Reply
We apologise to Dr. Maskarinec et al. for misinterpreting their work. We had originally intended to say that their study had observed a nonsignificant reduction in dense area and significant increase in percentage of density with higher intakes of soy foods. Unfortunately, while revising it in the light of referees’ reports, one of us (S.W.D.) misread a direction by the editor, and the error crept in.
As Dr. Maskarinec and Williams state, the results for a qualitative classification of breast patterns do not directly correspond to those for a quantitative scale, but Tabar patterns IV and V are typically >50% dense. As regards Dr. Maskarinec’s and Williams’ point about the absolute amounts, the quartile cutoffs for total soy protein intakes in gram/day were 2.9, 4.7, and 7.5. Thus, our “low risk” group has intakes that are relatively high in western terms but not particularly so in a Chinese population.
Contrary to the interpretation of Dr. Maskarinec and Williams, we did not suggest in our study that adjustment for BMI and energy intake would remove the confounding, which made fat intake appear to have a protective effect. The unadjusted effect of fat intake is of borderline significance, which is no longer seen when adjusted for soy intake. Our point about BMI was its negative association with breast density. Because of the negative confounding of BMI and breast density, we submit that the increased risk of breast cancer associated with breast density will be underestimated if not adjusted for BMI. Of course, if high BMI had a causal effect on breast density, which in turn produced the observed effect of high BMI on breast cancer risk, one could not assume the converse, because density would be part of the causal pathway of the effect of BMI on risk. However, since the two factors seem to act in opposite directions, at least in postmenopausal women, it seems likely that the effect of BMI on risk of breast cancer is operating by other pathways than breast density.