Abstract

The disposable soma theory proposes a trade-off between fertility and longevity but existing findings on this association have been mixed. This study used data from 15,622 twins born between 1901 and 1925 ascertained from the population-based Swedish Twin Registry to test the child-longevity association and whether it is accounted for by individual-level factors or by genetic and environmental factors shared by family members. Based on survival analysis, both women and men with children had significantly longer survival relative to the childless, with a slightly higher relative advantage in men. Adjustments for demographic factors and cotwin fertility did not mediate the parenting–survival association, indicating that this association is attributable to individual-level factors associated with fertility rather than family-level environmental or genetic factors shared by cotwins. These results, derived from a large, population-based sample, are inconsistent with the disposable soma theory as applied to modern human populations.

SINCE the early 20th century, there have been marked increases in life expectancy that have been accompanied by decreases in family size (1,2). It is clear that many of these shifts are associated with improved health and better access to health care, as well as changes in use of birth control and family planning (3). The correlations between family size and longevity have been observed at the population level; it is generally assumed that the increases in longevity drive decreases in family size. What is less clear is what occurs at the individual level and whether number of children affects an individual’s longevity. The possible trade-off between fertility and longevity has been the subject of debate for decades. Evidence for the existence and direction of the relationship between fertility and longevity is mixed, as indicated by existing reviews of studies investigating the fertility–longevity association (4–7).

Multiple studies of human populations have documented a negative association between fertility and longevity (8–14). For example, Westendorp and Kirkwood (8) found support for a trade-off between fertility and longevity using historic data on approximately 1,900 female members of the British aristocracy born between the 17th and 19th centuries. Number of offspring was negatively associated with longevity among postmenopausal (aged 60 and older) women in this sample (likelihood ratio [LR] = 24.2, df = 3l, p < .001). Another study of British peerage aged 50 and older who were born at the beginning of the 17th century found a 3.8% annual increase in mortality risk associated with each additional child for lifetime parities of two or more among women (15).

Other studies do not support any relationship or found a positive relationship between fertility and longevity. In their study of a 17th-century French–Canadian cohort (N > 3,500), Le Bourg and colleagues (16) found no relationship between number of offspring and the longevity of women. These findings are consistent with other studies that found no fertility–longevity relationship (16–18). In contrast to Le Bourg and colleagues’ (16) findings, another study of women from the same French–Canadian data found a 1.4% decrease in relative mortality risk with each additional child in postreproductive women aged 50 and older compared with childless women. Increased longevity was particularly marked among postmenopausal women with more than eight children and women who had children later in life (19). One reason why findings in the French–Canadian studies differ from British peerage studies may be due in part to differences in data quality. A study by Gavrilova and colleagues (18) investigating the accuracy of data used in Westendorp and Kirkwood’s (8) study found that inaccuracies in the British data, including underreporting of number of children, could have resulted in spurious negative associations between fertility and longevity. Variability in findings among studies that investigated the same or similar data (eg, studies of the French–Canadian data) may be attributable to differences in variables studied, analytic approaches, and variables controlled.

Some studies have suggested an inverted “U” or a “J”-shaped relationship between fertility and longevity, whereby childless individuals and individuals with a very high number of offspring have lower longevity relative to individuals with fewer children. A study of mortality in Austrian (N = 1,254,153) and English and Welsh women (N = 56,164) using census data from 1971 and 1981 indicated that mortality risk was higher for childless and high parity women compared with those with only one or two children (13). A study of 937 Old Order Amish in Pennsylvania born between 1749 and 1912 found evidence for increased longevity of 0.32 years per additional child for up to 14 offspring; however, there was a decrease of 4 years per additional child beyond the 14th (20).

Evolutionary Explanations for Fertility–Longevity Relationship

One hypothesis for the negative association between fertility and longevity found in some studies is the disposable soma theory (DST). The DST posits a trade-off between reproduction and longevity, as the organism is faced with optimal resource allocation in the presence of competing energy demands (21). Given that death is inevitable, the only solution to mortality pressures is reproduction. Therefore, investments in reproduction take precedence over somatic maintenance. Reproduction must occur within the time window permitted by mortality constraints, referred to by some researchers as the “essential life span” or “warranty period” (22–24).

The DST, like many other evolutionary theories of aging which predate it (e.g., mutation accumulation [25], antagonistic pleiotropy [26]), has been conceptualized at the species level. However, research on the DST has also been applied to predict individual differences. Based on the DST, individuals who have more offspring are expected to live shorter lives.

The DST suggests that there may be sex differences in reproductive investments. Depending on the species, males and females may both incur reproductive costs. For example, costs incurred by males include indirect costs associated with seeking mates and caring for offspring. Among mammals, females have higher direct costs of reproduction attributable to pregnancy and other energy demands such as milk production and greater responsibility in caring for offspring. Consequently, the DST predicts that females will experience the majority of the adverse consequences of parenthood and the impact on longevity will increase with number of offspring.

Alternative Explanations for Fertility–Longevity Association

Social perspectives also provide explanations for how offspring may adversely affect longevity. Higher fertility has been associated with increased poverty in families as multiple offspring may strain economic resources available (27,28). Increased poverty has been associated with a variety of adverse outcomes, including higher mortality (29,30).

However, other social explanations for the relationship between offspring and longevity frame children as a valuable commodity contributing to the survival of the family. The utility of children is particularly evident in agrarian settings and impoverished countries (31–33). It is estimated that during the Great Depression, child labor prevented nearly 7% of families in the United States from falling below the poverty line (34). The child’s ability to work becomes a part of an adaptive strategy as external economic forces become interwoven with family responsibilities and survival.

Critiques of Literature on Fertility and Longevity

Reviews of empirical studies on longevity and fertility in humans have identified multiple methodological issues that may contribute to divergence of findings in this literature, including the use of selected samples from historic populations (4–7). Using selected samples such as the British peerage may be biased because of their restricted socioeconomic range. Another source of differences comes from data analysis approaches, such as only modeling linear relations between number of offspring and longevity and not investigated possible U- or J-shaped trends. The quality and accuracy of such historic data sets, including the underreporting of offspring, have also been questioned (5,18). Another shortcoming is the lack of consideration of other variables that may affect both longevity and fertility, including socioeconomic factors.

Furthermore, lack of detailed investigations of mechanisms underlying the fertility-longevity association, such as genetic and environmental contributions, limit our understanding of whether and why an association exists. Gögele and colleagues’ (35) study of individuals living in three isolated villages in South Tyrol, Italy is one of few studies that investigate genetic and environmental contributions in the fertility-longevity association. Findings of this study suggest that longevity is positively correlated and shares a common genetic background with reproduction. Another study of preindustrial Finns also found genetic correlations between longevity and reproduction (36). Additional research that focuses on environmental and genetic influences on the fertility-longevity association is required to advance our understanding of whether an association exists and factors contributing to the association.

