Types of studies

We will include all relevant randomized controlled trials, in which individual participants are randomized. 

We will exclude quasi‐experimental studies, given the lack of random assignment at the individual‐level. Given the nature of the interventions in question, we will exclude cross‐over trials from our review, since both the comparator and the experimental interventions involve removal of bladder tumors, and this will present a serious carry‐over effect (Higgins 2011). Cluster‐randomized trials are also an inappropriate study design in the context of our research question, as individual allocation of the interventions is both feasible and desirable. Given that individually randomized trials are more appropriate, we will exclude cluster‐randomized trials from our review (Puffer 2005).

Types of participants

The eligible population is defined as adults, aged 18 and over, with a suspected or established diagnosis of primary or recurrent urothelial carcinoma of the bladder, based on one of the following:

• bladder mass or abnormal bladder mucosa findings per clinic‐based WLC

• bladder mass on cross‐sectional imaging, such as bladder filling defects and hydronephrosis

• positive or atypical urinary cytology

• positive fluorescence in situ hybridization test

We will exclude participants undergoing NBI‐guided surveillance, as we are only considering the use of NBI in the treatment setting. We will also exclude participants with distant metastatic disease. 

If studies include multiple participant groups or interventions, we will only include the subset of participants of interest. If multiple articles are published by the same group with the same participant cohort, we will merge and analyze relevant data from each article as one study.

Types of interventions

We plan to investigate the following comparisons. Concomitant interventions will have to be the same in the experimental and comparator groups to establish fair comparisons.

Types of outcome measures

Measurement of outcomes assessed in this review will not be used as an eligibility criterion.

Electronic searches

We will search for relevant studies in the following ten databases from their respective date of inception, using the search strategies outline in the Appendices. 

1. Cochrane Library (via Wiley); Appendix 1

2. International Pharmaceutical Abstracts (via Ovid); Appendix 2

3. MEDLINE (via Ovid); Appendix 3

4. Embase (via embase.com); Appendix 4

5. Web of Science Core Collection (via Clarivate); Appendix 5

6. Scopus (via Scous.com); Appendix 6

7. Latin American and Caribbean Health Science Information database (LILACS; lilacs.bvsalud.org/en/)

8. Open Grey (www.opengrey.eu/); Appendix 7

9. ClinicalTrials.gov (www.clinicaltrials.gov/); Appendix 8

10. World Health Organization (WHO) International Clinical Trials Registry Platform (apps.who.int/trialssearch/); Appendix 9

We will incorporate additional relevant key words found during these searches into our search strategies, and document changes accordingly.

Searching other resources

We will attempt to identify other potentially eligible studies by searching the reference lists of included publications. We will also contact study authors of included publications to identify any further studies that we may have missed. We will contact device manufacturers for ongoing or unpublished trials.

Selection of studies

After removing duplicate records, two review authors (LL, ST) will independently scan the titles and abstracts of studies identified by the electronic search for eligibility. The two review authors (LL, ST) will then screen the full‐text reports for all potentially eligible studies, according to predefined criteria. They will resolve any discrepancies through consensus or arbitration of a third review author (GL). For studies identified in trial registries, we will contact the authors or institutions recorded in the registry for trial reports. We will translate papers that are published in languages other than English to assess for eligibility. We will present a PRISMA flowchart, showing study selection, including reasons for exclusion of studies (Liberati 2009).

Data extraction and management

Two review authors (LL; ST or GL) will conduct data extraction using an extraction form developed, based on the recommendations of the Cochrane Handbook for Systematic Reviews of Interventions (Li 2020). We will resolve any disagreements by consensus, or by arbitration of a third review author (GL or LH) if needed.

