Negrier 2011

Methods	Study design: 3‐arm, parallel‐group, open‐label RCT. Study dates: randomization from May 2008 to May 2009, median follow‐up 23.2 months, study completion date February 2012. Setting: multicentre (24 centres), national, phase II. Country: France.
Participants	Inclusion criteria: histologically confirmed mRCC of all histological subtypes except papillary carcinomas; aged ≥ 18 years; ECOG Performance Status 0 to 2; measurable metastases; liver, renal and haematological functions in the range of 1.5 to 2 times above or below normal values; normal lipid and glycaemic concentrations; normal cardiac function within 6 weeks before randomization. Exclusion criteria: brain metastases, hypertension, systemic treatment for the disease, history of arterial or venous thrombosis in the past 6 months. Sample size: 171. Age (years, median with range): group 1: 62 (40 to 79); group 0a: 62 (33 to 83); group 0b: 61 (33 to 83). Sex (M/F, %): group 1: 66/34; group 0a: 74/26; group 0b: 76/24. Prognostic factors (all randomized participants): performance status (ECOG 0 or 1/2, %): 88/12; prior nephrectomy (%): 87; prior systemic therapies: excluded; risk prognosis (good/intermediate/poor, %): 34/53/13.
Interventions	Group 1 (n = 41): IFN‐α + bevacizumab IFN‐α 9 mIU SC 3 times/week + bevacizumab 10 mg/kg IV every 2 weeks. Group 0 (n = 42): sunitinib Sunitinib 50 mg/day for 4 weeks, followed by 2 weeks off. Group 0a (n = 88): temsirolimus + bevacizumab (excluded, not standard treatment). Temsirolimus 25 mg IV weekly + bevacizumab 10 mg/kg IV every 2 weeks. Treatments continued until disease progression, unacceptable toxicity or protocol violation. Cointerventions (standard‐of‐care according to local practice): not specified.
Outcomes	OS (secondary endpoint) How measured: time from randomization to death from any cause. Time points measured: during follow‐up. Time points reported: 12‐months OS (study is ongoing for long‐term OS). Subgroups: not reported. AEs, grade ≥ 3 (secondary endpoint) How measured:participants on study medication were assessed (NCI‐CTCAE v.3.0), data safety monitoring committee. Time points measured: at day 15 and then at least every 6 weeks over 48 weeks. Time points reported: main types of AEs (all grades and grade ≥ 3), frequency of AEs and SAEs ≥ 3. Subgroups: not reported. QoL not evaluated PFS (primary endpoint) How measured: time from randomization to disease progression or death from any cause (central reviewed data), 4 follow‐up CT scans according to RECIST 1.0. Time points measured: baseline and then every 12 weeks over 48 weeks with 4 follow‐up CT scans. Time points reported: Kaplan‐Meier survival curves over up to 30 weeks, median PFS with 95% CI. Subgroups: not reported. Tumour remission (secondary endpoint) How measured: thoracic, abdominal and pelvic CT scan, brain MRI or CT and bone scan. Time points measured: baseline and then every 12 weeks. Time points reported: best response. Subgroups: not reported.
Funding sources	French Ministry of Health and Wyeth Pharmaceuticals.
Declarations of interest	SN: honoraria from Novartis, Wyeth, Pfizer, GlaxoSmithKline and Roche; research funding from Wyeth, Roche and Novartis. DP: honoraria from Bayer, Eli Lilly and Roche. JOB: honoraria from Amgen; consultant with Novartis. LG, BL: honoraria from Novartis. BE: honoraria from Bayer, Roche, Pfizer, Genentech, Novartis, GlaxoSmithKline and Aveo; consultant with Bayer, Pfizer and Roche. All other authors declared no conflicts of interest.
Notes	Registration: NCT00619268 (TORAVA).
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Low risk of selection bias assumed, computer‐generated list, permutated blocks, stratification by participating centre and performance status.
Allocation concealment (selection bias)	Low risk	Low risk of selection bias assumed due to central allocation.
Blinding of participants and personnel (performance bias) Subjective outcomes	High risk	Participants and investigators were unmasked.
Blinding of participants and personnel (performance bias) Objective outcomes	Low risk	No performance bias on OS assumed.
Blinding of outcome assessment (detection bias) Subjective outcomes	Low risk	Masked central review of CT scans done in 89% of all randomized participants.
Blinding of outcome assessment (detection bias) Objective outcomes	Low risk	No detection bias on OS assumed.
Incomplete outcome data (attrition bias) (OS and PFS)	Low risk	No attrition bias assumed due to high completeness of, and similar censoring in, different treatment groups during follow‐up.
Incomplete outcome data (attrition bias) (safety)	Low risk	No attrition bias assumed, all participants who received at least 1 dose of the study drug were included.
Incomplete outcome data (attrition bias) (tumour remission)	Low risk	No attrition bias assumed, response was reported for all participants.
Incomplete outcome data (attrition bias) (QoL)	Unclear risk	Not evaluated.
Selective reporting (reporting bias)	High risk	Information on long‐term OS and QoL not published despite planning in protocol.
Other bias	High risk	Blocked randomization in centres in an unblinded trial, differences in second‐line treatment after study treatment failure because of toxicity or progression with lower rates of second‐line therapies with sunitinib (48%) compared to 68% to 69% in other groups.