Study Goals

In light of the methodological concerns with research cited as supporting the DST and the existence of alternative explanations for the relationship between fertility and longevity, we conducted this study. We used data from a large population-based sample of twins to (a) test the hypothesis that having offspring is associated with decreased longevity and (b) test if the relationship between fertility and longevity is best explained by individual-specific factors (i.e., consistent with a causal explanation such as the DST) or by genetic or environmental factors shared by family members (consistent with an indirect fertility–longevity association). Twin data have been used previously to study genetic contributions to fertility (37,38) and to longevity (39–42). To our knowledge, this is the first time twin data have been used to evaluate the relationship between reproduction and longevity. Twins studies permit a thorough investigation of environmental and genetic factors that was not possible in non-twin genealogical studies of trade-offs (35,36), particularly with regard to the degree of genetic resemblance shared by identical twins.

Methods

Study Population

Participants.—The Swedish Twin Registry is a population-based registry established in the late 1950s based on birth records and contains health and psychosocial data on twins born in Sweden since 1886 (43–45). The registry contains data on more than 99% of twins in the Swedish population born since 1901 and is regularly updated by matching against national health and death records, regardless of participation in Swedish Twin Registry-initiated data collection. This study used data collected from monozygotic (MZ) and dizygotic (DZ) like-sex pairs born between 1901 and 1925 who completed questionnaires administered in 1967 (Q67) and/or 1970 (Q70). Only selected participants were offered Q70 depending on nonresponse to Q67 or nonresponse to specific items in Q67.

We excluded individuals born in 1886–1900 from the current analyses because only a small proportion of individuals born in these years completed the Q67 or Q70 questionnaires and have data available on number of offspring. We thus opted to exclude the entire cohort, as including those surviving to 1967 could have introduced biases. We chose the birth year 1925 as the upper limit for inclusion because twins born later received somewhat different versions of the questionnaires. Additionally, some were young enough in 1967 to have had additional children after the questionnaire data were obtained.

For individuals who responded to both Q67 and Q70, information on offspring was compared to assess accuracy and to update possible changes in number of children that occurred between questionnaires. Responses to Q67 and Q70 were highly correlated for reported number of children (r = .95) indicating good reliability. For discrepancies of one or two children between questionnaire responses, the larger offspring value was selected for analyses. Participants were excluded from all analyses if there was a discrepancy in number of offspring of three or more (n = 18), invalid values (n = 2), conflicting information on whether their children were adopted or biological (n = 5), and missing values for number of children due to offspring items that were neither completed in Q67 nor Q70 (N = 97). Another 21 twins were lost to follow up and have unknown death age or survival status so they were treated as censored in the analyses.

After excluding the aforementioned participants from the original sample, the sample size was 15,622 individuals, including 8,812 women and 6,810 men. There were 7,264 complete twin pairs and 566 men and 528 women from incomplete pairs.

Whether pairs were identical (MZ) or fraternal (DZ) was determined using the responses of both twins to questionnaire items asking about physical similarity. When compared with the analyses of blood markers, responses to these questions have correctly determined zygosity in 99% of twin pairs (43). Approximately 96% of twin pairs were reared together throughout childhood. The complete male pairs included 1,065 MZ, 1,937 DZ, and 120 with unidentified zygosity. Complete female pairs included 1,396 MZ, 2,598 DZ, and 148 with unidentified zygosity.

Measures

Offspring.—We coded two variables based on the Q67 and Q70 questionnaire items inquiring how many children a participant had. Parenting status was coded as 1 if the individual had any biological children and 0 otherwise. Number of children ranged from 0 to 7, with individuals having more than 7 children assigned a value of 7 when the data were originally computerized. Only 127 individuals (0.8%) had responses of 7, so it is unlikely that much information was lost by truncating this variable at 7. The offspring item does not specifically ask participants to distinguish between number of biological and adopted children. However, if participants offered this information, then it was distinctly coded as “adopted,” and these individuals were excluded from our analyses.

Survival Age.—Dates of death of Swedish Twin Registry participants are determined through regular matches to the Swedish National Cause of Death Registry. Individuals who do not match are presumed alive. Age at death or last known age were calculated using records updated through 2008 and were used for survival analyses. The DST makes predictions about deaths from natural causes. Therefore, twins who died from accidents (as identified in cause of death data available through 2006) were included in the analysis with age at death treated as censored.

Demographic Covariates

Birth year, father’s occupation, and urban, small town, or rural residence were included as covariates in the analyses to test whether any of the fertility–longevity association was attributable to demographic factors. Father’s occupation was used to represent the pair’s socioeconomic status (SES) early in life. This was collected in Q67 and coded as upper (1), middle (2), and lower (3) class. When there was disagreement between ratings within a twin pair, the average of their responses was calculated and used in the analyses. Data on residence in adulthood were coded based on urban, rural, or town residency in Q67 and Q70. If there was disagreement within participants’ responses between Q67 and Q70, the more urban option was selected. As we had no clear a priori hypotheses about ordering of effects, we coded residence using two binary variables: rural = 1 if yes to rural and 0 otherwise; urban = 1 if yes to urban and 0 otherwise.

Analyses

Our primary approach to data analysis was to use survival analysis to test for associations between fertility (i.e., parenting status or number of offspring) and longevity. Given the wide age range of our sample and possible cohort differences in longevity and number of children, all models included birth year to adjust for linear cohort differences in longevity (in preliminary analyses, we tested for nonlinear effects and these were negligible). Whenever fertility effects were detected, we also tested for interactions of these effects with birth year. We initially analyzed men and women combined but observed sex × cohort differences in survival, which would have necessitated including three-way interactions to test some of our key hypotheses (eg, sex × birth year × parenting status). To clarify any sex differences, all subsequent model results are presented separately for men and women.

A first series of nested models used individual-level data. We tested the hypothesis that having offspring is associated with decreased longevity by including parenting status (Parent = 1/0). We then tested whether higher parity is associated with decreased longevity, as suggested by the DST. We used a nested model, keeping the “Parent” variable and adding a multiple child (MultChild) variable (coded as the number of children minus one). This allows the interpretation of “Parent” to remain similar across models (the effect of 0 vs 1 child) and the effect of “MultChild” to be interpreted as the impact associated with each additional child beyond one.

Some prior studies have reported nonlinear (U- or J-shaped) effects of number of children (i.e., higher risk of death associated with no children or many children relative to a few children). We tested for nonlinearity by fitting models with linear and squared terms for number of offspring (“#Kids,” range 1–7). Because the mechanisms contributing to having any children may differ from those influencing how many children (given that one has any), these analyses were conducted within the portion of the sample with at least one child.

After estimating the impact of the fertility variables, we added urban/rural residence and early SES as covariates. The intent of these analyses was to estimate if any observed fertility effects were mediated by these demographic factors.

Cotwin Survival Models

A second series of survival models included cotwin fertility information to test if genetic and/or environmental factors shared by family members partially mediate the association between fertility and longevity. MZ pairs are presumed to share all genetic variation, DZ twins share approximately half of their segregating genetic variation, and both groups are assumed to share the same familial environmental backgrounds (i.e., environmental factors shared by family members [46]).

