Sandborn 2008.

Methods	Randomized, double‐blind, placebo‐controlled multi‐center trial with a cross‐over design in the USA Patients were randomized 1:1:1:1 to an intravenous or subcutaneous arm, each with a placebo group Within each arm patients crossed over to the treatment or control group after eight weeks
Participants	Participants with clinically active Crohn's disease (N = 104) Inclusion criteria: CDAI 220‐450, age ≥ 18 years. Patients had received at least one of the following in the past: 5‐ASA, antibiotics, corticosteroids, azathioprine, 6‐mercaptopurine, methotrexate, submaximal infliximab doses or regimens, or other anti‐TNF‐α‐agents Exclusion criteria: > 20 mg prednisolone, recent treatment with any investigational agent or an anti‐TNF‐α‐agents, infections, cancer, short‐bowel syndrome, ostomy, obstructive symptoms with strictures
Interventions	Ustekinumab was given at 90 mg s.c. per week over four weeks (n = 25) or 4.5 mg/kg body weight i.v. once (n = 26), each compared to s.c. placebo (n = 26) or i.v. placebo (n = 27)
Outcomes	Primary outcome: Clinical response at week 8 (CDAI decrease ≥ 75 and ≥ 25%) Secondary outcomes: Clinical remission (CDAI < 150), laboratory results including CRP value, adverse events, anti‐drug antibodies, adherence to therapy. Endoscopic response / remission and quality of life were not assessed.
Duration of follow‐up	8 weeks until cross‐over and 20 weeks thereafter
Notes	Only the first part of the cross‐over design was evaluated for this review The placebo controlled study was accompanied by an unblinded study with participants who were non‐responsive to infliximab The unblinded sub‐study was not included in this review since it was not a placebo‐controlled or active comparator study
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer generated adaptive randomization stratified by investigative site
Allocation concealment (selection bias)	Low risk	Centralized randomization scheme
Blinding (performance bias and detection bias) All outcomes	Low risk	Blinding of patients and investigators Identical placebo
Incomplete outcome data (attrition bias) clinical relapse / remission	Low risk	Intention‐to‐treat analysis with worst case assumption Drop‐outs were balanced across interventions with similar reasons for withdrawal
Incomplete outcome data (attrition bias) adverse events	Low risk	Drop‐outs were balanced across interventions with similar reasons for withdrawal
Selective reporting (reporting bias)	Low risk	All assessed outcomes were reported
Other bias	Low risk	The study appears to be free of other sources of bias