The death of a close relative has been linked to a variety of adverse health consequences among bereaved family members. An excess risk of morbidity and mortality was observed among individuals who experienced the loss of a spouse (Prior et al., 2018; Brenn and Ytterstad, 2016), parent (Bylund Grenklo et al., 2013; Rostila et al., 2016), or child (Hendrickson, 2009; Li et al., 2003). Nevertheless, although the sibling tie represents one of the most enduring kin relations across the life course (Bedford, 2012), sibling loss has received relatively scant attention in the scientific literature. A recent Danish population-based study compared the relative weight of the different types of bereavement (loss of child, spouse, sibling, and parents) with regard to the subsequent risk of suicide, deliberate self-harm and psychiatric illness (Guldin et al., 2017). The results revealed that, although loss of a sibling led to less obvious short-term rate elevations (compared to the loss of a child or spouse), sibling bereavement was associated with the greatest long-term risk elevations of the studied adverse psychiatric consequences.

By virtue of being born at the same time, both monozygotic and dizygotic twins tend to have more in common (e.g., early life experiences) and report stronger emotional bond than other sibling-pairs (Bank and Kahn, 1997; Neyer, 2002). Moreover, the 100% genetic share between monozygotic twins may contribute to even closer emotional ties between them than dizygotic twins (Neyer, 2002; Cassell, 2011; Fortuna et al., 2010). Consequently, losing a co-twin by death may be a particularly devastating life stressor with considerable health implications for the surviving twins (Segal and Bouchard, 1993). Nevertheless, scientific studies on this topic are relatively rare. With regard to psychiatric reactions to a co-twin death, only a handful of descriptive studies with small sample sizes have been conducted, mainly using self-reported measures of grief (i.e. sorrow or general sadness) as an outcome (Segal et al., 1995; Segal and Ream, 1998; Rosendahl and Björklund, 2013). Evidence on a potential rise in rates of clinically diagnosed psychiatric disorders associated with the loss of a co-twin is, however, totally lacking. Moreover, given the shared genetic background of twins, the psychiatric consequences after a co-twin loss need to be estimated with caution (Segal, 2009; Pompili et al., 2006), ideally using family comparison with additional vigorous control of previous psychiatric morbidities and the causes of the co-twin’s death.

Due to the sharp rise of medically assisted reproduction and delayed childbearing, the twinning rate, particularly the rate for dizygotic twins, has increased dramatically in all developed countries since 1970 s (Pison et al., 2015) which calls for a vigorous documentation of the potential health consequences suffered by surviving twins after a co-twin loss. Therefore, taking advantage of nationwide health registers in Sweden and the unique Swedish Twin Registry, we conducted a nationwide population- and sibling-matched cohort study to estimate the short- and long-term association between loss of a co-twin and subsequent risk of first-recorded psychiatric disorders.

Here, we included all Swedish-born twins that lost a co-twin by death between 1973 and 2013 (exposed twins), their full siblings, together with a group of age- and sex-matched unexposed twins who were randomly selected from the twin population and did not experience such a loss. While comparing twins exposed to a co-twin loss to twins without such an experience allowed control for pregnancy-, birth- and some psychosocial aspects of being a twin (the matched twin cohort), the comparison of exposed twins with their full siblings within the same family enabled a direct comparison between two types of stressors, that is loss of a co-twin and loss of a full sibling, while controlling for familial factors (the twin-sibling family cohort). See details in ‘Materials and methods’, and Figure 1 and Supplementary file 1-Table 1.

Figure 1

Download asset Open asset

In total, the matched twin cohort (i.e., 4528 exposed twins and 22,640 matched unexposed twins) accrued 263,215 person-years, with an approximately 8 year average duration of follow-up. The median age at co-twin loss was 59 years and around 50% of the exposed twins were males (Table 1). While there were only small differences in family income, educational level, and marital status, exposed and unexposed twins differed somewhat with respect to history of severe somatic diseases and family history of psychiatric disorders (19.7% vs 11.8%, 49.9% vs 37.3%, respectively). As expected, owing to the shared familial background, we observed more similarities between exposed twins and their full siblings on all studied characteristics (the ‘twin-sibling cohort’, Table 1).

During the follow-up, we identified 2047 twins with first-recorded psychiatric disorders in the matched twin cohort, including 526 among the exposed and 1521 among the unexposed twins, with a crude incidence rate (IR) of 12.29 and 6.90 per 1000 person-years, respectively (Table 2). The corresponding age- and sex-adjusted hazard ratio (HR) was 1.80 (95% confidence interval ①CI] 1.63–2.00), and decreased to 1.65 (95% CI 1.48–1.83) after further adjustment for socioeconomic factors, history of severe somatic disease, and family history of psychiatric disorders. Notably, the association was considerably stronger within the first month after a co-twin loss (HR = 7.16, 95% CI 3.07–16.70) than thereafter. The HR was 1.59 (95% CI 1.31–1.92) for the period 10 years and onwards. The twin-sibling cohort (i.e. exposed twins compared to their full siblings who also experienced a full sibling loss) yielded similar results, with a fully adjusted HR of 1.55 (95% CI 1.31–1.82) (Table 2).

Subgroup analyses, in both the matched twin and the twin-sibling cohorts, indicated that the association between loss of a co-twin and subsequent risk of psychiatric disorders did not differ by sex and family history of psychiatric disorders (Supplementary file 1-Table 2). Of note, although higher HRs were observed after loss of a co-twin due to unnatural death (e.g., death with any external cause), loss of a co-twin due to natural causes was also associated with a significantly increased risk of first-recorded psychiatric disorders among the surviving twins (e.g. HR = 1.49, 95% CI 1.32–1.69, based on analyses of matched twin cohort). In addition, an age-dependent risk pattern was found, suggesting that after age of 40, there was a decline in HR with increasing age at the index date (Figure 2 and Supplementary file 1-Table 2).

