Li 2020.

Study characteristics
Methods	Accrual: June 2011 to December 2015 Single centre Phase of trial: 3 Study design: open‐label RCT Country or countries where the trial was conducted: China Median follow‐up: 57.3 months
Participants	Age: median 49, range 22–64 years Nodal status of breast cancer: 37% node positive, 63% node negative Adjuvant or neoadjuvant: adjuvant Notable exclusion criteria: none
Interventions	Arm 1: paclitaxel 150 mg/m2 + carboplatin AUC3 every 2 weeks for 8 cycles Arm 2: epirubicin 80 mg/m2 and cyclophosphamide 600 mg/m2 every 2 weeks for 4 cycles followed by paclitaxel 175 mg/m2 every 2 weeks for 4 cycles
Outcomes	Primary 3‐year DFS rate, defined as the date of randomisation to the date of the first local/distant recurrence (in the absence of other primary malignancies) Secondary OS, defined as the time from randomisation to death due to any cause Toxicity, according to NCI‐CTCAE, version 3.0
Notes	Trial registration record: NCT01378533 All randomised participants were included in analysis. Study did not report assessing the proportional hazards assumption. Funding considerations: funded by the National Key Research and Development Program of China and the Chinese Academy of Medical Science Initiative for Innovative Medicine.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "Simple randomization was conducted … using random allocation sequence" (p.487) but no details provided as to how random sequence was generated.
Allocation concealment (selection bias)	Unclear risk	No information provided at single site.
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Participants and personnel were aware of treatment allocation. This may have been associated with some performance bias but it was not judged to be of serious concern given types of outcomes collected.
Blinding of outcome assessment (detection bias): DFS	Low risk	Lack of blinding unlikely to influence this outcome.
Blinding of outcome assessment (detection bias): OS	Low risk	Lack of blinding unlikely to influence this outcome.
Blinding of outcome assessment (detection bias): toxicity	Low risk	Toxicity outcomes graded using the CTCAE. Although the study was open‐label, grading symptoms using the CTCAE is standardised and involved regular laboratory tests, etc. at the end of each cycle. Therefore, knowing treatment allocation may have had minimal effect on the grading of outcomes.
Selective reporting (reporting bias)	Low risk	All outcomes listed in trial registration record were reported in trial publication. OS was not prespecified but collected later and was a primary outcome for this review.
Incomplete outcome data (attrition bias) All outcomes	Low risk	All participants randomised were included in the analysis irrespective of outcome being measured or treatment actually received.
Other bias	Low risk	None identified.