Protective Effects of Familial Longevity Decrease With Age and Become Negligible for Centenarians

Abstract

It is known that biological relatives of long-lived individuals demonstrate lower mortality and longer life span compared to relatives of shorter-lived individuals, and at least part of this advantage is likely to be genetic. Less information, however, is available about effects of familial longevity on age-specific mortality trajectories. We compared mortality patterns after age 50 years for 10 045 siblings of US centenarians and 12 308 siblings of shorter-lived individuals (died at age 65 years). Similar comparisons were made for sons and daughters of longer-lived parents (both parents lived 80 years and more) and shorter-lived parents (both parents lived less than 80 years) within each group of siblings. Although relatives of longer-lived individuals have lower mortality at younger ages compared to relatives of shorter-lived individuals, this mortality advantage practically disappears by age 100 years. To validate this observation further, we analyzed the survival of 3 408 US centenarians born in 1890–1897 with known information on maternal and paternal life span. We found using the Cox proportional hazards model that both maternal and paternal longevity (life span 80+ years) is not significantly associated with survival after age 100 years. The results are compatible with the predictions of reliability theory of aging suggesting higher initial levels of system redundancy (reserves) in individuals with protective familial/genetic background and hence lower initial mortality. Heterogeneity hypothesis is another possible explanation for the observed phenomena.

Studies of familial longevity have deep historic roots. In 1899, the founder of biometrics Karl Pearson and his student Mary Beeton found a positive correlation between parent and child ages at death based on English genealogies dating back to the seventeenth century (1). Following the work by Raymond Pearl that ancestors of long-living persons (nonagenarians) had a substantially higher life span compared to a control population (2), numerous studies found that biological relatives of longer-lived individuals have a substantial survival advantage compared to relatives of shorter-lived individuals (3–10).

Although a strong familial clustering of longevity is now a well-established fact, heritability estimates for life span using standard methods of quantitative genetics turn out to be low or moderate (11–16). These estimates were obtained using both twin and family-linked data. The assumption of linear dependence between offspring and parental traits is fundamental in quantitative genetics because both the theory of quantitative genetics and its applications are based on this assumption. However, a study of heritability in European aristocratic families demonstrated that narrow-sense heritability of life span is relatively small (0.25) when data are analyzed for the whole range of parental life spans, while for parents living more than 75 years heritability estimates are significantly higher: 0.50 and more (17,18). These findings were later independently confirmed by other researchers (19–21). It was found that familial resemblance in life span between parents and children is very small when parents live shorter lives (30–70 years) and very strong in the case of longer-lived parents (80+), suggesting an unusual nonlinear pattern of life-span inheritance (17). The age of changing point for heritability increase was called a longevity threshold (17). Another way to estimate a longevity threshold is to use 10%–15% survival, which corresponds approximately to 80–90 years (22).

These observations may explain the existing longevity paradox: Although the heritability estimates for life span were reported to be rather low (11), it is well known that cases of extreme longevity have a strong familial clustering (21,23). Recently, researchers proposed to distinguish heritability of life span, which is relatively low from heritability of longevity, which ensures the clustering of longer-lived individuals in families (15). It was suggested that longevity can be defined as the age at top 10% survivors (22). Sebastiani et al. (21,24) advocate to measure longevity as percentile of survival for the cohort in question and define exceptional longevity as living past the age at which less than 1% of the cohort survived. This is another way to define long life, although in other studies of familial longevity persons living 80 years and more are considered to be long-lived (5,6,17), although in other studies of familial longevity, persons living 80 years and more are considered to be long-lived (5,6,17).

Most researchers believe that familial differences in longevity are at least partially caused by genetic differences (25–27). The candidate gene studies consistently found 2 loci associated with longevity: the apolipoprotein E (APOE) locus and the forkhead box O3 (FOXO3) locus (16). APOE is a protein involved in cholesterol transport that binds to low-density lipoprotein receptor and is crucial to blood cholesterol levels (27,28). FOXO3 is a part of the insulin–IGF1 signaling pathway (29). Its association with longevity was first found in 2008 (30) and now a total of 17 single-nucleotide polymorphisms (SNPs) in FOXO3 were found to be associated with longevity (29). These associations were later confirmed using a genome-wide association study approach (27). In addition to specific longevity genes, longevity may be a polygenic trait influenced by multiple genes and individuals with a higher polygenic risk score for longevity live longer (31).

