Shadyab and LaCroix Respond

Thank you to Adler for her letter on our article. The lack of significance for women aged 30 years or older is likely attributable to lower power resulting from a smaller sample size after multivariable adjustment, as several variables had missing data. (The age-adjusted model had a sample size of 15 082, whereas the fully adjusted model had a sample size of 11 748.) If there were substantial bias, the odds ratio for the oldest age group would likely have been much higher than that for the younger age group; however, they were similar in magnitude, suggesting a threshold effect. Accordingly, when the comparison was binary (younger than 25 years vs 25 years or older), the association was significant, and a dose–response association was supported by a significant P value for a linear trend.

In a prospective study, exposure data are collected at baseline, and participants are followed to determine development of outcomes. This does not preclude studies that do not follow participants from the time of exposure onward from being classified as prospective studies. Most prospective studies of longevity begin when participants are in midlife, at which time information on lifetime behaviors is assessed and then evaluated in relation to aging outcomes years later.^1–3 It is rare to have data on women from the beginning of their reproductive years into very late life, because that would require 60 to 70 years of follow-up. Our study was designed in a manner similar to that of previous aging studies.^1–3

The association between older age at first childbirth and longevity does not suggest that older age at first childbirth is a protective factor later in life; rather, it may reflect other underlying factors. Perls et al. contended that maximizing a woman’s childbearing years is one of the driving selective forces of the human life span, implying that women who can bear children at older ages may be aging slowly.⁴ Crawford proposed that a common set of genes may extend both menopause and longevity.⁵ It is possible that the timing of reproductive events may reflect a woman’s overall health, but older age at first childbirth may also reflect other factors that influence longevity, including socioeconomic status and childhood characteristics.⁵

We acknowledge that our study may have involved selection bias owing to a healthier population of older first-time mothers and that, similar to previous aging studies, our findings are conditional on women having survived to an older age.^1–3 However, any influence on our findings of earlier death due to maternal causes would be minimal.^6,7 Most women in our study gave birth for the first time between 1940 and 1950. After application of age-specific US maternal mortality rates from 1945 among women giving birth at age 20 to 24, 25 to 29, and 30 to 34 years to the corresponding age-at-first-childbirth categories in our study, an estimated total of only 22 women would be expected to have died from maternal causes before entering our study.⁷

We conducted an additional analysis among the youngest group of women (aged 50–59 years) who enrolled in the Women’s Health Initiative (WHI) and observed a significantly lower risk of death among those aged 25 to 29 years (hazard ratio [HR] = 0.82; 95% confidence interval [CI] = 0.76, 0.88) and 30 years or older (HR = 0.83; 95% CI = 0.74, 0.93) at first childbirth than among those younger than 25 years. This result lends support to our findings because it suggests that even younger women enrolled in the WHI did not appear to die of competing causes associated with later age at first childbirth.

Given that age at first childbirth is increasing in the United States, understanding how this trend is related to women’s long-term health is important. It appears that later age at first childbirth is an indicator of not only underlying health but also many factors throughout the life course. Future studies involving life course data are warranted to confirm and extend our findings.