We will extract the following data:

• Study information: author, title, source, publication date, publication type, language, duplicate publications, source of funding, authors’ conflict of interest

• Study characteristics: study design, randomization method, number of study center(s), country of study center(s), inclusion and exclusion criteria, subgroup analysis, statistical methods, period of enrollment, follow‐up period

• Participants characteristics: number of participants, number of participants per study arm, age, gender, ethnicity, clinical stage of disease (presentation, focality, tumor size), number of participants recruited, allocated, evaluated

• Intervention and comparator information: name, frequency, duration of treatment, adjuvant therapy, re‐intervention, follow‐up

• Outcomes: according to the review's predefined primary and secondary outcomes (including tumor stage and grade), events of intervention and comparator, timing of outcome measurement, number of re‐resections

We will extract relevant outcomes data needed to calculate summary statistics and measures of variance. For dichotomous outcomes, we will obtain numbers of events and totals to populate a 2 x 2 table, and calculate summary statistics with corresponding measures of variance. For continuous outcomes, we will obtain means and standard deviations or data necessary to calculate this information. For time‐to‐event outcomes, we will extract hazard ratios with corresponding measures of variance or data necessary to calculate this information.

Potentially relevant ongoing studies will be presented in the table ’Characteristics of ongoing studies’.

Assessment of risk of bias in included studies

Two review authors (LL, ST) will independently assess the risk of bias of each included study, using the Cochrane 'Risk of bias' assessment tool (Higgins 2011). They will resolve disagreements by consensus, or through arbitration by a third review author (GL). We will determine if risk of bias is low, high, or unclear, and present our findings in a 'Risk of bias' summary figure. We will assess for the following biases: 

• Random sequence generation (selection bias);

• Allocation concealment (selection bias);

• Blinding of participants and personnel (performance bias);

• Blinding of outcome assessment (detection bias);

• Incomplete outcome data (attrition bias);

• Selective reporting (reporting bias);

• Other sources of bias.

We will evaluate the risk of performance bias and detection bias separately for each outcome. We will group outcomes according to whether they were measured subjectively or objectively in the 'Risk of bias' tables.

Performance bias ‐ susceptible

• We will consider that all outcomes are similarly susceptible to performance bias

Detection bias ‐ susceptible

• Time to disease recurrence (time‐to‐event outcome) 

• Time to disease progression (time‐to‐event outcome)

• Time to death from bladder cancer (time‐to‐event outcome)

Detection bias ‐ not susceptible

• Adverse event, minor: (Clavien‐Dindo grade I or II; dichotomous outcome)

• Adverse event, major: (Clavien‐Dindo grade III, IV, or V; dichotomous outcome)

We will assess Incomplete outcome data, or attrition bias on an outcome‐specific basis. We will consider the rate of attrition as low (less than 10%), unclear (between 10% and 20%), and high (greater than or equal to 20%).

We will assess selective reporting bias on a study‐specific basis. We will only consider the risk of selective reporting bias as low if we can identify an a priori protocol, and if the analyses and outcomes match what the investigators preplanned.

We will summarize the risk of bias within and across outcomes and studies in graphs. We will use study‐specific risk of bias assessments to inform the preplanned sensitivity analyses. 

Measures of treatment effect

We will express outcomes with dichotomous data with risk ratios, with 95% confidence intervals (CIs). We will express outcomes with time‐to‐event data with hazard ratios with 95% CIs.

Unit of analysis issues

The unit of analysis will be the individual participant. For studies with repeated outcome measures, we will perform time to event analysis. If this is not possible, we will plan to define outcome measures as short term (<12 months) versus long term (13‐24 months) (Higgins 2020).

Dealing with missing data

We will contact study investigators for missing data. We will only analyze available data; we will not impute missing data. We will investigate attrition rates (e.g. dropouts, losses to follow‐up, and withdrawals), and critically appraise issues of missing data and imputation methods (e.g. last observation carried forward, if used by the study authors). We will address the impact of missing data on the findings of the review in the Discussion section (Deeks 2020).

We will conduct intention‐to‐treat analysis whenever possible. If intention‐to‐treat analysis is not possible, we will conduct as‐treated and per‐protocol population analyses. We will consider this a potential source of bias.

Assessment of heterogeneity

In the event of substantial clinical, methodological, or statistical heterogeneity, unexplained by subgroup analyses, we will not include such results in the pooled effect estimate in meta‐analysis. Instead, we will provide a narrative description of the results of each study. 

