This Article Abstract Full Text (PDF) Services Download to citation manager Citing Articles Citing Articles via HighWire PubMed PubMed Citation

Fertility and Life Span

Late Children Enhance Female Longevity

^a Departments of Statistics, University of California, Davis
^b Departments of Entomology, University of California, Davis
^c Division of Biostatistics and Bioinformatics, National Health Research Institutes, Taipei, Taiwan

Hans-Georg Müller, Department of Statistics, University of California, Davis, One Shields Avenue, Davis, CA 95616 E-mail: mueller{at}wald.ucdavis.edu.

Decision Editor: John A. Faulkner, PhD

	Abstract

Top Abstract Methods Results Discussion References

 
The relation between fertility and postmenopausal longevity is investigated for a sample of 1635 women from a historical (17th to 18th century) French-Canadian cohort who lived past the age of 50 years. We find that increased fertility is linked to increased rather than decreased postreproductive survival. Postreproductive life expectancy extension is found to be tied to late births. This finding sheds new light on the cost of reproduction and may be viewed as supporting a new paradigm that states that reproductive potential drives remaining longevity. The emerging reproductive potential concept complements the well-established cost of reproduction hypothesis. Alternative explanations for the observed association are also explored. A specific finding is that the degree to which mortality increases for 50-year-old mothers as a result of senescence is closely tied to the logarithm of the age of their youngest child. For example, 50-year-old mothers experience a mortality decrease of 38% and an increase of remaining lifetime of 3.93 years for every 10-fold decrease in the age of their youngest child. This amount of gain in remaining life expectancy would apply to a mother with a two-year-old child as compared with a mother with a 20-year-old offspring. We also find evidence for the existence of vulnerable periods in human life history that are characterized by phases of heightened mortality and are found to be tied to reproduction and senescence.

AN increase in the number of offspring has been traditionally associated with decreased longevity. The cost of reproduction hypothesis postulates that a trade-off occurs between resources available for reproduction on the one hand and for maintenance and life extension on the other ⁽¹⁾⁽²⁾. Extensive reproductive activity drains resources that would otherwise be available for maintenance, and as a consequence increased reproduction is assumed to be associated with shortened life span according to this hypothesis. The cost of reproduction concept is by now well established, and it was, for example, applied to an analysis of fertility data for the British royalty ⁽³⁾.

Recent studies of individual mortality and reproduction in the Mediterranean fruit fly (medfly) provided evidence for the existence of alternative mechanisms that lead to associations between reproduction and longevity. Medflies were found to regulate their remaining life span based on future reproductive potential on an individual basis. Such a strategy tends to maximize overall reproductive success ⁽⁴⁾⁽⁵⁾. These findings are related to the concept of reproductive determinism ⁽⁶⁾⁽⁷⁾ and prompted the present investigation. In this paper we take another look at the relationship between fertility and longevity for human cohorts in the light of these recent developments. For an overview on data and studies on the cost of reproduction for human longevity, we refer to Gavrilov and Gavrilova ⁽⁸⁾⁽⁹⁾. We note that one of their studies ⁽⁸⁾ provides an alternative perspective for the data analysis presented in the study by Westendorp and Kirkwood ⁽³⁾.

Specifically, we examine the relationship of women's fertility and longevity in a historical cohort of French-Canadian women of the 17th to 18th century. This cohort is characterized by high average fertility. We find that increased female fertility is associated with increased rather than decreased postreproductive survival. The latter would have been predicted by the cost of reproduction hypothesis. In order to explain this finding, we postulate that postreproductive life extension is triggered by late births and the associated presence of young children in the postreproductive period. We show that the data in fact confirm that specifically the presence of young children, or equivalently, late timing of the last birth, is associated with increased postreproductive longevity. Our findings thus support the reproductive potential hypothesis, which holds that life-span regulation has evolved in such a way as to maximize individual reproductive success.

	Methods

Top Abstract Methods Results Discussion References

 
When the relations between fertility and longevity for women are investigated, it is desirable to study cohorts with high average fertility. This excludes many modern cohorts, but it is an important prerequisite when the upper ranges of human fertility are studied. Highly fertile cohorts will allow one to observe a large range of variation in individual fertility, especially late fertility, and will bring the observations closer to presumed precontraceptive patterns. For these reasons, it is preferable to use well-documented data on a historical cohort. An ideal cohort would be narrowly positioned in space and time so as to minimize confounding and biases caused by changes in reproductive patterns that may occur over time or space. Such changes are notoriously hard to control and may have contributed to methodological problems in previous reports on the relation between fertility and longevity, such as those that occurred in the study by Westendorp and Kirkwood ⁽³⁾.

