Parental age at death is associated with age at first birth in offspring

Purpose: People age at different rates and the available evidence suggests that the rate of aging is partly inherited from previous generations. This heterogeneity in aging is evident already in midlife, but to what extent aging is associated with the timing of events earlier in life is not fully known. Here we aim to shed light on this topic by investigating the trade-off between reproduction and aging postulated by evolutionary theories of aging. Methods: Drawing on the inheritance of aging we use parental age at death as a proxy for aging-rates in the offspring, and study how age at first birth depends on this variable. We use data from an almost complete Swedish birth cohort comprising 92,359 individuals. Accelerated failure time models are used to estimate the association between parental age at death and age at first birth while adjusting for parental occupational class, educational attainment, and income. Results: Longer parental lifespans were consistently associated with older age at first births, both in men and women. Conclusion: Our findings suggest that aging-related processes may be interrelated with the processes underlying the timing of reproduction and are in general agreement with evolutionary theories of aging.


Introduction
Aging is a heterogeneous trait.People of the same chronological age may differ in how "aged" they are as evidenced by differential risk of disease and death (Edwards and Tuljapurkar, 2005;Elliott et al., 2021).This heterogeneity in aging is to some extent inherited; indeed, previous research has found robust and graded associations between parental age at death and risk of disease and death in offspring (Atkins et al., 2016;Christensen et al., 2020;Dutta et al., 2013;Terry et al., 2004;Thalén and Ledberg, 2024;Vågerö et al., 2017).This association is present at least in midlife (Thalén and Ledberg, 2024) and possibly even earlier (Christensen et al., 2020), likely a reflection of the many continuous processes constituting aging (López-Otín et al., 2013, 2023).The ways in which aging-related processes interrelate with development and other transformative processes during the life course is not fully known, but such knowledge may provide important insights into the principles under which life unfolds.Evolutionary accounts of aging postulate a trade-off between reproduction and aging, suggesting that interindividual differences in aging should be related to the onset of reproduction (Kirkwood, 1977;Williams, 1957).Prior research has shown a negative association between age at first birth and future risk of death (Grundy and Kravdal, 2008;Westendorp and Kirkwood, 1998), providing some support for this suggestion.However, an association between age at first birth and future health outcomes could reflect both a causal effect of young parenthood on future health, as well as the effect of other factors influencing both the timing of reproduction and later health.Previous studies conflate these two accounts.Here we draw on the known inheritance of lifespans (Hjelmborg et al., 2006;Timmers et al., 2019) and use parental age at death as a proxy for aging-related processes in the offspring.Thereby we are able to investigate if aging and reproduction are associated also when health consequences of early reproduction are ruled out.In particular, we investigate how interindividual variation in aging, as indexed by parental age at death, is related to age at first birth in an almost complete birth cohort of Swedish men and women.

Study population
We utilized data from the longitudinal research programme "Reproduction of Inequality through Linked Lives" (RELINK-53), which consists of all children born in 1953and residing in Sweden in 1960, 1965, or 1968, and their family members (Almquist et al., 2020).Data on three generations available within RELINK-53, G0 (parents, N = 213,883), G1 (index population, N = 110,006), and G2 (firstborn children, N = 88,355) were utilized.Several exclusion criteria were applied.We only included index persons where we had information on both parents.Index persons with parents that died before turning 55 years were excluded (around 10 % of total population) as such deaths were considered less likely to be related to aging.Parents who were older than 115 years were excluded as such extreme ages are likely due to overcoverage in the population registry (total 21 persons).The final study sample included 92,359 index persons.Approximately 18 % (n = 16,694) had no children and in around 4 % (n = 2981) of cases two index persons were parents to the same child.See Supplemental Fig. S1 for more information on the sample selection.

