Celldetect test performance in monitoring bladder cancer recurrence after transurethral resection

Abstract

Background

Non-muscle-invasive bladder cancer (NMIBC) has a high recurrence rate, necessitating rigorous surveillance primarily reliant on invasive cystoscopy. While urine cytology offers high specificity, its sensitivity for low-grade tumors is unacceptably low. CellDetect is a novel histochemical staining test based on differential metabolic activity between normal and malignant cells, which is often referred to as the Warburg effect. This study aimed to evaluate its real-world diagnostic performance for detecting NMIBC recurrence.

Methods

This retrospective cohort study included 243 patients under surveillance after transurethral resection of bladder tumor (TURBT). All patients underwent both the CellDetect test and cystoscopy within 24 months post-operatively. Using cystoscopy and histopathology as the reference standard, the sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy of CellDetect were evaluated. Subgroup analyses were performed based on tumor grade, stage, and European Organization for Research and Treatment of Cancer (EORTC) risk category.

Results

Among the 52 patients (21.4%) with histologically confirmed recurrence, the overall sensitivity of the CellDetect test was 76.9%, specificity was 78.0%, PPV was 52.6%, NPV was 91.2%, and overall accuracy was 77.8%. Sensitivity was significantly higher for high-grade (HG) tumors (93.8%) than for low-grade (LG) tumors (69.4%). In the high-risk EORTC subgroup, sensitivity was 88.9% and NPV was 95.2%. Furthermore, among 18 patients with “discordant” findings (positive CellDetect/ negative cystoscopy), 6 (33.3%) were subsequently diagnosed with recurrence upon random biopsy.

Conclusion

In this retrospective analysis, the CellDetect test demonstrated encouraging diagnostic performance in the post-TURBT surveillance setting, particularly its high NPV and sensitivity for HG tumors. “Discordant” results may hold predictive value for imminent recurrence. Further prospective validation is warranted. CellDetect represents a potential tool for developing a more personalized, risk-adapted, and patient-friendly follow-up strategy for NMIBC.

Introduction

Bladder cancer (BC) ranks among the most prevalent malignancies worldwide, with non-muscle-invasive bladder cancer (NMIBC) accounting for approximately 75% of initial diagnoses [1]. A hallmark of NMIBC is its high recurrence rate, necessitating lifelong, rigorous surveillance. Current guidelines recommend periodic cystoscopic examinations, which, while effective, are invasive, costly, and associated with patient discomfort and potential complications [2]. Urine cytology, the most widely used non-invasive adjunct, offers high specificity but suffers from unacceptably low sensitivity, particularly for low-grade (LG) tumors, which constitute the majority of recurrences. This significant diagnostic gap has driven the development of numerous urinary biomarkers over the past decades. However, factors such as variable accuracy, high cost, technical complexity, or susceptibility to confounding by inflammation have prevented their widespread adoption into routine clinical practice [3].

CellDetect^® (Zetiq Technologies Ltd., Israel) is a novel, histochemical staining platform designed to overcome these limitations [4]. Its diagnostic principle is based on differential metabolic activity between normal and malignant urothelial cells, often referred to as the Warburg effect. The test utilizes a proprietary plant extract combined with specific dyes, resulting in a distinct colorimetric discrimination: normal cells typically stain green/blue, while neoplastic cells exhibit red/purple nuclei. This approach integrates morphological assessment with a distinct metabolic indicator, permitting more objective and sensitive detection of cancer cells directly in voided urine samples. Previous multi-center validation studies have demonstrated promising performance for CellDetect, with varying sensitivity reports depending on the study population and design. For instance, in a comparative study, the sensitivity per protocol was 67% (95% CI: 46.0-83.5), during follow-up was 47% (95% CI: 22.9–72.2), and when combined with urine cytology during follow-up was 59% (95% CI: 32.9–81.6) [5]. Other tests in the same study, such as Bladder EpiCheck and Xpert BC Monitor, showed different sensitivities (e.g., up to 86% and ~ 79–80%, respectively).

