Familial Loss of a Loved One and Biological Aging

This cohort study evaluates associations between losing a loved one and accelerated biological aging.


Findings
Meaning These findings suggest that loss can accelerate biological aging even before midlife and that frequency of losses may compound this, potentially leading to earlier chronic diseases and mortality.

Introduction
The relationship between bereavement and health across the life course is well-established and enduring. 1However, there are indications that certain life stages may be more susceptible to the health and mortality risks associated with loss.7][8] Repeated family losses over one's life further compound these health risks. 6e impacts of loss may persist or become apparent long after the event. 5,6,8,9e mechanisms connecting loss to poor health and mortality remain unclear.There are at least 2 pathways, each with intermediate processes.One pathway begins with deficits in material and social resources formerly provided by the person who dies.Another begins with the psychological distress of bereavement.Both can ultimately result in dysregulation of biological systems due to changes in health behaviors related to bereavement-related distress and stress from economic constraints.One framework that integrates the impacts of these diverse pathways and intermediate processes on human health is emerging research on biological aging. 10ological aging refers to the progressive loss of integrity and resilience capacity in our cells, tissues, and organs as we grow older. 11While there are no universally accepted methods to measure biological aging in humans, the best-validated of these methods are a family of DNA-methylation algorithms known as epigenetic clocks, including several that show robust evidence for prediction of future morbidity and mortality. 10,12,13However, few studies have empirically explored the association between loss across developmental periods and DNA methylation markers of aging, and rarely have these investigations been conducted in racially and ethnically diverse population-based studies.
To address existing gaps, we investigated if familial loss and quantity of loss, occurring either in early life or adulthood, were associated with biological aging in a diverse, population-based sample.
Acknowledging prior research on racial disparities in loss experiences and variability in exposure histories, 6,14 we also examined the interaction between loss and race on biological aging.We analyzed survey responses from waves 1 to 5 of the National Longitudinal Study of Adolescent to Adult Health (Add Health), coupled with new epigenetic data from wave 5.

Methods
We used data from wave 1 to wave 5 of Add Health, which has been previously described 15 and is a US nationally representative cohort, following up participants since the 1994 to 1995 school year.Wave 5 took place between 2016 and 2018 and completed interviews with 12 300 participants.During wave 5, participants were invited for an additional home examination where a venous blood sample was gathered.Of the 7995 individuals who agreed to the home examination, 5381 were successfully visited, and 4940 provided a blood sample.We used data from 3963 participants who had a blood sample, reported losses at each wave, and reported data on covariates.We used weighting to ensure representation of the Add Health cohort since wave 1. 15 The weights account for the original sampling design, attrition to the current wave 5, and differential consent to the in-home blood collection. 16rticipants and their parents or caregiver in childhood provided written consent at waves 1 through 2; at age 18 years, only the participant's consent was obtained (waves 3-5).The study was approved by the institutional review boards of Columbia University and the University of North Carolina, Chapel Hill.We followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines. 17

Epigenetic Clocks
From 2018 to 2024, we conducted epigenome-wide profiling and construction of clocks, using whole-blood DNA for 4700 participants at wave 5 using the Infinium MethylationEPIC BeadChip array (Illumina, Inc), as previously described. 18After standard quality control, we calculated 4 biological clocks, including GrimAge, PhenoAge, and Horvath, to assess epigenetic age acceleration, [19][20][21][22] and the DunedinPACE to assess pace of biological aging. 23,24For analysis, Horvath, GrimAge, and PhenoAge were first residualized on chronological age and residuals were z-transformed to allow comparability across measures.DunedinPACE is a measure of the pace of aging-how rapidly or slowly a person is aging. 24It is a rate, not an age, and so no residualization was conducted, but values were z-transformed for comparability.

Familial Loss
Familial loss and the timing of loss variables were derived from waves 1 to 5. We included the following from each wave: (1) parental death (biological and parent figures), (2) death of a sibling, ( death of a spouse or partner, and (4) death of a child.We calculated the total number of losses by pooling across 5 waves, using only new reports of loss.Number of losses were coded as 0, 1, or 2 or more losses.

Parental Loss Across Waves 1 to 5
Given research suggesting a strong association between parental loss, in particular, and the health of surviving children, [25][26][27] in addition to the fact that it was the most common type of loss in Add Health, we assessed parental loss at any time point across waves 1 to 5. The parental loss variable included the death of parents and/or parent figures at any wave.

Developmental Period of Loss
To classify loss at varying developmental periods, we used survey reports of parental and sibling loss reported on surveys from wave 1 to 2 by identifying whether the participant was aged less than 18 years at the time of the loss.To classify loss among participants aged less than 18 years, we used the month and year of parental and sibling deaths reported at wave 3 or 4 or as reported in the last 12 months on wave 5. We also used reports of death of a partner or spouse and child from waves 3 to 5.
We developed variables for any loss in childhood or adolescence and any loss in adulthood based on reported loss timing.Any loss in childhood or adolescence includes deaths of parents (biological and parental figures) and siblings before age 18 years.Any loss in adulthood encompasses deaths of parents, siblings, spouses or partners, and children when the participant was 18 or older.
Additionally, we created variables for parental loss in childhood or adolescence (death of a parent or parental figure before age 18 years) and parental loss in adulthood (death of a parent or parental figure at age 18 years or older).

