Drs. Sheffield and Larson ask whether differences in the level of physical activity over time between women assigned to hormone therapy and those assigned to placebo may have explained the reduction in incident diabetes we observed among women taking hormones in the Heart and Estrogen/progestin Replacement Study (HERS). Participants in HERS self-reported physical activity at each annual visit. However, addition of physical activity as a time-dependent covariate in our model did not alter the risk for incident diabetes (relative hazard, 0.65 [95% CI, 0.47 to 0.89]).

Drs. Girod and Brotman suggest that hormone therapy may have produced favorable changes in body composition, including fat mass or lean body mass, that may have influenced glucose homeostasis. They correctly note that we evaluated surrogate markers for total or abdominal adiposity as potential mediators in our study and therefore may have underestimated the effect of the change in these more specific measures of body composition.

Dr. McElduff states that without the measurement of 2-hour postchallenge glucose level, it is premature to conclude that hormone therapy affects incident diabetes. We did not measure postchallenge glucose levels in HERS; our analysis was based on 3 fasting glucose levels obtained at separate clinical examinations, self-report of a new diagnosis or complication of diabetes, or use of a diabetes medication. We did note the absence of postchallenge glucose levels in our analysis and the increased postchallenge levels among women assigned to any of the hormone therapy regimens reported from PEPI (1). It is possible that women in the hormone therapy group in HERS may have increased postchallenge glucose levels, which may alter the overall incidence of diabetes. However, since the 2-hour glucose tolerance test is more sensitive in older adults (2), we would expect to see increased detection of diabetes in both the hormone therapy and placebo groups.

A second point raised by Dr. McElduff was that separate calculations of NNT_B for women with normal glucose metabolism and those with impaired fasting glucose level would be more meaningful than an overall NNT_B. Eight women with impaired fasting glucose levels or 65 women with normal glucose levels would need to be treated for 4.1 years to prevent 1 case of diabetes. However, we do not recommend the use of postmenopausal hormones for diabetes prevention. Our results encourage the addition of glucose-specific end points to ongoing or future clinical trials of hormone therapy. The current weight of the evidence supports the use of hormone therapy only for the treatment of menopausal symptoms (3, 4).