Chemically-activated luciferase gene expression (CALUX) bioassay and gas chromatography/mass spectrometry (GC/MS) are used to determine dioxin levels in food and humans. Valid measures of the agreement between the two methods would improve interpretation of bioassay results. Paired breast milk samples from 250 mothers, as 11 pooled samples, were analysed by GC/MS for total WHO-TEQ (7 polychlorinated dibenzo-para-dioxins, 10 polychlorinated dibenzofurans and 12 dioxin-like polychlorinated biphenyls) and as individual samples by CALUX. Mean difference between total WHO-TEQ (weighted by TEF system derived in 1997) and mean CALUX-TEQ in each pool was 1.6 pg/g fat (95% CI: 0.7, 2.4), indicating a statistically significant overestimation of CALUX-TEQ compared to WHO-TEQ, probably due to the presence of Ah-receptor agonists. CALUX estimated toxicity of 13 pg/g fat was greater than the WHO-TEQ by 0.9, 3.1 and 0.3 pg/g fat for mothers from Hong Kong, mainland China and overseas territories, respectively. When the 2005 TEF system was applied, a reduction of 14-26% in the WHO-TEQ and a larger but less disperse discrepancy between WHO-TEQ and CALUX-TEQ (3.9 pg/g fat, 95% CI: 3.5, 4.4) were observed. Our study suggested that the mothers' place of residence explained the discrepancy between CALUX-TEQ and WHO-TEQ and should be considered in inter-country comparisons for CALUX-TEQ. For regulatory purposes bioassays for detecting quantitative dioxin contents in any setting must be combined with adequate extraction, clean-up and validation with WHO-TEQs. The larger difference between the two measurements after using the new TEF system warrants further investigation.