To the Editor: Dr Schairer and colleagues1 address concerns over bias introduced by estimating
age at menopause2 by repeating their analyses
excluding women with unknown age at menopause, ostensibly leaving only women
with natural menopause and bilateral oophorectomy in their sample.
They fail to consider, however, another possible source of bias. Women
with bilateral oophorectomy, who also tend to have undergone hysterectomy,
were likely overrepresented among the estrogen-only HRT-exposed group. Controlling
for differences in age of menopause may leave residual confounding if oophorectomy
reduces breast cancer risk through pathways other than its strong association
with earlier age at menopause (eg, nonestrogen-mediated biologic effects,
selection effects related to factors leading to reproductive surgery, or both).
If this residual confounding is strong enough, the difference observed in
the RRs for estrogen-only HRT and estrogen-progestin HRT could conceivably