- Email: [email protected]
Mortality, cancer incidence, and survival in parents after bereavement
Annals of Epidemiology 26 (2016) 115e121
Contents lists available at ScienceDirect
Annals of Epidemiology journal homepage: www.annalsofepidemiology.org
Original article
Mortality, cancer incidence, and survival in parents after bereavement Limor Schorr RN, MSc a, b, Ayala Burger PhD b, Hagit Hochner PhD b, Ronit Calderon MD, MPH b, Orly Manor PhD b, Yechiel Friedlander PhD b, *, Gabriella M. Lawrence MPH b, Ora Paltiel MDCM, MSc b, c a b c
School of Nursing, HadassaheHebrew University, Jerusalem, Israel School of Public Health, HadassaheHebrew University, Jerusalem, Israel Hematology Department, Hadassah University Hospital, Jerusalem, Israel
a r t i c l e i n f o
a b s t r a c t
Article history: Received 15 June 2015 Accepted 20 December 2015 Available online 12 January 2016
Purpose: The study objective was to investigate whether child loss is related to mortality, cancer incidence, and cancer survival in parents. Methods: We used a population-based birth cohort (1964e1976) in Jerusalem and ascertained mortality (average follow-up of 39.1 years) and any cancer (average follow-up of 35.6 years) among parents who lost a child (2838 mothers and 2532 fathers) and among nonbereaved parents (38,212 mothers and 36,433 fathers). We also assessed mortality among parents with cancer. Time-dependent Cox models were used to estimate hazard ratios (HRs) with 95% confidence intervals (CIs). Results: Overall mortality rates among bereaved parents were modestly increased when compared with nonbereaved parents (HR ¼ 1.18, 95% CI: 1.05e1.32 in mothers; HR ¼ 1.10, 95% CI: 1.01e1.20 in fathers). Hazard models indicated a significant relationship between bereavement and deaths from coronary heart disease in mothers (HR ¼ 1.90, 95% CI: 1.23e2.95) and circulatory causes in both parents (HR ¼ 1.69; 95% CI: 1.22e2.34 in mothers and HR ¼ 1.25; 95% CI: 1.02e1.54 in fathers). Bereavement was not associated with parental risk of cancer disease and with survival from cancer. The association between bereavement and parental overall mortality was similar in the different parental sociodemographic characteristics. We observed a decrease in HRs for parental mortality associated with bereavement, with increasing time since the death of the child (HRs ¼ 9e10, 0e3 years; HRs ¼ 0.9e1.0, 9þ years; Pheterogeneity 3 1032). A similar decrease in HRs was observed for parental survival from cancer (HRs ¼ 6.7e8.7, 0e3 years; HRs ¼ 0.9e1.0, 9þ years). Conclusions: Our study suggests that child loss was associated with slightly increased risk of all-cause and circulatory mortality in parents but not with incidence of cancer and cancer survival. The considerable increased parental mortality during a short period after child loss support the involvement of pathways related to psychological stress. Ó 2016 Elsevier Inc. All rights reserved.
Keywords: Mortality Cancer Parental bereavement
Introduction Growing evidence suggests that chronic diseases and mortality are influenced by psychological stress [1,2]. Bereavement, an acute life event and its association with mortality and disease morbidity, has been studied in various populations, and a number of studies demonstrate excess mortality among widowers [3,4]. Child loss is a particularly stressful life event [5,6], even more so than the death of a spouse [7e9]. * Corresponding author. Unit of Epidemiology, Hebrew UniversityeHadassah Braun School of Public Health, POB 12272, Jerusalem 91120, Israel. Tel.: þ972-26777805; fax: þ972-2-6431086. E-mail address: [email protected] (Y. Friedlander). http://dx.doi.org/10.1016/j.annepidem.2015.12.008 1047-2797/Ó 2016 Elsevier Inc. All rights reserved.
Stressful events affect various physiological systems, such as the sympathetic nervous system, the hypothalamic-pituitary-adrenal axis, the neuroendocrine systems, and the immune systems [10]; we, therefore, predict that child loss may be associated with diseases such as cardiovascular and cancer and/or mortality in parents. Stressful life events raise the risk of various psychological disorders, which may lead to adverse lifestyle behaviors [11,12] and affect unnatural and natural deaths [13,14]. A study in Israel examined mortality in a group of bereaved parents of soldiers 18e40 years of age who were killed at war in 1973 and parents of offspring 18e30 years of age who died in accidents compared with the general population [15]. The 10-year follow-up indicated no excess mortality among bereaved parents;
116
L. Schorr et al. / Annals of Epidemiology 26 (2016) 115e121
however, when the follow-up was extended to 20 years, the findings showed an effect of stress on incidence of malignancies for selected sites and accelerated demise among parents bereaved after a cancer diagnosis, but not among those bereaved before such a diagnosis [16]. Other investigations, however, indicate an overall association between child loss and mortality in parents [17]. The Jerusalem Perinatal Study (JPS) provides a unique opportunity to explore the relation between child loss and all-causee and cause-specific mortality, cancer morbidity, and survival in parents. Methods The JPS is a prospective population-based birth cohort of 92,408 infants born (1964e1976) to residents of Jerusalem and their parents. Through the Israeli Population Registry (IPR), we verified identities of 41,050 mothers and 38,965 fathers (96.2% and 92.5%, respectively) in the original cohort and updated their vital status. We undertook a long follow-up study to determine the association between parental bereavement and mortality and cancer morbidity and survival. Non-Jewish parents, comprising 2.3% of the cohort, were included in analyses. Study variables Data included demographic and socioeconomic characteristics of parents and information on offspring including sex and birth date. Information was abstracted from birth certificates or maternity ward logbooks. Primary outcomes in parents were risk of all-cause mortality and cause-specific mortality, including mortality due to coronary heart disease (CHD), circulatory disorders, cancer, unnatural and other causes, incidence of any cancer and mortality in parents with cancer diagnosis. The underlying cause of death was categorized according to International Classification of Diseases, Ninth Revision (ICD-9), codes until 1997 and in 1998 according to ICD-10 codes [18,19]. Causes of deaths were categorized into the following groups: CHD (ICD-9 codes: 410e414, 427.4, and 427.5; ICD-10 codes: I20eI25, I46, and I49); all circulatory conditions (ICD-9 codes: 390e459; ICD-10 codes: I0eI99); all neoplasms (ICD-9: 140e239; ICD-10: C0eC99 and D0eD48); death due to unnatural causes (ICD-9: 80e99; ICD-10: S0eT88, V0eY99); and death due to all other causes taken together. In Israel, nearly 100% of deaths are reported to the IPR, along with the date and cause of death. Linkage to the IPR occurred in 2011 at which point the registry had been fully updated through December 2010 for 89.6% of the mothers and 88.4% of the fathers. The remaining parents had updated information through April 30, 2005. Cancer incidence was ascertained via the Israel National Cancer Registry which was established in 1961 [20]. Linkage occurred in 2010 at which point the registry had been updated through July 31, 2008. For parents diagnosed with cancer, mortality was assessed via the IPR, and they were determined to have died from cancer regardless of the actual cause of death, that is, “cancer mortality.” The following parental characteristics were included in the analyses: parents’ ethnic origindclassified according to the parents’ country of birth, or if born in Israel, that of the grandfather’s birth country (categorized as Israel, other West Asia, North Africa, Europe or America or other industrialized countries, and non-Jewish), a social class scale (socioeconomic status [SES]: from 1 ¼ highest to 3 ¼ lowest) based on father’s occupation, parity (number of children born before 1964, as reported by the mother, plus those born within the cohort [categorized as 1, 2, and 3]), having a stillbirth in the family (1, yes; 0, no), parents’ age at child’s birth (<30, 30e39,
40 years), parents’ age at child’s death (<30, 30 years), and parents’ education level (0e4, 5e8, 9e12, 13, unknown). In addition, we examined the following characteristics related to the deceased child: number of deceased children in the family (1, 2), child’s age at death (<1, 1e17, 18 years), and time passed since child loss (0e3, 4e8, 9 years). Statistical analysis Time was counted from birth of the deceased child or earliest birth in the cohort when parents had multiple live offspring until death of parent or study closure (December 31, 2010 or April 30, 2005). The Cox proportional hazards regression model, with child’s death as a time-dependent exposure variable, was used to evaluate associations between this life event and parental mortality after controlling for possible confounders. If a child died during the study period, the surviving parent(s) changed status from nonexposed (nonbereaved) to exposed (bereaved). All parents, even those who died before the child loss, were included in the mortality analysis. Hazard ratios (HRs) and 95% confidence intervals (CIs) were calculated, and competing hypotheses were tested by the likelihood ratio test. For the investigation of cancer incidence, parents with cancer diagnosed within 5 years after child loss were eliminated from the analysis. However, we performed a series of sensitivity analyses to test the robustness of our cancer results: (1) by including parents with cancer diagnosed 3 or more years after bereavement and (2) by including parents diagnosed 1 year after bereavement. We also compared survival in parents whose cancer was diagnosed before their loss (and survived at least till after the child loss) and those whose cancer was diagnosed after their loss. Adjustments for parents’ education level, parents’ age at child’s birth and death, and child’s age at death in all models were also assessed as continuous variables and yielded HRs similar to those obtained using age and education as categorical variables. Analyses were stratified by characteristics of parents and the deceased child, which may