We have previously used a cotwin survival model to examine the basis of individual- versus family-level predictors on onset of Alzheimer’s disease (47). This is an adaption of cotwin control logistic regression models described by Sham and colleagues (48). The logic for this approach is that under the DST (or other causal hypotheses), survival should be a function of one’s own fertility and should not be related to the fertility of siblings. Thus, inclusion of cotwin fertility information as a predictor is not expected to increase the predictability of survival. In contrast, if the fertility–survival association is an indirect effect, due to overlapping genetic factors or to family environmental factors (such as childhood nutrition and health care), the fertility of one’s cotwin would be correlated with one’s own health status and would be expected to improve the prediction of survival. Environmental factors shared by family members were modeled by predicting one twin’s longevity from cotwin parenting status. Genetic factors were modeled by including an interaction term composed of cotwin parenting status with pair genetic resemblance (MZ = 1 and DZ = 0.5).

Model Testing Approach

Variables were added to survival models sequentially. The fits of nested models were compared by the LR (difference in the model likelihoods) relative to differences in degrees of freedom (df). Given the large sample size, we evaluate the models using effect sizes as well as conventional significance levels. We used a model comparison strategy rather than evaluating the parameters estimates from a single model because of the expected non-negligible covariance among the model terms and our goal of evaluating mediation of the fertility effects by genetic and family background variables.

Data were analyzed at the individual level, ignoring the fact that the observations of twin pairs may be correlated. This yields unbiased estimates of the survival function but (to the extent twins are positively correlated) may underestimate the standard errors of these survival parameters. We consider this issue when evaluating our model estimates. Survival analyses were conducted with the LIFEREG procedure using Weibull distributions in SAS software (version 9.2 [49]).

Bivariate Variance Component Models

Given the importance of investigating genetic variance and covariance (50–52), we also conducted standard twin bivariate analyses to estimate the genetic and environmental sources of covariance between longevity and fertility (coded as having any offspring or number of offspring). For ~15% of twins who were still alive as of the last update of the twin registry against the national death registry (and whose age at death is thus unknown), we substituted age at last update. This approach treats age as continuous and does not account for the censored ages among the surviving twins. However, unlike the survival models, this approach allows us to estimate genetic and environmental correlation between longevity and fertility and to obtain parameter estimates adjusted for the nested nature of the twin data. These analyses were conducted using structural equation modeling with Mplus software (version 6.12 [53]). Details of fitting bivariate twin models with categorical data are described elsewhere (54).

Results

Table 1 displays descriptive information, grouped by sex and zygosity, including death status and number of children among those who had at least one child. These values were similar across zygosity groups within men and women. As expected, women lived longer than men. These differences are addressed in subsequent analyses.

Table 1.

Sample Descriptives: Parenting Status, Number of Children, Percentage Deceased, and Last Recorded Age by Sex and Twin Pair Zygosity

   Number of Children*  Age at Death or Last Contact (y) 
N Childless (%) M (SD) Median Deceased (%) M (SD
Women 
 All 8,812 22.5 2.3 (1.3) 2.0 78.2 81.9 (9.1) 
 In MZ pairs 2,931 21.6 2.3 (1.3) 2.0 78.1 82.0 (9.2) 
 In DZ pairs 5,563 22.9 2.3 (1.3) 2.3 78.0 81.9 (9.1) 
Men 
 All 6,810 24.7 2.2 (1.2) 2.2 87.9 78.1 (9.7) 
 In MZ pairs 2,284 23.0 2.2 (1.2) 2.0 88.1 78.5 (9.7) 
 In DZ pairs 4,260 25.5 2.3 (1.2) 2.2 87.8 78.0 (9.6) 
   Number of Children*  Age at Death or Last Contact (y) 
N Childless (%) M (SD) Median Deceased (%) M (SD
Women 
 All 8,812 22.5 2.3 (1.3) 2.0 78.2 81.9 (9.1) 
 In MZ pairs 2,931 21.6 2.3 (1.3) 2.0 78.1 82.0 (9.2) 
 In DZ pairs 5,563 22.9 2.3 (1.3) 2.3 78.0 81.9 (9.1) 
Men 
 All 6,810 24.7 2.2 (1.2) 2.2 87.9 78.1 (9.7) 
 In MZ pairs 2,284 23.0 2.2 (1.2) 2.0 88.1 78.5 (9.7) 
 In DZ pairs 4,260 25.5 2.3 (1.2) 2.2 87.8 78.0 (9.6) 

Notes. DZ = dizygotic; MZ = monozygotic.

*Among 11,876 individuals with one or more children.

A total of 318 women and 266 men with unknown zygosity are included in the all-male/all-female categories.

An initial set of survival models was run to estimate the importance of birth year, sex, and their interaction. Each of these factors significantly improved model fit (birth year [compared with a baseline model with no variables]: LR = 27.01, df = 1; adding sex: LR = 684.62, df = 1 [with women living longer than men]; and adding sex × birth year interaction: LR = 35.60, df = 1). Because of the evidence for sex differences, the results of subsequent survival models are presented stratified by sex.

Survival and Parenting Status

Figure 1 presents survival distributions by gender and parenting status. Survival was lower among males compared to females and, within gender, childless individuals experienced lower survival compared to individuals with children.

Table 2 presents the results of the individual survival analyses. Parenting status was significantly associated with survival (Model 1 vs baseline—women: LR = 15.17, df = 1; men: LR = 44.11, df = 1). Contrary to the pattern hypothesized by the DST, being a parent was associated with increased survival. Men who were parents experienced somewhat more of a survival advantage. Based on the estimates from this model, the hazard of dying in the next year among fathers was 26% lower than the risk for same-aged childless men, and the hazard for mothers was 16% lower than for same-aged childless women.

Figure 1.

Survival plot, stratified by gender and parenting status.

Figure 1.

Survival plot, stratified by gender and parenting status.

Table 2.

Results of Individual-level Analyses Predicting Survival from Parenting Status, Number of Children,* and Demographic Variables, Stratified by Sex