Still, many questions related to familial longevity remain unresolved. In particular, it is not clear whether protective effects of familial longevity operate at both younger and older ages. By effects here we consider the strength of relationship between familial longevity and survival. Some studies suggest that the survival advantage from familial longevity is sustainable and life-long (9,32), but the exact answer to this question still remains unclear. In this study, we explored the patterns of familial effects on human life span using US data on siblings of shorter-lived and longer-lived individuals. Specifically, we tested prediction of the hypothesis based on reliability theory of aging that children born to longer-lived parents will experience survival advantage mostly in their younger adult ages (33), because they have more initial reserves in functional elements (cells) not yet exhausted over time. Alternative hypothesis suggests the life-long sustained mortality advantage for persons having protection of familial longevity (9,32). Another explanation of mortality trajectories comes from the heterogeneity hypothesis. It suggests that weaker individuals die first leaving more robust subpopulation alive at older ages. This stronger subpopulation has a lower risk of death leading to slower mortality growth, which is called mortality deceleration (16,34,35).

Method

Study Population

In this study, we compare siblings of centenarians born in the United States to siblings of shorter-lived controls, also born in the United States, but died at age 65 years. Data were collected through a screening of over 400 000 online family histories available at Rootsweb (36), which is one of the largest publicly available repositories of online genealogies. Large-scale data screening for centenarians and shorter-lived controls in the Rootsweb database was conducted with assistance of the web-automation technique, which allows researchers to run automated queries and search online databases for individuals with desired properties (persons who lived 100+ years or 65 years in our case).

The data set on US centenarians and their shorter-lived controls was developed as an integral part of our study on exceptional longevity, which has identified 3 408 records of centenarians and 4 000 records of their shorter-lived peers (died at age 65) born in the United States between 1890 and 1897. The control selection strategy in this study was to avoid bias arising from noncomparability between cases and controls in centenarian studies. In order to minimize this bias, controls were selected to be a representative sample of the population, which produced the cases (ie, are selected from computerized family histories). Both cases and controls that had the same birth year window (1890–1897) were randomly sampled from the same population universe (computerized family histories available at Rootsweb.com) using identical sampling procedures. Only records from family histories of good quality with available information on exact (day, month, year) birth dates and death dates for study participants and their parents were used in the sampling procedure. This approach helped to eliminate a majority of incomplete family histories. The age at death for controls was selected assuming that the majority of deaths at age 65 occur due to chronic age-related diseases rather than injuries or infectious diseases (36). Age of centenarians was validated using the Social Security Administration Death Master File (82% success rate). All records have information on life span of parents (to control for familial longevity) and extensive information on life span of biological (parents, siblings, grandparents) and nonbiological (spouses, siblings-in-law) relatives. Verification of birth dates was accomplished through linkage to the 1900 US census when the person was a child (when age exaggeration is less common compared to claims of exceptional longevity made at old age). Linkage to the 1900 US census resulted in a 96.3% success rate for centenarians and a 96.4% success rate for controls. We found only a small disagreement (about 1.5%) between family histories and the 1900 census in birth year reports. For controls, 250 records were randomly selected per each sex and birth year producing 4 000 records. For centenarians, 220 records were randomly selected per each sex and birth year producing 3 408 records (for some male birth cohorts, there were less than 220 cases). Life span of parents was available for all records of centenarians and controls. More details on the database of US centenarians can be found elsewhere (36,37).