Figure 2

Download asset Open asset

Figure 2—source data 1

Summary data for Figure 2.

https://cdn.elifesciences.org/articles/56860/elife-56860-fig2-data1-v1.xls

Download elife-56860-fig2-data1-v1.xls

To explore potential differences in relative risks by type of sibling loss, we did sub-analyses by zygosity (78% of included twins had such information). Compared to unexposed monozygotic twins, twins who lost a monozygotic co-twin were at a subsequently increased risk of psychiatric disorders (HR = 1.86, 95% CI 1.40–2.47), which was significantly higher than the risk elevation related to loss of a dizygotic co-twin among dizygotic twins (HR = 1.33, 95% CI 1.15–1.54, P for difference = 0.0395, Table 3). The analyses of twin-sibling cohort conveyed similar information, indicating a hierarchy of psychiatric disorder risk ―the HR for ‘loss of a monozygotic twin’, ‘loss of a dizygotic twin’, and ‘loss of a full sibling’ was 2.45, 1.29, and 1.00, respectively (Table 3). In addition, we observed increasing HRs with larger age gaps between twins and non-twin full siblings held as reference (P for trend <0.001, Supplementary file 1-Table 2).

Based on the matched twin cohort, we separately estimated HRs for six subtypes of psychiatric disorders (Figure 3). The association between loss of a co-twin and risk of psychiatric disorders was noted for all subtypes of psychiatric disorders, with the highest HRs observed for psychotic disorders (HR 3.30, 95% CI 2.02–5.38) and stress-related disorders (HR 2.17, 95% CI 1.65–2.84) during the entire follow-up period, and for stress-related disorders (HR 5.36, 95% CI 2.31–12.40) and mood disorders (HR 2.63, 95% CI 1.32–5.24) during the first year after loss.

Figure 3

Download asset Open asset

With slightly lower crude IRs, sensitivity analyses where the ascertainment of psychiatric disorders was merely based on the primary diagnoses in the National Patient Register (NPR) or the diagnoses from inpatient care in NPR revealed largely similar estimates (Supplementary file 1-Table 3).

To our knowledge, this is the first nationwide population-based and sibling-matched cohort study testing the association between loss of a co-twin and the subsequent risk of psychiatric disorders. Compared to twins without such loss experience or full siblings of the bereaved twins (who also experienced a loss of a full sibling), twins exposed to a co-twin loss were at considerably elevated risk of psychiatric disorders, especially during the first month after loss and particularly when the loss occurred before the age of 40. The excess risk was most pronounced after loss of a monozygotic co-twin and was higher for loss of a dizygotic co-twin than loss of a full sibling. In addition, we observed evident risk elevation after co-twin loss through either natural or unnatural deaths.

The twin bond has been suggested to be the closest and most enduring human social relationship (Segal et al., 1995; Segal, 2000). Besides the genetic relatedness, the unique relationship between twins is characterized by distinct intimacy and ambiguous identity boundaries. The genetic relatedness, shared early life experiences, and attachment have been suggested to contribute to the development of the shared twin identity (Schave, 1983; Macdonald, 2002). Consequently, when exposed to a co-twin death, both the grief (i.e., emotional reactions) and confusion of identity may contribute to a profound and long-lasting vulnerability of surviving twins. Indeed, in line with previous literature describing intense grief of surviving twins after co-twin loss (Segal and Ream, 1998; Rosendahl and Björklund, 2013), in the present study, we showed that the subsequent risk of developing first-diagnosed psychiatric disorders increased by 55–65% in the twin population, after carefully controlling for familial factors. Of note, such risk elevation was most obvious during the first month (7-fold), but persistently existed more than 10 years after the loss. In addition to the aforementioned factors (i.e., grief and identity confusion) which might underlie both the short- and long-term rise in the risk of psychiatric disorders, the long-term mental health decline among the surviving twins could also be attributable to the impairment of beneficial social support due to the co-twin loss and possible life-style changes, such as substance abuse or sleep problems (Sinha, 2001).

Moreover, consistent with prior theories that twin attachment changes across developmental stages, we found particularly strong risk increases of first-recorded psychiatric disorders when twins were exposed to co-twin loss in childhood and early adulthood (Neyer, 2002; Segal and Bouchard, 1993). Indeed, young people seem more vulnerable to various health consequences of trauma. We have previously reported higher excess risk of life-threatening infections and cardiovascular disease among individuals with stress-related disorders after trauma at an earlier age (Song et al., 2019; Song et al., 2018). Long-lasting biological interruptions, for example promoted inflammatory reactions (Danese et al., 2011) and neuropsychological/cognitive impairment (De Bellis and Zisk, 2014; Kennedy et al., 2018), have been associated with childhood exposure to trauma.

Without any comparable data, our finding that individuals suffering the loss of a monozygotic co-twin were at higher risk of psychiatric disorders compared to individuals suffering loss of co-dizygotic twin gains support from previous reports indicating a greater grief intensity (Segal and Bouchard, 1993), together with a less grief reduction (Segal and Ream, 1998), after loss of a monozygotic co-twin relative to loss of a dizygotic co-twin. Interpretations for this phenomenon include evolutionary theories linking more intense grief after the bereavement to greater genetic relatedness to the deceased (Neyer, 2002; Segal and Ream, 1998; Parkes, 1993). Alternatively, monozygotic twins have been reported to be emotionally closer and more dependent on their co-twin than dizygotic twins (Fortuna et al., 2010; Penninkilampi-Kerola et al., 2005). Thus, it is plausible that losing a monozygotic co-twin can result in more severe psychiatric consequences, compared to losing a dizygotic co-twin.

The abovementioned explanations may also apply to our next finding indicating that the loss of a full sibling is associated with lower risk of psychiatric disorder compared to loss of a co-twin. Particularly, with respect to dizygotic twins, the twin pair shares equal genetic relatedness as a twin-sibling pair. Therefore, even with an extreme assumption that the increase in the risk of psychiatric disorders after loss of a monozygotic co-twin than loss of a dizygotic co-twin was solely due to genetic factors, the 29% increased rate of psychiatric disorders observed after loss of a dizygotic co-twin than loss of a full sibling should be largely attributable to bereavement. Furthermore, the stronger associations observed with larger age gaps between twins and non-twin siblings supports the notion that siblings closer in age who share more life experiences are more likely have a closer relationship (Conger and Little, 2010) and hence might experience more difficult bereavement after a sibling loss. Similarly, other studies have also suggested a 15–20% excess risk of psychiatric disorders among individuals exposed to loss of a full sibling, as compared to the general population (Guldin et al., 2017; Rostila et al., 2019).