 Models for Women Models for Men 
All Women (N = 8,765) Mothers (N = 6,785) Women w/Demographic Information (N = 7,542)†  All Men (N = 6,772) Fathers (N = 5,091) Men w/ Demographic Information (N = 5,873)†  
2a 2a 
Parameter estimates (standard errors) 
 Birth year 0.0012(0.0002) 0.0006(0.0004) 0.0006(0.0004) 0.0014(0.0002) 0.0014(0.0002) 0.0005(0.0004) 0.0005(0.0004) −0.0001(0.0002) −0.0007(0.0004) −0.0007(0.0004) 0.0002(0.0002) 0.0002(0.0002) −0.0005(0.0004) −0.0005(0.0004) 
 Parent‡ 0.0113(0.0029) 0.0124(0.0029) 0.0131(0.0032) — — 0.0127(0.0031) 0.0128(0.0031) 0.0221(0.0033) 0.0217(0.0033) 0.0209(0.0037) — — 0.0202(0.0036) 0.0202(0.0036) 
 Parent × birth year — 0.0008(0.0004) 0.0008(0.0004) — — 0.0010(0.0005) 0.0010(0.0005) — 0.0009(0.0005) 0.0009(0.0005) — — 0.0008(0.0005) 0.0008(0.0005) 
 MultChild§ — — −0.0005(0.0011) — — — — — — 0.0007(0.0014) — — — — 
 # Kids|| — — — −0.0005(0.0011) 0.0069(0.0037) — — — — — 0.0007(0.0014) 0.0073(0.0046) — — 
 # Kids2††  — — — — −0.0012(0.0006) — — — — — — −0.00110.0007 — — 
 Early SES¶ — — — — — — −0.0084(0.0023)    — — — −0.0071 (0.0028) 
Intercept 4.4621(0.0025) 4.4609(0.0026) 4.4609(0.0026) 4.4745(0.0029) 4.4658(0.0051) 4.4627(0.0027) 4.4826(0.0062) 4.4028(0.0028) 4.4030(0.0028) 4.4030(0.0028) 4.4234(0.0035) 4.4156(0.0063) 4.4056(0.0031) 4.4229(0.0075) 
Scale 0.1011(0.0010) 0.1011(0.0010) 0.1011(0.0010) 0.1007(0.0012) 0.1007(0.0012) 0.1017(0.0011) 0.1017(0.0011) 0.1101(0.0012) 0.1101(0.0012) 0.1101(0.0012) 0.1096(0.0014) 0.1095(0.0014) 0.1108(0.0013) 0.1108(0.0013) 
Weibull shape 9.8914(0.0998) 9.8892(0.0998) 9.8893(0.0998) 9.9310(0.1158) 9.9353(0.1159) 9.8284(0.1076) 9.8363(0.1076) 9.0849(0.0961) 9.0841(0.0961) 9.0842(0.0961) 9.1277(0.1127) 9.1288(0.1127) 9.0217(0.1031) 9.0261(0.1031) 
Model fit information 
 LR# 15.17 3.93 0.19 0.19 4.21 — 13.04 44.11 3.93 0.23 0.22 2.25 — 6.53 
 Compare to Baseline** Baseline** — 2a Baseline** Baseline** — 2a 
 Models for Women Models for Men 
All Women (N = 8,765) Mothers (N = 6,785) Women w/Demographic Information (N = 7,542)†  All Men (N = 6,772) Fathers (N = 5,091) Men w/ Demographic Information (N = 5,873)†  
2a 2a 
Parameter estimates (standard errors) 
 Birth year 0.0012(0.0002) 0.0006(0.0004) 0.0006(0.0004) 0.0014(0.0002) 0.0014(0.0002) 0.0005(0.0004) 0.0005(0.0004) −0.0001(0.0002) −0.0007(0.0004) −0.0007(0.0004) 0.0002(0.0002) 0.0002(0.0002) −0.0005(0.0004) −0.0005(0.0004) 
 Parent‡ 0.0113(0.0029) 0.0124(0.0029) 0.0131(0.0032) — — 0.0127(0.0031) 0.0128(0.0031) 0.0221(0.0033) 0.0217(0.0033) 0.0209(0.0037) — — 0.0202(0.0036) 0.0202(0.0036) 
 Parent × birth year — 0.0008(0.0004) 0.0008(0.0004) — — 0.0010(0.0005) 0.0010(0.0005) — 0.0009(0.0005) 0.0009(0.0005) — — 0.0008(0.0005) 0.0008(0.0005) 
 MultChild§ — — −0.0005(0.0011) — — — — — — 0.0007(0.0014) — — — — 
 # Kids|| — — — −0.0005(0.0011) 0.0069(0.0037) — — — — — 0.0007(0.0014) 0.0073(0.0046) — — 
 # Kids2††  — — — — −0.0012(0.0006) — — — — — — −0.00110.0007 — — 
 Early SES¶ — — — — — — −0.0084(0.0023)    — — — −0.0071 (0.0028) 
Intercept 4.4621(0.0025) 4.4609(0.0026) 4.4609(0.0026) 4.4745(0.0029) 4.4658(0.0051) 4.4627(0.0027) 4.4826(0.0062) 4.4028(0.0028) 4.4030(0.0028) 4.4030(0.0028) 4.4234(0.0035) 4.4156(0.0063) 4.4056(0.0031) 4.4229(0.0075) 
Scale 0.1011(0.0010) 0.1011(0.0010) 0.1011(0.0010) 0.1007(0.0012) 0.1007(0.0012) 0.1017(0.0011) 0.1017(0.0011) 0.1101(0.0012) 0.1101(0.0012) 0.1101(0.0012) 0.1096(0.0014) 0.1095(0.0014) 0.1108(0.0013) 0.1108(0.0013) 
Weibull shape 9.8914(0.0998) 9.8892(0.0998) 9.8893(0.0998) 9.9310(0.1158) 9.9353(0.1159) 9.8284(0.1076) 9.8363(0.1076) 9.0849(0.0961) 9.0841(0.0961) 9.0842(0.0961) 9.1277(0.1127) 9.1288(0.1127) 9.0217(0.1031) 9.0261(0.1031) 
Model fit information 
 LR# 15.17 3.93 0.19 0.19 4.21 — 13.04 44.11 3.93 0.23 0.22 2.25 — 6.53 
 Compare to Baseline** Baseline** — 2a Baseline** Baseline** — 2a 

Notes. SES = socioeconomic status. Degrees of freedom (df) for all comparisons = 1.

*N = 85 excluded from these analyses because number of children was unknown.

† Models based on subset of participants who had data on demographic information (specifically early SES and residency).

‡“Parent” indicates parenting status, coded as 0 if no children and 1 if any children.

§“MultChild” indicates number of children, coded as number of children − 1 (range 0–6).

||“#Kids” indicate number of children. Only includes parents with one or more children.

†“#Kids2” indicates nonlinear effect of number of children. Only includes parents with one or more children.

¶Early SES represents SES in childhood.

#LR = likelihood ratio relative to the comparison model. All comparisons have df = 1.

**Baseline model used for comparison in model 1 consisted of birth year alone and included all women and men: −2LL for baseline model is −5352.49 for women and −5160.45 for men. Baseline model used for comparison in model 4 consisted of birth year for mothers and fathers only: −2LL for baseline model is −3793.04 for mothers and −3747.38 for fathers.

Given the wide age range of our sample and possible cohort differences in parenting, we estimated whether the effect of having children on survival varied across birth cohorts. A parenting × birth year interaction improved model fit only slightly (Model 2 vs 1—women: LR = 3.93, df = 1; men: LR = 3.93, df = 1), with the positive effect of parenting being somewhat greater for individuals born more recently.

Survival and Number of Offspring

We then tested whether higher parity is associated with decreased longevity, as suggested by the DST. The effect of having multiple offspring on longevity was trivial for both men and women (Model 3 vs 2—women: LR = 0.19, df = 1; men: LR = 0.23, df = 1). We also examined the evidence for differences in longevity associated with number of offspring only among parents (i.e., excluding those with no children). There was virtually no evidence for a linear effect of children on survival (Model 4 vs baseline—mothers and fathers: LRs < 0.3, df = 1). However, a model including both linear and quadratic terms significantly improved model fit among women (Model 5 vs 4—LR = 4.21, df = 1) but not men (LR = 2.25, df = 1). These effects associated with number of children among parents did not vary by birth year (b = −0.0001, SE = 0.0002 for both sexes; not shown in Table 2).