Accordingly, we use data from a well-documented 17th- to 18th-century prebirth control cohort of native-born French-Canadian women. Dates of birth and death of the women and birth dates of their children are based on church records and are deemed reliable. Further details on this cohort and the nature of the data can be found in the report by Le Bourg and colleagues ⁽¹⁰⁾; also see the study by Nault and colleagues ⁽¹¹⁾. For other examples detailing the demography and interpretation of historical cohorts, we refer to Bideau ⁽¹²⁾ and Knodel ⁽¹³⁾.

In order to avoid confounding of mortality by birthing, which was associated with heightened mortality in the 17th to 18th century, and to also avoid confounding with marriage status, we base our analysis on the n = 1635 women in the cohort who were married, parous, and lived to at least the age of 50 years. Mortality and longevity are studied exclusively in the postreproductive period past age 50. Postreproductive longevity is defined as remaining life expectancy past age 50. Some characteristic features of this sample are described in the next section.

The statistical methods used include Cox proportional hazards regression for remaining lifetimes ⁽¹⁴⁾. In addition, we use nonparametric curve estimation techniques. These are based on the kernel method ⁽¹⁵⁾ for the nonparametric estimation of probability density functions with the aim of describing distributions of age at death, and on transformation methods for nonparametric hazard function estimation as described by Müller and colleagues ^(16)(17)(18). Such methods let the data speak for themselves and do not impose requirements and shape restrictions typically needed for parametric modeling. Such restrictions would be difficult to meet given the complex structure of density and hazard functions (mortality rates) that we observe for these data.

	Results

Top Abstract Methods Results Discussion References

 
It is remarkable that of the 1635 women in this cohort, 78.3% have at least 4, 54.5% have at least 8, and 21.6% have at least 12 children. These numbers clearly demonstrate that this cohort is useful for studying the effects of higher-end fertility on longevity. A bar chart demonstrating the distribution of the number of births in this cohort can be found in Fig. 1. The distribution of fertility for this cohort can be ascertained from Fig. 1. The histogram presented there indicates that numbers of births as a function of the age of the mother increase rapidly from age 14 to 26 years, where they reach a sustained high-level plateau that extends to age 35, declining rapidly over the next 11 years, and tapering off to 0 at age 49.

View larger version (31K): [in this window] [in a new window]   Figure 1. A, Percentage of 1635 parous women with particular numbers of births. B, Number of births per year of age of mothers for 1635 parous women.

In an additional analysis, we compare a group of women with lower fertility (one to seven births) with a group with higher fertility (eight or more births), dividing the cohort into two approximately equal large groups. Then the average odds ratio for the probability of dying in a given year within the age range of 50–80 years was found to be 1.22 (SD = 0.34, n = 31, and p = .00012) for women who had one to seven children as compared with women who had eight or more children, confirming the findings obtained with the Cox model.

Nonparametric estimates of the probability densities of remaining lifetimes after the age of 50 years for the low- and high-fertility groups are provided in Fig. 2. These estimates provide valuable information about the distribution of remaining lifetime (or remaining life expectancy). The shift toward longer remaining lifetimes for the higher-fertility group is quite obvious. Remaining lifetimes are shifted toward higher ages; this effect tapers off for the oldest old, where beginning at remaining lifetimes of 35+ (or ages of 85+) years the two densities overlap. These results clearly imply that very high fertility is associated with overall lower late-life mortality.

View larger version (13K): [in this window] [in a new window]   Figure 2. A, Smoothed probability densities of remaining lifetimes for 1635 women at age 50, for high-fertility (women with 8+ births) and low-fertility (women with 1–7 children) groups. B, Women whose last child is younger or older than 6 years of age, respectively. The shift in the dotted densities to the right illustrates enhanced longevity for highly fertile women and women with late births.

Testing this prediction, we find that indeed mortality increases by 38% as log₁₀(age) of the youngest child increases by 1 (p < .005 in a Cox model). This means that as the age of the youngest child decreases by a factor of 10, mortality declines by 38%. For example, a mother with a 2-year-old child at age 50 will have 18% less mortality than a mother with a 6-year-old child. The slope of the predictor log₁₀(age) in the Cox model is .327 with a SD of 0.116.