modify parents’ ability to cope with the acute life event. Owing to sample size limitations, stratification was not carried out on cause-specific mortality. We used heterogeneity tests for stratified data to test for interactions between the exposure and modifier variables. The significance of effect modification was assessed using Higgins and Thompson’s heterogeneity index [21] which under the null hypothesis is distributed as c2(I1) where I indicates the number of strata. Most analyses were completed using Stata, version 12.0 (StataCorp, College Station, TX), and heterogeneity tests were calculated using WinPepi, version 11.42 (J.H. Abramson, Jerusalem, Israel) [22]. All tests were two-sided, and an alpha set at 0.05. This study was approved by the Institutional Review Boards of HadassaheHebrew University Medical Center in Jerusalem. Results Table 1 shows baseline characteristics of bereaved (2838 mothers and 2532 fathers) and nonbereaved (38,212 mothers and 36,433 fathers) parents. There was no relation between bereavement and parents’ age at birth. About 77% of mothers and 60% of fathers gave birth before the age of 30 years. Compared with the unexposed group, exposed Jewish parents tended to be of Asian and North African origin. Exposed parents were less educated, exhibited lower SES, and have more children. About 9.5% of parents were exposed to death of two children or more, and more than 50% of the deceased children died during their first year of life. During the follow-up, mothers contributed 1,608,772 personyears (average: 39.1 person-years) for overall mortality and 1,451,323 person-years (average: 35.6 person-years) for cancer
L. Schorr et al. / Annals of Epidemiology 26 (2016) 115e121 Table 1 Baseline characteristics of parents Characteristic
Mothers
Fathers
Nonbereaved
Bereaved
Nonbereaved
Bereaved
n ¼ 38,212
n ¼ 2838
n ¼ 36,433
n ¼ 2532
59.9 30.2 9.9
59.2 32.3 8.4
14.2 29.1 19.7 35.7 1.2
14.2 34.3 22.3 26.1 3.1
3.8 18.3 36.6 38.3
6.6 23.3 36.9 31.5
32.7 38.4 28.9
23.6 35.5 40.9
21.6 25.9 52.5
5.1 14.6 80.3
d d
90.3 9.7
d d d
54.3 17.3 28.4
Age at first birth (%) <30 76.5 77.8 30e39 21.4 20.7 40 2.1 1.5 Origin (%) Israel 13.6 12.1 Asia 27.4 32.1 North Africa 20.7 24.8 Europe or West 37.0 27.8 Non-Jews 1.3 3.1 Education (%) 0e4 y 7.1 11.4 5e8 y 19.4 28.0 9e12 y 37.0 33.2 13 y 33.5 25.5 SES (%) High 38.2 32.5 Middle 38.5 38.4 Low 23.2 29.1 No. of children in family (%) 1 22.7 7.4 2 25.7 15.3 3 51.5 77.4 No. of deceased children per family (%) 1 d 90.6 2 d 9.4 Age of the deceased child (%) <1 y d 54.7 1e18 y d 16.3 19 y d 29.0
morbidity. During this period, 4359 mothers (10.6%) died, and 5123 (12.5%) developed cancer. Mothers who developed cancer contributed 184,998 person-years of follow-up (average: 36.1 person-years), and 2032 of the 5123 diagnosed mothers (39.7%) died during follow-up. The 38,965 fathers contributed 1,350,576 person-years (average: 34.7 person-years) of mortality follow-up and 1,316,350 personyears (average: 34.1 person-years) of cancer follow-up, with 8750 (22.5%) and 5146 (13.2%) deaths and cancer events, respectively. Fathers who developed cancer contributed 146,604 person-years of follow-up (average: 28.5 person-years), and 2767 of the 5146 diagnosed fathers (53.8%) died during follow-up. Table 2 presents HRs among parents for all-cause and causespecific mortality associated with child loss. The adjusted mortality rate among bereaved parents was modestly yet significantly higher (HR ¼ 1.10e1.18) when compared with nonbereaved parents (P .03). In addition, the adjusted mortality rate due to CHD and circulatory disorders among bereaved mothers was significantly higher when compared with nonbereaved mothers (HR ¼ 1.90 and HR ¼ 1.69, respectively). Among fathers, only adjusted mortality
117
rates due to circulatory disorders were significantly associated with bereavement (HR ¼ 1.25). Table 3 presents HRs among parents for overall mortality according to characteristics that may affect ability to cope with acute stress. Parent’s age at entry and at child’s death did not modify the hazard associated with child loss. The analyses by categories of education and SES provided no evidence of heterogeneity in the associations. Non-Jewish mothers and Jewish mothers born in Israel showed higher HRs for overall mortality (P for heterogeneity ¼ .038) as compared to mothers from other origin groups. We noted a decrease in HRs for total mortality by parity, and mothers who lost two children had an HR of 1.58 for overall mortality (P for heterogeneity ¼ .027). We noted a significantly increased hazard (9.0- to 10.2-fold HR) for overall mortality in parents, within a short period (0e3 years) after child loss. Our analysis indicated a significant decrease in hazard with increasing time since the child loss (HRs ¼ 3.2e3.3 within a period of 4e8 years and no increased risk [HRs ¼ 0.9e1.1] after a longer period; P for heterogeneity 3 1032 to <1033). Table 4 presents HRs among parents for cancer morbidity and mortality after cancer diagnosis associated with child loss. No association was observed between offspring’s death and parent’s cancer morbidity (HR ¼ 0.89e0.98). Similar results were obtained when parents with cancer diagnosed 3 or more years after bereavement were included (HR ¼ 0.93e0.99, data not shown), as well as when parents with cancer diagnosed 1 year after bereavement were included (HR ¼ 0.96e1.04, data not shown). Bereavement was not associated with the risk of mother’s survival from cancer (HR ¼ 1.00) and similar results were obtained in mothers diagnosed with cancer before their loss (HR ¼ 0.85, 95% CI ¼ 0.44, 1.65, data not shown) and after their loss (HR ¼ 1.00, 95% CI ¼ 0.83, 1.21, data not shown). Among fathers, all diagnosed with cancer after their child loss, the HR for mortality was modestly increased (HR ¼ 1.15), though not statistically significant. Table 4 also presents HRs among parents for cancer incidence and cancer survival according to characteristics of parents and the deceased child. No differences were observed in HRs by parent’s age at entry. With respect to cancer incidence in mother and cancer survival in fathers, the adjusted HR tended to be higher among parents who were younger than 30 years at their child’s death compared with those who were older. In mothers, analyses by education and SES level provided no evidence of heterogeneity in associations. Our data suggested higher HRs for cancer mortality in fathers belonging to the highest education and SES groups. Non-Jewish mothers and Jewish mothers born in Israel showed higher HRs for mortality from cancer (P for heterogeneity ¼ .02) as compared to mothers from other origin groups. Mortality from cancer tended to have the strongest association with bereavement in fathers who were born in Europe/West countries. We also noted a decrease in HRs for mortality from cancer, with increased number of children in the family. Fathers who lost two children had an HR of 1.25 for cancer incidence and 0.85 among fathers who lost one child (P for heterogeneity ¼ .023).
Table 2 Associations* between death of a child and overall mortality and cause-specific mortality in parents Cause of death
Mothers HR
95% CI
P value
HR
95% CI
P value
All-cause mortality Coronary heart disease Circulatory disorders Cancer Unnatural causes Any other cause
1.18 1.90 1.69 1.16 1.39 1.12
1.05e1.32 1.23e2.95 1.22e2.34 0.86e1.57 0.64e3.05 0.98e1.28
.004 .004 .002 .34 .41 .08
1.10 1.12 1.25 1.23 0.94 1.06
1.01e1.20 0.86e1.47 1.02e1.54 0.94e1.61 0.48e1.85 0.96e1.17
.03 .39 .03 .14 .86 .22
*
Fathers
Models adjusted for parental age at entry, education, SES, ethnic origin, family size, and stillbirth in family.
118
L. Schorr et al. / Annals of Epidemiology 26 (2016) 115e121
Table 3 Associations* between death of a child and overall mortality in parents, according to basic characteristics Characteristic
Mothers HR
Total group 1.18 Parental age at entry <30 1.12 30e39 1.08 40 1.38 Test for heterogeneity d Age of parent when child died <30 1.27 30 1.14 Test for heterogeneity d Parental education 0e4 1.12 5e8 1.16 9e12 1.14 13 1.40 Test for heterogeneity d Parental SES High 1.18 Middle 1.30 Low 1.09 Test for heterogeneity d Parental origin Israel 1.47 Asia 1.24 North Africa 0.89 Europe 1.32 Non-Jews 1.61 Test for heterogeneity d Family size 1 child 1.59 2 children 1.65 3 children 1.15 Test for heterogeneity d No. of children who died 1 child 1.12 2 children 1.58 Test for heterogeneity d Age of child at death, y <1 1.22 1e18 0.96 19 1.27 Test for heterogeneity d Length of time from death of child, y 0e3 10.24 4e8 3.33 9 1.05 Test for heterogeneity d
Fathers 95% CI
P value
HR
95% CI
P value
1.05e1.32
.004
1.10
1.01e1.20
.02
0.96e1.31 0.91e1.29 0.90e2.12 c2(2) ¼ 1.08
.16 .37 .14 .583
1.01 1.10 1.11 d
0.87e1.18 0.97e1.25 0.93e1.32 c2(2) ¼ 0.90
.86 .13 .24 .639
1.04e1.56 1.00e1.30 c2(1) ¼ 0.77
.02 .05 .381
1.04 1.11 d
0.84e1.30 1.02e1.22 c2(1) ¼ 0.29
.71 .02 .589
0.89e1.41 0.95e1.42 0.90e1.44 1.07e1.84 c2(3) ¼ 1.83
.32 .14 .27 .01 .608
1.22 0.97 1.10 1.19 d
0.98e1.53 0.83e1.13 0.95e1.27 0.99e1.43 c2(3) ¼ 4.07
.08 .67 .21 .06 .255
0.94e1.48 1.07e1.56 0.91e1.30 c2(2) ¼ 1.77
.16 .007 .36 .412
1.33 1.06 1.06 d
1.08e1.64 0.91e1.23 0.94e1.19 c2(2) ¼ 3.79
.01 .45 .35 .151
1.08e2.01 1.02e1.50 0.71e1.12 1.04e1.67 0.95e2.74 c2(4) ¼ 10.16
.02 .03 .31 .02 .08 .038
1.07 1.10 1.15 1.13 0.76 d
0.84e1.37 0.96e1.26 0.98e1.35 0.94e1.36 0.43e1.35 c2(4) ¼ 2.01
.59 .18 .09 .19 .35 .734
0.96e2.62 1.18e2.33 1.02e1.30 c2(2) ¼ 5.00
.07 .004 .02 .082
1.47 1.23 1.10 d
0.85e2.55 0.91e1.65 1.00e1.20 c2(2) ¼ 1.47
.17 .18 .04 .479
0.99e1.27 1.20e2.09 c2(1) ¼ 4.92
.06 .001 .027
1.10 1.07 d
1.01e1.21 0.84e1.37 c2(1) ¼ 0.04
.03 .57 .835
1.06e1.40 0.73e1.27 1.00e1.62 c2(2) ¼ 2.71
.004 .79 .05 .258
1.09 1.07 1.22 d
0.98e1.21 0.89e1.29 1.02e1.46 c2(2) ¼ 1.33
.12 .46 .03 .514
6.99e15.00 2.24e4.94 0.93e1.18 c2(2) ¼ 145.17
3.7 1033 2.47 109 .46 3.0 1032
8.99 3.15 0.94 d
6.92e11.69 2.44e4.06 0.86e1.03 c2(2) ¼ 304.99
<1033 6.3 1019 .20 <1033
* Models adjusted for maternal age at entry, parental education, SES, ethnic origin, family size, and stillbirth in family. In any stratified analysis, the model was adjusted for other covariables (e.g., in stratified models on parental age groups, we excluded from the Cox model the stratified variabledparental age).
