Vienna Yearbook of Population Research 2013 (Vol. 11), pp. 295–323

Determinants of exceptional human longevity: new ideas and findings
Leonid A. Gavrilov and Natalia S. Gavrilova∗

Abstract
Studies of centenarians are useful in identifying factors leading to long life and avoidance of fatal diseases. In this article we consider several approaches to study effects of early-life and midlife conditions on survival to advanced ages: use of non-biological relatives as controls, the within-family analysis, as well as a sampling of controls from the same population universe as centenarians. These approaches are illustrated using data on American centenarians, their relatives and unrelated shorterlived controls obtained from the online genealogies. The within-family analysis revealed that young maternal age at person’s birth is associated with higher chances of exceptional longevity. Comparison of centenarians and their shorter-lived peers (died at age 65 and sampled from the same pool of online genealogies) confirmed that birth timing in the second half of the calendar year predicts survival to age 100. Parental longevity as well as some childhood and midlife characteristics also proved to be significant predictors of exceptional longevity.

1 Introduction
Studies of centenarians (people living to 100 and older) could be useful in identifying factors leading to long life and avoidance of fatal diseases. Even if some individual characteristics have a moderate protective effect on the risk of death, people with this trait/condition should be accumulated among long-lived individuals, because of cumulative survival advantage. Thus, studying centenarians may be a sensitive way to find genetic, familial, environmental and life-course factors associated with lower mortality and better survival. Most studies of centenarians in the United States are focused on either genetic (Hadley et al. 2000; Murabito et al. 2012; Perls and Terry 2003; Sebastiani et al. 2012; Zeng et al. 2010) or psychological (Adkins et al. 1996; Hagberg et al. 2001; Margrett et al. 2010; Martin et al. 2010; Murabito et al. 2012) aspects of survival
∗

Leonid A. Gavrilov (correspondence author), Center on Aging, NORC at the University of Chicago, USA. Email: gavrilov@longevity-science.org Natalia S. Gavrilova, Center on Aging, NORC at the University of Chicago, USA.

DOI: 10.1553/populationyearbook2013s295

296

Determinants of exceptional human longevity: new ideas and findings

to advanced ages. On the other hand, several studies suggest that early-life events and conditions may have significant long-lasting effects on survival to advanced ages (Barker 1998; Costa and Lahey 2005; Elo and Preston 1992; Finch and Crimmins 2004; Fogel and Costa 1997; Gavrilov and Gavrilova 2001a; Gavrilov and Gavrilova 2003a; Gavrilov and Gavrilova 2003b; Gavrilov and Gavrilova 2006; Hayward and Gorman 2004; Kuh and Ben-Shlomo 1997; Smith et al. 2009b). Studies of centenarians can provide important contributions to this area of research. However, they require serious work on age validation (Jeune and Vaupel 1999; Poulain 2010; Poulain 2011) and careful design including the choice of an appropriate control group. Living to age 100 is a rare event, and it is very difficult to obtain a representative population sample of centenarians for a large country as the United States. Nevertheless, use of general population as a control group is one of the most popular approaches in centenarian studies (Jarry et al. 2012a; Montesanto et al. 2011; Perls et al. 2007; Willcox et al. 2006). We used this approach in our prior study of early-life conditions and longevity (Gavrilova and Gavrilov 2007). In that investigation we assumed that persons in computerised family histories did not differ from the general population and found some agreement with other studies (Hill et al. 2000; Preston et al. 1998). However, there is still a possibility of obtaining biased results if this assumption is not true. For this reason, better research approaches to centenarian studies should be developed as they are described below. In this article we consider several approaches to study effects of early-life conditions on survival to advanced ages which use more appropriate control groups compared to the population-based control. These approaches include: (1) use of nonbiological relatives, such as siblings-in-law, as a control group, (2) the within-family analysis and (3) the between-family analysis with selection of controls from the same population universe. These approaches are used in the study of early-life and midlife predictors of exceptional longevity in the United States. For our studies we use family histories (genealogies) which proved to be a good source of information for different types of relatives and have successfully been used in historical demography (Adams and Kasakoff 1984; Anderton et al. 1987; Bean et al. 1992) and biodemography (Caselli et al. 2006; Gavrilov et al. 2002; Gavrilova et al. 1998; Smith et al. 2009a; Smith et al. 2009b).