Covariates
9][30] Chronological age at blood draw was included as a continuous variable.Participant self-identified race and ethnicity was based on the race or ethnicity that the participant indicated that they most strongly identified with at the wave 5 survey, including Asian, Black, Hispanic, American Indian, other, Pacific Islander, or White.Participants that responded as other did not identify with any race or ethnic categories or identified as multiracial only.In a small proportion of cases (9 participants) where wave 5 information was missing or unavailable, wave 1 was used to identify self-reported race and ethnicity.We present loss for all racial and ethnic groups where sample sizes permit.In our regression models, we categorized race as Black, Hispanic, and a combined category for all other racial groups due to limited sample sizes.Respondent's report on sex assigned at birth at wave 5 was used to categorize participants as female or male.To account for environmental or neighborhood influences on epigenetic aging in early life, we included a variable that represented the proportion of households with an income below the poverty line in 1989 based on participant addresses at wave 1.We also adjusted for the number of household members in wave 1, to account for the increased likelihood of loss among those with greater members.We also

JAMA Network Open | Genetics and Genomics
Familial Loss of a Loved One and Biological Aging included parental educational attainment based on the highest attainment of either parent at wave 1 (high school degree or less, some college, and college or higher).Smoking was based on parent's self-report of smoking at wave 1.We adjusted for epigenetic assay batch (batch 1, 41.34%; batch 2, 56.51%; batch 3, 1.15%).We also conducted sensitivity analyses by adjusting for cell counts as described in eTable 1, eTable 2, eTable 3, and eTable 4, and for duration since loss in eTable 5, eTable 6, eTable 7, and eTable 8 in Supplement 1.

Statistical Analysis
We used survey linear regression models for assessing the associations between each number of losses and each of the biological aging measures, adjusting for all covariates described previously.
[37][38][39] In models investigating timing of loss (any loss or parental loss), we included a variable indicating loss during childhood or adolescence and a variable indicating loss during adulthood so that we could assess the association of childhood or adolescence and adulthood loss independently of one another.Following Add Health guidelines, 16,38 analyses accounted for complex sampling features (ie, probability weights, nesting of individuals in primary sampling units, no siblings, and stratification) so that results represent the population of individuals who were enrolled in grades 7 to 12 in the US in 1994 to 1995. 16We used 95% CIs to evaluate statistical significance.
5][46][47][48] Furthermore, clocks, mainly developed in predominantly White samples, show varying sensitivities to social and environmental exposures across different races. 49Thus, we presented exposure to loss by race and included an interaction term between loss and race in our biological aging models to explore effect modification by race. 44,47,48Interaction analyses focused on Black (and White participants due to the smaller number of other racial and ethnic categories) (see Table ).We applied the same models as in the overall sample, adding an interaction term between loss and race, considering interaction terms significant at a 2-sided P < .10.Data were analyzed from Janaury 2022 to July 2024.

Results
The analytical sample included 3963 Add Health participants who had data on all variables in the analysis (

Number of Losses
The mean z score and SEs for each biological clock by number of losses are presented in Figure 1.As the number of losses increased, GrimAge, PhenoAge, and DunedinPACE increased.There were no discernible differences in the Horvath clock by number of losses (β = −0.08;95% CI, −0.23 to 0.06).
In adjusted models, participants who experienced only 1 loss vs no loss had older biological ages for GrimAge (β = 0.

JAMA Network Open | Genetics and Genomics
Familial Loss of a Loved One and Biological Aging

Timing of Loss
Participants who experienced any losses during childhood or adolescence (adjusting for later life loss)   We found that adult losses showed greater associations with biological aging markers than childhood or adolescence losses, a pattern also seen in a smaller sample of older adults in the Ireland study. 51This may be due to stronger adult reactions to traumatic events, impacting physiological risk, or greater epigenetic recovery from losses that occurred in the more distant past.Understanding the distinct impacts of loss at different life stages requires studies with repeated measures of loss and DNA methylation over time.
1][42][43] Consistent with recent work by Donnelly et al, 14 we noted that Hispanic participants experienced a higher number of losses than White non-Hispanic participants.American Indian and Pacific Islanders also experienced greater loss, but the sample size for these estimates were small.Our interaction analysis suggests that there were no significant differences between loss and epigenetic aging across Black and White participants.A previous study noted higher instances and impacts of loss on GrimAge among Black participants. 54 were unable to stably test for effect modification across the smaller samples of racial and ethnic groups.Future research should examine these associations among American Indian, Hispanic, and Pacific Islanders, as well as individuals of multiracial and other identities.

