David John Manton1, Filip Van Kove1,
Martin D. Pickles1, Lindsay W. Turnbull1
1Yorkshire Cancer Research
Centre for MR Investigations,
The accuracy, compared to pathology, of contrast-enhanced MRI-derived, pre-operative maximum tumour diameters was assessed across a range of clinically relevant sub-sets (e.g. tumour type and Her2 status) in a retrospective cohort of women who had undergone neoadjuvant chemotherapy for locally advanced breast cancer. Bland-Altman limits of agreement (LOAs) were not statistically significant within sub-sets so the data-set was analysed as a whole. MRI demonstrated a low false complete response rate (13%), a small bias (-4.85 mm on average) but quite large LOAs (95% of the differences lying between 38.1 and -47.8 mm) thus lending more credibility to its clinical utility.