These sex differences and cohort effects are illustrated in Figure 2, which shows the estimated hazard of death at 80 years based on the results of Models 3 (including childless) and 5 (parents only). The estimated hazard of dying in the next year among individuals surviving to age 80 is about 8% in men and 5% in women. For both sexes, individuals with children have lower hazards than those with no children. This effect is more pronounced for individuals in more recent birth cohorts (i.e., born 1925) versus those born earlier. Among men with at least one child, there is no discernible difference in the hazard by birth cohort. In contrast, among women, the advantage associated with children is larger for those born in more recent cohorts. The effect of number of children is very slight and reliably different only for women (as indicated by Model 5). The hazard decreases for women with more than one child, is lowest for women with three to four children, and shows a small increase for five or more.

Figure 2.

Predicted hazard of death at age 80 by sex, birth year, and number of children.

Figure 2.

Predicted hazard of death at age 80 by sex, birth year, and number of children.

Demographic Covariates: Residence and Early SES

Table 2 also presents results of testing whether demographic variables may mediate some of the association between fertility and longevity. These models were fit using the subset of individuals (~86%) who had data on these variables. For both sexes, adding early SES to a model with the parenting status effect (Model 6) improved model fit (Model 6 vs 2a—women: LR = 13.04, df = 1; men: LR = 6.53, df = 1). Higher status was associated with increased longevity (particularly in women). A model testing the effects of urban and rural residence indicated that these variables were not related to longevity (results not presented in Table 2; women: LR = 1.99, df = 2; men: LR = 1.70, df = 2).

To evaluate whether early SES mediated the association between fertility and survival, we examined the parameter estimates for the parenting status effect with (Model 6) and without (Model 2a) SES included in the models. The regression weight for parenting status was essentially unchanged in both men and women (women: b = 0.0127 vs b = 0.0128, SE = 0.0031 [same SE in both models]; men: remained b = 0.0008, SE = 0.0005 in both models), suggesting that the survival advantage associated with having offspring is not due to differences between parents and childless based on early SES (at least as reflected by our measure).

Cotwin Survival Analyses

The results from the analyses including cotwin fertility information are summarized in Table 3. Model analyses were conducted with the subset of the sample with known cotwin fertility. Results were essentially identical for men and women, so they are presented combined. Model fits are compared with a baseline model consisting of birth year, sex, and the birth year x sex interaction. As in the analyses based on the full individual-level sample (Table 2, Model 1), parenting status is strongly associated with survival (Table 3, Model C1, LR = 52.22, df = 1). In contrast, including cotwin parenting status did not improve fit (Model C2, LR = 1.84, df = 1), indicating that unmeasured environmental factors shared by family members do not explain the association between parental status and longevity. Similarly, including the interaction between cotwin parenting status and zygosity did not improve the fit (Model C3, LR = 0.48, df = 1), indicating that genetic factors do not contribute to the association.

Table 3.

Results of Analyses Predicting Survival from Parenting Status of Individuals and their Cotwins*

 Model # 
C1 C2 C3 C4 
Parameter estimates (standard errors) 
 Birth year 0.0007 (0.0001) 0.0007 (0.0002) 0.0008 (0.0001) 0.0007 (0.0002) 
 Sex† 0.0249 (0.0010) 0.0248 (0.0010) 0.0248 (0.0010) 0.0249 (0.0010) 
 Birth year × sex 0.0008 (0.0001) 0.0008 (0.0001) 0.0008 (0.0001) 0.0008 (0.0001) 
 Individual parent‡ 0.0168 (0.0023)   0.0167 (0.0023) 
 Cotwin parent‡  0.0031 (0.0023)  0.0016 (0.0028) 
 Cotwin parent x zygosity§   −0.0014 (0.0020) −0.0033 (0.0024) 
Model fit information 
 LR|| 52.22 1.84 0.48 54.17 
df¶ 1  1  
 Model # 
C1 C2 C3 C4 
Parameter estimates (standard errors) 
 Birth year 0.0007 (0.0001) 0.0007 (0.0002) 0.0008 (0.0001) 0.0007 (0.0002) 
 Sex† 0.0249 (0.0010) 0.0248 (0.0010) 0.0248 (0.0010) 0.0249 (0.0010) 
 Birth year × sex 0.0008 (0.0001) 0.0008 (0.0001) 0.0008 (0.0001) 0.0008 (0.0001) 
 Individual parent‡ 0.0168 (0.0023)   0.0167 (0.0023) 
 Cotwin parent‡  0.0031 (0.0023)  0.0016 (0.0028) 
 Cotwin parent x zygosity§   −0.0014 (0.0020) −0.0033 (0.0024) 
Model fit information 
 LR|| 52.22 1.84 0.48 54.17 
df¶ 1  1  

*N = 13,862 with cotwin information available. Model C1 is the same as Model 1 in Table 2, but was fit to data for subset of sample (N = 13,862).

†Sex: men = 5,953 and women = 7,909. Sex coded as −1 if men and 1 if women.

‡“Parent” indicates parenting status, coded as 0 if no children, 1 if any children. “Individual parent” and “cotwin parent” indicate the parenting status of the individual and cotwin.

§Zygosity: dizygotic (DZ) = 0.5; monozygotic (MZ) = 1.

||LR = likelihood ratio relative to the comparison model. Baseline demographic model includes birth year, sex, and birth year × sex had −2LL = 9198.34.

df = degrees of freedom.

To look for evidence of familial mediation, estimates for the individual parenting status effect were compared before and after adding cotwin effects. The regression estimate for parenting status is unchanged in Model C4 (b = 0.0167, SE = 0.0023), compared with Model C1 (b = 0.0168, SE = 0.0023), indicating that individual-level processes are responsible for the association between having children and longer survival. These results suggest that the fertility-suvival association is not mediated by familial environmental or genetic factors.

Bivariate Variance Component Analysis of Longevity and Fertility

The estimated within-person correlations between age at death (or last contact) with parenting status were as follows: women, r = .04; men, r = .10, and with number of children were as follows: women, r = .03; men, r = .06 (all SE = 0.01). The correlations between twins in a pair for parenting status, age at death, and across variables (Twin 1 parenting status—Twin 2 death age; Twin 2 parenting status—Twin 1 death age) did not differ from zero for MZ or DZ pairs (range −0.03 to 0.01). Consistent with these values and with the results of the cotwin survival models, the results of the bivariate twin models indicated that virtually all of the fertility-longevity covariance was due to within-person sources. There was no evidence of common environmental sources contributing to the association. The estimated genetic correlation between parenting and longevity was r = .00 (SE > 10) for women and was r = 0.01 (SE > 10) for men.

Discussion

For both men and women in this sample, having any children was associated with higher longevity relative to childless individuals, with male parents experiencing somewhat more of a survival advantage. The effect of being a parent relative to being childless was particularly pronounced among individuals born more recently compared with those born earlier. There was only a small nonlinear association between number of children and survival, with little change in mortality risk as number of children increased. The results also indicated more of a survival advantage associated with having one or more children in later cohorts relative to earlier cohorts among women but not men. Individuals of higher SES in early life had increased longevity relative to those of lower status, but this did not mediate the association between parenting and longevity. Including information on cotwin fertility and longevity enabled us to rule out that this association was due to genetic or environmental factors shared by family members and indicated that the association was due to factors at the individual level.