While we attempted to evaluate the influence of co-twin loss on the potential rise in psychiatric disorders, genetic confounding, namely the genetic susceptibility to psychiatric disorders and premature death shared within a twin pair, especially the monozygotic twins, is a concern in the present study. We made every effort to address this issue in the design of our study (e.g. excluding twins with past history of psychiatric disorders, controlling for family history of psychiatric disorders, and comparisons with full siblings) as well as in our analytic strategies (e.g. disentangling natural from unnatural deaths and performing a time-trend analysis). Although we cannot completely exclude the possibility of residual confounding, the results of the abovementioned analyses do argue against a pure explanation by genetic factors. Also, the time-trend analyses demonstrated – above and beyond possible genetic confounding - a distinct bereavement-response pattern, illustrated by the dramatic risk rise of first onset psychiatric disorders during the first months after the loss of a co-twin. Since we had limited information on prenatal/neonatal factors and behavior-related factors (e.g., smoking and alcohol consumption), residual confounding due to these factors remains another concern. Further research with detailed data on birth characteristics and lifestyle is warranted to clarify the role of such factors on the reported associations. Additionally, the absence of data from primary care as well as late inclusion of outpatient specialist care records in the NPR might have led to an underestimated number of individuals with milder forms of psychiatric disorders.

The major merit of our study is the use of population-based cohort design, including 4528 surviving twins with a complete follow-up of up to 40 years, and the between twin-full sibling comparison which enabled a direct comparison between two types of stressors, that is loss of a co-twin and loss of a full sibling, while controlling for potential familial confounders (e.g., genetic background was fully controlled for monozygotic twins, and partially controlled for dizygotic twins). As the largest cohort study on co-twin loss so far, we had sufficient statistical power to perform all planned subgroup analyses. Information bias was minimized since the registration and diagnosis of exposure and outcome was compiled prospectively and independently. Furthermore, the availability of rich information on sociodemographic and medical factors enabled considerations of a wide range of important confounding factors.

In conclusion, in the Swedish population, death of a co-twin was associated with a subsequently elevated rate of first-recorded psychiatric disorders among the surviving twins, particularly among monozygotic twins but also dizygotic twins as compared to their non-twin full siblings. Given that dizygotic twins share equal genetic relatedness to the twin-sibling pairs, this dose-response pattern suggests that both genetic relatedness and early life attachment may contribute to the subsequent risk of psychiatric disorders among surviving twins after co-twin loss.

From the Total Population Register, we identified all individuals born in Sweden between 1932 and 2013. We then linked these individuals to other registers in Sweden, including the Multi-Generation Register, the Causes of Death Register, the Swedish Twin Registry, and the National Patient Register (NPR), utilizing the personal identification numbers that are uniquely assigned to all Swedish residents.

From the Multi-Generation Register which includes almost complete familial link for all individuals born in Sweden since 1932, we identified all twin pairs as full siblings (i.e., with the same biological father and mother) that were born on the same day (+ / - 1 day), after excluding multiple births (more than two babies), and defined individuals with same biological father and mother but born on different dates as normal full siblings.

Two source data files have been provided. Original data is held by Swedish National Board of Health and Welfare, Statistics Sweden and the Swedish Twin Registry. Due to Swedish law on data protection and the ethical approval of the current study, we cannot make the data publicly available. However, any researcher can access the data by obtaining an ethical approval from a regional ethical review board and thereafter request the original data from the Swedish National Board of Health and Welfare, Statistics Sweden, and the Swedish Twin Register. Detailed information on data application can be found at https://www.registerforskning.se/en/ and https://ki.se/en/research/the-swedish-twin-registry.

1. Book

   1. Bank S
   2. Kahn M

   (1997)

   The Sibling Bond

   New York: Basic Books.

   • Google Scholar
2. Book

   1. Bedford VH

   (2012)

   Sibling relationships from midlife to old age

   In: Blieszner R. B. V, editors. Handbook of Families and Aging (2nd ed). Santa Barbara: Praeger. pp. 125–151.

   • Google Scholar
3. 1. Brenn T
   2. Ytterstad E

   (2016) Increased risk of death immediately after losing a spouse: cause-specific mortality following widowhood in Norway

   Preventive Medicine 89:251–256.

   https://doi.org/10.1016/j.ypmed.2016.06.019

   • Google Scholar
4. 1. Bylund Grenklo T
   2. Kreicbergs U
   3. Hauksdóttir A
   4. Valdimarsdóttir UA
   5. Nyberg T
   6. Steineck G
   7. Fürst CJ

   (2013) Self-injury in teenagers who lost a parent to Cancer: a nationwide, population-based, long-term follow-up

   JAMA Pediatrics 167:133–140.

   https://doi.org/10.1001/jamapediatrics.2013.430

   • PubMed
   • Google Scholar
5. Conference

   1. Cassell S

   (2011)

   Examining the twin bond: a look at the psychological development of twins and the differences in individuality and identity

   Differences Between Fraternal and Identical Same-Sex Twins 2.

   • Google Scholar
6. 1. Conger KJ
   2. Little WM

   (2010) Sibling Relationships During the Transition to Adulthood

   Child Development Perspectives 4:87–94.

   https://doi.org/10.1111/j.1750-8606.2010.00123.x

   • Google Scholar
7. 1. Danese A
   2. Caspi A
   3. Williams B
   4. Ambler A
   5. Sugden K
   6. Mika J
   7. Werts H
   8. Freeman J
   9. Pariante CM
   10. Moffitt TE
   11. Arseneault L

   (2011) Biological embedding of stress through inflammation processes in childhood

   Molecular Psychiatry 16:244–246.

   https://doi.org/10.1038/mp.2010.5

   • Google Scholar
8. 1. De Bellis MD
   2. Zisk A

   (2014) The biological effects of childhood trauma

   Child and Adolescent Psychiatric Clinics of North America 23:185–222.

   https://doi.org/10.1016/j.chc.2014.01.002

   • PubMed
   • Google Scholar
9. Book

   1. Fe H

   (2001)

   Regression Modeling Strategies, with Applications to Linear Models, Survival Analysis and Logistic Regression

   New York, USA: Springer-Verlag, New York, Inc.