Final data set for this study contains records on 3 408 centenarians, 4 000 controls died at age 65 years, 19 188 siblings of centenarians, and 23 385 siblings of shorter-lived controls. Thus, both centenarians and controls had 6 siblings on average listed in genealogies. 69% of records for siblings of centenarians and 70% for siblings of 65-year-old controls had information about life span. Percent of records with known life span was higher for adults having spouses: 82% for siblings of centenarians and 81% for siblings of 65-year olds, suggesting that significant portion of cases with unknown life span belongs to siblings died during childhood. About 10 045 siblings of centenarians and 12 308 siblings of controls lived 50 years and more, and these data were used for analyses (4 801 records for siblings born after 1900 were removed from the analyses in order to avoid truncation bias). Median birth year was equal to 1890 for both siblings of centenarians and siblings of controls. An earlier study found that childhood and adulthood socioeconomic conditions did not differ significantly for centenarians and controls (37), so it is reasonable to assume that siblings of centenarians and controls had similar socioeconomic conditions during childhood and most likely during their adulthood. On the other hand, centenarians (and their siblings) had significantly higher proportion of long-lived parents (80+ years) compared to 65-year-old controls (37).

Definition of Long Life and Short Life

This study uses 2 definitions of long and short life. In the first step, siblings of longer-lived centenarians (100+ years old) are compared to siblings of shorter-lived persons who died at age 65 years. In the second step, children of long-lived parents are compared to children of shorter-lived parents within each group of siblings: (a) siblings of centenarians and (b) siblings of shorter-lived persons died at age 65 years. Longevity of parents was defined in the following way. If both parents lived 80 years or more then it was considered that the person had longer-lived parents. If both parents lived less than 80 years then the person had shorter-lived parents. Selected age of 80 years agrees with the previously established longevity threshold (17) and used in other studies of longevity as well (15).

Statistical Methods

Hazard rate was estimated using standard statistical package Stata (StataCorp LLC, College Station, TX). Procedure ltable calculates hazard rate for discrete data in the following way as presented in the Stata Manual (https://www.stata.com/manuals/stltable.pdf). Let f_j= d_j/n_j is within-interval failure rate (where d_j is a number of deaths within interval j and n_j is a number alive at the beginning of interval j). Then the maximum likelihood estimate of the hazard rate for interval j is:

$${\displaystyle \begin{array}{l}{\displaystyle {\mu }_j=\frac{1}{\mathrm{\Delta }x}\frac{f_j}{1-\frac{f_j}{2}}}\end{array}}$$ (1)

where ∆x is the length of age interval j.

Hazard rates were calculated using annual age intervals and expressed in yearly units (year⁻¹).

Hazard rates were fitted using the Gompertz model (33,38,39):

$${\displaystyle \begin{array}{l}{\displaystyle {\mathrm{\mu }}_x=R\mathrm{e}\mathrm{x}\mathrm{p}(\alpha x)}\end{array}}$$ (2)

where x corresponds to age and α and R are parameters.

Parameters of the Gompertz model were calculated in the age interval 50–90 years using weighted nonlinear regression (Stata nlin command). Age-specific exposure values (person-years) were used as weights (40).

Parameters of the Compensation Effect of Mortality (CEM or Strehler–Mildvan correlation when the Makeham parameter is close to zero) were calculated using linear regression model, assuming that the external age-independent component of mortality (the Makeham parameter) in the studied groups of siblings is negligible (33,41):

$${\displaystyle \begin{array}{l}{\displaystyle \mathrm{I}\mathrm{n}(R)=\mathrm{I}\mathrm{n}M-B\alpha }\end{array}}$$ (3)

where ln(R) is a natural logarithm of the intercept (level) parameter R and α is the slope parameter of the Gompertz model; lnM and B are parameters of linear regression. Parameter M is called the species-specific mortality (hazard rate) and parameter B—the species-specific life span (33).

Effects of maternal and paternal longevity on mortality of centenarians were analyzed using the Cox proportional hazards model adjusted for birth year, gender, and sibship size.

All calculations were conducted using Stata statistical software, release 14.