2 Comparative study of biological and non-biological relatives
Numerous studies found that biological relatives of long-lived individuals have substantial survival advantage compared to relatives of shorter-lived individuals (Gavrilov and Gavrilova 2001b; Gavrilov et al. 2002; Kerber et al. 2001; Pearl and Pearl 1934; Perls et al. 2007; Smith et al. 2009a; Willcox et al. 2006), while relatively few studies have analysed the lifespan of non-biological relatives (Jarry et al. 2012a; Mazan and Gagnon 2007; Montesanto et al. 2011; Schoenmaker et al. 2006;

Leonid A. Gavrilov and Natalia S. Gavrilova

297

Table 1: Number of centenarians and their siblings at different stages of data collection and cleaning

Type of records All initial non-duplicate records for centenarians born in 1880–1895 with names of parents available Centenarians having detailed information on birth and death dates of their parents Centenarians having detailed information on birth and death dates of their parents and siblings Centenarians with confirmed death dates through the linkage to US Social Security Death Master File, DMF, having detailed information on birth and death dates of their parents and siblings

Number of centenarians Males Females Total 7,174 18,277 25,451

Number of siblings

6,370

16,757

23,127

172,091

707

2,127

2,834

21,893

398

1,313

1,711

13,295

Westendorp et al. 2009). Besides being an interesting population sub-group for the study of longevity, non-biological relatives may serve as a nonbiased environmental control group for centenarians in contrast to the general population. In general, studies of biological and non-biological relatives deal with relatively small case numbers. In our study we intended to create a large sample of centenarians and their siblings. Therefore, we conducted a large-scale search in many hundreds of online family histories using a technique known as ‘web automation’ (Sklar and Trachtenberg 2002). This technique allowed us to search online databases for people with exceptional longevity (and some other traits). Table 1 shows the steps of data collection and the number of records obtained in each stage of data collection. In a first step we scanned more than 300,000 publicly available online databases in the Rootsweb WorldConnect project (http://wc.rootsweb.ancestry.com). This search led to the identification of 25,451 non-duplicate records of alleged centenarians born between 1880 and 1895 with available information about names of their parents. The next step was to collect detailed data on the parents of the centenarians from computerised genealogies using the same web-automation technique. After this

298

Determinants of exceptional human longevity: new ideas and findings

procedure, we selected the most detailed genealogies where information on birth and death dates for both parents was available. Our prior experience in working with computerised genealogies suggests that this procedure selects-out the majority of genealogies with poor quality. As a result of this procedure, the total number of centenarian records decreased to 23,127. In the next step, we collected data on 172,091 siblings for those centenarians who had detailed data on parental birth and death dates. However, a significant proportion of these records did not contain information about the death dates, which created some difficulties for our withinfamily study of human longevity (described in the next section). Hence, the next step was to identify the most detailed family histories of families with complete information on birth and death dates for siblings. As a result of this identification procedure, we found 2,834 families for which information on birth and death dates was known for more than 80 per cent of siblings. Finally, the age at death of centenarians in our sample was verified using the US Social Security Administration Death Master File (DMF). This procedure confirmed the age for 1,711 centenarians born in the years 1880 to 1895.1 The final database of centenarians and their relatives contains information on lifespan for 398 male and 1,313 female centenarians, their 13,295 siblings, 1,307 spouses and 7,924 siblings-in-law. Lifespan information was collected for all siblings of centenarians who did not die in childhood. Completeness of lifespan information was 76% for spouses of centenarians and 82% for siblings-in-law. These data formed the basis of our comparative analysis of longevity using different types of biological and non-biological relatives. Although the studied sample of centenarians is not representative, it is useful for analytical studies of early-life conditions and longevity when appropriate control groups are available (Woodward 2005).2 Comparison of the mean lifespan for relatives who survived to age 50 reveals a survival advantage of brothers and sisters of centenarians compared to the same-sex siblings-in-law. As shown in Table 2, brothers lived 2.6 years longer and sisters lived even 2.9 years longer on average compared to siblings-in-law of the same sex, with the differences in lifespan being statistically significant (p< 0.001). Wives of centenarians tend to live 0.8 year less on average than married sisters of centenarians; however this difference is not statistically significant. Husbands of centenarians live 2.3 years less on average than married brothers of centenarians
1