If both log₁₀(age) and number of births are considered as predictors, only log₁₀(age) remains significant, indicating that the effect of increased fertility on reducing postmenopausal mortality is indeed mediated by a late birth of the last child. A simple linear regression analysis of remaining life span versus log₁₀(age) of the youngest child shows that life span increases by 3.93 years on the average per log₁₀-year decrease in the age of the last child (p = .002; the fitted regression line is y = 23.79 - 3.93x, with SD = 1.27 for the slope). For example, a mother with a child of 1 year at age 50 will live 2.75 years longer as compared with a mother with a child of 5 years.

These analyses point to a protective effect of a late birth for the mother. This protective effect is illustrated in Fig. 2. One notices an obvious shift in the distribution of remaining life expectancy toward higher ages for the group of women with children of age 6- (younger than 6) years as compared with women with children of age 6 years or older at age 50.

	Discussion

Top Abstract Methods Results Discussion References

 
On the methodological level, it is relevant to note that we restrict our study of remaining lifetime to women who are 50 years old and who are thus past the period of increased mortality that is associated with the reproductive phase. This restriction is necessary in order to avoid confounding deaths caused by childbearing with postreproductive mortality. Women that live longer have the opportunity to bear more children, and a direct analysis of the association between overall lifetime and number of offspring will lead to a correlation between these quantities that is purely an artifact and results from confounding.

A cost of reproduction is indeed manifesting itself during the time of reproduction, as is clearly evidenced in Fig. 3. In Fig. 3 the nonparametric density function estimates of lifetime or age at death for all N = 3055 women in the study are provided. This sample is much larger than the one considered herein, as it also includes the nulliparous women and in particular women that did not live to age 50. Here, the densities of the distribution of lifetime (age at death) are shown for three fertility groups of women in the overall sample, irrespective of whether they survived to age 50 or not. The densities are shown for the three groups of women with no offspring (n = 244), with one to seven births (n = 1218), and eight or more births (n = 1593). The corresponding hazard rate estimates (smoothed and transformed mortality rates, also known as trajectories of mortality) are displayed in Fig. 3.

View larger version (14K): [in this window] [in a new window]   Figure 3. A, Nonparametric density function estimates of lifetimes for 2088 women at age 50, for null parous (0 birth), low-fertility (1–7 births), and high-fertility (8+ births) groups. B, Nonparametric hazard function estimates corresponding to the three groups. The bandwidths used for density and hazard function estimates, respectively, the sample size, the mean lifetime, and its standard deviation are (8, 10, 244, 51.55, 24.15), (9.46, 10, 1102, 46.63, 21.17), and (5.33, 11, 1593, 63.95, 15.47) for the three groups.

We note that Fig. 3 also reveals a striking bimodality in the distribution of age at death. Although this is more evident in the density graphs of Fig. 3, it is also a clear pattern in the hazard rates of Fig. 3. A plethora of early deaths is likely caused by the challenges of reproduction, and the surge in late deaths is caused by senescence. This bimodality is particularly strongly expressed for the groups with lower reproduction. The left "reproduction" mode shifts to the right for women with a higher number of offspring, whereas the location of the right "senescence" mode remains surprisingly stable, irrespective of fertility level. Interestingly, bimodality in the distribution of age at death was also observed for medflies ⁽⁵⁾ and thus may point to a more general phenomenon occurring for various species.

These facts illustrate and corroborate two central notions: First, that increasing lifetime primarily has an enabling effect for reproduction; compare the study by Müller and colleagues ⁽⁴⁾ for further details on the relationship between remaining reproductive potential and mortality. If a woman dies early in life she is not provided with the opportunity to have any or more children. Second, there exist two clearly distinguished vulnerable periods in regard to mortality: The first vulnerable period occurs at the time of reproduction; a second vulnerable period occurs at around the age of 80 years for this historical sample. Such vulnerable periods were established as mortality patterns occurring in medflies ⁽¹⁸⁾, and our analysis shows that they are not limited to that species. Such vulnerable periods would primarily be observable in a "more natural" reproductive setting as provided by the historical French-Canadian cohorts but would less likely be observable in a modern cohort under the conditions of birth control and largely improved medical care.