At our hospital, the CellDetect test has been used successfully for several years and is considered an easy-to-use tool with satisfactory results. To formally evaluate its performance, we designed this retrospective study to assess its real-world diagnostic accuracy in patients undergoing follow-up (surveillance) after transurethral resection of bladder tumor (TURBT). This study aimed to evaluate the real-world diagnostic performance of the CellDetect test: its sensitivity, specificity, and predictive values overall and across critical subgroups defined by tumor grade, stage, and European Organization for Research and Treatment of Cancer (EORTC) risk category. Additionally, we sought to investigate the clinical implications of discordant findings, specifically discordant cases with positive CellDetect results alongside negative cystoscopy.

Materials and methods

Study design and patient population

This retrospective cohort study included patients who had undergone TURBT and were under surveillance for NMIBC recurrence, with age over 18 years. Between January 2022 and December 2024, all patients were followed according to the current European Association of Urology (EAU) guidelines [6]. Patients were included if they had undergone a CellDetect test and had a valid test result alongside scheduled white light cystoscopy during their follow-up visits within 24 months post-operatively. It should be noted that this follow-up period is relatively short for a disease characterized by potential late recurrences, such as NMIBC. This limitation restricts the ability to assess the long-term predictive value of the test and its utility in surveillance de-escalation strategies. Patients with concurrent upper tract urothelial cancer or progression to muscle-invasive bladder cancer (MIBC), or severe urinary tract infection were excluded from this study.

Sample collection and processing

A single voided urine sample was collected from each patient just prior to the scheduled cystoscopic examination. Crucially, samples were collected at least two weeks after the patient underwent bladder instillation therapy, to evaluate the test’s robustness in a real-world clinical setting. The sample was processed for the CellDetect test by skilled technicians who were blinded to the cystoscopic and histopathological results. This test was repeated if previously negative.

CellDetect test

Cell staining has been described in our previous article [4]. Briefly, according to manufacturer instructions, with use of the CellDetect kit, urine samples were processed. Following centrifugal separation and smear preparation, cellular staining was performed with the red and green dyes in the kit - nuclear staining with hematoxylin followed by differentiation in HCl/ ethanol. The following categories were included with regard to CellDetect staining results:

• ①positive: the nucleus of the malignant cell is stained in red/ purple, with the cytoplasm of cancer cells being stained in pink. In particular, cancer cells are arranged in clusters, with high nucleus/ cytoplasm ratio (Fig. 1A).
• ② negative: the nuclei of the normal urothelial cells are stained in green, blue or dark purple, with greenish cytoplasm (Fig. 1B). Inflammatory cells are usually stained in purple/ red, with good discrimination on the basis of the morphology.
• ③discordant: Nuclear-to-cytoplasmic ratio >0.5, with mildly irregular nuclei and hyperchromatic, as well as coarse chromatin.

Fig. 1.

A Positive urine smear with bladder cancer (×30). Malignant cells show red/purple nuclei and pink cytoplasm, arranged in clusters with high nucleus-to-cytoplasm ratio. B Negative urine smear of normal subject (×30). Normal urothelial cells show green/blue nuclei and greenish cytoplasm. C Urine smear of discordant subject (×30). Nuclear-to-cytoplasmic ratio >0.5, with mildly irregular nuclei and hyperchromatic, coarse chromatin.

Cystoscopy and histopathological evaluation

All patients underwent white light cystoscopy examination routinely. Any suspicious lesions or visible tumors were biopsied or resected transurethrally for histopathological evaluation. The pathological evaluation was based on the TNM staging system and the World Health Organization (WHO) grading criteria. The final diagnosis, determined from cystoscopic findings and histopathologic results, was considered as the gold standard for evaluating the performance of the CellDetect test.

Definition of outcomes

A positive recurrence was defined as the presence of histopathologically confirmed malignant tissue. A negative result was defined as either a normal cystoscopic appearance or a histopathologically nonmalignant result.

Patients with no visible disease on cystoscopy but positive CellDetect results were classified as having “discordant” findings. They were further evaluated with random bladder biopsies, as recommended by the EAU guidelines for cases with positive urine cytology and negative cystoscopy, in particular to address potential diagnosis of carcinoma in situ (CIS) [7]. The flow diagram of the study is seen in Fig. 2.

Fig. 2.