With respect to survival from cancer, mothers who lost two children had an HR of 1.45 for mortality and 0.94 among mothers who lost one child (P for heterogeneity ¼ .08), and the adjusted HR was higher among mothers whose child died at the age of 18 years or more compared with those whose child died before the age of 18 years (P for heterogeneity ¼ .05). Similar to overall mortality, we observed a significant heterogeneity in HRs for cancer survival by length of time since the child loss (P for heterogeneity 9.5 1011e2.4 1017). Because incidence of cancer did not include diagnosed cases within 5 years after child loss, we did not carry out this heterogeneity test for cancer morbidity. Discussion Overall and cause-specific mortality Results from our study (1964e2010) show a modest increased overall mortality in parents associated with child loss, yet a hugely
increased risk of death within the first 3 years after bereavement in both fathers and mothers. In addition, increased mortality due to CHD is seen among bereaved mothers, and circulation-related mortality was greater among bereaved parents compared with those who did not lose a child. This study joins a growing body of evidence showing an association between bereavement and parental morbidity and mortality. For example, a study in Denmark showed a significantly increased overall mortality rate in bereaved mothers (HR ¼ 1.43) and modest excess risk in bereaved fathers (HR ¼ 1.09) [17]. In a 22year follow-up in the Swedish population, authors found increased mortality risk (relative risk ¼ 1.31) among mothers after the death of a young offspring (ages 10e17 years) and attenuated risk (relative risk ¼ 1.15) among fathers [23]. Our results also suggest that excess mortality associated with bereavement in mothers is greater than that in fathers. Gender differences in the ability to cope with stress [24] leading to more adverse health among mothers than fathers after child loss [17,25,26], along with stronger relationships
L. Schorr et al. / Annals of Epidemiology 26 (2016) 115e121
119
Table 4 Associations* between death of a child and cancer morbidity and cancer survival in parents, according to basic characteristics Characteristic
Mothers
Fathers
Cancer morbidity HR
95% CI
Total group 0.98 0.86e1.11 Parental age at entry <30 1.00 0.86e1.16 30e39 0.84 0.66e1.07 40 0.48 0.15e1.52 Test for d c2(2) ¼ 2.78 heterogeneity Age of parent when child died <30 1.15 0.96e1.38 30 0.87 0.74e1.03 Test for d c2(1) ¼ 4.96 heterogeneity Parental education 0e4 0.75 0.50e1.11 5e9 1.00 0.79e1.26 9e12 0.95 0.76e1.19 13 1.11 0.88e1.39 c2(3) ¼ 2.96 Test for d heterogeneity Parental SES High 0.95 0.77e1.18 Middle 1.02 0.83e1.24 Low 0.95 0.75e1.20 Test for d c2(2) ¼ 0.30 heterogeneity Parental origin Israel 0.88 0.61e1.27 Asia 0.86 0.68e1.08 North Africa 1.14 0.89e1.45 Europe 1.00 0.80e1.25 Non-Jews 1.47 0.62e3.51 Test for d c2(4) ¼ 3.90 heterogeneity Family size 1 child 0.69 0.38e1.25 2 children 1.00 0.71e1.42 3 children 0.95 0.83e1.09 Test for d c2(2) ¼ 1.18 heterogeneity No. of children who died 1 child 0.96 0.84e1.09 2 children 1.15 0.82e1.61 Test for d c2(1) ¼ 0.96 heterogeneity Age of child at death, y <1 1.08 0.93e1.24 1e18 0.92 0.69e1.23 19 0.67 0.46e0.97 Test for d c2(2) ¼ 5.85 heterogeneity Length of time from death of child, y 0e3 d d 4e8 5þ y d 9 0.98 0.86e1.11 Test for d d heterogeneity
Cancer survival P value
Cancer morbidity
Cancer survival
HR
95% CI
P value
HR
95% CI
P value
HR
95% CI
P value
.71
1.00
0.83e1.20
1.00
0.89
0.79e1.01
.08
1.15
0.99e1.35
.07
.99 .15 .21 .249
1.03 0.96 0.54 d
0.82e1.30 0.71e1.30 0.11e2.54 c2(2) ¼ 0.72
.82 .79 .43 .697
0.88 0.84 1.00 d
0.72e1.07 0.69e1.02 0.72e1.37 c2(2) ¼ 0.82
.19 .08 .98 .662
1.19 1.06 1.14 d
0.91e1.57 0.84e1.35 0.82e1.58 c2(2) ¼ 0.41
.20 .61 .44 .815
.13 .10 .026
0.99 1.00 d
0.73e1.35 0.80e1.24 c2(1) ¼ 0.00
.97 .99 .958
0.77 0.92 d
0.58e1.04 0.81e1.06 c2(1) ¼ 1.18
.09 .27 .278
1.69 1.09 d
1.14e2.49 0.92e1.29 c2(1) ¼ 4.08
.01 .30 .043
.15 .98 .64 .39 .398
0.98 0.96 0.81 1.18 d
0.6e1.61 0.69e1.34 0.56e1.17 0.83e1.67 c2(3) ¼ 2.13
.93 .82 .26 .35 .547
1.23 0.79 0.88 0.90 d
0.82e1.85 0.61e1.01 0.71e1.09 0.71e1.15 c2(3) ¼ 3.31
.31 .06 .25 .41 .347
1.24 0.92 1.14 1.37 d
0.79e1.93 0.68e1.26 0.86e1.50 1.01e1.87 c2(3) ¼ 3.36
.35 .62 .37 .04 .339
.66 .86 .66 .862
0.90 0.98 1.09 d
0.64e1.26 0.73e1.33 0.79e1.50 c2(2) ¼ 0.65
.53 .91 .61 .721
0.79 0.88 0.94 d
0.59e1.06 0.72e1.09 0.78e1.13 c2(2) ¼ 0.98
.12 .26 .50 .612
1.47 1.43 0.94 d
1.02e2.14 1.09e1.87 0.75e1.17 c2(2) ¼ 7.30
.04 .01 .58 .026
.49 .20 .30 .98 .39 .420
1.58 1.16 0.60 1.03 2.10 d
0.98e2.53 0.84e1.61 0.40e0.91 0.74e1.44 0.65e6.81 c2(4) ¼ 11.68
.06 .37 .02 .84 .22 .02
0.85 0.99 0.90 0.82 0.39 d
0.59e1.22 0.80e1.22 0.70e1.17 0.64e1.05 0.11e1.36 c2(4) ¼ 3.11
.37 .91 .43 .11 .14 .539
0.90 1.15 1.02 1.64 0.27 d
0.54e1.50 0.89e1.49 0.76e1.37 1.20e2.24 0.03e2.17 c2(4) ¼ 8.16
.69 .29 .91 .0008 .22 .086
.22 1.00 .47 .555
2.08 1.42 0.96 d
0.97e4.43 0.89e2.29 0.78e1.17 c2(2) ¼ 5.41
.06 .15 .66 .067
0.92 0.68 0.89 d
0.44e1.94 0.43e1.08 0.77e1.01 c2(2) ¼ 1.23
.83 .10 .07 .541
2.99 2.00 1.10 d
1.09e8.15 1.22e3.28 0.94e1.30 c2(2) ¼ 8.36
.03 .01 .24 .015
.50 .42 .328
0.94 1.45 d
0.77e1.14 0.93e2.26 c2(1) ¼ 3.06
.54 .10 .08
0.85 1.25 d
0.74e0.97 0.92e1.69 c2(1) ¼ 5.16
.02 .16 .023
1.20 0.89 d
1.02e1.42 0.59e1.34 c2(1) ¼ 1.75
.03 .56 .185
.33 .58 .03 .054
0.89 0.94 1.50 d
0.70e1.12 0.61e1.45 1.05e2.13 c2(2) ¼ 5.99
.31 .79 .03 .05
0.96 0.82 0.76 d
0.83e1.13 0.62e1.10 0.56e1.03 c2(2) ¼ 2.26
.65 .19 .08 .324
1.15 1.12 1.20 d
0.95e1.40 0.80e1.58 0.85e1.69 c2(2) ¼ 0.08
.15 .51 .30 .961
6.73 2.53 0.88 d
3.70e12.25 1.31e4.88 0.73e1.08 c2(2) ¼ 46.16
d 5þ y 0.89 d
d d 0.79e1.01 d
d d .08 d
8.76 2.13 0.99 d
5.48e13.99 1.32e3.45 0.83e1.18 c2(2) ¼ 76.56
5.4 1020 .002 .92 2.4 1017
d d .71 d
4.2 1010 .006 .22 9.5 1011
* Models adjusted for parental age at entry, parental education, SES, ethnic origin, family size, and stillbirth in family. In any stratified analysis, the model was adjusted for other covariables (e.g., in stratified models on parental age groups, we excluded from the Cox model the stratified variabledparental age).
between children and their mothers than their fathers [27], may explain the stronger adverse impact on mother’s mortality risk. Other studies, however, showed no association between child loss and mortality in parents [15,28]. Although, in the Israeli Study [15], parents were exposed to a very stressful life event which did not allow for an attenuating effect of anticipatory grief (child killed in action or in accident) [29], it is possible that the null finding is a result of a short duration of follow-up and nonoptimal selection of comparison group.