Results

Table 1 presents sample sizes and mean life span (with 95% confidence intervals [CIs]) for studied groups of siblings who survived to age 50 years. Note that siblings of centenarians born to longer-lived parents have the highest mean life span at age 50 years compared to other groups of siblings (within corresponding gender). Reciprocally, siblings of shorter-lived persons born to shorter-lived parents have the lowest mean life span at age 50 years (Table 1). Male siblings of male centenarians have slightly higher mean life span at age 50 years compared to male siblings of female centenarians. This phenomenon is not observed for female siblings of centenarians. Having a sibling centenarian has a very strong survival advantage, so that siblings of centenarians with shorter-lived parents still live longer than siblings of shorter-lived persons with longer-lived parents (Table 1). Table 1 presents data for mean life span. As for the median life span, it is 6 years higher for siblings of centenarians who survived to age 50 years than for siblings of shorter-lived controls for both genders (80 vs 74 years for men and 85 vs 79 years for women). Overall, siblings of centenarians have much better survival compared to siblings of shorter-lived persons who died at age 65 years (Supplementary Figures 1 and 2).

Table 1.

Mean Life Span (MLS) of Siblings Survived to Age 50, by Sex and Studied Group

Group of Siblings	Men		Women
	N	MLS (95% CI)	N	MLS (95% CI)
Siblings of centenarians	4 446	78.69 (78.35–79.02)	3 689	82.75 (82.37–83.14)
Born to short-lived parents	1 434	77.32 (76.73–77.91)	1 157	81.30 (80.59–82.01)
Born to long-lived parents	952	80.50 (79.79–81.22)	780	83.96 (83.13–84.80)
Siblings of male centenarians	2 189	79.37 (78.89–79.84)	1 755	82.67 (82.11–83.23)
Siblings of female centenarians	2 257	78.03 (77.57–78.50)	1 934	82.83 (82.30–83.36)
Siblings of short-lived persons (died at age 65 years)	5 009	73.16 (72.86–73.46)	4 408	77.65 (77.30–78.00)
Born to short-lived parents	2 609	72.26 (71.85–72.68)	2 245	75.98 (75.50–76.48)
Born to long-lived parents	488	76.38 (75.46–77.30)	417	80.31 (79.26–81.37)

Figure 1 shows mortality as a function of age in a semi-log scale for siblings of longer-lived (centenarians) and siblings of shorter-lived persons who died at age 65 years (separately for men and women). Note that the relative differences in mortality between siblings of longer-lived and siblings of shorter-lived persons are higher at younger ages compared to relative mortality differences at older ages. This phenomenon of mortality convergence at ages 95–100 years is observed for both men and women.

Figure 1.

Mortality (hazard rate) convergence for siblings of centenarians and siblings of shorter-lived individuals died at age 65 years: brothers (top) and sisters (bottom). Lines correspond to a linear fit of the common (base 10) logarithm of hazard rate as a function of age (corresponding parameters of the Gompertz model are given in Table 2).

Figure 2 compares mortality as a function of age for sons and daughters of long-lived and short-lived parents for each group of siblings (siblings of centenarians and siblings of shorter-lived controls died at age 65 years). For both groups, children of long-lived parents have lower mortality at younger ages compared to children of shorter-lived parents. However, by age 95–100 years, this difference in mortality practically disappears. Also, note that mortality difference between children of shorter-lived and longer-lived parents is much smaller for siblings of centenarians who already have beneficial genetic background.

Figure 2.

Mortality (hazard rate) convergence for siblings born to long-lived parents (both parents lived 80+ years, white circles) and to short-lived parents (both parents lived <80 years, black circles). Lines correspond to a linear fit of the common (base 10) logarithm of hazard rate as a function of age (corresponding parameters of the Gompertz model are given in Table 2).

The observed mortality convergence for different groups of siblings and children is a manifestation of the CEM when mortality for different populations is converging at one point (or one relatively small area). CEM can be quantified by estimating parameters of the Gompertz model. Table 2 presents estimates of the Gompertz parameters and their standard errors for the studied groups of siblings. If survival advantage for siblings of centenarians is life-long then we may expect similar slope parameters for siblings of longer-lived and shorter-lived persons. However, the slope parameters for siblings of centenarians and siblings of shorter-lived persons (died at age 65 years) are quite different and their CIs do not overlap. The Gompertz slope parameter for mortality of the sisters of centenarians is equal to 0.101 (95% CI: 0.096–0.107) and for mortality of the sisters of shorter-lived individuals died at age 65 years—0.085 (95% CI: 0.081–0.089). For brothers, these values are 0.090 (95% CI: 0.087–0.094) and 0.079 (95% CI: 0.075–0.083), respectively. For children of longer-lived and shorter-lived parents, this difference in the slope parameters is smaller, but still the slope parameter for shorter-lived group is lower as expected (Table 2).