Flow diagram of the study procedures

Statistical analysis

Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and overall accuracy were calculated for the CellDetect test, using histopathology as the reference standard. Subgroup analyses were performed to stratify performance by tumor grade (low-grade vs. high-grade including CIS) and EORTC risk category. Receiver operating characteristic (ROC) curve analysis was also performed to assess the overall diagnostic performance, and the area under the curve (AUC) was calculated. A p-value of less than 0.05 was considered statistically significant. All analyses were performed using SPSS version 22.0 (IBM Corp., Armonk, NY, USA).

Results

Patient characteristics

A total of 243 patients under surveillance for NMIBC following TURBT were included in the final analysis. Of these, 176 (72.4%) were male and 67 (27.6%) were female, with a mean age of 68.5 years (range 42–85). Demographics are detailed in Table 1.

Table 1.

Patient demographics and baseline characteristics

Number of patients	243
Mean age (range; years)	68.5 (42–85)
Male/ female	176/67
Primary histological diagnosis
pTaG1	84
pTaG2	30
pTaG3	19
pT1G2	43
pT1G3	55
CIS	12
EORTC risk stratification
low-risk	112
intermediate-risk	78
high-risk	53

Regarding the primary histopathology, 84 cases (34.6%) were diagnosed with pTaG1 bladder cancer, 30 cases (12.3%) with pTaG2, 19 (7.8%) with pTaG3, 43 cases (17.7%) with pT1G2, 55 cases (22.6%) with pT1G3 and 12 cases (4.9%) with carcinoma in situ (CIS). A total of 65 patients (26.7%) were treated with an intravesical therapy of BCG and 93 (38.3%) with mitomycin-C. According to the EORTC risk stratification, 112 patients (46.1%) were classified as low-risk, 78 (32.1%) as intermediate-risk, and 53 (21.8%) as high-risk at the time of initial diagnosis (Table 1).

Recurrence rate

During the postoperative 24-month follow-up period, 52 patients (21.4%) experienced histologically confirmed NMIBC recurrence. Among these, 36 patients (69.2%) had low-grade (LG) recurrence and 16 (30.8%) had high-grade (HG) recurrence. The majority of recurrences were 37 with stage Ta (71.2%), followed by 14 with T1 (26.9%) and 1 with CIS (1.9%) (Table 2).

Table 2.

Grade, stage, and CellDetect results of the patients with recurrence

Category	Number of patients
Grade
Low Grade	36
High Grade	16
Stage
Ta	37
T1	14
CIS	1
CellDetect result
Positive	40
Negative	12

Diagnostic performance of celldetect test

As shown in Table 3, the overall sensitivity of the CellDetect test was 76.9% (40/52). In addition, the overall specificity was 78.0% (149/191). The positive predictive value (PPV) was 52.6% (40/76), and the negative predictive value (NPV) was 91.2% (149/167). The overall accuracy was 77.8% (189/243). The area under the ROC curve (AUC) was 0.77 (95% CI: 0.70–0.84), indicating moderate diagnostic performance.

Table 3.

Diagnostic performance of the CellDetect test, overall and stratified by tumor grade and stage

	Sensitivity (95% CI)	Specificity (95% CI)	PPV (95% CI)	NPV (95% CI)	Accuracy (95% CI)
overall	76.9% (64.8–85.6)	78.0% (71.7–83.2)	52.6% (41.5–63.4)	91.2% (85.7–94.6)	77.8% (72.3–82.5)
Grade
LG	69.4% (52.9–82.0)	92.4% (86.5–95.9)	64.1% (48.0-77.7)	94.6% (89.1–97.3)	81.7% (75.4–86.8)
HG	93.8% (71.7–98.9)	45.8% (33.5–58.6)	34.1% (21.6–49.3)	96.4% (82.3–99.4)	56.0% (44.7–66.8)
p-value	0.049*
Stage
Ta	73.0% (56.9–84.7)	77.9% (70.3–84.0)	47.4% (35.0–60.0)	91.4% (84.8–95.4)	76.9% (70.1–82.6)
T1	85.7% (60.1–96.0)	78.4% (65.4–87.5)	52.2% (32.4–71.6)	95.2% (84.2–98.7)	80.0% (68.7–88.0)
CIS	100% (20.7–100)	75.0% (30.1–95.4)	50.0% (9.5–90.5)	100% (43.9–100)	80.0% (37.6–96.4)

* Fisher’s Exact Test

CI confidence interval.

