Di Lorenzo 2010.

Study characteristics
Methods	Study design: multicenter, prospective, randomized, phase 2 trial Setting/country: multicenter in Italy Dates when study was conducted: June 2006 to May 2008
Participants	Ethnicity: NA Inclusion criteria High‐risk NMIBC, based on the European Organization for Research and Treatment of Cancer scoring system, failing BCG therapy, for whom radical cystectomy was indicated, but not conducted because of refusal or ineligibility because of age or comorbidities and high anesthesiologic risk Exclusion criteria Concurrent or previous muscle‐invasive disease Concurrent or previous tumor in the upper urinary tract or prostatic urethra Chronic urinary tract infection, cured or active tuberculosis Any other malignancy except basal cell carcinoma of skin Previous pelvic irradiation Creatinine, glutamate oxaloacetic transaminase, and glutamate pyruvic transaminase higher than twice the standard Pregnancy or lactation, and any other disease with immunodeficiency Total number of participants randomly assigned Screened: 92 Eligible: 80 Disease type: recurrent (1‐course BCG failure) high‐risk disease Intervention: gemcitabine Number of all participants randomly assigned: 40 Age: mean 69.3 (SD 8.4) years Gender (men/women): 27/13 Tumor T stage Ta: 10; T1: 30 Tumor grade (low/high) 11/29 Tumor size (< 3 cm/≥ 3 cm): 15/25 Number of tumors (1/> 1): 30/10 Comparator: BCG Number of all participants randomly assigned: 40 Age: mean 71.4 (SD 7.9) years Gender (men/women): 22/18 Tumor T stage Ta: 8; T1: 32 Tumor grade (low/high) 13/27 Tumor size (< 3 cm/≥ 3 cm): 17/23 Number of tumors (1/> 1): 8/32
Interventions	Intervention: gemcitabine Gemcitabine 2000 mg/50 mL saline Treatment after 4–6 weeks from the last TUR, twice weekly (days 1 and 4) for 6 consecutive weeks (induction course), and then weekly for 3 consecutive weeks at 3, 6, and 12 months Comparator: BCG BCG: Connaught strain, 81 mg/50 mL saline Treatment after 4–6 weeks from the last TUR, 6‐week induction course and then each week for 3 weeks, at 3, 6, and 12 months. Follow‐up: gemcitabine: median 15.2 months (range 6–22); BCG: median 15.8 months (range 7–21)
Outcomes	Primary outcomes Recurrence rate (percentage of participants with recurrence) at 1‐year follow‐up How measured: cystoscopy and pathologically confirmed after TUR Time point measured: cytology and cystoscopy performed at 3‐month intervals Time point reported: 1 year Secondary outcomes Time to recurrence; progression rate; time to progression; toxicity How measured Time to recurrence; progression rate; time to progression: cystoscopy and pathologically confirmed after TUR Toxicity: CTCAE version 3.0 Time point measured: 3‐month intervals after initial treatment Time point reported: likely cumulative Subgroup analysis: none
Funding Sources	NA
Declarations of interest	None
Notes	Protocol: NA Language of publication: English
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote from publication: "By using a central computer‐generated randomisation list." Comment: we considered this method of random sequence generation to have low risk of bias.
Allocation concealment (selection bias)	Low risk	Quote from publication: "By using a central computer‐generated randomisation list." Comment: central randomization; this method may ensure allocation concealment.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Quote from publication: "An open label study design was used, that is, patients and investigators were not masked as to the drugs they were assigned."
Blinding of outcome assessment (detection bias) Subjective outcomes	High risk	Quote from publication: "An open label study design was used, that is, patients and investigators were not masked as to the drugs they were assigned."
Blinding of outcome assessment (detection bias) Objective outcomes	Low risk	Comment: objective outcomes not likely affected by lack of blinding.
Incomplete outcome data (attrition bias) Time to recurrence	Low risk	Comment: no loss to follow‐up; all randomized participants were included in the analysis.
Incomplete outcome data (attrition bias) Time to progression	Low risk	Comment: the study did not investigate this outcome, but reported several progressions from all randomized participants. We approximated the effect estimates (HR) using an indirect method (Parmar 1998). Thus, low risk of attrition bias.
Incomplete outcome data (attrition bias) Grade III–V adverse events	Low risk	Comment: no loss to follow‐up; all randomized participants were included in the analysis.
Incomplete outcome data (attrition bias) Time to death from bladder cancer	Low risk	Comment: the study did not investigate this outcome, but reported several bladder cancer death from all randomized participants. We approximated the effect estimates (HR) using an indirect method (Tierney 2007). Thus, low risk of attrition bias.
Incomplete outcome data (attrition bias) Grade I or II adverse events	Low risk	Comment: no loss to follow‐up; all randomized participants were included in the analysis.
Selective reporting (reporting bias)	Unclear risk	Comment: the protocol was not published (author response); time to first recurrence was not defined in the method section.
Other bias	High risk	Comment: treatment schedule differed between groups; gemcitabine was given twice weekly for 6 weeks, then weekly for 3 consecutive weeks at 3, 6, and 12 months; BCG was given weekly for 6 weeks, then 3 weekly instillations at 3, 6, 9, and 12 months.