Similar to other studies (e.g., references [9,17,23]), we examined whether parental characteristics or those related to the deceased child modify the risk of overall mortality in parents associated with bereavement. The risk of mortality was clearly modified by time elapsed since the child loss, with strong increased hazard of parental mortality (HR ¼ 40.6 for mothers and HR ¼ 30.0 for fathers, data not shown) during the first year and no increased risk (HRs ¼ 0.8e1.05) after 9 years. Similarly, in another study, excess mortality risk was evident until approximately 9 years after child
120
L. Schorr et al. / Annals of Epidemiology 26 (2016) 115e121
loss [17]. Studies indicate that increased mortality due to loss of a close family member is divided into two groups: immediate effects, identifying bereavement as a grief-producing agent (e.g., stress, anxiety, depression), and late effects, suggesting that bereavement leads to disorganized life patterns and changes in lifestyle behaviors. Our findings support the role of short-period effects. It is possible that the psychological impact of bereavement leads to short-term mental illness and other mental/behavioral changes. A recent study found that during the 2 years after child-loss in a collision, bereaved parents had significant increase in risk of depression, anxiety disorders, marital breakup, and physician visits for mental illness as compared with nonbereaved parents [30]. Mental illness has been found to be associated with several specific causes of death, including suicide, cardiovascular, and cancer and with increased all-cause mortality rates [31,32]. In our study, the death of a child was associated with CHD/circulatory causes of death but not with an increased risk of unnatural causes. Studies also suggest that the effect of a child’s death may be modified by age of parent or the deceased child [9,17]. For example, Rostila et al. [23] found that the loss of an older child had a lower effect on parental mortality even during the first few years. The loss of a young child, strongly conflicting with life-cycle expectations, may expose parents to intense stress, leading to unusual psychophysiological states and abnormal health effects. In addition, many adult children have their own families, and their death may lead to stronger bonds between the grieving parent and the nuclear family of the deceased child, alleviating the negative effects after the death of the adult child. However, our study did not support this heterogeneity in bereavement effect by parental age or the child’s age at death. Family size also tends to modify the risk associated with bereavement [17]. In our study, parents who lost their only child showed the highest HR for overall mortality and those having three children and more had the lowest mortality HR. The Israeli population is characterized by high percent of families with three children and more [33]. Although the status of a bereaved parent never ceases, the effect of grief may differ in large families where parents’ grief is not intertwined with loss of social support, and where the presence of other remaining children provides motivation or incentive for parents to be mindful of their health and safety. Some studies have suggested that socioeconomic variables could also modify the ability of bereaved parents to cope with stress [34e36], yet other studies did not support this [17,37]. In this group of parents, marked differences in education and SES were observed by the different ethnic groups. Jewish mothers born in countries of North Africa and probably belonging to lower SES and education groups [38] tend to show the lowest HR associated with child loss. One possible explanation is related to differences in family size between parents born in countries of Western Asia or North Africa and those born in Israel or Europe [38]. Although, we were able to adjust for several potential confounders such as education, SES, ethnic origin, and family size, the risk of death associated with bereavement could be modified by ethnic groups because of differences in socioeconomic characteristics and family size. Cancer incidence and survival Our results, indicating no association between child loss and cancer morbidity, are in line with other studies [16,39e43]. For example, the cohort of Israeli parents, who lost an adult son in war or a car accident, showed no significant association between bereavement and incidence of any cancer [16]. Similar to other studies [16,41,44e46], we also failed to show any association between bereavement and cancer survival in parents, although it has been suggested that psychological stress may lead to reduced
cancer survival [47]. Similar results were obtained in the few mothers (n ¼ 50) who were diagnosed with cancer before their bereavement and in 268 mothers diagnosed after their loss. In our group of fathers, marked differences in HRs for cancer survival by SES groups were observed. In the Israeli study of parents whose offspring were killed in war, an increased risk of death was found among fathers whose cancer was diagnosed before bereavement [16]. Interestingly, this group of bereaved fathers was of a higher social class than the comparison group. Other studies have also suggested a possible link between social factors and cancer survival [48,49]. Both in mothers and fathers, family size tended to modify the cancer survival associated with bereavement. Again, this modification of cancer survival suggests the involvement of social support in modifying risk of mortality from cancer. In parents who lost their only child, grief is intertwined with loss of social support from close family members [50]. Finally, similar to the results on overall mortality, the risk of cancer mortality associated with bereavement was modified by time since the child loss. The increased parental HR for cancer mortality associated with bereavement was no longer significant after 9 years after child loss. In another study, no differences in HRs of cancer incidence and cancer mortality associated with bereavement by years of follow-up were observed [44]. Psychological stress associated with parental bereavement has been suggested to impact cancer initiation and progression through several pathways [1,51,52]. Because our findings suggest that bereaved parents who developed cancer were at higher risk for a rapid demise during a short period after the child loss, this risk probably arises from the decrease of immunologic defences, rather than because of changes in behavioral factors. It could also mean a delayed diagnosis among bereaved parents resulting in more detection of late stage cancer which is more likely to be fatal. Some advantages and limitations of the present study should be noted. We used register-based data of high quality abstracted from government and medical records, and all bereaved parents in the cohort were correctly enrolled into the exposed group. One of our study’s limitations relates to the restricted information available on the entire JPS cohort from birth notifications. Although we were able to adjust for several potential sociodemographic variables, it was impossible to control for basic health status and lifestyle characteristics of the parents. Finally, the large number of participants and the long follow-up provided us with sufficient outcome events to obtain reasonably precise statistical assessments of HR estimates and to examine possible heterogeneity by various characteristics. Conclusion In conclusion, this study provides evidence for slightly increased mortality in parents after bereavement. Specifically, an association was identified for mothers’ deaths from CHD and for both parents who died from circulatory causes. The effect of psychological stress related to bereavement on cancer incidence and survival from cancer is likely to be negligible, if any. The considerable increased mortality risk during a period of a few years after child loss merits further investigation to support the finding and elucidate the mechanisms underlying this association. References [1] Mcewen BS. Protective and damaging effects of stress mediators. Library 1998;338(3):171e9. [2] Cohen S, Janicki-Deverts D, Miller GE. Psychological stress and disease. JAMA 2007;298(14):1685e7.
L. Schorr et al. / Annals of Epidemiology 26 (2016) 115e121 [3] Stroebe M, Schut H, Stroebe W. Health outcomes of bereavement. Lancet 2007;370(9603):1960e73. [4] Hart CL, Hole DJ, Lawlor DA, Smith GD, Lever TF. Effect of conjugal bereavement on mortality of the bereaved spouse in participants of the Renfrew/ Paisley Study. J Epidemiol Community Health 2007;61(5):455e60. [5] Davies R. New understandings of parental grief: literature review. J Adv Nurs 2004;46(5):506e13. [6] American Psychiatric Association. Diagnostic and statistical manual of mental disorders, 5th edition (DSM-5). Washington, DC: American Psychiatric Association; 2013. [7] Levav I, Kransnoff L, Dohrenwend B. Israeli PERI life event scale: ratings of events by community sample. Isr J Med Sci 1981;17:176e83. [8] Dyregrov A. Parental reactions to the loss of an infant child: a review. Scand J Psychol 1990;31(4):266e80. [9] Rubin SS. The death of a child is forever: the life course impact of child loss. In: Handbook of bereavement: theory, research, and intervention. US: Cambridge University Press; 1993. p. 285e99 http://dx.doi.org/10.1017/CBO9780 511664076.020. [10] McEwen BS. Protective and damaging effects of stress mediators: the good and bad sides of the response to stress. Metabolism 2002;51(6 Suppl 1):2e4. [11] Vassilas CA, Morgan HG. Suicide in Avon. Life stress, alcohol misuse and use of services. Br J Psychiatry 1997;170:453e5. [12] Anda RF, Williamson DF, Escobedo LG, Mast EE, Giovino GA, Remington PL. Depression and the dynamics of smoking. A national perspective. JAMA 1990;264(12):1541e5. [13] Rehm J, Gmel G, Sempos CT, Trevisan M. Alcohol-related morbidity and mortality. Alcohol Res Health 2003;27(1):39e51. Available from: http://www. ncbi.nlm.nih.gov/pubmed/15301399. [14] Ezzati M, Lopez AD. Estimates of global mortality attributable to smoking in 2000. Lancet 2003;362(9387):847e52. [15] Levav I, Friedlander Y, Kark JD, Peritz E. An epidemiologic study of mortality among bereaved parents. N Engl J Med 1988;319(8):457e61. [16] Levav I, Kohn R, Iscovich J, Abramson JH, Tsai WY, Vigdorovich D. Cancer incidence and survival following bereavement. Am J Public Health 2000;90(10):1601e7. [17] Li J, Precht DH, Mortensen PB, Olsen J. Mortality in parents after death of a child in Denmark: a nationwide follow-up study. Lancet 2003; 361(9355):363e7. [18] Centers for Disease Control and Prevention. ICD title and code cross reference file. Available from: http://www.cdc.gov/nchs/data/statab/gmwki_98.pdf. Accessed December 7, 2015. [19] International Statistical Classification of Diseases and Related Health Problems 10th Revision, Volume 2, Second Edition. Available from: http://www.who.int/ classifications/icd/ICD-10_2nd_ed_volume2.pdf. Accessed December 7, 2015. [20] Calderon-Margalit R, Friedlander Y, Yanetz R, Deutsch L, Perrin MC, Kleinhaus K, et al. Preeclampsia and subsequent risk of cancer: update from the Jerusalem Perinatal Study. Am J Obstet Gynecol 2009;200(1):63.e1e5. [21] Higgins J, Thompson S. Quantifying heterogeneity in a meta-analysis. Stat Med 2002;21:1539e58. [22] Abramson JH. WINPEPI (PEPI-for-Windows): computer programs for epidemiologists. Epidemiol Perspect Innov 2004;1:6. [23] Rostila M, Saarela J, Kawachi I. Mortality in parents following the death of a child: a nationwide follow-up study from Sweden. J Epidemiol Community Health 2012;66(10):927e33. [24] Vance JC, Boyle FM, Najman JM, Thearle MJ. Gender differences in parental psychological distress following perinatal death or sudden infant death syndrome. Br J Psychiatry 1995;167(6):806e11. [25] Moriarty HJ, Carroll R, Cotroneo M. Differences in bereavement reactions within couples following death of a child. Res Nurs Health 1996;19(6):461e9. [26] Vance JC, Foster WJ, Najman JM, Embelton G, Thearle MJ, Hodgen FM. Early parental responses to sudden infant death, stillbirth or neonatal death. Med J Aust 1991;155(5):292e7.