Table 2.

Parameters of the Gompertz Model for Mortality of the Studied Groups of Siblings^*

Group of Siblings	Slope Parameter (α) ± Standard Error	Intercept Parameter (R) × 105 ± Standard Error
	Men
Siblings of centenarians	0.090 ± 0.002	4.98 ± 0.77
Born to short-lived parents	0.084 ± 0.003	9.08 ± 2.16
Born to long-lived parents	0.093 ± 0.004	3.41 ± 1.19
Siblings of short-lived persons (died at age 65 years)	0.079 ± 0.002	20.33 ± 3.24
Born to short-lived parents	0.079 ± 0.003	21.70 ± 5.29
Born to long-lived parents	0.079 ± 0.006	15.52 ± 7.33
	Women
Siblings of centenarians	0.101 ± 0.003	1.32 ± 0.31
Born to short-lived parents	0.084 ± 0.003	5.90 ± 1.71
Born to long-lived parents	0.135 ± 0.009	0.074 ± 0.056
Siblings of short-lived persons (died at age 65 years)	0.085 ± 0.002	8.18 ± 1.31
Born to short-lived parents	0.082 ± 0.002	11.04 ± 2.09
Born to long-lived parents	0.097 ± 0.008	2.64 ± 1.66

*Parameters of the Gompertz model were estimated in the age interval 50–90 years using weighted nonlinear regression with weights equal to age-specific exposure values.

As follows from Table 2, groups with higher life span have higher slope parameter (actuarial aging rate) of the Gompertz model and lower intercept parameter compared to groups with lower life span. Availability of the Gompertz parameters for 12 studied groups of siblings allows us to estimate parameters of linear regression between the natural logarithm of intercept parameter and the slope parameter of the Gompertz model (see eqn (3)). Run of linear regression on the Gompertz parameters presented in Table 2 resulted in the species-specific life span (B) equal to 96.06 ± 3.53 years, which is close to the species-specific life span obtained earlier for human populations: 95 ± 2 years (33). Thus, the observed parameter B for siblings with different levels of familial longevity agrees with earlier estimates of this CEM parameter (33).

Figures 1 and 2 suggest that mortality after age 100 years has a weak dependence on the level of familial longevity. To test this suggestion further, we analyzed effects of paternal and maternal longevity on survival of centenarians from the same database. Table 3 presents results of the Cox proportional hazards model describing mortality after age 100 years for 3 408 centenarians. This analysis demonstrates that gender is the only significant variable affecting survival after age 100 years. Both paternal longevity and maternal longevity (life span 80 years and more) had no significant effect on survival of centenarians. Effects of parental longevity on survival of centenarians remained nonsignificant when longevity threshold was elevated to 90 years (data not shown). These results confirm that protective effects of familial longevity are weak after age 100 years.

Table 3.

Parental Longevity Does Not Affect Survival After Age 100 Years

Variable	Hazard ratio (95% CI)	p
Paternal life span 80+ years	0.97 (0.91–1.04)	.405
Maternal life span 80+ years	0.95 (0.89–1.02)	.161
Female sex of centenarian	0.88 (0.83–0.95)	<.0001
Sibship size	1.00 (0.99–1.01)	.866
Birth year	1.01 (0.99–1.02)	.292

Note: Parameters of the Cox proportional hazards model for survival of 3 408 centenarians born in 1890–1897.

