3.2. Reference Standard Definition and Follow-Ups
Regarding all studies, the sensitivity and specificity of the BE were compared to a prespecified reference standard, defined as the result of cystoscopy, cytology, and/or histology. The definition of the reference standard and, consequently, of its result, is a key point in the evaluation of the positivity or negativity of the BE. Concerning the case of a positive cystoscopy (suspicious for recurrence), histology was performed, either an endoscopic cold-cup biopsy or a transurethral resection (TURB). Regarding the case of a positive/suspicious wash cytology, subsequent endoscopic mapping biopsies were planned.
To define a ‘negative’ patient, the white-light cystoscopy, wash cytology, and histology must be negative. Particularly, for cytological specimen classification, all studies used the Paris System for Reporting Urinary Cytology. During three of the studies [10
] the wash cytology was considered ‘negative’ when the result was ‘negative for high-grade urothelial carcinoma (HGUC)’ and ‘atypical urothelial cells’, whereas ‘suspicious for HGUC’, ‘HGUC’, and ‘low-grade (LG) intraepithelial neoplasia’ were clustered as ‘positive’. Concerning the case of positive spontaneous or wash cytology with negative cystoscopy and/or histology, the reference standard was evaluated at the subsequent follow-up cystoscopy with multiple random biopsies or targeted biopsies on a suspicious area [17
]. Concurrently, in the study by Pierconti et al. [16
], patients with a cytological diagnosis of ‘atypical urothelial cells’ also were considered ‘positive’ and underwent cystoscopy within three months, with multiple random biopsies in the case of a negative cystoscopy.
Notably, Witjes at al., [10
] used more rigid parameters in the reference standard definition, considering the pathology specimen as the key determinant. When the patient had a positive pathology, the sample was considered ‘positive’; when the pathology was negative, but the cytology was positive, the sample was considered inconclusive, and excluded from the final analysis. Similarly, samples with positive or equivocal cystoscopic or cytologic results without a subsequent confirmatory pathology also were classified as inconclusive and excluded from the final analysis.
Patient follow-ups after BE varied from six months [17
] to 12 months [11
]. Follow-up data was not available from one study [10
3.4. Results: Sensitivity, Specificity, and Predictive Values Analysis
The validation study by Wasserstrom et al. [15
] applied BE on a sample population of 222 patients. Histological diagnosis of bladder cancer was possible for 40 patients (18%), and BE correctly identified 36 malignancies, with an overall sensitivity of 90%. Notably, BE sensitivity changed based on staging and grading, with a higher sensitivity in higher stages (81%, 100%, 100%, 91%, in Ta, T1, T2 (tumor invading muscularis) and CIS, respectively), and in higher grades (84% in low-grade, and 95% in high-grade tumors). The overall BE specificity was 83%, correctly diagnosing 151/182 cancer-free patients, and the NPV was 97%.
Conversely, considering wash cytology as the reference standard, BE showed a greater sensitivity (90% versus 38%, in low and high-grade tumors), but a lower specificity (83% versus 96%).
This means that BE detected all cancers that also were detected by cytology, and that all cancers missed by BE also were missed by cytology. Finally, the authors highlighted that high histological grading was significantly associated with a higher mean EpiScore score, than was low grading (85 versus 69, respectively; p = 0.006).
Witjes et al. [10
] tested BE on 440 patients, performing statistical analysis on 353 of them (87 subjects excluded due to inconclusive diagnosis according to the reference standard, no BE results, or both). Forty-four out of three hundred and fifty-three patients (12.5%) were ‘positive’ according to the reference standard.