121
[27] Rossi A, Rossi P. Of human bonding: parent-child relations across the life course. New York: Aldine de Gruyter; 1990. [28] Werthmann J, Smits LJM, Li J. Parental mortality rates in a western country after the death of a child: assessment of the role of the child’s sex. Gend Med 2010;7(1):39e46. [29] Raphael B. The anatomy of bereavement. New York: Basic Books; 1986. [30] Bolton JM, Au W, Walld R, Chateau D, Martens PJ, Leslie WD, et al. Parental bereavement after the death of an offspring in a motor vehicle collision: a population-based study. Am J Epidemiol 2014;179(2):177e85. [31] Russ TC, Stamatakis E, Hamer M, Starr JM, Kivimaki M, Batty GD. Association between psychological distress and mortality: individual participant pooled analysis of 10 prospective cohort studies. BMJ 2012;345:e4933. [32] Batty GD, Whitley E, Gale CR, Osborn D, Tynelius P, Rasmussen F. Impact of mental health problems on case fatality in male cancer patients. Br J Cancer 2012;106(11):1842e5. [33] Dior UP, Hochner H, Friedlander Y, Calderon-Margalit R, Jaffe D, Burger A, et al. Association between number of children and mortality of mothers: results of a 37-year follow-up study. Ann Epidemiol 2013;23(1): 13e8. [34] Martikainen P, Valkonen T. Do education and income buffer the effects of death of spouse on mortality? Epidemiology 1998;9(5):530e4. [35] Kaplan GA, Keil JE. Socioeconomic factors and cardiovascular disease: a review of the literature. Circulation 1993;88(4 Pt 1):1973e98. [36] Mittleman MA, Maclure M, Nachnani M, Sherwood JB, Muller JE. Educational attainment, anger, and the risk of triggering myocardial infarction onset. The Determinants of Myocardial Infarction Onset Study Investigators. Arch Intern Med 1997;157(7):769e75. [37] Li J. Myocardial infarction in parents who lost a child: a nationwide prospective cohort study in Denmark. Circulation 2002;106(13):1634e9. [38] Harlap S, Davies AM, Deutsch L, Calderon-Margalit R, Manor O, Paltiel O, et al. The Jerusalem Perinatal Study cohort, 1964-2005: methods and a review of the main results. Paediatr Perinat Epidemiol 2007;21(3):256e73. [39] Li J, Johansen C, Hansen D, Olsen J. Cancer incidence in parents who lost a child: a nationwide study in Denmark. Cancer 2002;95(10): 2237e42. [40] McGee R. Does stress cause cancer? BMJ 1999;319(7216):1015e6. [41] Kvikstad A, Vatten LJ. Cancer risk and prognosis in Norway: comparing women in their first marriage with women who have never married. J Epidemiol Community Health 1996;50(1):51e5. [42] Johansen C, Olsen JH. Psychological stress, cancer incidence and mortality from non-malignant diseases. Br J Cancer 1997;75(1):144e8. [43] Jacobs JR, Bovasso GB. Early and chronic stress and their relation to breast cancer. Psychol Med 2000;30(3):669e78. [44] Li J, Johansen C, Olsen J. Cancer survival in parents who lost a child: a nationwide study in Denmark. Br J Cancer 2003;88(11):1698e701. [45] Jamison RN, Burish TG, Wallston KA. Psychogenic factors in predicting survival of breast cancer patients. J Clin Oncol 1987;5(5):768e72. [46] Tross S, Herndon 2nd J, Korzun A, Kornblith AB, Cella DF, Holland JF, et al. Psychological symptoms and disease-free and overall survival in women with stage II breast cancer. J Natl Cancer Inst 1996;88(10):661e7. [47] Funch DP, Marshall J. The role of stress, social support and age in survival from breast cancer. J Psychosom Res 1983;27(1):77e83. [48] Ramirez AJ, Craig TK, Watson JP, Fentiman IS, North WR, Rubens RD. Stress and relapse of breast cancer. BMJ 1989;298(6669):291e3. [49] Forsen A. Psychosocial stress as a risk for breast cancer. Psychother Psychosom 1991;55(2-4):176e85. [50] Susser M. Widowhood: a situational life stress or stressful life event? Am J Public Health 1981;71:793e5. [51] Cohen S, Rabin BS. Psychologic stress, immunity, and cancer. J Natl Cancer Inst 1998;90(1):3e4. [52] Kiecolt-Glaser JK, Glaser R. Psychoneuroimmunology and cancer: fact or fiction? Eur J Cancer 1999;35(11):1603e7.
Contents lists available at ScienceDirect
Annals of Epidemiology journal homepage: www.annalsofepidemiology.org
Original article
Mortality, cancer incidence, and survival in parents after bereavement Limor Schorr RN, MSc a, b, Ayala Burger PhD b, Hagit Hochner PhD b, Ronit Calderon MD, MPH b, Orly Manor PhD b, Yechiel Friedlander PhD b, *, Gabriella M. Lawrence MPH b, Ora Paltiel MDCM, MSc b, c a b c
School of Nursing, HadassaheHebrew University, Jerusalem, Israel School of Public Health, HadassaheHebrew University, Jerusalem, Israel Hematology Department, Hadassah University Hospital, Jerusalem, Israel
a r t i c l e i n f o
a b s t r a c t
Article history: Received 15 June 2015 Accepted 20 December 2015 Available online 12 January 2016
Purpose: The study objective was to investigate whether child loss is related to mortality, cancer incidence, and cancer survival in parents. Methods: We used a population-based birth cohort (1964e1976) in Jerusalem and ascertained mortality (average follow-up of 39.1 years) and any cancer (average follow-up of 35.6 years) among parents who lost a child (2838 mothers and 2532 fathers) and among nonbereaved parents (38,212 mothers and 36,433 fathers). We also assessed mortality among parents with cancer. Time-dependent Cox models were used to estimate hazard ratios (HRs) with 95% confidence intervals (CIs). Results: Overall mortality rates among bereaved parents were modestly increased when compared with nonbereaved parents (HR ¼ 1.18, 95% CI: 1.05e1.32 in mothers; HR ¼ 1.10, 95% CI: 1.01e1.20 in fathers). Hazard models indicated a significant relationship between bereavement and deaths from coronary heart disease in mothers (HR ¼ 1.90, 95% CI: 1.23e2.95) and circulatory causes in both parents (HR ¼ 1.69; 95% CI: 1.22e2.34 in mothers and HR ¼ 1.25; 95% CI: 1.02e1.54 in fathers). Bereavement was not associated with parental risk of cancer disease and with survival from cancer. The association between bereavement and parental overall mortality was similar in the different parental sociodemographic characteristics. We observed a decrease in HRs for parental mortality associated with bereavement, with increasing time since the death of the child (HRs ¼ 9e10, 0e3 years; HRs ¼ 0.9e1.0, 9þ years; Pheterogeneity 3 1032). A similar decrease in HRs was observed for parental survival from cancer (HRs ¼ 6.7e8.7, 0e3 years; HRs ¼ 0.9e1.0, 9þ years). Conclusions: Our study suggests that child loss was associated with slightly increased risk of all-cause and circulatory mortality in parents but not with incidence of cancer and cancer survival. The considerable increased parental mortality during a short period after child loss support the involvement of pathways related to psychological stress. Ó 2016 Elsevier Inc. All rights reserved.
Keywords: Mortality Cancer Parental bereavement
Introduction Growing evidence suggests that chronic diseases and mortality are influenced by psychological stress [1,2]. Bereavement, an acute life event and its association with mortality and disease morbidity, has been studied in various populations, and a number of studies demonstrate excess mortality among widowers [3,4]. Child loss is a particularly stressful life event [5,6], even more so than the death of a spouse [7e9]. * Corresponding author. Unit of Epidemiology, Hebrew UniversityeHadassah Braun School of Public Health, POB 12272, Jerusalem 91120, Israel. Tel.: þ972-26777805; fax: þ972-2-6431086. E-mail address: [email protected] (Y. Friedlander). http://dx.doi.org/10.1016/j.annepidem.2015.12.008 1047-2797/Ó 2016 Elsevier Inc. All rights reserved.