Discussion

We found that the survival advantage of biological relatives of long-living individuals vanishes at older ages, suggesting that protective effects of longevity assurance genes become weaker at age 95–100 years. We compared survival patterns of 10 045 siblings of US centenarians with survival of a control group of 12 308 siblings of shorter-lived individuals (died at age 65 years). Survival analysis after age 50 years was conducted separately for male and female siblings of centenarians and controls born in 1890–1897. Although siblings of longer-lived individuals have lower mortality at younger ages compared to siblings of shorter-lived individuals, their actuarial aging rate (rate of mortality growth with age) is consistently higher, so that their survival advantage practically disappears at older ages. It is possible that siblings of centenarians and siblings of controls may differ from each other not only by familial longevity, but also by other environmental factors including wealth and income. However, our earlier study of centenarians and shorter-lived controls born in 1890–1891 (a portion of data used here) demonstrated that familial longevity is the main factor of difference between these 2 groups. The only other difference between centenarians and shorter-lived controls was higher proportion of farmers among male centenarians while childhood environmental and economic conditions were not significantly different between centenarians and controls (37). It is interesting that the recent study also found no influence of environmental covariates on the association between parental longevity and offspring survival (22).

To validate these findings further, we analyzed data on survival of 3 408 US centenarians from the same database born in 1890–1897 with known information on maternal and paternal life span. We found that indeed both maternal and paternal longevity (life span 80+ years) have negligible protective effect on survival after age 100 years. Our findings are compatible with the predictions of the reliability theory of aging suggesting higher initial levels of system redundancy (reserves) in individuals with protective familial/genetic background (33,42,43). This phenomenon can also be explained by the heterogeneity hypothesis, although in this case mortality deceleration at advanced ages should be expected (however, not observed in this study) (34,35).

Earlier studies of mortality trajectories for siblings of centenarians and siblings of shorter-lived controls reported a life-long sustainable mortality advantage for siblings of centenarians (9,32). In our study, mortality advantage of siblings of centenarians gradually decreased with age and almost disappeared by age 100 years. We believe that the reason for this difference between our study and earlier results can be explained by relatively small samples of siblings used for mortality analysis in earlier studies (2 092 siblings of centenarians in the New England Study and 1 142 siblings of Okinawan centenarians). As a result, both age-specific mortality estimates and age-specific relative risks of mortality in these studies were subject to significant variations. High variation of estimates for the relative risk of mortality did not allow researchers to make definite conclusions about true age trends of this indicator. Authors of both articles admit that mortality curves for different human populations have a tendency to converge at old ages for such dimensions as gender, race, education, physical activity, and other factors (9,32). The authors came to a conclusion that the stability of relative risk of mortality over the wide age range may suggest that the advantage is attributable more to genetic rather than environmental factors (9,32). Our study with a larger sample size of siblings demonstrates that familial longevity affects mortality in the same manner as do other major factors of mortality (gender, race, education, and others) and is compatible with the CEM (33).

Most researchers believe that the effects of familial longevity on survival of relatives have a significant genetic component (27,44,45). In our study, longer life of the siblings of centenarians compared to the siblings of shorter-lived persons who died at age 65 years is likely caused by genetic factors as well, because environmental factors (socioeconomic conditions of childhood and most likely of adulthood) do not significantly differ between cases and controls (37). Lack of significant effect of parental longevity on survival after age 100 years suggests that genes responsible for longevity advantage of centenarians operate earlier in life rather than later in life. Reliability theory of aging explains this phenomenon suggesting that organisms of longer-lived individuals have more reserves of functional cells and other biological components (33,42,43). At advanced ages, these initial reserves become exhausted and individuals with favorable familial background lose their advantage. This general conclusion comes from the inspection of shapes of mortality trajectories for longer-lived and shorter-lived individuals. Indeed, studies of centenarians found that centenarian genomes are depleted of SNPs associated with age-related diseases and enriched with protective SNPs (31,46,47). The observed mortality convergence can also be explained by heterogeneity hypothesis when siblings of shorter-lived persons have higher heterogeneity and more deleterious mutations compared to siblings of centenarians. Over time, persons with deleterious mutations die out leaving healthier individuals and more homogeneous sample. In this case, we may expect mortality deceleration at advanced ages, although in our sample this phenomenon is not readily visible (16,34,35).