   • Google Scholar
10. 1. Fortuna K
    2. Goldner I
    3. Knafo A

    (2010) Twin relationships: a comparison across monozygotic twins, dizygotic twins, and nontwin siblings in early childhood

    Family Science 1:205–211.

    https://doi.org/10.1080/19424620.2010.569367

    • Google Scholar
11. 1. Guldin MB
    2. Ina Siegismund Kjaersgaard M
    3. Fenger-Grøn M
    4. Thorlund Parner E
    5. Li J
    6. Prior A
    7. Vestergaard M

    (2017) Risk of suicide, deliberate self-harm and psychiatric illness after the loss of a close relative: a nationwide cohort study

    World Psychiatry 16:193–199.

    https://doi.org/10.1002/wps.20422

    • PubMed
    • Google Scholar
12. 1. Hendrickson KC

    (2009) Mortality, and parental grief: a review of the literature on the relationship between the death of a child and the subsequent health of parents

    Palliative & Supportive Care 7:109–119.

    https://doi.org/10.1017/S1478951509000133

    • PubMed
    • Google Scholar
13. 1. Howe N
    2. Ross HS

    (1990) Socialization, perspective-taking, and the sibling relationship

    Developmental Psychology 26:160–165.

    https://doi.org/10.1037/0012-1649.26.1.160

    • Google Scholar
14. 1. Kennedy B
    2. Chen R
    3. Valdimarsdóttir U
    4. Montgomery S
    5. Fang F
    6. Fall K

    (2018) Childhood bereavement and lower stress resilience in late adolescence

    Journal of Adolescent Health 63:108–114.

    https://doi.org/10.1016/j.jadohealth.2018.02.002

    • PubMed
    • Google Scholar
15. 1. Li J
    2. Precht DH
    3. Mortensen PB
    4. Olsen J

    (2003) Mortality in parents after death of a child in Denmark: a nationwide follow-up study

    The Lancet 361:363–367.

    https://doi.org/10.1016/S0140-6736(03)12387-2

    • Google Scholar
16. 1. Macdonald AM

    (2002) Bereavement in twin relationships: an exploration of themes from a study of twinship

    Twin Research 5:218–226.

    https://doi.org/10.1375/twin.5.3.218

    • PubMed
    • Google Scholar
17. 1. Neyer FJ

    (2002) Twin relationships in old age: a developmental perspective

    Journal of Social and Personal Relationships 19:155–177.

    https://doi.org/10.1177/0265407502192001

    • Google Scholar
18. 1. Parkes CM

    (1993) Bereavement and adaptation: a comparative study of the aftermath of death

    Journal of Psychosomatic Research 37:679–680.

    https://doi.org/10.1016/0022-3999(93)90067-P

    • Google Scholar
19. 1. Penninkilampi-Kerola V
    2. Moilanen I
    3. Kaprio J

    (2005) Co-twin dependence, social interactions, and academic achievement: a population-based study

    Journal of Social and Personal Relationships 22:519–541.

    https://doi.org/10.1177/0265407505054521

    • Google Scholar
20. 1. Pison G
    2. Monden C
    3. Smits J

    (2015) Twinning rates in developed countries: trends and explanations

    Population and Development Review 41:629–649.

    https://doi.org/10.1111/j.1728-4457.2015.00088.x

    • Google Scholar
21. 1. Pompili M
    2. Galeandro PM
    3. Lester D
    4. Tatarelli R

    (2006) Suicidal behavior in surviving co-twins

    Twin Research and Human Genetics 9:642–645.

    https://doi.org/10.1375/twin.9.5.642

    • PubMed
    • Google Scholar
22. 1. Prior A
    2. Fenger-Grøn M
    3. Davydow DS
    4. Olsen J
    5. Li J
    6. Guldin MB
    7. Vestergaard M

    (2018) Bereavement, multimorbidity and mortality: a population-based study using bereavement as an Indicator of mental stress

    Psychological Medicine 48:1437–1443.

    https://doi.org/10.1017/S0033291717002380

    • PubMed
    • Google Scholar
23. 1. Rosendahl SP
    2. Björklund A

    (2013) Images of sorrow—Experiences of losing a co-twin in old age

    Health 5:64–73.

    https://doi.org/10.4236/health.2013.512A009

    • Google Scholar
24. 1. Rostila M
    2. Berg L
    3. Arat A
    4. Vinnerljung B
    5. Hjern A

    (2016) Parental death in childhood and self-inflicted injuries in young adults-a national cohort study from Sweden

    European Child & Adolescent Psychiatry 25:1103–1111.

    https://doi.org/10.1007/s00787-016-0833-6

    • PubMed
    • Google Scholar
25. 1. Rostila M
    2. Berg L
    3. Saarela J
    4. Kawachi I
    5. Hjern A

    (2019) Experience of sibling death in childhood and risk of psychiatric care in adulthood: a national cohort study from Sweden

    European Child & Adolescent Psychiatry 28:1581–1588.

    https://doi.org/10.1007/s00787-019-01324-6

    • PubMed
    • Google Scholar
26. Book

    1. Schave BCJ

    (1983)

    Identify and Intimacy in Twins

    New York: Praeger Publication.

    • Google Scholar
27. 1. Segal NL
    2. Wilson SM
    3. Bouchard TJ
    4. Gitlin DG

    (1995) Comparative grief experiences of bereaved twins and other bereaved relatives

    Personality and Individual Differences 18:511–524.

    https://doi.org/10.1016/0191-8869(94)00174-Q

    • Google Scholar
28. Book

    1. Segal NL

    (2000)

    Entwined Lives: Twins and What They Tell Us About Human Behavior

    New York: Plume.

    • Google Scholar
29. 1. Segal NL

    (2009) Suicidal behaviors in surviving monozygotic and dizygotic co-twins: is the nature of the co-twin's cause of death a factor?

    Suicide and Life-Threatening Behavior 39:569–575.

    https://doi.org/10.1521/suli.2009.39.6.569

    • PubMed
    • Google Scholar
30. 1. Segal NL
    2. Bouchard TJ

    (1993)

    Grief intensity following the loss of a twin and other relatives: test of kinship genetic hypotheses

    Human Biology 65:87–105.