The results are inconsistent with the DST, which predicts a negative effect of parenthood on survival, primarily in women, and predicts an increasing effect with greater number of children. The results of this study are consistent with individual fertility having a causal relation with longevity or with a third variable or process operating at the individual level that contributes to both fertility and longevity.

Possible Mechanisms Underlying Increased Longevity in Parents

It is important to consider mechanisms underlying longevity of individuals who are parents, though many possible mechanisms could not be tested by this study. There may be direct and indirect mechanisms by which having offspring relates to longevity. Adult offspring may contribute directly to the well-being of their older parents through financial and social support (55–57). Indirect effects on longevity may occur via greater life satisfaction among those with children (29,30). Other possible sources of the association include religiosity, financial stability, and social networks (58–62).

Although both male and female parents experienced higher longevity relative to the childless, there were some interesting sex differences. Men experienced a slightly higher survival advantage associated with being a parent. One explanation for this is the positive effect of marriage rather than the effect of having children. Unmarried men engage in risk-taking health behaviors (eg, alcohol abuse) more often than unmarried women, and the decline of such behaviors after marriage is more dramatic among men compared with women (63). A decline in risk-taking behaviors may partially account for the survival advantage observed in male parents.

The cotwin design allowed us to rule out unmeasured genetic and family background factors as explaining the association between parenting and longevity. This indicates there are other individual-level factors underlying the association. The finding that the beneficial effect on longevity was from parenting status rather than number of children suggests the mechanism is indirect, due to some factor correlated with being a parent. One possibility is that individuals who have children may have better overall health relative to childless individuals. Individuals with chronic health and mental health problems are less likely to have children and they also have lower longevity (64–66). Another possible mechanism is marriage. Benefits of marriage, such as economic support, higher life satisfaction, and reduced stress, may have a direct positive effect on longevity (63,67,68). There may also be an indirect effect through selection, with healthier individuals being more desirable potential spouses (69,70). In our sample, having children was nearly perfectly correlated with being married, so we were unable to evaluate these separately.

The finding of an individual-level association between parenting and survival is counter to the trend observed at the aggregate level, with many countries demonstrating decreased birthrates accompanied by increased longevity (1,2). The influences on childbearing in the early 20th century may differ from those in more recent times. Consistent with this, we found that the positive effect of being a parent on survival was more pronounced for individuals born more recently. For older cohorts, the protective effect associated with children may have been outweighed by the relatively larger negative effects of infectious disease and other illnesses that had less impact on younger cohorts. For example, greater exposure to sources of inflammation among older cohorts could have led to health problems and increased mortality (71). Improvements in medicine and standards of living in later cohorts may have decreased the degree to which health problems interfered with detecting the magnitude of the effect of being a parent on survival, thereby allowing the effect of parenting status to be more pronounced in later cohorts. We also found sex-based cohort differences among parents with one or more children, with mothers in later cohorts having lower mortality risk than mothers in older cohorts. These cohort differences in women may indicate some form of selection: perhaps the mechanisms associated with havin g kids or getting married differ for mothers in younger cohorts.

Implications for the DST

The present findings of increased longevity in parents compared with childless individuals regardless of number of offspring and sex contrasts with the DST’s prediction that female longevity is expected to decrease with additional offspring. Our results indicate that the fertility–survival association occurs due to factors at the level of the individual. Furthermore, the results from cotwin analyses enable us to conclude that genetic and family environmental factors do not account for the survival–parenting association. Some evolutionary models would predict a negative genetic correlation between longevity and fertility (52). Our finding of essentially a zero genetic correlation suggests that if this process did once exist, it is not pertinent to modern populations.

Some characteristics of our study may help explain why the results differ from past findings that have been cited as evidence supporting the DST. The Swedish Twin Registry provides a more representative sample based in the general population. It is possible that historical changes have reduced the impact of childbearing on longevity. For example, methods of birth control introduced in the 20th century permit women greater regulation over pregnancy (72). The regulation of reproduction in modern populations could interfere with natural fertility conditions seen in historic populations. Birth control could have particularly benefitted women who experienced difficulties and health problems during pregnancy and childbirth, preventing early deaths due to pregnancies.

Limitations and Strengths

The sample’s homogeneous composition (i.e., native-born Swedes) may impact the generalizability of the results to other populations. However, homogeneity can be considered beneficial, as heterogeneity could introduce confounding effects on fertility and longevity. It is possible that this birth cohort may be atypical. For example, the childlessness rate of 20% may reflect specific influences on health or fertility that do not pertain to subsequent generations or other populations (eg, food shortages during WWI and WWII, or the 1917–1918 influenza epidemic).

Another weakness is the absence of data on other variables relevant to childbearing. Ideally, a thorough test of the DST would include information on miscarriages, abortions, and children who died in addition to number of surviving offspring. However, the absence of this information is unlikely to have altered the conclusion that the findings are inconsistent with the DST.

Individual variability in energy investments was not investigated in this study. A multitude of factors influence how taxing reproduction will be on an individual, including but not limited to nutrition, specific characteristics of the environment, and menstrual cycle patterns (6). Other evidence suggests the gender of the child influences energy costs (73,74). Although this study controlled for some relevant variables, an exhaustive inspection of other reproductive costs and the overall energy budget is generally not possible in epidemiological studies. The large number of factors that influence reproductive costs serves as a reminder that the DST cannot be applied simplistically to humans.

The study also has important strengths. It adds to a small set of others that have used large, relatively modern, population-based samples. Unlike some prior studies, we had sufficient statistical power to test the predictions of the DST. To our knowledge, this is the first study to test the DST using data from twins. This design allowed us to rule out genetic and shared family environmental factors as explanations for the association between parenting and longevity. The consistency in findings across the survival analyses and the bivariate twin models increases our confidence that individual-level factors primarily account for the observed association between parenting and longevity.

Conclusions

These findings provide evidence for a positive association between longevity and being a parent, regardless of number of offspring, in both men and women. Although additional research is needed to identify specific mechanisms, our novel approach using a large population-based sample of twins permits a confident rejection of the DST, at least in modern, western populations.

Funding

The Swedish Twin Registry is supported by a grant from the Ministry of Higher Education and the Swedish Research Council.

Acknowledgments

We thank the staff at the Karolinska Institutet for administrative support and the twins for their participation in the Registry. We thank Tuck Finch, Frank Manis, Lewina Lee, Nicole Sintov, Kelly Young-Wolff, Archana Jajodia, Emily Fine-Foster, and anonymous reviewers for their comments on earlier versions of this paper.