Our findings of a positive association of high fertility with lower postreproductive mortality are in line with recent work ⁽¹⁹⁾ concerning a relative increase in the frequency of a late birth among centenarian women compared with shorter-lived women; also see the study by Perls and Fretts ⁽²⁰⁾. Our findings are compatible with the reproductive potential hypothesis, namely that increased reproductive potential is associated with prolonged life ⁽⁴⁾. A mother who has children late in life would be associated with higher reproductive potential.

An evolutionary explanation for this positive association is provided by the hypothesis that the presence of a young child has a life-prolonging effect on the mother. We have shown here that indeed the postmenopausal mortality declines with declining age of the youngest child. Moreover, age of the youngest child was identified as the primary predictor for the decline in postmenopausal mortality.

A plausible explanation is that a high level of fertility increases the chance that young offspring need to be cared for after menopause of the mother, and that the probability of their survival is likely to be linked to the presence of the mother as a caregiver. This outlines a possible pathway of how the positive association of postmenopausal longevity with fertility has evolved, essentially through enhancing the likelihood for the presence of a caregiver for late children. Thus, it is plausible that extended longevity confers a selective evolutionary advantage to women who have had children later in life. This hypothesis is in accordance with recent findings of an interspecies analysis for primates regarding the relationship between the period of caregiving required to raise offspring and longevity, in which it was found that increased caregiving requirements are indeed associated with extended life span ⁽²¹⁾.

An alternative explanation could be that both longevity and increased fertility as well as late births are associated with a third factor that influences both simultaneously in such a way as to create a positive association. This could be a genetic, socioeconomic, or environmental factor, or it could be a combination of these. That certain longevity genes may extend both the childbearing period and the longevity of individual women has been suggested before by Doblhammer ⁽²²⁾.

More specifically, one could think of a subject-specific frailty that is randomly assigned to each woman, based on biological or environmental factors. Increased frailty of a woman would decrease the chance to survive a given birth. If surviving any birth is the outcome of an independent Bernoulli trial with the probability of survival depending on a random frailty as above, then a mathematical consequence is that the conditional expected frailty of 50-year-old women with many births is lower than that of women with fewer births. Assuming that lower frailty is then also associated with increased remaining life span, a positive association between number of births and remaining life span may result. There also exists some evidence that early life events have an impact on late life mortality ⁽²³⁾. It is of interest here to note that age of the last child is a more significant predictor of remaining lifetime than the number of births, a fact that may slightly favor the caregiver hypothesis of postmenopausal life extension.

Our findings point to a life-prolonging effect for women with late children under conditions of naturally regulated and quite high baseline fertility, coupled with historic conditions of health care. The observed association leads to several interesting questions. First, does the association hold under modern conditions with increasingly late childbearing that occurs under birth control conditions and an altered environment that includes largely improved health care in comparison to the historical cohorts? If this were the case, late childbearing in modern women would still serve as a pointer toward extended remaining lifetime, irrespective of whether late childbearing is causal for the observed life-span extension.

A second question that looms large is indeed that of causality. An association exists, as we have demonstrated in this article, but the fact remains that this finding is derived from an observational study, and thus it is subject to all the caveats that one must entertain in the interpretation of such studies. In particular, causality cannot be established from such data. The caregiver hypothesis provides a plausible causative mechanism, within the intriguing framework of the reproductive potential hypothesis. However, alternative models such as the above-described selection through frailty also may explain the observations. One suggestion would be to seek experimental evidence from animal studies that could shed further light on some of these issues.

	Acknowledgments

 
This research was supported in part by the National Institute on Aging Grant AG-0876 and by NSF Grant DMS-9971602.

We thank Dr. Bertrand Desjardins for providing us access to the longevity and fertility data on Canadian-born French-Canadian women. We also thank two referees for detailed comments on a previous version of this paper.