Performance stratified by tumor grade and stage

Stratification by tumor grade revealed a critical performance differential. The CellDetect test demonstrated a sensitivity of 69.4% for patients with LG tumors and a significantly higher sensitivity of 93.8% for HG tumors (including CIS) (p = 0.049, Fisher’s Exact Test). When classified by stage, the CellDetect test detected 73.0% (27/37) of Ta tumors and 85.7% (12/14) of T1 tumors. The single case of CIS was also correctly identified by the CellDetect test (100%). The results are also shown in Table 3.

Subgroup analysis by EORTC risk category

The performance of the CellDetect test was further analyzed within EORTC risk strata to evaluate its utility across different risk profiles. Given our endeavors of de-escalating surveillance in low-risk patients and ensuring the safety of high-risk patients, our analysis focused on comparing these two most distinct groups. In the high-risk subgroup (n = 53), where detecting recurrence is most critical, the CellDetect test showed a sensitivity of 88.9% and an NPV of 95.2%. Conversely, in the low-risk patients (n = 112), the test demonstrated a sensitivity of 66.7% with a maintained high NPV of 92.3%. The intermediate-risk group was excluded from this comparative analysis to provide a clearer contrast between the management goals for the lowest and highest-risk patients. The diagnostic performance for these two EORTC risk groups is detailed in Table 4.

Table 4.

Diagnostic performance of the CellDetect test stratified by EORTC risk category

	Sensitivity (95% CI)	Specificity (95% CI)	PPV (95% CI)	NPV (95% CI)	Accuracy (95% CI)
Low-risk	66.7% (38.8–86.3)	92.3% (86.0–96.0)	66.7% (38.8–86.3)	92.3% (86.0–96.0)	87.5% (80.5–92.3)
High-risk	88.9% (56.5–98.0)	45.5% (31.5–60.3)	25.0% (13.1–42.2)	95.2% (77.3–99.2)	52.8% (39.3–66.0)

CI confidence interval.

Discordant findings and their clinical implications

There were 18 patients (7.4% of the cohort) with negative cystoscopy but positive CellDetect results. These patients underwent random bladder biopsies during the same clinical visit or at a scheduled follow-up visit within 3 months, as per our protocol for discordant findings. Of these discordant patients, 6 patients (33.3%) were found to have histologically confirmed recurrence upon biopsy; none of these six were diagnosed with CIS. This finding underscores the potential clinical significance of a discordant CellDetect result as a predictor of potentially imminent recurrence, warranting intensified surveillance. It also suggests that a positive CellDetect test in the absence of cystoscopically visible lesions might in some cases represent molecular-level disease activity preceding visible tumor formation.

Comparative performance with other diagnostic methods

Table 5 provides a comparative overview of the diagnostic performance of CellDetect against standard surveillance methods and other commercially available urinary biomarkers in the context of NMIBC surveillance. This comparison highlights the relative strengths and limitations of each approach, though direct head-to-head comparisons within our cohort were not performed.

Table 5.

Comparative diagnostic performance of various tests for NMIBC surveillance (based on literature data)

Test/Method	Sensitivity	Specificity	PPV	NPV	Level of Evidence*
Cystoscopy (white light)	~ 70–90%	High	High	Moderate	1a
Urine Cytology	~ 30–50% (LG)	> 90%	Moderate-high	Moderate	1a
	~ 60–80% (HG)
CellDetect (this study)	76.9% (overall)	78.0%	52.6%	91.2%	2b
	69.4% (LG)
	93.8% (HG)
Xpert BC Monitor	~ 79–80%	~ 75–85%	~ 50–60%	~ 90–95%	1b
Bladder EpiCheck	Up to 86%	~ 85–90%	~ 60–70%	~ 90–95%	1b
ADX-Bladder	~ 70–80%	~ 75–85%	~ 50–60%	~ 85–90%	2a
UBC	~ 60–70%	~ 70–80%	~ 40–50%	~ 80–85%	2a

*Level of evidence according to Oxford Centre for Evidence-Based Medicine: 1a = systematic review of Randomized controlled trials (RCTs); 1b = individual RCT; 2a = systematic review of cohort studies; 2b = individual cohort study.