In a previous study we found that the ages of centenarians confirmed by linking to the DMF are confirmed through the linkage to early US censuses in more than 98 per cent of cases (Gavrilova and Gavrilov 2007). Thus, there is not a strong need to carry out any additional verification with the early census records. 2 For instance, the sibship size of the centenarian families was larger than that of the general population. This can be explained by the fact that genealogies are more likely to be compiled for larger families and that longer-lived individuals in the United States were born more often in rural areas with higher fertility (Gavrilova and Gavrilov 2007; Preston et al. 1998). However, it should be noted that this difference in sibship size with the general population is not critical for the within-family design of the study when the shorter-lived siblings raised in the same family or spouses are used as control group.

Leonid A. Gavrilov and Natalia S. Gavrilova

299

Table 2: Mean lifespan conditional on survival to age 50 (LS50) with 95% confidence intervals (95% CI) for relatives of centenarians (compared to the 1900 US birth cohort)

Male relatives Relatives of centenarians Parents All siblings Married siblings Spouses Siblings-in-law 1900 US birth cohort Sample size, N 1590 5324 3221 876 2349 LS50 (95% CI), years 76.2 (75.7–76.8) 77.6 (77.3–77.9) 77.7 (77.3–78.1) 75.4 (74.6–76.1) 75.0 (74.6–75.5) 73.3

Female relatives Sample size, N 1557 4877 3028 283 2407 LS50 (95% CI), years 77.2 (76.7–77.8) 82.4 (82.0–82.7) 82.2 (81.8–82.6) 81.4 (80.1–82.7) 79.5 (79.0–79.9) 79.4

(p < 0.001). Although fathers of centenarians were born about 30 years earlier than brothers-in-law of centenarians they still have a longer lifespan conditional on survival to age 50 than later-born non-biological relatives such as siblings-in-law (p < 0.001) and husbands of centenarians (p = 0.04). On the other hand, mothers of centenarians who survived to age 50 have the shortest lifespan among all relatives— 77.2 years on average. This is, however, most likely related to the physiological burden of childbearing in these large families, as mothers of many children have been shown to have increased mortality (Dior et al. 2013; Gagnon et al. 2009; Riley 2003). Thus, excluding the mothers of centenarians we can conclude that siblings-in-law have a shorter lifespan than biological relatives and spouses born in a similar time period. This result probably reflects the fact that siblings-in-law are lacking certain genetic or environmental advantages that other relatives have. At the same time, the lifespan of siblings-in-law is still higher than the mean lifespan of the general population, i.e. the 1900 US birth cohort (Bell et al. 1992). This difference is particularly high for men—1.7 years (p < 0.001) while for women it is not statistically significant (Table 2). This finding is particularly important because it indicates that comparing the survival of siblings or other biological relatives of centenarians to the general population may overstate their survival advantage and hence overestimate the genetic contribution to lifespan. Thus, the using the general population to compare survival of relatives of long-lived individuals may be inappropriate, especially for males. Therefore siblings-in-law appear to be a more appropriate control group than the general population. Although the positive association of a person’s longevity with better survival of biological relatives is well documented, little is known about the effects of a centenarian’s gender on longevity of biological and non-biological relatives. Table 3 shows that fathers of male centenarians lived significantly longer than fathers of female centenarians: the mean lifespans conditional on survival to age

300

Determinants of exceptional human longevity: new ideas and findings

Table 3: Mean lifespan conditional on survival to age 50 (LS50) for biological and non-biological relatives of centenarians, by gender of centenarian

Relatives of male centenarians Type of relative Parents Fathers Mothers Siblings Brothers Sisters Siblings-in-law Brothers-in-law Sisters-in-law
† p-values

Relatives of female centenarians Sample size N 1216 1195 4056 3806 1857 1796 LS50, years 75.93 77.03 77.09 82.45 74.55 79.55 p-value† 0.023 0.087