Received July 16, 2001

Accepted January 25, 2002

	References

Top Abstract Methods Results Discussion References

 

1. Partridge L, Barton NH, 1993. Optimality, mutation and the nature of ageing. Nature. 362:305-311. [Medline]
2. Kirkwood TBL, Austad SN, 2000. Why do we age?. Nature. 408:233-238. [Medline]
3. Westendorp RGJ, Kirkwood TBL, 1998. Human longevity at the cost of reproductive success. Nature. 396:743-746. [Medline]
4. Müller HG, Wu D, Carey J, Liedo P, Vaupel J, 2001. Reproductive potential predicts longevity of female Mediterranean fruit flies. Proc R Soc London Ser B. 268:445-450. [Medline]
5. Carey J, Liedo P, Müller HG, Wang JL, Chiou JM, 1998. Relationship of age patterns of fecundity to mortality, longevity and lifetime reproduction in a large cohort of Mediterranean fruit fly females. J Gerontol Biol Sci. 53A:B245-B251. [Abstract]
6. Bell G, 1984. Measuring the cost of reproduction. The correlation structure of the life tables of five freshwater invertebrates. Evolution 38:300-313.
7. Bell G, Koufopanou V, 1986. The cost of reproduction. Dawkins R, Ridley M, , ed.Oxford Surveys in Evolutionary Biology Vol. 3 83-131. Oxford University Press, Oxford.
8. Gavrilov LA, Gavrilova NS, 1999. Is there a reproductive cost for human longevity?. J Anti-Aging Med. 2:121-123.
9. Gavrilov LA, Gavrilova NS, 1999. Data resources for biodemographic studies on familial clustering of human longevity. Demograph Res. 1:4
10. Le Bourg E, Thon B, Legare J, Desjardins B, Charbonneau H, 1993. Reproductive life of French-Canadians in the 17th–18th centuries—a search for a trade-off between early fecundity and longevity. Exp Gerontol. 28:217-232. [Medline]
11. Nault F, Desjardins B, Legare J, 1990. Effects of reproductive behavior on infant mortality of French-Canadians during the seventeenth and eighteenth centuries. Populat Stud. 44:273-285.
12. Bideau A, 1986. Fertility and mortality at ages 45 and over—the contribution of research in historical demography. Population. 41:59-72.
13. Knodel J. Demographic Behavior in the Past: German Village Populations in the 18th and 19th Centuries. Cambridge: Cambridge University Press; 1988.
14. Cox DR, 1972. Regression models and life-tables (with discussion). J R Stat Soc B. 34:187-220.
15. Müller HG, 1997. Density estimation. Kotz S, Read CB, Banks DL, , ed.Encyclopedia of Statistical Science 185-200. Wiley, New York.
16. Müller HG, Wang JL, Capra WB, 1997. From lifetables to hazard rates: the transformation approach. Biometrika. 84:881-892.
17. Wang JL, Müller HG, Capra WB, 1998. Analysis of oldest-old mortality: lifetables revisited. Ann Stat. 26:126-163.
18. Müller HG, Wang JL, Capra WB, Liedo P, Carey JR, 1997. Early mortality surge in protein-deprived females causes reversal of sex differential of life expectancy in Mediterranean fruit flies. Proc Natl Acad Sci USA. 94:2762-2765. [Abstract/Free Full Text]
19. Perls T, Alpert L, Fretts R, 1997. Middle aged mothers live longer. Nature. 389:133[Medline]
20. Perls T, Fretts R, 2001. The evolution of menopause and human life span. Ann Human Biol. 28:237-245. [Medline]
21. Allman J, Rosin A, Kumar R, Hasenstaub A, 1998. Parenting and survival in anthropoid primates: caretakers live longer. Proc Natl Acad Sci USA. 95:6866-6869. [Abstract/Free Full Text]
22. Doblhammer G, 2000. Reproductive history and mortality later in life: a comparative study of England, Wales and Austria. Pop Stud J Demog. 54:169-176.
23. Doblhammer G, Vaupel JW, 2001. Lifespan depends on month of birth. Proc Natl Acad Sci USA. 98:2934-2939. [Abstract/Free Full Text]

This article has been cited by other articles:

				E. Grundy and O. Kravdal Reproductive History and Mortality in Late Middle Age among Norwegian Men and Women Am. J. Epidemiol., February 1, 2008; 167(3): 271 - 279. [Abstract] [Full Text] [PDF]

				P. F. McArdle, T. I. Pollin, J. R. O'Connell, J. D. Sorkin, R. Agarwala, A. A. Schaffer, E. A. Streeten, T. M. King, A. R. Shuldiner, and B. D. Mitchell Does having children extend life span? A genealogical study of parity and longevity in the amish. J. Gerontol. A Biol. Sci. Med. Sci., February 1, 2006; 61(2): 190 - 195. [Abstract] [Full Text] [PDF]

This Article Abstract Full Text (PDF) Services Download to citation manager Citing Articles Citing Articles via HighWire PubMed PubMed Citation