Types of studies

We included randomised and quasi‐randomised trials. We planned to consider individually randomised trials and cluster‐randomised trials, as well as cross‐over trials (providing separate data were available from the period prior to cross‐over). We excluded non‐randomised trials.

We included studies irrespective of language of publication and publication status.

Types of participants

We included studies of adults (over 18 years of age) of either gender with histologically confirmed Ta and T1 superficial bladder cancer, with or without CIS, treated with TUR.

Participants with NMIBC were categorised as low, intermediate, or high risk in accordance with the EORTC and the European Association of Urology (EAU) classification (Babjuk 2017).

We planned to include studies that evaluated only a subset of relevant participants if separate data were available for the relevant subsets.

Types of interventions

We investigated the following comparisons of experimental versus comparator interventions.

Types of outcome measures

We did not use the measurement of the outcomes included in this review as an eligibility criterion for considering studies.

If we were unable to derive time‐to‐event information, we attempted to assess a dichotomous outcome (e.g. recurrence, mortality) instead. We considered outcomes measured up to and including six months after randomisation as short term, later than six months up to and including 24 months as mid term, and later than 24 months as long term.

Electronic searches

We searched the following sources from their respective inceptions to the current date on 14 March 2016. The date of last search of all databases was 25 August 2016.

• Cochrane Central Register of Controlled Trials (CENTRAL) (Issue 8, 2016) (for the search strategy, see Appendix 1)

• MEDLINE (OvidSP) (1946 to 2016) (Appendix 2)

• Embase (OvidSP) (1974 to 2016) (Appendix 3)

We also searched the following trials registers on 14 March 2016 and again on 25 August 2016.

• ClinicalTrials.gov (www.clinicaltrials.gov/; Appendix 4)

• World Health Organization International Clinical Trials Registry Platform (WHO ICTRP) (apps.who.int/trialsearch/; Appendix 5)

If we detected additional relevant keywords during any of the electronic or other searches, we modified the electronic search strategies to incorporate these terms and documented the changes.

Searching other resources

We attempted to identify other potentially eligible trials or ancillary publications by searching the reference lists of retrieved included studies, reviews, meta‐analyses, and health technology assessment reports. We also contacted study authors of included trials in order to identify any further studies that we may have missed. We contacted intervention manufacturers for ongoing or unpublished trials.

We handsearched abstract proceedings of relevant meetings from the last three years, including: the International Bladder Cancer Network, the American Urological Association, the European Association of Urology, the American Society of Clinical Oncology, and the European Society for Radiotherapy & Oncology.

Selection of studies

We used reference management software to identify and remove potentially duplicate records (EndNote). Two review authors (ARHS, ES) independently scanned the abstracts, titles, or both, of the remaining records retrieved to determine which studies should be assessed further. Two review authors (ARHS, ES) independently investigated all potentially relevant records as full text, mapped records to studies, and classified studies as included, excluded, awaiting classification, or ongoing, in accordance with the criteria for each provided in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011a). We resolved any discrepancies through consensus or recourse to the third review author (NRB). Where resolution of a disagreement was not possible, we designated the study as 'awaiting classification' and contacted the study authors for clarification. We documented reasons for the exclusion of studies that may have reasonably been expected to be included in the review in the Characteristics of excluded studies table. We presented a PRISMA flow diagram showing the process of study selection (Liberati 2009).

Data extraction and management

We developed a dedicated data extraction form that we pilot tested ahead of time.

Two review authors (ARHS, ES) independently abstracted the following information for studies that fulfilled the inclusion criteria.

• Study design

• Study dates (if dates were not available, then that was reported as such)

• Study settings and country

• Participant inclusion and exclusion criteria

• Participant details, baseline demographics

• The number of participants by study and by study arm

• Details of relevant experimental and comparator interventions, such as dose, route, frequency, and duration

• Definitions of relevant outcomes, and method and timing of outcome measurement as well as any relevant subgroups

• Study funding sources

• Declarations of interest by primary investigators

• Study characteristics relevant to 'Risk of bias' assessment

We extracted outcome data relevant to this review as needed for calculation of summary statistics and measures of variance. For dichotomous outcomes, we obtained number of events in each group, total number of participants in each group, summary statistic, and 95% confidence intervals (CI). We attempted to obtain log hazard ratios (HR) with standard errors, and HR with 95% CI for time‐to‐event outcomes. For continuous outcomes, we planned to obtain means and standard deviations (SD), or data necessary to calculate this information.

We resolved any disagreements by discussion, or, if required, by consulting the third review author (NRB).

We provided information about potentially relevant ongoing studies, including the trial identifier, in the Characteristics of ongoing studies table.

We attempted to contact authors of included studies to obtain key missing data as needed.

Assessment of risk of bias in included studies

Two review authors (ARHS, ES) independently assessed the risk of bias of each included study. We resolved disagreements by consensus, or, if required, by consulting the third review author (NRB).