    • PubMed
    • Google Scholar
31. 1. Segal NL
    2. Ream SL

    (1998) Decrease in grief intensity for deceased twin and non-twin relatives: an evolutionary perspective

    Personality and Individual Differences 25:317–325.

    https://doi.org/10.1016/S0191-8869(98)00050-6

    • Google Scholar
32. 1. Sinha R

    (2001) How does stress increase risk of drug abuse and relapse?

    Psychopharmacology 158:343–359.

    https://doi.org/10.1007/s002130100917

    • PubMed
    • Google Scholar
33. 1. Song H
    2. Fang F
    3. Tomasson G
    4. Arnberg FK
    5. Mataix-Cols D
    6. Fernández de la Cruz L
    7. Almqvist C
    8. Fall K
    9. Valdimarsdóttir UA

    (2018) Association of Stress-Related disorders with subsequent autoimmune disease

    Jama 319:2388–2400.

    https://doi.org/10.1001/jama.2018.7028

    • PubMed
    • Google Scholar
34. 1. Song H
    2. Fall K
    3. Fang F
    4. Erlendsdóttir H
    5. Lu D
    6. Mataix-Cols D
    7. Fernández de la Cruz L
    8. D’Onofrio BM
    9. Lichtenstein P
    10. Gottfreðsson M
    11. Almqvist C
    12. Valdimarsdóttir UA

    (2019) Stress related disorders and subsequent risk of life threatening infections: population based sibling controlled cohort study

    BMJ 30:l5784.

    https://doi.org/10.1136/bmj.l5784

    • Google Scholar

Acceptance summary:

Using the excellent Swedish registry system, this manuscript shows that being bereaved of a co-twin puts you at an increased risk of being diagnosed with a psychiatric disorder later in life. This is an interesting and novel finding that contributes to the literature on health consequences of sibling loss.

Decision letter after peer review:

Thank you for submitting your article "Risk of psychiatric disorders among the surviving twins after a co-twin loss" for consideration by eLife. Your article has been reviewed by three peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Eduardo Franco as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Maria H. Chahrour (Reviewer #3).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Summary:

This manuscript aimed to estimate the association between loss of a co-twin and the risk psychiatric disorder using a nationwide population-and sibling-matched cohort study. Leveraging the excellent Swedish registry system, they analyzed data from a cohort of ~4,500 twins who had lost a co-twin, their ~4,900 sibs, and ~23,000 non-bereaved twins. The authors reported that being bereaved of a co-twin puts you at an increased risk of being diagnosed with a psychiatric disorder later in life. The relative risk of the bereaved twins was increased compared to their siblings, after loss of a monozygotic (the highest risk) or a dizygotic co-twin. This well-written manuscript, which used a high quality and unique twin data, is a nice addition to the literature on mental health consequences by sibling loss. However, there are a few concerns.

Essential revisions:

• While studying monozygotic twins and comparing them to dizygotic ones may give indications about the significance of relationship for health consequences by bereavement, it also opens up the possibility that much of the association is due to genetic confounding and a genetic predisposition for psychiatric health problems especially because the strongest association is found in monozygotic twins. The authors should elaborate more on this possibility and also be critical about whether associations reflect a bereavement effect or confounding/underlying risk factors (residual confounding). The possibility of confounding should be highlighted in the Introduction of the paper. In addition, the authors should discuss the analytical strategies used to reduce and estimate the risk of confounding in the Discussion section. For example, to what extent does the authors believe that their results indicate a "bereavement effect"?

• The authors could be clearer about the expected mental health consequences by twin loss compared to non-twin sibling loss by comparing these two events. Why is twin loss more detrimental? A possibility is to empirically compare the two types of sibling losses. Is it possible to study how twin loss influence bereaved twin/non-twin siblings within the same family? Then, it would be necessary to study families with twins where there is another non-twin sibling (or several non-twins). Twin siblings are also the exact same age and the association could therefore reflect that the siblings have a closer relationship because they naturally have shared their lives with one another. Is the loss of non-twin siblings with a small age differences more detrimental?

• Previous research has suggested closer relationships between female siblings and that women have higher quality relationships with family members across the life course. Why did the authors decide not to focus on the age at loss? or gender differences? How about having additional siblings?

• A paper studying the risk of mental health problems following sibling loss is missing. Rostila et al., 2019. It could be interesting to compare the results from the two studies.

• The authors study short term and long term effects by twin loss. However, what could explain these short or long term consequences on psychiatric disorder? Complicated grief and post-traumatic stress could be, for instance, underlie short-term effects.

• How can the stronger association by loss of monozygotic compared to loss of dizygotic twin be explained? They should have an equally close relationship given that they follow each other across life from birth to death. This should be discussed much more thoroughly in the paper and also the comparison to non-twin siblings.

• The different follow-up periods and the rationale behind these are not mentioned in the Introduction or the Results but are discussed in the Discussion section. The Discussion section also includes other results that are not previously presented in the paper. Please check so that new findings are discussed.

• Twins who lost a co-twin before the age of 2 were removed in order to establish an attachment relationship. Why did the authors use this cut off?

• The authors used the Cox proportional hazard modelling; however, they did not mention if the proportional hazard assumption was met for any of the models. While this does not invalidate their findings, it should be described in the Materials and methods and reported in the Results. The proportional hazard assumption checking is probably implied with the age-time interaction spline modelling (see Hess, https://doi.org/10.1002/sim.4780131007) that was used in the Materials and methods and shown in Figure 2 (with the HR dropping off from age 40). However, the authors need to explicitly connect this to proportional hazard. Even better, they could formally check these assumptions using K-M curves or the Schonfield residuals.

• The paper would benefit from a comprehensive discussion of the implications of their finding; the hazard ratios drops off precipitously from when the surviving twin hits the age of 40. This was given scant to no attention in the Discussion section.

Essential revisions:

• While studying monozygotic twins and comparing them to dizygotic ones may give indications about the significance of relationship for health consequences by bereavement, it also opens up the possibility that much of the association is due to genetic confounding and a genetic predisposition for psychiatric health problems especially because the strongest association is found in monozygotic twins. The authors should elaborate more on this possibility and also be critical about whether associations reflect a bereavement effect or confounding/underlying risk factors (residual confounding). The possibility of confounding should be highlighted in the Introduction of the paper. In addition, the authors should discuss the analytical strategies used to reduce and estimate the risk of confounding in the Discussion section. For example, to what extent does the authors believe that their results indicate a "bereavement effect"?