Using forest plots,

We will identify heterogeneity by assessing overlaps in CIs in forest plots. We will also assess the impact of heterogeneity on the meta‐analysis using the I² statistic (Higgins 2002; Higgins 2003). We will interpret the I² statistic as follows (Deeks 2020):

• 0% to 40%: may not be important

• 30% to 60%: may indicate moderate heterogeneity

• 50% to 90%: may indicate substantial heterogeneity

• 75% to 100%: considerable heterogeneity

The importance of the observed the I² statistic will depends on the magnitude and direction of effects, and the strength of evidence for heterogeneity. If we identify heterogeneity, we will attempt to determine possible reasons for it by examining individual study and subgroup characteristics.

Assessment of reporting biases

We will attempt to obtain study protocols to assess for selective outcome reporting. If at least 10 studies are available for meta‐analysis, we will assess for publication bias by creating and visually inspecting funnel plots.

Data synthesis

If there are sufficient studies available for meta‐analysis, we will conduct a meta‐analysis with a random‐effects model for pooling data (Wood 2008). We will conduct statistical analyses according to the statistical guidelines contained in the Cochrane Handbook for Systematic Reviews of Interventions (Li 2020). We will analyze dichotomous outcomes with the Mantel‐Haenszel method and continuous outcomes with the inverse variance method. We will analyze time‐to‐event outcomes using the generic inverse variance method. We will use Review Manager 5 software to conduct all analyses (Review Manager 2020).

Subgroup analysis and investigation of heterogeneity

Certain tumor characteristics may impact outcomes. If there are sufficient data, we will conduct the following subgroup analyses:

• Setting: primary versus recurrent bladder cancer;

• Multifocality: solitary versus multiple lesions of bladder cancer;

• Tumor size: 3 cm or less versus larger than 3 cm;

• Stage: positive cytology, or history of carcinoma in situ (CIS; in the case of recurrent disease), or both, versus negative cytology, or the absence of history of CIS, or both.

The rationale underlying these subgroup analyses is as follows:

• Setting: per the European Organization for Research and Treatment of Cancer (EORTC) criteria, the setting of primary versus recurrent bladder cancer (primary with or without one recurrence, versus  > one recurrence) can affect the risk of recurrence and progression (Sylvester 2006)

• Multifocality: per the EORTC criteria, the number of tumors (1, 2 to 7, versus ≥ 8) can affect the risk of recurrence and progression (Sylvester 2006)

• Tumor size: per the EORTC criteria, the size of tumors (< 3 cm versus ≥ 3 cm) can affect the risk of recurrence and progression (Sylvester 2006)

• Stage: compared to other histological types, the detection of CIS is particularly difficult due to its flat growth within the cell level (Sylvester 2006)

Sensitivity analysis

If there are sufficient studies, we will conduct sensitivity analyses to evaluate differences in methodology that could impact the results of meta‐analyses. We will conduct sensitivity analysis by 1) excluding studies with high risk of bias (Deeks 2020), and 2) by excluding studies in which all participants underwent re‐resection on a routine basis, since routine re‐resection may mitigate potential benefits of NBI.

Summary of findings and assessment of the certainty of the evidence

We will present the overall quality of the evidence for each outcome according to the GRADE approach, which takes into account criteria related to internal validity (risk of bias, inconsistency, imprecision, publication bias), and external validity (directness of results; (Guyatt 2008). Two review authors (LL or GL) will independently rate the quality of evidence for each outcome as high, moderate, low, or very low using GRADEpro GDT (GRADEpro GDT). We will resolve any discrepancies by consensus, or if needed, by arbitration with a third review author (PD). We will present a summary of the evidence in a 'Summary of findings' table, which provides key information about the best estimate of the magnitude of the effect in relative terms and absolute differences; numbers of participants and studies addressing each important outcome; and the rating of the overall confidence in effect estimates for each outcome (Guyatt 2011; Schünemann 2019). If meta‐analysis is not possible, we will present results in a narrative 'Summary of findings' table.

We will present a 'Summary of findings' table reporting the following outcomes listed according to a priority rating established by the clinicians on our team with the input of external experts.

1. Time to disease recurrence 

2. Time to disease progression 

3. Adverse event, major

4. Time to death from bladder cancer 

5. Adverse event, minor