Discussion

This study evaluated the diagnostic performance of a novel urinary marker- the CellDetect test- in the surveillance of patients following TURBT for NMIBC. CellDetect is a visually-based, non-invasive diagnostic test that utilizes a proprietary dual-staining technology to distinguish between normal and malignant urothelial cells in voided urine samples [8]. Our findings demonstrated that CellDetect, in a real-world clinical setting, exhibits encouraging diagnostic performance, particularly its high NPV and superior sensitivity for HG recurrences, providing a substantive basis for further investigation of its potential role in optimizing follow-up strategies.

The most pivotal finding of our study was the high overall NPV of 91.2% for the CellDetect test. This indicates that a negative CellDetect result could reliably exclude the presence of recurrent disease with over 90% accuracy in this cohort. In clinical practice, a high NPV is of importance as it may provide the urologist with increased confidence to consider extending the interval between scheduled cystoscopies for patients with consistently negative tests, particularly in those classified as low or intermediate-risk. This could lead to a decrease in the number of invasive procedures, thereby potentially lowering healthcare costs and minimizing patient discomfort. This finding is consistent with other studies, such as the multi-institutional validation by Davis et al. [9], who also reported a high NPV. The ability to reliably rule out recurrence non-invasively could complement the current surveillance paradigm. Future studies should explore the feasibility, safety, and cost-effectiveness of integrating CellDetect into standardized risk-adapted protocols.

However, it is crucial to balance this potential benefit with the implications of the modest PPV (52.6%) observed in our study. A positive CellDetect test in this context would be false in nearly half of the cases, potentially leading to additional biopsies, and increased resource utilization if applied uncritically. Therefore, positive results should be interpreted with caution and integrated into a structured follow-up protocol rather than triggering immediate aggressive intervention.

In our study, the CellDetect test demonstrated a markedly higher sensitivity for HG tumors (93.8%) compared to LG tumors (69.4%). This performance profile is clinically relevant [10]. HG tumors (including CIS) carry a significant risk of progression [11]. The ability of CellDetect to detect the majority of these high-risk lesions addresses a critical need in NMIBC follow-up, acting as a potential safety net, particularly for flat lesions like CIS. The even higher sensitivity (88.9%) and NPV (95.2%) within the high-risk patients further highlight its potential value in monitoring this vulnerable group.

While the sensitivity for LG tumors (69.4%) was lower than for HG tumors, it remains substantially superior to the historically poor sensitivity of conventional urinary cytology for low-grade disease (often reported as < 30%) [12]. This enhanced ability to detect LG recurrences, which constitute the majority of recurrent cases, fills a major diagnostic gap. A notable and substantial limitation of the present study is the lack of a direct, contemporaneous comparison with urine cytology within the same cohort. Urine cytology remains the clinical standard non-invasive comparator in bladder cancer surveillance, and its absence prevents a definitive assessment of whether CellDetect provides additive, complementary, or substitutive diagnostic value in real-world clinical settings. Therefore, conclusions regarding the potential role of CellDetect in surveillance de-escalation or clinical decision-making should be moderated and considered preliminary. Furthermore, studies comparing CellDetect to other markers like FISH (e.g., Shang et al. [11]) have shown comparable sensitivity, with CellDetect potentially offering advantages in technical simplicity [13, 14].

Our results align well with two key studies. The multi-institutional validation by Davis et al. [8] reported a similar overall sensitivity (84%) and confirmed the test’s high sensitivity for high-grade tumors (93%), similar to our findings. Furthermore, our observation that 33.3% of “discordant” cases later recurred is consistent with their finding that 21% of initial false positives developed cancer during follow-up, suggesting CellDetect may act as an “anticipatory” marker. The prospective UroFollow trial [5], which focused on low/intermediate-risk patients, reported a sensitivity range for CellDetect (47–67%) that is comparable to our performance in the low-grade subgroup (69.4%), and also highlighted the potential value of combining the test with cytology. These comparisons reinforce the clinical utility of CellDetect test in NMIBC surveillance.

A strength of our study methodology was the collection of urine samples without special timing regarding intravesical therapy, reflecting real-world conditions. The fact that the CellDetect test maintained robust performance metrics under these conditions is a significant clinical advantage. It suggests that the test’s interpretation may be less confounded by inflammatory changes induced by agents like BCG or chemotherapy, a common pitfall for some other urinary markers [15, 16].