We certainly agree that possibility of genetic confounding needs to be adequately addressed in our study. In our original manuscript, several approaches have been used for controlling for familial factors:

1) When designing the study, we added a twin-sibling cohort where we compared rates of psychiatric disorders between twins exposed to co-twin loss and their full siblings (i.e., a comparison within a family). Since both the surviving twin and their non-twin full siblings were affected by the same bereavement of a twin loss, this cohort enabled a direct comparison between two types of stressors, i.e., loss of a co-twin and loss of a full sibling, while controlling for shared familial factors between siblings. Particularly, with regard to dizygotic twins, the twin pair shares equal genetic relatedness as the twin-sibling pair. Therefore, even with an extreme assumption that the increase in the risk of psychiatric disorders after loss of a monozygotic co-twin than loss of a dizygotic co-twin was solely due to a greater genetic relatedness to the deceased, the 29% increased rate of psychiatric disorders observed after loss of a dizygotic co-twin than loss of a full sibling, as shown in Table 3, should be largely attributable to bereavement. Importantly, given the reference group here is also a cluster of bereaved individuals (due to loss of a full sibling), who had also an 15-20% increased risk of developing psychiatric disorder after a loss of sibling(according to literature) , the ‘29% increase’ is reasonably a conservative estimate.

2) The sub-analyses focusing on the cause of the co-twin’s death demonstrated a clearly elevated rate of psychiatric disorders among twins losing a co-twin due to natural death (increase by 49% compared to matched unexposed twins, by 62% compared to their full siblings). Given that a natural death is less likely affected by a liability to psychiatric disorders shared by the twin pair, this result provides some relief to the concern that the observed association is mainly driven by the shared genetic liability.

3) Importantly, all analyses were limited to twins or sibling pairs without previous history of psychiatric disorders. Additional adjustment for family history of psychiatric disorders (in the twin cohort) attenuated the point estimates somewhat but marginally – see Table 2, the adjusted HR decreased from 1.71 to 1.65 after additionally adjusting for family history of psychiatric disorder).

4) The results from our time-trend analysis indicate considerable “emotional shock” effect as observed by the considerable elevated HRs during the first month after loss of a co-twin (HR=7.16, see Table 2). This phenomenon held true both after loss of monozygotic twins (HR=9.47, see Table 3) and dizygotic twins (HR=4.20), also in comparison with their full bereaved sibling (HR=7.21, see Table 2). If only genetic factors contributed to the increased risks of psychiatric disorders after co-twin loss, we would have expected a monotone rise across time after the co-twin loss.

The analytic strategies we used were stated in our Discussion section. We have now highlighted the importance of controlling for genetic confounding in the Introduction part, and thoroughly discussed this concern in the Discussion of the revised manuscript.

In the revised manuscript, we added:

‘Introduction section: “Moreover, given the shared genetic background of twins, the psychiatric consequences after a co-twin loss need to be estimated with caution, ideally using family comparison with additional vigorous control of previous psychiatric morbidities and the causes of the co-twin’s death.”

Discussion section: “Particularly, with respect to dizygotic twins, the twin pair shares equal genetic relatedness as a twin-sibling pair. Therefore, even with an extreme assumption that the increase in the risk of psychiatric disorders after loss of a monozygotic co-twin than loss of a dizygotic co-twin was solely due to genetic factors, the 29% increased rate of psychiatric disorders observed after loss of a dizygotic co-twin than loss of a full sibling should be largely attributable to bereavement.”

Discussion section: “While we attempted to evaluate the influence of co-twin loss on the potential rise in psychiatric disorders, genetic confounding, namely the genetic susceptibility to psychiatric disorders and premature death shared within a twin pair, especially the monozygotic twins, is a concern in the present study. […] Also, the time-trend analyses demonstrated – above and beyond possible genetic confounding – a distinct bereavement-response pattern illustrated by the dramatic risk rise of first-onset psychiatric disorders during the first months after the loss of a co-twin.”

Discussion section: “’Since we had limited information on prenatal/neonatal factors and behavior-related factors (e.g., smoking and alcohol consumption), residual confounding due to these factors remains another concern.”

• The authors could be clearer about the expected mental health consequences by twin loss compared to non-twin sibling loss by comparing these two events. Why is twin loss more detrimental? A possibility is to empirically compare the two types of sibling losses. Is it possible to study how twin loss influence bereaved twin/non-twin siblings within the same family? Then, it would be necessary to study families with twins where there is another non-twin sibling (or several non-twins).

Thank you for this important comment. We fully agree that studying how twin loss influence bereaved twin vs. non-twin siblings within the same family is very informative on our hypothesis. This is exactly why we performed the twin-sibling cohort where bereaved twins are compared with bereaved non-twin full siblings within the same family.

The finding that loss of a co-twin led to a higher risk of psychiatric disorders compared to loss of a full sibling (HR=1.55, 95% CI 1.31-1.82), can both be explained by greater genetic relatedness, in case of monozygotic twins, and a closer relationship between a twin pair than a pair of full sibling.

In our opinion, probably one of the strongest indications for a “the twin-related bereavement effect” in our data is that bereaved dizygotic twins have 29% increased risk of first-onset psychiatric disorders compared to their bereaved full siblings. As dizygotic twins share equal genetic relatedness to the deceased twin as their full siblings, this finding indicates that the co-twin bereavement constitutes an additional burden to full-sibling loss.

Please refer to the statement in the manuscript:

Materials and methods section: “We further constructed a twin-sibling cohort where we compared rates of psychiatric disorders between twins exposed to co-twin loss and their full siblings, if any, who were also exposed to the same loss. Since both the surviving twin and other non-twin full siblings were affected by the same bereavement of a twin loss, this cohort enabled a direct comparison between two types of stressors, i.e., loss of a co-twin and loss of a full sibling, while controlling for familial factors.”

