline or remained negative, showed a higher probability of experiencing any bladder cancer event (56.5% overall risk of tumour event vs 0–9.7%, respectively). This underscores that the variation in BE status pre/post‐BCG plays a significant role and provides additional insights into the likelihood of recurrence. Furthermore, variations in the EpiScore value were also strongly associated with any bladder cancer event (OR 14.4; P = 0.001) or BCG failure (OR 7.1; P < 0.05). In particular, an increase in the EpiScore value, regardless of whether it becomes positive, represents an increased risk of unsuccessful BCG treatment. This highlights that predictions can be made using both the binary result of the BE test and EpiScore values.

Additionally, we estimate the risk probability of any tumour event based on three independent predictive factors: BE status post‐BCG, EpiScore value pre/post‐BCG changes, and cystoscopy results post‐BCG induction. The calculated risk probabilities were able to outperform CUETO table predictions, suggesting that BE analysis can enhance the ability to predict the behaviour of an individual tumour beyond clinicopathological parameters and potentially equivocal cystoscopy results. Interestingly, and in line with several studies [17, 20], the proposed nomogram also indicated that there might be patients at low risk of bladder cancer events (<8%) in whom the BE test can be used to reduce the burden of follow‐up cystoscopies. Moreover, in our study, the benefit of cystoscopy appears to be more evident in discriminating patients at higher risk when the two BE predictors do not follow the same trend. Yet, we do not believe that the BE test should replace cystoscopy but rather as a complementary tool to reduce the frequency of cystoscopies. Finally, even though BCG maintenance could decrease the risk probabilities of tumour recurrence, it should be noted that the nomogram is designed to predict the risk of a tumour event or BCG failure after induction treatment, which is the optimal time for considering alternative therapies.

Our study has multiple strengths. The most important is that we included a diverse group of patients receiving BCG treatment, representing a real‐world practice. During the year of follow‐up, 24.6% and 6.2% of patients developed tumour recurrence and progression, respectively, indicative of the high‐risk status. To date, this is the first series of patients recruited prospectively to evaluate the BE's utility in predicting BCG response.

We must also acknowledge some study limitations. First, it is a single‐centre study on a relatively small cohort with limited number of events. However, it has been demonstrated that the events per variable <10 could still produce valid results, particularly when predictors are not highly correlated [21]. In our study, we relied on a high‐quality database, and the three selected variables were not correlated (P = 0.22). Second, performing two BE tests is necessary to accurately estimate the risk of a bladder cancer event, this might increase costs in bladder cancer management mainly in hospitals that do not use the BE test in routine practice. Even though, we must bear in mind that a cost analysis of the BE strategy (using the BE test to rule out cystoscopies) demonstrated reduced costs in NMIBC follow‐up [22]. Third, cytology was not available for all patients and could not be included in the analysis. Validation of our results in a larger and independent series would be necessary to strengthen the role of the BE in predicting BCG response and to evaluate its predictive capacity across sub‐categories of NMIBC (Ta, T1, and CIS).

Conclusions

The BE post‐BCG status and variations in EpiScore values can help us to identify patients at higher risk of any bladder cancer event and BCG failure promptly. Thus, the BE test is likely to refine conventional surveillance techniques used after BCG therapy. These data can have an impact on disease prognosis as patients could benefit from alternative treatment strategies.

Disclosure of Interests

Lourdes Mengual certifies that all conflicts of interest, including specific financial interests and relationships and affiliations relevant to the subject matter or materials discussed in the manuscript (e.g., employment/affiliation, grants or funding, consultancies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: Lourdes Mengual is a consultant for Nucleix Ltd. Antonio Alcaraz has received honorarium as speaker or part of advisory boards from the following companies: Astellas, Janssen, Ipsen, Bayer, Sanofi, Olympus and Intuitive Surgical. The remaining authors have nothing to disclose.

Funding

This research was partially sponsored by Nucleix and private funding grants to Antonio Alcaraz and Lourdes Mengual. The sponsors played no direct role in the study.

Supporting information

Fig. S1. Box plots illustrating the difference in risk probability means between patients with and without a bladder cancer event according to (A) proposed nomogram, (B) nomogram with BE predictors alone, and (C) CUETO risk tables.

Fig. S2. Performance of the proposed nomogram, nomogram including only BE results, and CUETO risk tables. (A) Receiver operating characteristic curves showing the higher area under the curve achieved by the proposed nomogram compared with the nomogram including only BE results, and CUETO risk tables. (B) Overall models quality. A model with a value <0.5 indicates the model is no better than a random prediction.