Stressful events affect various physiological systems, such as the sympathetic nervous system, the hypothalamic-pituitary-adrenal axis, the neuroendocrine systems, and the immune systems [10]; we, therefore, predict that child loss may be associated with diseases such as cardiovascular and cancer and/or mortality in parents. Stressful life events raise the risk of various psychological disorders, which may lead to adverse lifestyle behaviors [11,12] and affect unnatural and natural deaths [13,14]. A study in Israel examined mortality in a group of bereaved parents of soldiers 18e40 years of age who were killed at war in 1973 and parents of offspring 18e30 years of age who died in accidents compared with the general population [15]. The 10-year follow-up indicated no excess mortality among bereaved parents;
116
L. Schorr et al. / Annals of Epidemiology 26 (2016) 115e121
however, when the follow-up was extended to 20 years, the findings showed an effect of stress on incidence of malignancies for selected sites and accelerated demise among parents bereaved after a cancer diagnosis, but not among those bereaved before such a diagnosis [16]. Other investigations, however, indicate an overall association between child loss and mortality in parents [17]. The Jerusalem Perinatal Study (JPS) provides a unique opportunity to explore the relation between child loss and all-causee and cause-specific mortality, cancer morbidity, and survival in parents. Methods The JPS is a prospective population-based birth cohort of 92,408 infants born (1964e1976) to residents of Jerusalem and their parents. Through the Israeli Population Registry (IPR), we verified identities of 41,050 mothers and 38,965 fathers (96.2% and 92.5%, respectively) in the original cohort and updated their vital status. We undertook a long follow-up study to determine the association between parental bereavement and mortality and cancer morbidity and survival. Non-Jewish parents, comprising 2.3% of the cohort, were included in analyses. Study variables Data included demographic and socioeconomic characteristics of parents and information on offspring including sex and birth date. Information was abstracted from birth certificates or maternity ward logbooks. Primary outcomes in parents were risk of all-cause mortality and cause-specific mortality, including mortality due to coronary heart disease (CHD), circulatory disorders, cancer, unnatural and other causes, incidence of any cancer and mortality in parents with cancer diagnosis. The underlying cause of death was categorized according to International Classification of Diseases, Ninth Revision (ICD-9), codes until 1997 and in 1998 according to ICD-10 codes [18,19]. Causes of deaths were categorized into the following groups: CHD (ICD-9 codes: 410e414, 427.4, and 427.5; ICD-10 codes: I20eI25, I46, and I49); all circulatory conditions (ICD-9 codes: 390e459; ICD-10 codes: I0eI99); all neoplasms (ICD-9: 140e239; ICD-10: C0eC99 and D0eD48); death due to unnatural causes (ICD-9: 80e99; ICD-10: S0eT88, V0eY99); and death due to all other causes taken together. In Israel, nearly 100% of deaths are reported to the IPR, along with the date and cause of death. Linkage to the IPR occurred in 2011 at which point the registry had been fully updated through December 2010 for 89.6% of the mothers and 88.4% of the fathers. The remaining parents had updated information through April 30, 2005. Cancer incidence was ascertained via the Israel National Cancer Registry which was established in 1961 [20]. Linkage occurred in 2010 at which point the registry had been updated through July 31, 2008. For parents diagnosed with cancer, mortality was assessed via the IPR, and they were determined to have died from cancer regardless of the actual cause of death, that is, “cancer mortality.” The following parental characteristics were included in the analyses: parents’ ethnic origindclassified according to the parents’ country of birth, or if born in Israel, that of the grandfather’s birth country (categorized as Israel, other West Asia, North Africa, Europe or America or other industrialized countries, and non-Jewish), a social class scale (socioeconomic status [SES]: from 1 ¼ highest to 3 ¼ lowest) based on father’s occupation, parity (number of children born before 1964, as reported by the mother, plus those born within the cohort [categorized as 1, 2, and 3]), having a stillbirth in the family (1, yes; 0, no), parents’ age at child’s birth (<30, 30e39,
40 years), parents’ age at child’s death (<30, 30 years), and parents’ education level (0e4, 5e8, 9e12, 13, unknown). In addition, we examined the following characteristics related to the deceased child: number of deceased children in the family (1, 2), child’s age at death (<1, 1e17, 18 years), and time passed since child loss (0e3, 4e8, 9 years). Statistical analysis Time was counted from birth of the deceased child or earliest birth in the cohort when parents had multiple live offspring until death of parent or study closure (December 31, 2010 or April 30, 2005). The Cox proportional hazards regression model, with child’s death as a time-dependent exposure variable, was used to evaluate associations between this life event and parental mortality after controlling for possible confounders. If a child died during the study period, the surviving parent(s) changed status from nonexposed (nonbereaved) to exposed (bereaved). All parents, even those who died before the child loss, were included in the mortality analysis. Hazard ratios (HRs) and 95% confidence intervals (CIs) were calculated, and competing hypotheses were tested by the likelihood ratio test. For the investigation of cancer incidence, parents with cancer diagnosed within 5 years after child loss were eliminated from the analysis. However, we performed a series of sensitivity analyses to test the robustness of our cancer results: (1) by including parents with cancer diagnosed 3 or more years after bereavement and (2) by including parents diagnosed 1 year after bereavement. We also compared survival in parents whose cancer was diagnosed before their loss (and survived at least till after the child loss) and those whose cancer was diagnosed after their loss. Adjustments for parents’ education level, parents’ age at child’s birth and death, and child’s age at death in all models were also assessed as continuous variables and yielded HRs similar to those obtained using age and education as categorical variables. Analyses were stratified by characteristics of parents and the deceased child, which may modify parents’ ability to cope with the acute life event. Owing to sample size limitations, stratification was not carried out on cause-specific mortality. We used heterogeneity tests for stratified data to test for interactions between the exposure and modifier variables. The significance of effect modification was assessed using Higgins and Thompson’s heterogeneity index [21] which under the null hypothesis is distributed as c2(I1) where I indicates the number of strata. Most analyses were completed using Stata, version 12.0 (StataCorp, College Station, TX), and heterogeneity tests were calculated using WinPepi, version 11.42 (J.H. Abramson, Jerusalem, Israel) [22]. All tests were two-sided, and an alpha set at 0.05. This study was approved by the Institutional Review Boards of HadassaheHebrew University Medical Center in Jerusalem. Results Table 1 shows baseline characteristics of bereaved (2838 mothers and 2532 fathers) and nonbereaved (38,212 mothers and 36,433 fathers) parents. There was no relation between bereavement and parents’ age at birth. About 77% of mothers and 60% of fathers gave birth before the age of 30 years. Compared with the unexposed group, exposed Jewish parents tended to be of Asian and North African origin. Exposed parents were less educated, exhibited lower SES, and have more children. About 9.5% of parents were exposed to death of two children or more, and more than 50% of the deceased children died during their first year of life. During the follow-up, mothers contributed 1,608,772 personyears (average: 39.1 person-years) for overall mortality and 1,451,323 person-years (average: 35.6 person-years) for cancer
L. Schorr et al. / Annals of Epidemiology 26 (2016) 115e121 Table 1 Baseline characteristics of parents Characteristic
Mothers
Fathers
Nonbereaved
Bereaved
Nonbereaved
Bereaved
n ¼ 38,212
n ¼ 2838
n ¼ 36,433
n ¼ 2532
59.9 30.2 9.9
59.2 32.3 8.4
14.2 29.1 19.7 35.7 1.2
14.2 34.3 22.3 26.1 3.1
3.8 18.3 36.6 38.3
6.6 23.3 36.9 31.5
32.7 38.4 28.9
23.6 35.5 40.9
21.6 25.9 52.5
5.1 14.6 80.3
d d
90.3 9.7
d d d
54.3 17.3 28.4
Age at first birth (%) <30 76.5 77.8 30e39 21.4 20.7 40 2.1 1.5 Origin (%) Israel 13.6 12.1 Asia 27.4 32.1 North Africa 20.7 24.8 Europe or West 37.0 27.8 Non-Jews 1.3 3.1 Education (%) 0e4 y 7.1 11.4 5e8 y 19.4 28.0 9e12 y 37.0 33.2 13 y 33.5 25.5 SES (%) High 38.2 32.5 Middle 38.5 38.4 Low 23.2 29.1 No. of children in family (%) 1 22.7 7.4 2 25.7 15.3 3 51.5 77.4 No. of deceased children per family (%) 1 d 90.6 2 d 9.4 Age of the deceased child (%) <1 y d 54.7 1e18 y d 16.3 19 y d 29.0
morbidity. During this period, 4359 mothers (10.6%) died, and 5123 (12.5%) developed cancer. Mothers who developed cancer contributed 184,998 person-years of follow-up (average: 36.1 person-years), and 2032 of the 5123 diagnosed mothers (39.7%) died during follow-up. The 38,965 fathers contributed 1,350,576 person-years (average: 34.7 person-years) of mortality follow-up and 1,316,350 personyears (average: 34.1 person-years) of cancer follow-up, with 8750 (22.5%) and 5146 (13.2%) deaths and cancer events, respectively. Fathers who developed cancer contributed 146,604 person-years of follow-up (average: 28.5 person-years), and 2767 of the 5146 diagnosed fathers (53.8%) died during follow-up. Table 2 presents HRs among parents for all-cause and causespecific mortality associated with child loss. The adjusted mortality rate among bereaved parents was modestly yet significantly higher (HR ¼ 1.10e1.18) when compared with nonbereaved parents (P .03). In addition, the adjusted mortality rate due to CHD and circulatory disorders among bereaved mothers was significantly higher when compared with nonbereaved mothers (HR ¼ 1.90 and HR ¼ 1.69, respectively). Among fathers, only adjusted mortality
117
rates due to circulatory disorders were significantly associated with bereavement (HR ¼ 1.25). Table 3 presents HRs among parents for overall mortality according to characteristics that may affect ability to cope with acute stress. Parent’s age at entry and at child’s death did not modify the hazard associated with child loss. The analyses by categories of education and SES provided no evidence of heterogeneity in the associations. Non-Jewish mothers and Jewish mothers born in Israel showed higher HRs for overall mortality (P for heterogeneity ¼ .038) as compared to mothers from other origin groups. We noted a decrease in HRs for total mortality by parity, and mothers who lost two children had an HR of 1.58 for overall mortality (P for heterogeneity ¼ .027). We noted a significantly increased hazard (9.0- to 10.2-fold HR) for overall mortality in parents, within a short period (0e3 years) after child loss. Our analysis indicated a significant decrease in hazard with increasing time since the child loss (HRs ¼ 3.2e3.3 within a period of 4e8 years and no increased risk [HRs ¼ 0.9e1.1] after a longer period; P for heterogeneity 3 1032 to <1033). Table 4 presents HRs among parents for cancer morbidity and mortality after cancer diagnosis associated with child loss. No association was observed between offspring’s death and parent’s cancer morbidity (HR ¼ 0.89e0.98). Similar results were obtained when parents with cancer diagnosed 3 or more years after bereavement were included (HR ¼ 0.93e0.99, data not shown), as well as when parents with cancer diagnosed 1 year after bereavement were included (HR ¼ 0.96e1.04, data not shown). Bereavement was not associated with the risk of mother’s survival from cancer (HR ¼ 1.00) and similar results were obtained in mothers diagnosed with cancer before their loss (HR ¼ 0.85, 95% CI ¼ 0.44, 1.65, data not shown) and after their loss (HR ¼ 1.00, 95% CI ¼ 0.83, 1.21, data not shown). Among fathers, all diagnosed with cancer after their child loss, the HR for mortality was modestly increased (HR ¼ 1.15), though not statistically significant. Table 4 also presents HRs among parents for cancer incidence and cancer survival according to characteristics of parents and the deceased child. No differences were observed in HRs by parent’s age at entry. With respect to cancer incidence in mother and cancer survival in fathers, the adjusted HR tended to be higher among parents who were younger than 30 years at their child’s death compared with those who were older. In mothers, analyses by education and SES level provided no evidence of heterogeneity in associations. Our data suggested higher HRs for cancer mortality in fathers belonging to the highest education and SES groups. Non-Jewish mothers and Jewish mothers born in Israel showed higher HRs for mortality from cancer (P for heterogeneity ¼ .02) as compared to mothers from other origin groups. Mortality from cancer tended to have the strongest association with bereavement in fathers who were born in Europe/West countries. We also noted a decrease in HRs for mortality from cancer, with increased number of children in the family. Fathers who lost two children had an HR of 1.25 for cancer incidence and 0.85 among fathers who lost one child (P for heterogeneity ¼ .023).