We assessed risk of bias using the Cochrane tool for assessing risk of bias (Higgins 2011b). We assessed the following 'Risk of bias' domains.

• 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 judged the domains as 'low risk', 'high risk', or 'unclear risk', and evaluated individual bias items as described in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011b).

We presented a 'Risk of bias' summary figure to illustrate these findings.

For performance bias (blinding of participants and personnel) and detection bias (blinding of outcome assessment), we evaluated the risk of bias separately for each outcome, and grouped outcomes according to whether they were measured subjectively or objectively when reporting our findings in the 'Risk of bias' tables. We defined the following endpoints as subjective outcomes: time‐to‐recurrence, time‐to‐progression, discontinuation of therapy due to adverse events, disease‐specific survival, adverse events, and disease‐specific quality of life. We defined the following endpoint as objective: time‐to‐death.

We also assessed attrition bias (incomplete outcome data) on an outcome‐specific basis, and presented the judgement for each outcome separately when reporting our findings in the 'Risk of bias' tables.

We further summarised the risk of bias across domains for each outcome in each included study, as well as across studies and domains for each outcome.

Measures of treatment effect

For dichotomous data, we calculated risk ratios (RR) with 95% CIs. We calculated HR and corresponding 95% CI to assess the intervention effect for time‐to‐event outcomes. We planned to express continuous data as mean differences (MD) with 95% CIs, unless different studies used different measures to assess the same outcome, in which case we planned to express data as standardised mean differences (SMD) with 95% CIs. We analysed the data using Review Manager 5 software (RevMan).

Unit of analysis issues

The unit of analysis was the individual participant. We handled trials with more than two intervention groups for inclusion in the review in accordance with guidance provided in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011c).

Dealing with missing data

We attempted to obtain missing data from study authors to perform intention‐to‐treat analyses; if data were not available, we performed available‐case analyses. If possible, we planned to investigate attrition rates (e.g. dropouts, losses to follow‐up, and withdrawals) and critically appraise issues of missing data. We did not impute missing data.

Assessment of heterogeneity

In the event of excessive heterogeneity unexplained by subgroup analyses, we planned not to report outcome results as the pooled effect estimate in a meta‐analysis but instead to provide a narrative description of the results of each study.

We identified heterogeneity (inconsistency) through visual inspection of the forest plots to assess the amount of overlap of CIs, and the I² statistic, which quantifies inconsistency across studies, to assess the impact of heterogeneity on the meta‐analysis (Higgins 2002; Higgins 2003). We interpreted I² as follows.

• 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 value of I² depended on magnitude and direction of effects, as well as strength of evidence for heterogeneity (e.g. P value from a Chi² test) (Deeks 2011).

When we found heterogeneity, we attempted to determine the possible reasons for it by examining individual study and subgroup characteristics (see Subgroup analysis and investigation of heterogeneity).

Assessment of reporting biases

We attempted to obtain study protocols to assess for selective outcome reporting.

In future updates of this review, if we include 10 or more studies investigating a particular outcome, we will use funnel plots to assess small‐study effects. There are several possible explanations for the asymmetry of a funnel plot, including true heterogeneity of effect with respect to trial size, poor methodological design (and hence bias of small trials), and publication bias, therefore we will interpret results carefully.

Data synthesis

We summarised data using a random‐effects model. We interpreted random‐effects meta‐analyses with due consideration of the whole distribution of effects. In addition, we performed statistical analyses according to the statistical guidelines contained in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011a). For dichotomous outcomes, we used the Mantel‐Haenszel method; for continuous outcomes, we planned to use the inverse‐variance method; and for time‐to‐event outcomes, we used the inverse‐variance method. We used Review Manager 5 software to perform analyses (RevMan).

Subgroup analysis and investigation of heterogeneity

We expected the following characteristics to introduce clinical heterogeneity, therefore where sufficient data were available, we planned to perform the following predefined subgroup analyses.

• Risk (low risk versus intermediate risk versus high risk) according to the EORTC/EAU risk classification system (Babjuk 2017)

• Dose and schedule of BCG (e.g. standard dose (81 mg weekly for six weeks) versus low dose (27 mg weekly for six weeks) versus very low dose (13.5 mg weekly for six weeks))

• Dose and schedule of IFN‐α (e.g. higher dose (100 million units (MU) weekly for six weeks) versus lower dose (50 MU weekly for six weeks))

If EAU risk categories were not available, and if sufficient data were available, we planned to perform subgroup analyses based on:

• number of tumours (one versus more than one)

• tumour size (< 3 cm versus ≥ 3 cm)

• tumour stage (Ta versus T1)

• presence of CIS (absent or present)

• tumour grade (Grade 1 versus Grades 2 and 3)

• primary versus recurrent disease

Where there were sufficient studies, we used the test for subgroup differences in Review Manager 5 to compare subgroup analyses (RevMan).

Sensitivity analysis

We planned to perform sensitivity analyses based on risk of bias by excluding studies judged to be at 'high risk' or 'unclear risk' of bias for the particular outcome.