Our findings are in line with previous studies that analyzed effects of specific longevity genes on mortality. It was found that carrying the ε2ε2 or ε2ε3 APOE genotypes is associated with significantly reduced risk of death compared with carrying the genotype ε3ε3. However, reduction of risk of death was modest beyond an age reached by less than 1% of the population corresponding to 96 years for men and 100 years for women (24). Somewhat more controversial results were obtained when comparing ε3ε4 and ε4ε4 APOE genotypes with ε3ε3 genotypes. Ewbank (48) reported mortality convergence at older ages for carriers of ε4 allele and carriers of ε3ε3 genotype in Finnish and Swedish cohorts. On the other hand, the study of long-lived Danes found an increasing influence of APOE ε4 allele on mortality at advanced ages (49). In another study, researchers analyzed FOXO3 11 known SNPs associated with human longevity using survival analysis. These analyses confirmed previous associations of common variants of FOXO3 with older age, but these common variants did not modify risk for mortality at ages beyond the oldest 1% age of survival (29). Thus, most studies demonstrate that protection of APOE and FOXO3 longevity genotypes is weaker at advanced ages.

We need to mention some weaknesses of this study. The sample of centenarians and controls represents predominantly White population, so the results of this study may suffer from the lack of generalizability. Comparison of records for centenarians and controls with the IPUMS 1% 1900 US Census sample revealed that computerized genealogies have more data on families living in rural settings compared to the general population. However, no differences were found between centenarians and controls in the socioeconomic characteristics described in the 1900 census suggesting that there was no bias related to longevity. Also, life-span information is available not for all records of siblings. If participation into the cohort is associated with better survival and if this “healthy participant” survival advantage differs between siblings of centenarians and siblings of 65-year olds, respectively, then this fact could influence the results. If this hypothesis is correct, then we may expect that longer-lived individuals are more represented in genealogies and hence there would be more siblings of centenarians compared to siblings of controls in the sample. This hypothesis was not confirmed, because both centenarians and shorter-lived controls have an equal number of siblings (6) on average. If long-lived individuals are more likely to be included in the sample, then we may expect lower number of records with unknown life span among siblings of centenarians who have higher proportion of long-lived individuals compared to controls. However, the proportion of records with unknown life span is almost identical for both groups (31% for siblings of centenarians and 30% for siblings of controls; 18% and 19% for adult siblings of centenarians and controls, respectively), suggesting that there is no specific selection into cohort based on longevity.

Rather weak effects of familial longevity on mortality at later ages may be a challenge for those studies that look for longevity assurance genes in centenarians and other longer-lived individuals. It turns out that mortality advantage of longer-lived individuals associated with better genetic background is observed mostly at younger ages. This observation does not mean that there are no genes that may operate at very old age. For example, Sebastiani et al. (50) identified genetic signatures correlated with better survival of centenarians after age 100 years. Thus, looking for longevity genes that operate at later years may be a promising approach in potential life extension. Of particular interest are genes that are able to decrease the actuarial aging rate (Gompertz slope parameter) without increasing the initial mortality level (Gompertz intercept parameter). Such genes would break the rules of the CEM and bring more hopes for significant extension of human life.

Conclusion

We analyzed effects of familial longevity on human mortality using data on 2 types of biological relatives: siblings and parents. In both cases, longevity of relatives resulted in lower mortality and higher life expectancy after age 50 years (when deaths are caused predominantly by aging-related diseases). The highest differences in life expectancy were found between siblings of centenarians and siblings of shorter-lived individuals (who died at age 65 years). Individuals having longer-lived relatives (centenarian siblings or parents living 80 years or more) have lower mortality at younger ages while this mortality advantage significantly decreases by age 100 years. This observation is a challenge for a search of longevity assurance genes and means that genes more common in centenarians are able to ensure survival to ages 90–100 years while mortality reduction at extreme old ages may require different approaches.

Funding

This study was supported in part by the grant from the National Institute on Aging of the National Institutes of Health (R01 AG028620 to L.A.G.).

Conflict of Interest

None declared.

Author Contributions

N.S.G. designed the study, conducted statistical analyses, and prepared the manuscript. L.A.G. analyzed and interpreted results, and edited the manuscript.