Table 2 Associations* between death of a child and overall mortality and cause-specific mortality in parents Cause of death
Mothers HR
95% CI
P value
HR
95% CI
P value
All-cause mortality Coronary heart disease Circulatory disorders Cancer Unnatural causes Any other cause
1.18 1.90 1.69 1.16 1.39 1.12
1.05e1.32 1.23e2.95 1.22e2.34 0.86e1.57 0.64e3.05 0.98e1.28
.004 .004 .002 .34 .41 .08
1.10 1.12 1.25 1.23 0.94 1.06
1.01e1.20 0.86e1.47 1.02e1.54 0.94e1.61 0.48e1.85 0.96e1.17
.03 .39 .03 .14 .86 .22
*
Fathers
Models adjusted for parental age at entry, education, SES, ethnic origin, family size, and stillbirth in family.
118
L. Schorr et al. / Annals of Epidemiology 26 (2016) 115e121
Table 3 Associations* between death of a child and overall mortality in parents, according to basic characteristics Characteristic
Mothers HR
Total group 1.18 Parental age at entry <30 1.12 30e39 1.08 40 1.38 Test for heterogeneity d Age of parent when child died <30 1.27 30 1.14 Test for heterogeneity d Parental education 0e4 1.12 5e8 1.16 9e12 1.14 13 1.40 Test for heterogeneity d Parental SES High 1.18 Middle 1.30 Low 1.09 Test for heterogeneity d Parental origin Israel 1.47 Asia 1.24 North Africa 0.89 Europe 1.32 Non-Jews 1.61 Test for heterogeneity d Family size 1 child 1.59 2 children 1.65 3 children 1.15 Test for heterogeneity d No. of children who died 1 child 1.12 2 children 1.58 Test for heterogeneity d Age of child at death, y <1 1.22 1e18 0.96 19 1.27 Test for heterogeneity d Length of time from death of child, y 0e3 10.24 4e8 3.33 9 1.05 Test for heterogeneity d
Fathers 95% CI
P value
HR
95% CI
P value
1.05e1.32
.004
1.10
1.01e1.20
.02
0.96e1.31 0.91e1.29 0.90e2.12 c2(2) ¼ 1.08
.16 .37 .14 .583
1.01 1.10 1.11 d
0.87e1.18 0.97e1.25 0.93e1.32 c2(2) ¼ 0.90
.86 .13 .24 .639
1.04e1.56 1.00e1.30 c2(1) ¼ 0.77
.02 .05 .381
1.04 1.11 d
0.84e1.30 1.02e1.22 c2(1) ¼ 0.29
.71 .02 .589
0.89e1.41 0.95e1.42 0.90e1.44 1.07e1.84 c2(3) ¼ 1.83
.32 .14 .27 .01 .608
1.22 0.97 1.10 1.19 d
0.98e1.53 0.83e1.13 0.95e1.27 0.99e1.43 c2(3) ¼ 4.07
.08 .67 .21 .06 .255
0.94e1.48 1.07e1.56 0.91e1.30 c2(2) ¼ 1.77
.16 .007 .36 .412
1.33 1.06 1.06 d
1.08e1.64 0.91e1.23 0.94e1.19 c2(2) ¼ 3.79
.01 .45 .35 .151
1.08e2.01 1.02e1.50 0.71e1.12 1.04e1.67 0.95e2.74 c2(4) ¼ 10.16
.02 .03 .31 .02 .08 .038
1.07 1.10 1.15 1.13 0.76 d
0.84e1.37 0.96e1.26 0.98e1.35 0.94e1.36 0.43e1.35 c2(4) ¼ 2.01
.59 .18 .09 .19 .35 .734
0.96e2.62 1.18e2.33 1.02e1.30 c2(2) ¼ 5.00
.07 .004 .02 .082
1.47 1.23 1.10 d
0.85e2.55 0.91e1.65 1.00e1.20 c2(2) ¼ 1.47
.17 .18 .04 .479
0.99e1.27 1.20e2.09 c2(1) ¼ 4.92
.06 .001 .027
1.10 1.07 d
1.01e1.21 0.84e1.37 c2(1) ¼ 0.04
.03 .57 .835
1.06e1.40 0.73e1.27 1.00e1.62 c2(2) ¼ 2.71
.004 .79 .05 .258
1.09 1.07 1.22 d
0.98e1.21 0.89e1.29 1.02e1.46 c2(2) ¼ 1.33
.12 .46 .03 .514
6.99e15.00 2.24e4.94 0.93e1.18 c2(2) ¼ 145.17
3.7 1033 2.47 109 .46 3.0 1032
8.99 3.15 0.94 d
6.92e11.69 2.44e4.06 0.86e1.03 c2(2) ¼ 304.99
<1033 6.3 1019 .20 <1033
* Models adjusted for maternal age at entry, parental education, SES, ethnic origin, family size, and stillbirth in family. In any stratified analysis, the model was adjusted for other covariables (e.g., in stratified models on parental age groups, we excluded from the Cox model the stratified variabledparental age).
With respect to survival from cancer, mothers who lost two children had an HR of 1.45 for mortality and 0.94 among mothers who lost one child (P for heterogeneity ¼ .08), and the adjusted HR was higher among mothers whose child died at the age of 18 years or more compared with those whose child died before the age of 18 years (P for heterogeneity ¼ .05). Similar to overall mortality, we observed a significant heterogeneity in HRs for cancer survival by length of time since the child loss (P for heterogeneity 9.5 1011e2.4 1017). Because incidence of cancer did not include diagnosed cases within 5 years after child loss, we did not carry out this heterogeneity test for cancer morbidity. Discussion Overall and cause-specific mortality Results from our study (1964e2010) show a modest increased overall mortality in parents associated with child loss, yet a hugely
increased risk of death within the first 3 years after bereavement in both fathers and mothers. In addition, increased mortality due to CHD is seen among bereaved mothers, and circulation-related mortality was greater among bereaved parents compared with those who did not lose a child. This study joins a growing body of evidence showing an association between bereavement and parental morbidity and mortality. For example, a study in Denmark showed a significantly increased overall mortality rate in bereaved mothers (HR ¼ 1.43) and modest excess risk in bereaved fathers (HR ¼ 1.09) [17]. In a 22year follow-up in the Swedish population, authors found increased mortality risk (relative risk ¼ 1.31) among mothers after the death of a young offspring (ages 10e17 years) and attenuated risk (relative risk ¼ 1.15) among fathers [23]. Our results also suggest that excess mortality associated with bereavement in mothers is greater than that in fathers. Gender differences in the ability to cope with stress [24] leading to more adverse health among mothers than fathers after child loss [17,25,26], along with stronger relationships
L. Schorr et al. / Annals of Epidemiology 26 (2016) 115e121
119
Table 4 Associations* between death of a child and cancer morbidity and cancer survival in parents, according to basic characteristics Characteristic
Mothers
Fathers
Cancer morbidity HR
95% CI
Total group 0.98 0.86e1.11 Parental age at entry <30 1.00 0.86e1.16 30e39 0.84 0.66e1.07 40 0.48 0.15e1.52 Test for d c2(2) ¼ 2.78 heterogeneity Age of parent when child died <30 1.15 0.96e1.38 30 0.87 0.74e1.03 Test for d c2(1) ¼ 4.96 heterogeneity Parental education 0e4 0.75 0.50e1.11 5e9 1.00 0.79e1.26 9e12 0.95 0.76e1.19 13 1.11 0.88e1.39 c2(3) ¼ 2.96 Test for d heterogeneity Parental SES High 0.95 0.77e1.18 Middle 1.02 0.83e1.24 Low 0.95 0.75e1.20 Test for d c2(2) ¼ 0.30 heterogeneity Parental origin Israel 0.88 0.61e1.27 Asia 0.86 0.68e1.08 North Africa 1.14 0.89e1.45 Europe 1.00 0.80e1.25 Non-Jews 1.47 0.62e3.51 Test for d c2(4) ¼ 3.90 heterogeneity Family size 1 child 0.69 0.38e1.25 2 children 1.00 0.71e1.42 3 children 0.95 0.83e1.09 Test for d c2(2) ¼ 1.18 heterogeneity No. of children who died 1 child 0.96 0.84e1.09 2 children 1.15 0.82e1.61 Test for d c2(1) ¼ 0.96 heterogeneity Age of child at death, y <1 1.08 0.93e1.24 1e18 0.92 0.69e1.23 19 0.67 0.46e0.97 Test for d c2(2) ¼ 5.85 heterogeneity Length of time from death of child, y 0e3 d d 4e8 5þ y d 9 0.98 0.86e1.11 Test for d d heterogeneity
Cancer survival P value
Cancer morbidity
Cancer survival
HR
95% CI
P value
HR
95% CI
P value
HR
95% CI
P value
.71
1.00
0.83e1.20
1.00
0.89
0.79e1.01
.08
1.15
0.99e1.35
.07
.99 .15 .21 .249
1.03 0.96 0.54 d
0.82e1.30 0.71e1.30 0.11e2.54 c2(2) ¼ 0.72
.82 .79 .43 .697
0.88 0.84 1.00 d
0.72e1.07 0.69e1.02 0.72e1.37 c2(2) ¼ 0.82
.19 .08 .98 .662
1.19 1.06 1.14 d
0.91e1.57 0.84e1.35 0.82e1.58 c2(2) ¼ 0.41
.20 .61 .44 .815
.13 .10 .026
0.99 1.00 d
0.73e1.35 0.80e1.24 c2(1) ¼ 0.00
.97 .99 .958
0.77 0.92 d
0.58e1.04 0.81e1.06 c2(1) ¼ 1.18
.09 .27 .278
1.69 1.09 d
1.14e2.49 0.92e1.29 c2(1) ¼ 4.08
.01 .30 .043
.15 .98 .64 .39 .398
0.98 0.96 0.81 1.18 d
0.6e1.61 0.69e1.34 0.56e1.17 0.83e1.67 c2(3) ¼ 2.13
.93 .82 .26 .35 .547
1.23 0.79 0.88 0.90 d
0.82e1.85 0.61e1.01 0.71e1.09 0.71e1.15 c2(3) ¼ 3.31
.31 .06 .25 .41 .347
1.24 0.92 1.14 1.37 d
0.79e1.93 0.68e1.26 0.86e1.50 1.01e1.87 c2(3) ¼ 3.36
.35 .62 .37 .04 .339
.66 .86 .66 .862
0.90 0.98 1.09 d
0.64e1.26 0.73e1.33 0.79e1.50 c2(2) ¼ 0.65
.53 .91 .61 .721
0.79 0.88 0.94 d
0.59e1.06 0.72e1.09 0.78e1.13 c2(2) ¼ 0.98
.12 .26 .50 .612
1.47 1.43 0.94 d
1.02e2.14 1.09e1.87 0.75e1.17 c2(2) ¼ 7.30
.04 .01 .58 .026
.49 .20 .30 .98 .39 .420
1.58 1.16 0.60 1.03 2.10 d
0.98e2.53 0.84e1.61 0.40e0.91 0.74e1.44 0.65e6.81 c2(4) ¼ 11.68
.06 .37 .02 .84 .22 .02
0.85 0.99 0.90 0.82 0.39 d
0.59e1.22 0.80e1.22 0.70e1.17 0.64e1.05 0.11e1.36 c2(4) ¼ 3.11
.37 .91 .43 .11 .14 .539
0.90 1.15 1.02 1.64 0.27 d
0.54e1.50 0.89e1.49 0.76e1.37 1.20e2.24 0.03e2.17 c2(4) ¼ 8.16
.69 .29 .91 .0008 .22 .086
.22 1.00 .47 .555
2.08 1.42 0.96 d
0.97e4.43 0.89e2.29 0.78e1.17 c2(2) ¼ 5.41
.06 .15 .66 .067
0.92 0.68 0.89 d
0.44e1.94 0.43e1.08 0.77e1.01 c2(2) ¼ 1.23
.83 .10 .07 .541
2.99 2.00 1.10 d
1.09e8.15 1.22e3.28 0.94e1.30 c2(2) ¼ 8.36
