In reply:

Abstract

In reply: We previously reported that in a large, nondemented, community cohort of older, white women, both current users of estrogen replacement therapy (ERT) and never users had similar declines in cognitive test scores across a 6-year period1 and that high education attainment was, after age, the most powerful predictor of good cognitive function. Drs. Cunningham, O'Neill, and Rowan suggest that the protective effect of ERT on cognitive decline may have been obscured by incomplete follow-up of the cohort. Although follow-up was quite reasonable considering the women's age, nearly 7% died between examinations, and an additional 16% were not re-examined. Those who participated were younger, more educated, more likely to use ERT, and performed better on the cognitive tests at the initial examination relative to participants. We noted (p. 522; 1) that “As a consequence, we could only test our hypotheses in a better functioning group.” It is not possible to know how differential loss to follow-up affects our results. Survival may be better among both those who use ERT and those with preserved cognitive function,2-4 with the result that women using ERT and those having better cognitive function were being more likely to be alive and able to participate in our follow-up examination. This type of bias would lead to an overestimation of a protective effect of ERT. On the other hand, if women who are having memory problems are being prescribed ERT by their physicians or are reporting incorrectly that they are on ERT, then women with low cognitive function would be in the ERT group, leading to an underestimation or negative effect of ERT. Note that the prospective studies5, 6 showing a protective effect of ERT did not have complete follow-up of women because the eligibility criteria included follow-up examinations. Our study did not use the full Mini-Mental Status Examination, and scores ranged from 0 to 26 rather than 0 to 30. Therefore, comparisons of our findings with others based on cutoffs of less than 24 cannot be made. The studies5, 6 showing a protective effect of ERT examined risk of developing Alzheimer's disease (AD), not change in cognitive scores as in our study. The biological determinants of AD may differ from the determinants of cognitive decline in older people. Two large cohort studies did show that ERT users exhibited less cognitive decline over time in unadjusted analyses.7, 8 However, one did not adjust for potential confounders,7 and the effect for ERT became nonsignificant in the second after adjustment for education.8 Our conclusion is that the most powerful intervention for protecting older women from cognitive decline may be education, not the use of ERT. Clinical trials are needed to evaluate both the effects of late-life education and the use of ERT.