Gulanikar 1991.

Methods	Study design: parallel RCT Time frame: 1982 to 1992 Follow‐up period: 5 years
Participants	Country: Canada Setting: multicentre (14) First and subsequent cadaveric or living kidney transplant; functioning graft 90 days after transplantation, with SCr < 2.5 mg/d Number (randomised/analysed): withdrawal group (260/260); maintenance group (263/263) Mean age ± SD (years): withdrawal group (39 ± 1); maintenance group (40 ± 1) Sex (female): withdrawal group (35%); maintenance group (41%) Donor source (% living donors): not reported Exclusion criteria: acute rejection in previous 2 weeks; malignancy
Interventions	Treatment group Steroid withdrawal after at least 90 days Control group Steroid maintenance Baseline Immunosuppression CsA: twice daily adjusted to 12‐h trough levels between 75 to 200 ng/mL Steroids Prednisone: from day 1 after transplantation 1 mg/kg on alternate days, reduced by 5 mg (when clinical conditions allowed) until a dosage of 0.3 mg/kg
Outcomes	Mortality Graft loss NODAT Infection CMV infection Malignancy Cardiovascular event SCr (mg/dL) CrCl (mL/min)
Notes	Did not report number screened for eligibility Number of patients discontinued treatment Withdrawal group: 143 patients; cessation by physician (45), decoded on request (34), no test drug given (33), CsA stopped (15), noncompliance (15), technical withdrawal (1) Maintenance group: 123 patients; because of cessation by physician (33), decoded on request (32), no test drug given (25), CsA stopped (18), noncompliance (14), technical withdrawal (1)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Stated 'randomised blocks of various sizes were generated and used to attain a balanced, restricted randomization according to treatment centre. The order of randomization did not have a repeating sequence'
Allocation concealment (selection bias)	Low risk	Stated 'Physicians did not know the randomization number until the patient was enrolled, and the code was not broken until the analysis'
Blinding (performance bias and detection bias) All outcomes	Low risk	Stated '...the code was not broken until the analysis. Patients were randomly assigned at 90 days to receive either a placebo or prednisone by means of a process that prevented prior knowledge of their treatment group'
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Stated 'The study was doubly blinded. The placebo and prednisone were prepared in an indistinguishable form and dispensed as coded therapy'
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Double‐blinded placebo controlled, outcomes are objective hard endpoints
Incomplete outcome data (attrition bias) All outcomes	Low risk	Stated 'No patients were excluded after entry (as distinct from withdrawals in the survival analysis) or lost to follow‐up.'; ITT analysis performed
Selective reporting (reporting bias)	High risk	Acute rejection not reported
Other bias	High risk	This work was supported by Sandoz Ltd., Basel, Switzerland, Sandoz Canada Inc., Dorval, Que., Upjohn Ltd., Kalamazoo, Mich., the Richard and Jean Ivey Fund, London, Ont., the Michael Fung Endowment Fund, London, Ont., the Claudine Keown Endowment Fund, London, Ont., the University Hospital Transplant Research Fund, London, Ont., Robarts Research Institute endowment funds and the City of London, Ont