We discuss this issue in more detail when exploring potential mechanisms for the observed hierarchy in the rate elevation of psychiatric disorders following ‘loss of a monozygotic twin’, ‘loss of a dizygotic twin’, and ‘loss of a non-twin full sibling’ in the revised manuscript:

Discussion section: “Without any comparable data, our finding that individuals suffering the loss of a monozygotic co-twin were at higher risk of psychiatric disorders compared to individuals suffering loss of co-dizygotic twin gains support from previous reports indicating a greater grief intensity (Segal and Bouchard, 1993), together with a less grief reduction (Segal and Roam, 1998), after loss of a monozygotic co-twin relative to loss of a dizygotic co-twin. […] Thus, it is plausible that losing a monozygotic co-twin can result in more severe psychiatric consequences, compared to losing a dizygotic co-twin.”

Discussion section: “The abovementioned explanations may also apply to our next finding indicating that the loss of a full sibling is associated with lower risk of psychiatric disorder compared to loss of a co-twin. […] Furthermore, the stronger associations observed with larger age gaps between twins and non-twin siblings supports the notion that siblings closer in age who share more life experiences are more likely have a closer relationship (Conger and Little, 2010) and hence might experience more difficult bereavement after a sibling loss.”

Twin siblings are also the exact same age and the association could therefore reflect that the siblings have a closer relationship because they naturally have shared their lives with one another. Is the loss of non-twin siblings with a small age differences more detrimental?

This is an excellent point. We have now performed subgroup analyses by age difference (i.e., birth year difference between twins and bereaved non-twin full siblings) in the twin-sibling cohort. Please find the results in Supplementary file 1—supplementary table 2. The results indicate that the difference in rates of first-onset psychiatric disorders between bereaved twins and non-twin siblings becomes greater with increasing age differences. Namely, the bereaved non-twin full siblings who were at a similar age of the deceased twin suffered more similar risk of psychiatric disorder as the surviving twin after loss. This finding supports the notion, as suggested by the reviewer, that the excess risk in psychiatric disorders among surviving twins is partly due to the proximity in age (indicating more shared experiences and, perhaps, closeness of relationship).

We have added this analysis and the corresponding result to the revised manuscript.

Abstract: “Similarly, compared to non-twin full siblings, the adjusted relative risks were significantly increased after loss of monozygotic co-twin (2.45-fold), and loss of a dizygotic co-twin (1.29-fold), with higher HR observed with greater age gaps between twins and non-twin siblings.”

Materials and methods section: “Given that the age difference between siblings may affect the degree of their shared experiences and their relationship, we further tested the impact of age difference on the risk of psychiatric disorder after a loss by calculating HRs by age gaps between twins and non-twin full siblings (0-2, 3-5, and >5 years intervals).”

Results section: “In addition, we observed increasing HRs with larger age gaps between twins and non-twin full siblings held as reference (P for trend <0.001, in Supplementary file 1—supplementary table 2).”

Discussion section: “Furthermore, the stronger associations observed with larger age gaps between twins and non-twin siblings supports the notion that siblings closer in age who share more life experiences are more likely have a closer relationship (Conger and Little, 2010) and hence might experience more difficult bereavement after a sibling loss.”

• Previous research has suggested closer relationships between female siblings and that women have higher quality relationships with family members across the life course. Why did the authors decide not to focus on the age at loss? or gender differences? How about having additional siblings?

Thank you for this comment. We did analyses by age and gender, shown in Supplementary file 1—supplementary table 2 (also quoted in Author response table 1). Briefly, we didn’t see a difference in HR by gender, but there was a clear trend of increasing HRs with lower age at loss.

Regarding the gender difference of dizygotic twin pairs, we did the analyses and showed the results in Table 3 in the manuscript. The HR was 1.43 (95% CI 1.13-1.82) and 1.33 (95% CI 1.10-1.62) after loss of a same-gender dizygotic twin and an opposite-gender dizygotic twin, respectively, implying no strong evidence for effect modification by gender.

According to the reviewer’s comments, we additionally performed sub-analyses for exposed twins with and without additional full sibling(s) at the time of loss and found similar estimates. Also, please note that exposed twins with addition full siblings were exactly the ones involved in the twin-sibling cohort. The crude incidence rates of psychiatric disorders among all exposed twins (12.29 per 1000 person-year) and exposed twins with additional full siblings (11.92 per 1000 person-year) have been shown in Table 2 in the original manuscript. We therefore didn’t add these results to the revised manuscript. But we are certainly willing to reconsider it on the editor/reviewer’s request.

Author response table 1. Hazard ratios (HRs) with 95% confidence intervals (CIs) for any psychiatric disorder among the surviving twins after co-twin loss, compared to matched unexposed twins or their full siblings, by different characteristics

^* Cox regression models were stratified by matching identifiers (birth year and sex, in matched twin cohort) or family identifiers (in twin-sibling cohort), and adjusted for covariates mentioned in the ‘model information’ column. Time since the index date was used as underlying time scale.

• A paper studying the risk of mental health problems following sibling loss is missing. Rostila et al., 2019. It could be interesting to compare the results from the two studies.

Thank you for your comments. We have added a sentence, discussing the results of the mentioned paper.

In the revised manuscript:

‘Discussion section: “Similarly, other studies have also suggested a 15-20% excess risk of psychiatric disorders among individuals exposed to loss of a full sibling, as compared to the general population (Guldin et al., 2017; Rostila et al., 2019).”

• The authors study short term and long term effects by twin loss. However, what could explain these short or long term consequences on psychiatric disorder? Complicated grief and post-traumatic stress could be, for instance, underlie short-term effects.

As the reviewer suggested, we have now discussed in more details the underlying possible mechanisms for short-term and long-term effects of co-twin loss on the risk of psychiatric disorders.

In the revised manuscript:

Discussion section: “In addition to the aforementioned factors (i.e., grief and identity confusion) which might underlie both the short- and long-term rise in the risk of psychiatric disorders, the long-term mental health decline among the surviving twins could also be attributable to the impairment of beneficial social support due to the co-twin loss and possible life-style changes, such as substance abuse or sleep problems (Sinha, 2001).”

• How can the stronger association by loss of monozygotic compared to loss of dizygotic twin be explained? They should have an equally close relationship given that they follow each other across life from birth to death. This should be discussed much more thoroughly in the paper and also the comparison to non-twin siblings.

Thank you for this comment. We have now added a new paragraph in the revised manuscript, discussing the possible explanations for a stronger association observed after loss of a monozygotic twin than after loss of a dizygotic twin.