The overall sensitivity was 68% (30/44) and the overall specificity was 88% (272/309). The test performance did not vary with age, gender, smoking habits, occupational exposure, and treatment for recent recurrence (stopped or ongoing). Like the previous study, BE sensitivity varied according to the staging and grading, with a higher sensitivity in higher stages (52%, 100%, 100% for Ta, T1 and CIS, respectively) and grades (40% in low-grade, and 89% in high-grade tumors). NPV was high for the entire patient cohort (95%). Notably, BE could exclude the presence of high-grade NMIBCs with a NPV of 99%, and, conversely, it could detect the presence of high-grade NMIBCs with a sensitivity of 92%. The AUC was 0.82 for all tumors (including low-grade Ta), and 0.94 when excluding low-grade recurrences.
Lozano et al. [14
] recently updated the results of the previous study, with the addition of 382 patients, 304 of whom were recorded for results analysis (final cohort: 657 patients). Eighty patients were ‘positive’ as compared to the reference standard. The results were similar between the first and the second analyses (overall sensitivity: 62.5%; sensitivity when excluding low-grade NMIBCs: 86%; specificity: 86%). NPV was 94%. Notably, BE outperformed cytology in all categories (all-grade, low-grade, and high-grade NMIBCs). Moreover, in high-grade NMIBCs, BE showed a higher sensitivity than cystoscopy (86% versus 73%; p
= 0.113), although not significant. Conversely, BE sensitivity in the low-grade disease was significantly lower than cystoscopy (33% versus 94%; p
A secondary external independent analysis of the study by Witjes et al. [11
] tested BE performance on 357 patients, with 49/357 (13.7%) intravesical disease recurrences. The study showed an overall sensitivity and specificity of 67% and 88%, respectively. The sensitivity rose to 89% for the detection of high-grade NMIBCs. NPV was 94% for any-grade NMIBCs, and 99% for high-grade NMIBCs.
Considering univariable logistic regression analyses, a one-point increase in the EpiScore corresponded to a 4% increase in the risk of any-grade bladder cancer, and to an 8% increase in the risk of high-grade NMIBC. Using multivariable logistic regression analysis, a positive BE result was independently associated with the presence of any-grade (odds ratio, OR = 18.1, 95% CI 8.66–40.2; p < 0.001) and high-grade NMIBC (OR = 78.3, 95% CI 19.2–547; p < 0.001).
Finally, using explorative analysis, the performance of BE, evaluated by the AUC, was not affected by epidemiological features (age, gender, time from last recurrence, ongoing intravesical therapy, pathological stage and grade), and the addition of BE to these clinical variables significantly improved their AUC by 16% for the prediction of any-grade bladder cancer and 22% for the prediction of high-grade NMIBCs (BE AUC: 86% and 96% for any-grade and high-grade NMIBC, respectively).
During the study by Trenti et al. [12
], BE was tested on 243 patients, with 215 evaluable results (28 test failures excluded from the final analysis). Sixty-nine (32%) patients had an NMIBC recurrence. The overall BE sensitivity was 62%, which rose to 83% for high-grade NMIBCs. The overall BE specificity was 86%, and the NPV was 83%. The diagnostic efficacy of BE was good, with an AUC of 0.785. The lower accuracy and lower NPV of BE in this study was probably due to the higher prevalence of NMIBC recurrence.
A subsequent study by the same authors [13
] compared BE to another Real-Time Polymerase Chain Reaction-Based urinary marker (Xpert Bladder Cancer Monitor [18
]), which measures the level of five target messenger RNAs, upregulated in urine samples from patients with bladder cancer. The study enrolled 487 patients, 55 of whom had an invalid BE (total population: 432 patients). Twenty-one percent of the patients had an NMIBC recurrence. The test showed an overall sensitivity of 64% (66% for the Xpert Bladder Cancer Monitor test), which increased to 79% in high-grade NMIBCs (57% for low-grade NMIBCs). The overall specificity was 82% for the BE. The NPV was almost the same for both BE and Xpert Bladder Cancer Monitor (about 89% for both). BE demonstrated a high diagnostic efficacy, with an AUC of 74% (95% CI, 67.2–80.5%).