Procedures
Age at first birth (for men and women) was the dependent variable.Parental age at death was used as the main independent variable.However, some parents were still alive (∼ 10%) at the end of follow-up.Thus, we refer to the length of the parental lifespan as "parental attained age," which was either their age at death or, for living parents, their age in May 2021.Restricting the analysis to index persons whose parents are dead does not change the results substantially (Supplementary Fig. S2, Supplementary Table S1).In the descriptive analysis, the average of parents' attained age was divided into four groups: 55-70 years, 70-80 years, 80-90 years, 90-108 years.To account for differences in socioeconomic status in the regression models (see below), we included information on: parental socioeconomic index (SEI) in 1960, categorized as blue-collar workers, white-collar workers, or other; parental educational attainment in 1970, categorized as nine years of school or less, more than nine years of school, or education not specified; and parental earnings (income) in 1970, categorized in quartiles, plus a separate category of those with zero income.See Table A1 for a more detailed description of the variables.
In the analyses, follow-up started at the index persons' 13th birthday and continued until either: first childbirth, emigration, death, or end of follow-up which was set to their 55th birthday.As all index persons were born in the same year, this resulted in a follow-up time of maximally 42 years.Only 21 first births happened after index persons turned 55 years of age, and we reasoned that those births were not representative of the processes we are interested in here and hence were excluded from the analysis.

Statistical analysis
Probability densities of age at first birth were estimated using a kernel density estimator with default parameters.Cumulative probabilities of age at first birth were estimated using the Nelson-Aalen estimator, treating death and emigration as competing "risks".
To quantify the dependency between age at first birth and parental attained age, and in order to adjust for socioeconomic differences, we applied an accelerated failure time (AFT) framework (e.g., Kalbfleisch and Prentice, 2002).This implies assuming that the independent variables act on age at first birth multiplicatively, something that is supported by the descriptive analysis.In particular, we model the logarithm of age at first birth as a linear function of parental attained age and covariates, while accounting for censoring.We further assume that the distribution of age at first birth approximately follows a log-normal distribution, also in reasonable agreement with the descriptive results.The results from the AFT-models are not very sensitive to the distributional assumptions and remain similar when using a log-logistic distribution instead of a log-normal (not shown).Models were fit for men and women separately, and we used mother's and father's attained ages as continuous variables.In the supplemental material we also show results for analyses where each parent's attained age was modeled separately.We note that this statistical model (AFT) mirrors the more substantial assumption of different time scales; some people age faster because their time is accelerated, and we investigate if that also coincides with having children at a younger age.
Index persons who reached end of follow-up without having a child were excluded from the regression models as they are not providing any information pertaining to the main question investigated here.In the unadjusted models, 38,856 men and 39,643 women were included.In the adjusted models, we excluded any index person lacking information on any of the covariates (parental SEI, parental educational attainment, parental income).This resulted in models with 34,836 men and 35,242 women.See Supplemental Fig. S1 for more information on the sample selection.

Software
Data management and analyses were made in R (R Core Team, 2021) using the survival package (Therneau, 2021) for the survival models and ggplot2 for visualizations (Wickham, 2016).

Results
These results are based on data from 47,430 men and 44,929 women, all born in 1953, for whom we have almost complete reproductive history and know the attained age of both parents.The mean age at first childbirth was 25.0 for women and 28.2 for men.When index persons were divided into four groups according to the mean attained age of their parents, a graded relation between age at first child birth and attained age of parents became apparent (Fig. 1).Probability densities of age at first birth are more right-shifted for those with parents that live longer (Fig. 1A, B), and the cumulative probability of having given birth as a function of age is consequently a graded function of parental attained age (Fig. 1C, D).There were 8574 men and 5286 women who did not have children before end of follow-up.Being childless did not depend on parental attained age for women, but for men, having parents who lived longer was associated with a slight increase in the probability of having a child (Fig. 1D).
To quantify the dependence between parental attained age and age at first birth, while adjusting for potential confounders, we fitted accelerated failure time models to the data.This showed that age at first birth increased with parental attained age (Table 1).If β is the coefficient for parental attained age, the interpretation is that an increase in parental attained age by x years leads to a proportional slowing down of time by a factor of .In other words, a woman who had her first child at age 25, would have had that child 6 months later, had her mother lived for another 10 years: exp(0.0019⋅10)⋅25≃ 25.5.Adjustment for socioeconomic variables reduced the magnitudes of the associations, but they still remained statistically significant in the adjusted models (Table 1).The association was stronger for women than for men, and in women, the attained age of the mother had a stronger association than the attained age of the father.The overall magnitude was modest: increasing the average parental attained age with a decade is expected to increase the mean age at first birth by 2.5 months for men and 5 months for women (predictions from the adjusted models).

