Miller 2005.

Methods	Accrual: June 1999 to October 2002 Single‐centre, phase II randomised controlled trial conducted in US Neoadjuvant study Median follow‐up: not followed up (trial publication specifically states that participants were not followed for recurrence or OS)
Participants	Age: median 50 years, range: 30–64 years (arm 1); and 50 years, range: 36–65 years (arm 2) Inflammatory disease: 17% (arm 1), 23% (arm 2) Median tumour size: 5.5 cm (arm 1), 6.0 cm (arm 2) ER‐positive: 57% (arm 1), 57% (arm 2) HER2‐positive: 23% (arm 1), 17% (arm 2) Excluded: prior breast or chest wall radiation, chemotherapy or hormonal therapy
Interventions	Arm 1: docetaxel (40 mg/m²) weekly for 6 cycles followed by doxorubicin (75 mg/m²) every 2 weeks for 3 cycles with filgrastim on days 2–11 Arm 2: doxorubicin (75 mg/m²) every 2 weeks for 3 cycles with filgrastim on days 2–11 followed by docetaxel (40 mg/m²) weekly for 6 cycles
Outcomes	Trial publication did not appear to define outcomes as primary and secondary. Outcomes collected were: Serological and imaging markers of angiogenesis Tumour microvessel density Clinical complete response, defined as the complete disappearance of all evidence of disease by physical examination Clinical partial response, defined as ≥ 50% decrease in the product of the greatest perpendicular tumour diameters pCR, defined as no evidence of invasive malignancy in the breast and lymph node specimens at the time of definitive surgery Toxicities
Notes	Trial not powered to detect differences between treatment arms. Clinical trial registration record not found. Funding considerations: supported by American Cancer Society, American Society of Clinical Oncology, Breast Cancer Research Foundation, Walter Cancer Institute and unrestricted research grants from Aventis and Amgen.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "randomly assigned after stratification for tumour size (≤ 5 cm versus > 5 cm) and clinical axillary node status (positive versus negative)". Comment: there did not appear to be imbalances in baseline characteristics between groups.
Allocation concealment (selection bias)	Unclear risk	No details provided
Blinding of participants and personnel (performance bias) All outcomes	Low risk	No information provided about blinding in the trial publication. Participants in both groups would have received the same drugs though in different order. Therefore, unlikely that this would lead to a material bias in participants' and physicians' behaviours even if unblinded.
Blinding of outcome assessment (detection bias) Toxicity and treatment adherence	Unclear risk	Used formalised toxicity criteria (NCI CTC version 2) and measured a range of toxicity outcomes where in borderline cases some may be affected if unblinded (e.g. neuropathy). Dose‐reductions/delays were based on toxicity assessments. Overall, unblinding may have influenced the physicians' assessments for a subset of outcomes.
Blinding of outcome assessment (detection bias) Neoadjuvant studies only: pCR	Low risk	Study pathologist blinded to treatment.
Incomplete outcome data (attrition bias) All outcomes	Low risk	A CONSORT diagram was not provided. There did not appear to be missing data for the outcomes reported.
Selective reporting (reporting bias)	Low risk	Did not identify clinical trial registry record (trial started recruitment in June 1999 and completed in October 2002). However, the outcomes outlined in the Methods section were reported in the results section in the publication. Also the Methods section clearly stated that survival data were not collected.
Other bias	Low risk	None identified