Limitations
We acknowledge limitations.We had access to only 1 time point of epigenetic data, leaving open questions as to why childhood or adolescent losses exhibited lesser associations with epigenetic clock measures as compared with losses occurring during adulthood, closer to the time of DNA collection.It is also possible that individuals with parents who die prematurely may inherit familial health conditions that impact both parental loss and biological aging.However, we adjusted for time since loss and did not observe significant differences in model estimates.This study may not have had sufficient power to detect minor interaction effects by race.Given the significant differences in exposure to loss across racial and ethnic groups, future research will require larger and more diverse population samples.The consistency of our results across 3 different epigenetic clocks and multiple model specifications in a large, diverse, prospective, national cohort contribute important evidence that experiences of loss represent a marker of risk for accelerated biological aging.Future research should aim to replicate our innovative findings related to the quantity of loss and its variability by type.

Conclusions
In conclusion, our study sheds light on how loss may affect biological aging and ultimately health and mortality.We found that adults with a history of loss had higher biological ages than those without such experiences.More losses were associated with older biological age.These findings suggest that loss can accelerate biological aging even before midlife and frequency of losses may compound this, potentially leading to earlier chronic diseases and mortality.Future research should focus on identifying coping strategies and social support to lessen the negative effects of loss on aging, aiding clinical and public health approaches.
In a cohort study of 3963 participants from the National Longitudinal Study of Adolescent to Adult Health, nearly 40% experienced the loss of a close relation by adulthood.Participants who had experienced a greater number of losses exhibited significantly older biological ages compared with those who had not experienced such losses.

Figure
Figure 2. Number of Losses and Biological Aging and Parental Loss at Any Time Period (Childhood/Adolescence to Adulthood) and Biological Aging

Table .
Population Descriptive Statistics (Weighted Proportions and Unweighted Ns) (continued) a Estimate based on small sample size (eg, <30).bParticipantswhoself-identified as other did not identify with any other race or ethnic category or indicated multiracial only.cEstimatecannot be presented because of deductive disclosure risks based on Add Health restricted data use policy.
JAMA Network Open.2024;7(7):e2421869.doi:10.1001/jamanetworkopen.2024.21869(Reprinted) July 29, 2024 6/13 Downloaded from jamanetwork.comby guest on 08/04/2024 β = 0.22; 2. Number of Losses and Biological Aging and Parental Loss at Any Time Period (Childhood/Adolescence to Adulthood) and Biological Aging Figure 3. Any Loss in Either Childhood/Adolescence or in Adulthood and Biological Aging and Parental Loss in Either Childhood/Adolescence or in Adulthood and Biological Aging This is an open access article distributed under the terms of the CC-BY License.© 2024 Aiello AE et al.JAMA Network Open.Graf GH, Crowe CL, Kothari M, et al.Testing Black-White disparities in biological aging among older adults in the United States: analysis of DNA-methylation and blood-chemistry methods.Am J Epidemiol.2022;191(4): 613-625.doi:10.1093/aje/kwab28154.Holloway TD, Harvanek ZM, Xu K, Gordon DM, Sinha R. Greater stress and trauma mediate race-related differences in epigenetic age between Black and White young adults in a community sample.Neurobiol Stress.2023;26:100557.doi:10.1016/j.ynstr.2023.100557Number of Losses and Biological Aging eTable 2. Parental Loss at Any Time Period (Childhood to Adulthood) and Biological Aging eTable 3. Any Loss in Childhood and in Adulthood and Biological Aging eTable 4. Parental Loss in Childhood and in Adulthood and Biological Aging eTable 5. Count of Losses and Biological Aging eTable 6. Parental Loss at Any Time Period (Childhood to Adulthood) and Biological Aging eTable 7. Any Loss in Childhood and in Adulthood and Biological Aging eTable 8. Parental Loss in Childhood and in Adulthood and Biological Aging eTable 9. Number of Losses and Biological Aging eTable 10.Parental Loss at Any Time Period (Childhood to Adulthood) and Biological Aging eTable 11.Any Loss in Childhood and in Adulthood and Biological Aging eTable 12. Parental Loss in Childhood and in Adulthood and Biological Aging eTable 13.Number of Losses and Biological Aging eTable 14.Parental Loss at Any Time Period (Childhood to Adulthood) and Biological Aging eTable 15.Any Loss in Childhood and in Adulthood and Biological Aging eTable 16.Parental Loss in Childhood and in Adulthood and Biological Aging eTable 17.Parental Loss by Gender of Parent at Any Time Period (Childhood to Adulthood) and Biological Aging eTable 18. Interaction Between Any Loss in Childhood and in Adulthood and its Association With Biological eTable 19.Interaction Between Parental Loss in Childhood and in Adulthood and its Association With Biological Aging Corresponding Author: Allison E. Aiello, PhD, MS, Department of Epidemiology, Robert N Butler Columbia Aging Center, Mailman School of Public Health, Columbia University, 122 W 168th St, New York, NY 10032 (aea27@cumc.columbia.edu).53.