References

1.
United Nations
.
Report of the Second World Assembly on Ageing
 ,
2002
.
2.
United Nations Department of Economic and Social Affairs, Population Division
.
World Population Prospects
 .
2006
.
3.
World Health Organization Department of Health Statistics and Informatics
.
World Health Statistics
 .
2011
.
4.
Le Bourg
E
.
A mini-review of the evolutionary theories of aging: is it time to accept them?
Demogr Res
 .
2001
;
4
:
1
28
.
5.
Le Bourg
E
.
Does reproduction decrease longevity in human beings?
Ageing Res Rev
 .
2007
;
6
:
141
149
.
6.
Jasienska
G
.
Reproduction and lifespan: trade-offs, overall energy budgets, intergenerational costs, and costs neglected by research
.
Am J Hum Biol
 .
2009
;
21
:
524
532
.
7.
Hurt
LS
Ronsmans
C
Thomas
SL
.
The effect of number of births on women’s mortality: systematic review of the evidence for women who have completed their childbearing
.
Popul Stud (Camb)
 .
2006
;
60
:
55
71
.
8.
Westendorp
RG
Kirkwood
TB
.
Human longevity at the cost of reproductive success
.
Nature
 .
1998
;
396
:
743
746
.
9.
Gagnon
A
Mazan
R
Desjardins
B
Smith
KR
.
Post-reproductive longevity in a natural fertility population
.
PSC Discus Papers Ser
 .
2005
;
19
:
1
.
10.
Smith
KR
Mineau
GP
Bean
LL
.
Fertility and post-reproductive longevity
.
Soc Biol
 .
2002
;
49
:
185
205
.
11.
Gagnon
A
Smith
KR
Tremblay
M
Vézina
H
Paré
PP
Desjardins
B
.
Is there a trade-off between fertility and longevity? A comparative study of women from three large historical databases accounting for mortality selection
.
Am J Hum Biol
 .
2009
;
21
:
533
540
.
12.
Penn
DJ
Smith
KR
.
Differential fitness costs of reproduction between the sexes
.
Proc Natl Acad Sci USA
 .
2007
;
104
:
553
558
.
13.
Doblhammer
G
.
Reproductive history and mortality later in life: a comparative study of England and Wales and Austria
.
Popul Stud (Camb)
 .
2000
;
54
:
169
176
.
14.
Korpelainen
H
.
Fitness, reproduction and longevity among European aristocratic and rural Finnish families in the 1700s and 1800s
.
Proc Biol Sci
 .
2000
;
267
:
1765
1770
.
15.
Doblhammer
G
Oeppen
J
.
Reproduction and longevity among the British peerage: the effect of frailty and health selection
.
Proc Biol Sci
 .
2003
;
270
:
1541
1547
.
16.
Le Bourg
E
Thon
B
Légaré
J
Desjardins
B
Charbonneau
H
.
Reproductive life of French-Canadians in the 17-18th centuries: a search for a trade-off between early fecundity and longevity
.
Exp Gerontol
 .
1993
;
28
:
217
232
.
17.
Helle
S
Käär
P
Jokela
J
.
Human longevity and early reproduction in pre-industrial Sami populations
J Evol Biol
 .
2002
;
15
:
803
807
.
18.
Gavrilova
NS
Gavrilov
LA
Semyonova
VG
Evdokushkina
GN
.
Does exceptional human longevity come with a high cost of infertility? Testing the evolutionary theories of aging
.
Ann N Y Acad Sci
 .
2004
;
1019
:
513
517
.
19.
Müller
HG
Chiou
JM
Carey
JR
Wang
JL
.
Fertility and life span: late children enhance female longevity
.
J Gerontol A Biol Sci Med Sci
 .
2002
;
57
:
B202
B206
.
20.
McArdle
PF
Pollin
TI
O’Connell
JR
et al
Does having children extend life span? A genealogical study of parity and longevity in the Amish
.
J Gerontol A Biol Sci Med Sci
 .
2006
;
61
:
190
195
.
21.
Kirkwood
TB
.
Evolution of ageing
.
Nature
 .
1977
;
270
:
301
304
.
22.
Carnes
BA
.
What is lifespan regulation and why does it exist?
Biogerontology
 .
2011
;
12
:
367
374
.
23.
Carnes
BA
Olshansky
SJ
Grahn
D
.
Biological evidence for limits to the duration of life
.
Biogerontology
 .
2003
;
4
:
31
45
.
24.
Rattan
SI
.
Biogerontology: the next step
.
Ann N Y Acad Sci
 .
2000
;
908
:
282
290
.
25.
Medawar
P
.
An Unsolved Problem in Biology
 .
London
:
HK Lewis
;
1952
.
26.
Williams
GC
.
Pleiotropy, natural selection, and the evolution of senescence
.
Evolution
 .
1957
;
11
:
398
411
.
27.
Downey
DB
.
When bigger is not better: family size, parental resources, and children’s educational performance
Am Sociol Rev
 .
1995
;
60
:
746
761
.
28.
Eastwood
R
Lipton
M
.
Demographic transition and poverty: Effects via economica growth, distribution, and conversion
. In:
Birdsall
N
Kelley
AC
, eds.
Population Matters: Demographic Change, Economic Growth, and Poverty in the Developing World
 .
New York
:
Oxford University Press
;
2001
:
213
216
.
29.
Fiscella
K
Franks
P
Gold
MR
Clancy
CM
.
Inequality in quality: addressing socioeconomic, racial, and ethnic disparities in health care
.
JAMA
 .
2000
;
283
:
2579
2584
.
30.
Steenland
K
Hu
S
Walker
J
.
All-cause and cause-specific mortality by socioeconomic status among employed persons in 27 US states, 1984-1997
.
Am J Public Health
 .
2004
;
94
:
1037
1042
.
31.
Rosenzweig
MR
Evenson
R
.
Fertility, schooling, and the economic contribution of children of rural India: An econometric analysis
.
Econometrica
 .
1977
;
1065
1079
.
32.
Moen
P
Wethington
E
.
The concept of family adaptive strategies
Annu Rev Sociol
 .
1992
;
18
:
233
251
.
33.
Schmitz
CL
Traver
EK
Larson
D
Pieris
P
.
Introduction
. In:
Schmitz
CL
Traver
EK
Larson
D
, eds.
Child labor: A Global View
 .
Westport, CN
:
Greenwood Publishing Group
;
2004
:
2
30
.
34.
Lindenmeyer
K
.
2004
;
Child labor
. In:
Mink
G
O’Connor
AO
, eds.
Poverty in the United States: An Encyclopedia of History, Politics, and Policy
 .
Santa Barbara, CA
:
ABC-CLIO
;
156
159
.
35.
Gögele
M
Pattaro
C
Fuchsberger
C
Minelli
C
Pramstaller
PP
Wjst
M
.
Heritability analysis of life span in a semi-isolated population followed across four centuries reveals the presence of pleiotropy between life span and reproduction
.
J Gerontol A Biol Sci Med Sci
 .
2011
;
66
:
26
37
.
36.
Pettay
JE
Kruuk
LE
Jokela
J
Lummaa
V
.