A study by Righetto et al. [17
] on 88 patients and 72 BE results considered for the statistical analysis, showed a prevalence of recurrent NMIBC in 17.3% of the study population. The sensitivity of BE in all-grade NMIBCs was 76%, which rose to 100% (95% CI, 75.5–100%) for high-grade NMIBCs. The overall specificity of the BE was 90%, the NPV was 97% in all-grade NMIBCs, and 100% in high-grade NMIBCs.
Clinical data regarding the BE performance among the published studies are summarized in Table 1
3.5. Results: BE and Cytology
Few studies have compared the sensitivity, PPV and NPV for both cytology and BE, and only one evaluated the diagnostic accuracy of the combination of the two tests [12
Seen in all studies, the sensitivity of cytology was lower than the BE sensitivity (Table 2
), with percentages ranging from 27% [13
] to 38% maximum [15
]. Particularly, the cytology sensitivity was markedly low in low-grade NMIBCs (0–13% across studies), and increased, while remaining lower than BE sensitivity, in high-grade NMIBCs (50–67% across studies). Notably, the NPV of the BE was higher than cytology NPV in all studies. Thus, the high sensitivity and NPV of BE could be used as an additional tool to improve the sensitivity of cytology.
The EpiScore, as a continuous variable, also could stratify differently according to different results of cytology. This data suggests that different cytological categories also carry a distinct methylation (thus, molecular) pattern.
Pierconti et al. [16
] analyzed the EpiScore results of 374 patients with high-grade NMIBC (268 T1, 106 CIS) according to different cytological results (negative for high-grade urothelial carcinoma, NHGUC; atypical urothelial cells, AUC; suspicious for high-grade urothelial carcinoma, SHGUC; high-grade urothelial carcinoma, HGUC; low-grade urothelial neoplasia, LGUC; unsatisfactory/non-diagnostic), finding that differences between cytological categories also could be based on molecular disparities. Actually, the EpiScore increased from NHGUC to HGUC; an EpiScore <60 correlated with NHGUC (OR = 3.9, 95% CI 1.9–8.1, p
= 0.0003), while an EpiScore ≥60 correlated with SHGUC (OR = 3.8, 95% CI 1.6–8.9; p
= 0.0031) and HGUC (OR = 3.9, 95% CI 1.6–9.5; p
= 0.0027). Moreover, during a one year follow-up, the EpiScore had a sensitivity of 100%, a specificity of 89.9%, and a NPV of 100% in the NHGUC group while, in the HGUC group, the EpiScore reached a sensitivity of 98%, a specificity of 100%, and a NPV of 86%.
A study by Trenti et al. [12
] evaluated the possibility of combining BE and cytology results with the aim of increasing the diagnostic accuracy. The authors showed that the overall sensitivity of cytology (33%) could significantly rise in combination with BE (67% for the two tests combined). Particularly, the sensitivity for low-grade NMIBCs varied from 8% of cytology to 46% of BE, while their combination yielded an overall sensitivity of 56%. Conversely, the sensitivity for high-risk NMIBCs was 67% for cytology and 83% for BE alone, whereas the combination of the two tests obtained an overall sensitivity of 90%.
The authors did not find a significant advantage in the combination of the two tests regarding PPV and NPV. The PPV was higher for cytology than for BE (92% versus 68%), while the NPV was similar for both (75.8% for cytology, 82.9% for BE). The combination of the two tests failed to reach a significant improvement in PPV and NPV (overall PPV: 69%, overall NPV: 84.5%).
To conclude, the combination of both cytology and BE for the surveillance of patients with NMIBC could increase the sensitivity of the individual tests, especially in high-grade NMIBCs. Consequently, the combination of the two tests could potentially reduce the number and frequency of cystoscopies during follow-ups of patients with NMIBCs, especially for those with low-risk disease. However, we must consider that BE needs a dedicated and equipped laboratory to perform real time-PCR.