Discussion
We have shown a robust and graded association between parental attained age and age at first birth in a national Swedish birth cohort (Fig. 1).The associations were slightly more pronounced in women than in men, and became smaller when we adjusted for socioeconomic variables (Table 1).The magnitudes of the associations were modest; an increase of parental attained age of ten years would, according to the adjusted model, imply an expected postponement of the age of first birth by six month for women, and less for men.To the extent that parental lifespans reflect aging-related processes in the offspring (see below) our results provide support for a trade-off between aging and reproduction as has been suggested by evolutionary models of aging (Kirkwood, 1977; Mean Parental Attained Age A. Thalén and A. Ledberg Williams, 1957).Our results also show that age at first birth is dependent on parental occupational class, education, and income, in line with previous research showing that age at puberty is associated with socioeconomic conditions in childhood (Colich et al., 2020).
The association between parental attained age and age at first birth were seen both in men and women, suggesting that the mechanisms linking aging-related processes and reproduction are similar in both sexes.One way to conceptualize our findings is in terms of an underlying, and partly inheritable, rate of aging.According to this hypothesis, a higher rate of aging would imply a younger (chronological) age at sexual maturity and, by extension, a younger age at first birth, as well as a shorter lifespan on average.Previous research has found an inverse association between age at puberty and later adverse health outcomes (Day et al., 2015;Lakshman et al., 2009), supporting this simple model.Further support comes from a recent study showing that the genetic correlation between age at first birth and parental age at death is substantial (Long and Zhang, 2023), indicating that these two traits are shaped by overlapping genetic architectures.
We used age at first birth as the reproduction-related outcome in this study.There are a large number of factors and processes that influence at what age someone has their first child, the inheritable component we have investigated here is expected to contribute only a small fraction to the overall variability.The cohort we studied were all born in 1953 and birth control pills became available in Sweden in 1964, making it likely that cohort members exerted substantial control over both when they had their first child, and how many children they had in total.The tradeoff between aging and reproduction postulated by evolutionary theories of aging does not single out the onset of reproduction as a critical parameter, other reproduction-related outcomes of interest are: age at puberty onset, fecundity, and age at menopause.We do not have any data pertaining to the latter, and we believe total number of offspring to be shaped by too many factors for it to be a useful measure of fecundity.One advantage of using age at first birth is that it was possible to get almost complete coverage of an entire national birth cohort, however it only provides a partial view of reproduction.In future studies it would be interesting to investigate associations between parental age at deaths and other outcomes related to reproduction.
We have used parental attained age as a proxy for aging processes in the offspring.This is sensible given the ample evidence supporting the inheritance of lifespans, even if the purely genetic heritability may be modest (Hjelmborg et al., 2006;Ruby et al., 2018;Timmers et al., 2019).However, associations between parental attained age and age at first birth could possibly originate from other factors than aging-related processes.By only studying people born the same year we rule out confounding by secular fertility trends.By adjusting the regression models for parental education, income and occupational status we hope to have ruled out major childhood socioeconomic factors that could cause an association.However, it is not possible to completely rule out unmeasured confounding as contributing to the associations, and the results should therefore be interpreted with this caveat in mind.Another possible pathway through which parental attained age and age at first birth could become associated is through childhood adversity.There is a substantial literature showing that certain types of childhood adversity are associated with measures of biological aging, including age at puberty (Colich et al., 2020).Losing a parent at a young age could certainly be such an adversity, and the association we have described herein might hence, in part, be due to an adaptation to the stress induced by losing a parent (cf, Belsky et al., 1991).In Supplementary Table S4 we show that the associations we find remain very similar after we exclude all index persons whose parents died before they turned 21 years of age.This does not rule out a contribution of childhood-adversity-induced developmental adaptation, leading to younger age at reproduction, but show that such adversity may not be directly linked to the death of a parent.