Discussion section: “Without any comparable data, our finding that individuals suffering the loss of a monozygotic co-twin were at higher risk of psychiatric disorders compared to individuals suffering loss of co-dizygotic twin gains support from previous reports indicating a greater grief intensity (Segal and Bouchard, 1993), together with a less grief reduction (Segal and Roam 1998), after loss of a monozygotic co-twin relative to loss of a dizygotic co-twin. […] Thus, it is plausible that losing a monozygotic co-twin can result in more severe psychiatric consequences, compared to losing a dizygotic co-twin.”

• The different follow-up periods and the rationale behind these are not mentioned in the Introduction or the Results but are discussed in the Discussion section. The Discussion section also includes other results that are not previously presented in the paper. Please check so that new findings are discussed.

Thank you. We have closely checked and ensured all estimates presented in the Discussion section have been properly presented in the Results section.

• Twins who lost a co-twin before the age of 2 were removed in order to establish an attachment relationship. Why did the authors use this cut off?

Thank you for your question. Our reasons of using 2 years as a cut-off point include:

a) Although the attachment bond between children and their parents has been reported to be formed as early as 8-12 months, the social interaction relevant to sibship-relationship formation depends on the activation of exploratory system, which may start around the age of 2-3. (https://hal-univ-tlse2.archives-ouvertes.fr/hal-01498767/document)

b) Given that a very early death of a twin baby (e.g., infant mortality) may be due to difficult birth or congenital defects, which can be also highly related to the physical condition of the surviving twin (including perhaps also psychiatric disorders), we chose to exclude loss of co-twin at very early age (less than 2 years) as exposure in the present study.

In our study population, among a co-twin loss within the first two years of life, more than 70% of the losses occurred within 60 days after twin birth. We found 787 co-twin deaths within 60 days, 332 between 61 days and 1 year, and 122 between 1 and 2 years after twin birth. Therefore, restricting to ‘1 year old’ age at loss gains little additional data power.

In the revised manuscript, we added:

Materials and methods section: “To enable the establishment of attachment relationship (Dufoyer, 1980) , as well as to reduce the possibility of a co-twin loss due to poor birth condition or congenital defects that shares within a twin pair, twins who lost a co-twin before age of 2 years were removed from the exposed cohort (n=1,241).”

• The authors used the Cox proportional hazard modelling; however, they did not mention if the proportional hazard assumption was met for any of the models. While this does not invalidate their findings, it should be described in the Materials and methods and reported in the Results. The proportional hazard assumption checking is probably implied with the age-time interaction spline modelling (see Hess, https://doi.org/10.1002/sim.4780131007) that was used in the Materials and methods and shown in Figure 2 (with the HR dropping off from age 40). However, the authors need to explicitly connect this to proportional hazard. Even better, they could formally check these assumptions using K-M curves or the Schonfield residuals.

Thank you for pointing out this important point. During the analyses, we checked the proportional hazards assumption graphically and by Schoenfeld’s partial residuals and found no evidence of assumption violation. Please note that since the exposed and unexposed twins were individually matched by age at the index date (i.e., therefore, age at loss was stratified in the COX model as a matching identifier), the change of HRs with age at the index does not imply an violence of proportional hazard assumption.

In the revised manuscript, we added:

Materials and methods section: “Throughout our analyses, we checked the proportional hazards assumption graphically and by Schoenfeld’s partial residuals; and neither of them revealed any indication of violation of this assumption”.

• The paper would benefit from a comprehensive discussion of the implications of their finding; the hazard ratios drops off precipitously from when the surviving twin hits the age of 40. This was given scant to no attention in the Discussion section.

Thank you for your comment. In the revised manuscript, we discussed the more pronounced association observed among individuals that experienced a co-twin loss at an earlier age and have now added this to the Discussion.

Discussion section: “Moreover, consistent with prior theories that twin attachment changes across developmental stages, we found particularly strong risk increases of first-recorded psychiatric disorders when twins were exposed to co-twin loss in childhood and early adulthood (Neyer, 2002; Segal and Bouchard, 1993). […] Long-lasting biological interruptions, e.g., promoted inflammatory reactions (24) and neuropsychological/cognitive impairment (25, 26), have been associated with childhood exposure to trauma.”

Author details

1. Huan Song

   1. West China Biomedical Big Data Center, West China Hospital, Sichuan University, Chengdu, China
   2. Center of Public Health Sciences, Faculty of Medicine, University of Iceland, Reykjavík, Iceland
   3. Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden

   Contribution

   Conceptualization, Data curation, Formal analysis, Validation, Investigation, Methodology, Writing - original draft, Writing - review and editing

   For correspondence

   songhuan@wchscu.cn

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0003-3845-8079

2. Henrik Larsson

   1. Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden
   2. School of Medical Sciences, Örebro University, Örebro, Sweden

   Contribution

   Resources, Data curation, Writing - review and editing

   Competing interests

   No competing interests declared

3. Fang Fang

   Institute of Environmental Medicine, Karolinska Institutet, Stockholm, Sweden

   Contribution

   Conceptualization, Supervision, Methodology, Writing - review and editing

   Competing interests

   No competing interests declared

4. Catarina Almqvist

   1. Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden
   2. Astrid Lindgren Children’s Hospital, Karolinska University Hospital, Stockholm, Sweden

   Contribution

   Supervision, Methodology, Writing - review and editing

   Competing interests

   No competing interests declared

5. Nancy L Pedersen

   1. Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden
   2. Department of Psychology, University of Southern California, Los Angeles, United States

   Contribution

   Supervision, Methodology, Writing - review and editing

   Competing interests

   No competing interests declared

6. Patrik KE Magnusson

   Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden

   Contribution

   Conceptualization, Resources, Data curation, Supervision, Writing - review and editing

   Competing interests

   No competing interests declared

7. Unnur A Valdimarsdóttir

   1. Center of Public Health Sciences, Faculty of Medicine, University of Iceland, Reykjavík, Iceland
   2. Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden
   3. Department of Epidemiology, Harvard TH Chan School of Public Health, Boston, United States

   Contribution

   Conceptualization, Resources, Data curation, Formal analysis, Funding acquisition, Methodology, Writing - review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-5382-946X

• 24,699

  Page views

• 189

  Downloads

• 4

  Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.