Katz et al. Respond

We appreciate the careful attention to our article by Prentice et al., and the thoughtful details incorporated into their rejoinder.

We must begin our response, however, by highlighting an apparent misunderstanding. Our article is not critical of the Women’s Health initiative (WHI),¹ neither explicitly, nor implicitly. The WHI was, of course, an extremely important and well-conducted trial, and the source of the data² in our own analysis. The WHI demonstrated varying effects of hormone replacement on health outcomes based on the particulars of both the hormones and the women. Our article highlights the cost of failing to translate those details properly into our prevailing cultural understanding and practices. We are faulting the translational step, not the trial.³

As for our formula: it is very simple, conservative, and quite robust. As we indicate in the article, we subjected our analysis both to the most conservative constraints, and a sensitivity analysis to generate a range for our estimates.

There seems to be particular concern about the significance of the excess mortality reported by LaCroix et al.² in women aged 50 to 59 years, and our application of that finding. First, we respectfully refute the notion that all-cause mortality is anything but an outcome of primary interest, whether it was a primary outcome measure a priori.

Second, our article was explicit about the age group to which our analysis pertained: women aged 50 to 59 years. We acknowledged that estrogen replacement was associated with a higher mortality in older women. But we must also point out that death within a decade of age 50 to 59 years is very different from death within a decade of age 70 to 79 years. The former implies as much as two full decades cut from the average life expectancy among women in the United States⁴; the latter implies death at or after average life expectancy is achieved.

As for our mortality figure, it was calculated directly from the data published by LaCroix et al. and represents a standard approach to the 95% confidence interval^5,6; the range was from six to 20 excess deaths per 10 000 per year. We calculated this interval ourselves only after two letters to the corresponding author went unanswered. As a further safeguard, we recommended WHI investigators, including LaCroix, as peer reviewers of our article. Finally, we note the consistency of this figure with findings reported by others,^7,8 and with reported trends at the population level.⁹

The WHI is an enormously important trial. Applications of it, however, have conferred both benefits and harms. We stand by our article and our estimates of the mortality toll of estrogen avoidance among those most apt to benefit from its use.