The management of patients with a positive urine test but a negative cystoscopy is a classic clinical dilemma [17, 18]. Our data provided insight: 33.3% of patients with “discordant” findings developed a confirmed recurrence. This finding suggests that a positive CellDetect result with a negative cystoscopy may in some cases be an “anticipatory positive” signal. However, this interpretation requires caution. The 52.6% PPV indicates that nearly half of all positive tests were false positives in this study, which could lead to unnecessary patient anxiety, additional biopsies, and increased resource utilization if applied uncritically. Therefore, such a result warrants a cautious and structured follow-up protocol rather than immediate aggressive intervention. This phenomenon has been observed in other studies with CellDetect [8].

Our findings should also be contextualized within the evolving landscape of risk-adapted NMIBC management and bladder-preserving strategies. Recent literature, such as the work by Busetto et al. on risk-adapted strategies and Finati et al. on bladder preservation after BCG failure, underscores the importance of developing reliable tools to personalize follow-up and optimize outcomes [19, 20]. CellDetect, with its high NPV and good sensitivity for high-grade disease, could potentially contribute to such personalized approaches, though prospective validation is needed.

Limitations

This study has several important limitations. Its retrospective, single-center design with a relatively small cohort and short follow-up period of 24 months makes it vulnerable to selection bias and limits the statistical power of subgroup analyses. The 24-month follow-up is particularly limiting for assessing the long-term predictive value of the test and its utility in surveillance de-escalation strategies for a disease with potential late recurrences. The results might reflect local patient demographics and clinical practices, potentially limiting generalizability. Not all patients with a positive CellDetect test underwent random biopsy, which introduces a potential verification bias, particularly in the assessment of discordant cases. Furthermore, the absence of a direct comparison with urine cytology or other newer biomarkers in this cohort represents a major limitation, as it prevents objective assessment of CellDetect’s additive or comparative value. We did not evaluate critical practical aspects such as inter-observer variability in test interpretation, the learning curve for technicians, detailed cost-analysis, or long-term clinical outcomes of a CellDetect-guided strategy. External validation through prospective, multi-center studies with longer follow-up and head-to-head comparisons is warranted to confirm these findings and establish the test’s role in clinical pathways.

Conclusion

Our retrospective study suggests that the CellDetect test is a potentially useful non-invasive tool in post-TURBT surveillance. Its high NPV might help clinicians safely consider de-escalating surveillance intensity for some patients, thereby reducing the burden of cystoscopy. Its high sensitivity for HG tumors and CIS could provide a useful safeguard for high-risk patients. The test’s performance appears unimpeded by intravesical therapy, and its discordant results may hold predictive value. However, the modest PPV and limitations inherent to the study design- particularly the short follow-up and lack of direct comparison with standard cytology- necessitate cautious interpretation. The CellDetect test represents a promising candidate for advancing towards a more personalized, risk-adapted follow-up strategy for NMIBC, but its integration into routine practice requires further prospective validation and assessment of cost-effectiveness and reproducibility.

Abbreviations

NMIBC

Non-muscle-invasive bladder cancer

TURBT

Transurethral resection of bladder tumor

PPV

Positive predictive value

NPV

Negative predictive value

EORTC

European Organization for Research and Treatment of Cancer

HG

High-grade

LG

Low-grade

BC

Bladder cancer

MIBC

Muscle-invasive bladder cancer

EAU

European Association of Urology

CIS

Carcinoma in situ

WHO

World Health Organization

ROC

Receiver operating characteristic

AUC

Area under the curve

CI

Confidence interval

FISH

Fluorescent in situ hybridization

BCG

Bacille Calmette-Guérin

UBC

Urinary bladder cancer antigen test

RCT

Randomized controlled trial

Authors’ contributions

Writing, data collection and conceptualization: Wenbo Gao.; methodology and review: Jingjing Yu.

Funding

This project received no funding.

Data availability

All data and materials involved in this study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This study was conducted in compliance with the Declaration of Helsinki. The study protocol was approved by the Medical Ethics Committee of Ningbo Urology and Nephrology Hospital. Informed consent was obtained from all individual participants included in the study. Patient data were anonymized to ensure confidentiality.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.