Heritability and genetic constraints of life-history trait evolution in preindustrial humans
.
Proc Natl Acad Sci USA
 .
2005
;
102
:
2838
2843
.
37.
Kohler
HP
Rodgers
JL
Christensen
K
.
Is fertility behavior in our genes? Findings from a Danish twin study
.
Popul Dev Rev
 .
1999
;
25
:
253
288
.
38.
Rodgers
JL
Kohler
HP
Kyvik
KO
Christensen
K
.
Behavior genetic modeling of human fertility: findings from a contemporary Danish Twin Study
.
Demography
 .
2001
;
38
:
29
42
.
39.
Herskind
AM
McGue
M
Holm
NV
Sørensen
TI
Harvald
B
Vaupel
JW
.
The heritability of human longevity: a population-based study of 2872 Danish twin pairs born 1870-1900
.
Hum Genet
 .
1996
;
97
:
319
323
.
40.
vB Hjelmborg
J
Iachine
I
Skytthe
A
et al
Genetic influence on human lifespan and longevity
.
Hum Genet
 .
2006
;
119
:
312
321
.
41.
Ljungquist
B
Berg
S
Lanke
J
McClearn
GE
Pedersen
NL
.
The effect of genetic factors for longevity: a comparison of identical and fraternal twins in the Swedish Twin Registry
.
J Gerontol A Biol Sci Med Sci
 .
1998
;
53
:
M441
M446
.
42.
Iachine
IA
Holm
NV
Harris
JR
et al
How heritable is individual susceptibility to death? The results of an analysis of survival data on Danish, Swedish and Finnish twins
.
Twin Res
 .
1998
;
1
:
196
205
.
43.
Lichtenstein
P
De Faire
U
Floderus
B
Svartengren
M
Svedberg
P
Pedersen
NL
.
The Swedish Twin Registry: a unique resource for clinical, epidemiological and genetic studies
.
J Intern Med
 .
2002
;
252
:
184
205
.
44.
Lichtenstein
P
Sullivan
PF
Cnattingius
S
et al
The Swedish Twin Registry in the third millennium: an update
.
Twin Res Hum Genet
 .
2006
;
9
:
875
882
.
45.
Pedersen
NL
Lichtenstein
P
Svedberg
P
.
The Swedish Twin Registry in the third millennium
.
Twin Res
 .
2002
;
5
:
427
432
.
46.
Neale
MC
Cardon
LR
.
Methodology for Genetic Studies of Twin Families
 .
Dordrecht
:
Kluwer
;
1992
.
47.
Gatz
M
Prescott
CA
Pedersen
NL
.
Lifestyle risk and delaying factors
.
Alzheimer Dis Assoc Disord
 .
2006
;
20
(
3 Suppl 2
):
S84
S88
.
48.
Sham
PC
Walters
EE
Neale
MC
Heath
AC
MacLean
CJ
Kendler
KS
.
Logistic regression analysis of twin data: estimation of parameters of the multifactorial liability-threshold model
.
Behav Genet
 .
1994
;
24
:
229
238
.
49.
SAS Institute
.
SAS Software Version 9.2
 .
Cary, NC
:
SAS Institute
;
2002
.
50.
Reznick
D
Perry
E
Travis
J
.
Measuring the cost of reproduction: a comment on papers by Bell
.
Evolution
 .
1986
;
40
:
1338
1344
.
51.
Reznick
D
.
Costs of reproduction: an evaluation of the empirical evidence
.
Oikos
 .
1985
;
44
:
257
267
.
52.
Blomquist
GE
.
Trade-off between age of first reproduction and survival in a female primate
.
Biol Lett
 .
2009
;
5
:
339
342
.
53.
Muthen
LK
Muthen
BO
.
Mplus User’s Guide
 .
6th ed.
Los Angeles, CA
:
Muthen & Muthen
;
1998
.
54.
Prescott
CA
.
Using the Mplus computer program to estimate models for continuous and categorical data from twins
.
Behav Genet
 .
2004
;
34
:
17
40
.
55.
Brubaker
TH
.
Families in later life: A burgeoning research area
J Marriage Fam
 .
1990
;
523
:
959
981
.
56.
O’Neill
C
Sorensen
ES
.
Home care of the elderly: a family perspective
.
ANS Adv Nurs Sci
 .
1991
;
13
:
28
37
.
57.
McMullin
JA
Marshall
VW
.
Family, friends, stress, and well-being: Does childlessness make a difference?
Can J Aging
 .
1996
;
15
:
355
373
.
58.
Beckman
LJ
Houser
BB
.
The consequences of childlessness on the social-psychological well-being of older women
.
J Gerontol
 .
1982
;
37
:
243
250
.
59.
Keith
PM
.
A comparison of the resources of parents and childless men and women in very old age
Fam Relat
 .
1983
;
32
:
403
409
.
60.
Rempel
J
.
Childless elderly: What are they missing?
J Marriage Fam
 .
1985
;
47
:
343
348
.
61.
Zhang
Z
Hayward
MD
.
Childlessness and the psychological well-being of older persons
.
J Gerontol B Psychol Sci Soc Sci
 .
2001
;
56
:
S311
S320
.
62.
Koropeckyj-Cox
T
.
Loneliness and depression in middle and old age: are the childless more vulnerable?
J Gerontol B Psychol Sci Soc Sci
 .
1998
;
53
:
S303
S312
.
63.
Umberson
D
.
Family status and health behaviors: social control as a dimension of social integration
.
J Health Soc Behav
 .
1987
;
28
:
306
319
.
64.
Auquier
P
Lançon
C
Rouillon
F
Lader
M
Holmes
C
.
Mortality in schizophrenia
.
Pharmacoepidemiol Drug Saf
 .
2006
;
15
:
873
879
.
65.
McGrath
JJ
Hearle
J
Jenner
L
Plant
K
Drummond
A
Barkla
JM
.
The fertility and fecundity of patients with psychoses
.
Acta Psychiatr Scand
 .
1999
;
99
:
441
446
.
66.
Anderson
BJ
Wolf
FM
.
Chronic physical illness and sexual behavior: psychological issues
.
J Consult Clin Psychol
 .
1986
;
54
:
168
175
.
67.
Pearlin
LI
Johnson
JS
.
Marital status, life-strains and depression
.
Am Sociol Rev
 .
1977
;
42
:
704
715
.
68.
Waite
LJ
.
Does marriage matter?
Demography
 .
1995
;
32
:
483
507
.
69.
Waldron
I
Hughes
ME
Brooks
TL
.
Marriage protection and marriage selection–prospective evidence for reciprocal effects of marital status and health
.
Soc Sci Med
 .
1996
;
43
:
113
123
.
70.
Goldman
N
.
Marriage selection and mortality patterns: inferences and fallacies
.
Demography
 .
1993
;
30
:
189
208
.
71.
Finch
CE
Crimmins
EM
.
Inflammatory exposure and historical changes in human life-spans
.
Science
 .
2004
;
305
:
1736
1739
.
72.
Rengel
M
.
Encyclopedia of Birth Control
 .
Santa Barbara, CA
:
Greenwood
;
2000
;
221
234
.
73.
Beise
J
Voland
E
.
Effect of producing sons on maternal longevity in premodern populations
Science
 .
2002
;
298
:
317
.
74.
Helle
S
Lummaa
V
Jokela
J
.
Sons reduced maternal longevity in preindustrial humans
.
Science
 .
2002
;
296
:
1085
.

Author notes

Decision Editor: Rafael de Cabo, PhD