Bang 2010.

Methods	International RCT (122 centers in 24 countries), Phase III
Participants	Number of participants: 594 (September 2005 to June 2010 ) Number randomized: Experimental group (Trastuzumab in combination with fluoropyrimidine and cisplatin): 298 Control group (fluoropyrimidine and cisplatin): 296 Number evaluated: Experimental group: 294, age (mean ± SD): 59.4 ± 10.8 years Control group: 290, age (mean ± SD): 58.5 ± 11.2 years Diagnosis: Patients with gastric or gastro‐esophageal junction cancer were eligible for inclusion if their tumors showed over‐expression of HER2 protein by immunohistochemistry or gene amplification by fluorescence in situ hybridization Inclusion: eligible patients should: be ≥ 18 years of age; histologically‐confirmed inoperable locally‐advanced, recurrent, or metastatic adenocarcinoma of the stomach or gastro‐esophageal junction; ECOG performance status 0 – 2; adequate organ function; and measurable or non‐measurable disease; tumors were centrally tested for HER2 status with immunohistochemistry (HercepTest, Dako, Denmark) and fluorescence in situ hybridization (FISH; HER2 FISH pharmDx, Dako)
Interventions	Experimental group : D1 trastuzumab 8 mg/kg i.v. for cycle 1, and then 6 mg/kg i.v. Q3W until disease progression; D1 ‐ 5 800 mg/m² fluorouracil i.v., Q3W for 6 cycles; D1 80 mg/m² cisplatin i.v., Q3W for 6 cycles; D1 ‐ 15 1000 mg/m² capecitabine by mouth twice daily, every 3 weeks for 6 cycles Control group: D1 ‐ 5 800 mg/m² fluorouracil i.v., every 3 weeks for 6 cycles; D1 80 mg/m² cisplatin i.v., every 3 weeks for 6 cycles; D1 ‐ 15 1000 mg/m² capecitabine by mouth twice daily, every three weeks for 6 cycles
Outcomes	Duration of follow‐up (median): 18.6 months (IQR 11 ‐ 25) for experimental group, 17.1 (IQR 9 ‐ 25) for control group OS, FPS, overall response, adverse events
Notes	Sponsor: Hoffmann‐La Roche The sponsor was involved in the study design, data collection, data analysis, results interpretation, and manuscript preparation Study referred to as "ToGA"
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"The randomisation sequence was created by F Hoff mann‐La Roche"
Allocation concealment (selection bias)	Low risk	"Treatment was assigned by use of a randomised block design with block sizes of four patients, via a central interactive voice recognition system (by telephone)"
Blinding of participants and personnel (performance bias) All outcomes	High risk	The open‐label design may introduce performance bias (e.g. through stress‐related mechanisms)
Blinding of outcome assessment (detection bias) OS, Serious adverse events	Low risk	These outcomes were unlikely to be affected by the blinding status of assessors
Blinding of outcome assessment (detection bias) PFS, Response, quality of life & adverse events	Low risk	"Efficacy and safety data were monitored by an independent data monitoring committee."
Incomplete outcome data (attrition bias) All outcomes	Low risk	584 randomized participants were included in analysis (98.3%)
Selective reporting (reporting bias)	Low risk	Prior protocol was available
Other bias	Low risk	No obvious potential source of bias