.03 .01 .24 .015
.50 .42 .328
0.94 1.45 d
0.77e1.14 0.93e2.26 c2(1) ¼ 3.06
.54 .10 .08
0.85 1.25 d
0.74e0.97 0.92e1.69 c2(1) ¼ 5.16
.02 .16 .023
1.20 0.89 d
1.02e1.42 0.59e1.34 c2(1) ¼ 1.75
.03 .56 .185
.33 .58 .03 .054
0.89 0.94 1.50 d
0.70e1.12 0.61e1.45 1.05e2.13 c2(2) ¼ 5.99
.31 .79 .03 .05
0.96 0.82 0.76 d
0.83e1.13 0.62e1.10 0.56e1.03 c2(2) ¼ 2.26
.65 .19 .08 .324
1.15 1.12 1.20 d
0.95e1.40 0.80e1.58 0.85e1.69 c2(2) ¼ 0.08
.15 .51 .30 .961
6.73 2.53 0.88 d
3.70e12.25 1.31e4.88 0.73e1.08 c2(2) ¼ 46.16
d 5þ y 0.89 d
d d 0.79e1.01 d
d d .08 d
8.76 2.13 0.99 d
5.48e13.99 1.32e3.45 0.83e1.18 c2(2) ¼ 76.56
5.4 1020 .002 .92 2.4 1017
d d .71 d
4.2 1010 .006 .22 9.5 1011
* Models adjusted for parental age at entry, parental education, SES, ethnic origin, family size, and stillbirth in family. In any stratified analysis, the model was adjusted for other covariables (e.g., in stratified models on parental age groups, we excluded from the Cox model the stratified variabledparental age).
between children and their mothers than their fathers [27], may explain the stronger adverse impact on mother’s mortality risk. Other studies, however, showed no association between child loss and mortality in parents [15,28]. Although, in the Israeli Study [15], parents were exposed to a very stressful life event which did not allow for an attenuating effect of anticipatory grief (child killed in action or in accident) [29], it is possible that the null finding is a result of a short duration of follow-up and nonoptimal selection of comparison group.
Similar to other studies (e.g., references [9,17,23]), we examined whether parental characteristics or those related to the deceased child modify the risk of overall mortality in parents associated with bereavement. The risk of mortality was clearly modified by time elapsed since the child loss, with strong increased hazard of parental mortality (HR ¼ 40.6 for mothers and HR ¼ 30.0 for fathers, data not shown) during the first year and no increased risk (HRs ¼ 0.8e1.05) after 9 years. Similarly, in another study, excess mortality risk was evident until approximately 9 years after child
120
L. Schorr et al. / Annals of Epidemiology 26 (2016) 115e121
loss [17]. Studies indicate that increased mortality due to loss of a close family member is divided into two groups: immediate effects, identifying bereavement as a grief-producing agent (e.g., stress, anxiety, depression), and late effects, suggesting that bereavement leads to disorganized life patterns and changes in lifestyle behaviors. Our findings support the role of short-period effects. It is possible that the psychological impact of bereavement leads to short-term mental illness and other mental/behavioral changes. A recent study found that during the 2 years after child-loss in a collision, bereaved parents had significant increase in risk of depression, anxiety disorders, marital breakup, and physician visits for mental illness as compared with nonbereaved parents [30]. Mental illness has been found to be associated with several specific causes of death, including suicide, cardiovascular, and cancer and with increased all-cause mortality rates [31,32]. In our study, the death of a child was associated with CHD/circulatory causes of death but not with an increased risk of unnatural causes. Studies also suggest that the effect of a child’s death may be modified by age of parent or the deceased child [9,17]. For example, Rostila et al. [23] found that the loss of an older child had a lower effect on parental mortality even during the first few years. The loss of a young child, strongly conflicting with life-cycle expectations, may expose parents to intense stress, leading to unusual psychophysiological states and abnormal health effects. In addition, many adult children have their own families, and their death may lead to stronger bonds between the grieving parent and the nuclear family of the deceased child, alleviating the negative effects after the death of the adult child. However, our study did not support this heterogeneity in bereavement effect by parental age or the child’s age at death. Family size also tends to modify the risk associated with bereavement [17]. In our study, parents who lost their only child showed the highest HR for overall mortality and those having three children and more had the lowest mortality HR. The Israeli population is characterized by high percent of families with three children and more [33]. Although the status of a bereaved parent never ceases, the effect of grief may differ in large families where parents’ grief is not intertwined with loss of social support, and where the presence of other remaining children provides motivation or incentive for parents to be mindful of their health and safety. Some studies have suggested that socioeconomic variables could also modify the ability of bereaved parents to cope with stress [34e36], yet other studies did not support this [17,37]. In this group of parents, marked differences in education and SES were observed by the different ethnic groups. Jewish mothers born in countries of North Africa and probably belonging to lower SES and education groups [38] tend to show the lowest HR associated with child loss. One possible explanation is related to differences in family size between parents born in countries of Western Asia or North Africa and those born in Israel or Europe [38]. Although, we were able to adjust for several potential confounders such as education, SES, ethnic origin, and family size, the risk of death associated with bereavement could be modified by ethnic groups because of differences in socioeconomic characteristics and family size. Cancer incidence and survival Our results, indicating no association between child loss and cancer morbidity, are in line with other studies [16,39e43]. For example, the cohort of Israeli parents, who lost an adult son in war or a car accident, showed no significant association between bereavement and incidence of any cancer [16]. Similar to other studies [16,41,44e46], we also failed to show any association between bereavement and cancer survival in parents, although it has been suggested that psychological stress may lead to reduced
cancer survival [47]. Similar results were obtained in the few mothers (n ¼ 50) who were diagnosed with cancer before their bereavement and in 268 mothers diagnosed after their loss. In our group of fathers, marked differences in HRs for cancer survival by SES groups were observed. In the Israeli study of parents whose offspring were killed in war, an increased risk of death was found among fathers whose cancer was diagnosed before bereavement [16]. Interestingly, this group of bereaved fathers was of a higher social class than the comparison group. Other studies have also suggested a possible link between social factors and cancer survival [48,49]. Both in mothers and fathers, family size tended to modify the cancer survival associated with bereavement. Again, this modification of cancer survival suggests the involvement of social support in modifying risk of mortality from cancer. In parents who lost their only child, grief is intertwined with loss of social support from close family members [50]. Finally, similar to the results on overall mortality, the risk of cancer mortality associated with bereavement was modified by time since the child loss. The increased parental HR for cancer mortality associated with bereavement was no longer significant after 9 years after child loss. In another study, no differences in HRs of cancer incidence and cancer mortality associated with bereavement by years of follow-up were observed [44]. Psychological stress associated with parental bereavement has been suggested to impact cancer initiation and progression through several pathways [1,51,52]. Because our findings suggest that bereaved parents who developed cancer were at higher risk for a rapid demise during a short period after the child loss, this risk probably arises from the decrease of immunologic defences, rather than because of changes in behavioral factors. It could also mean a delayed diagnosis among bereaved parents resulting in more detection of late stage cancer which is more likely to be fatal. Some advantages and limitations of the present study should be noted. We used register-based data of high quality abstracted from government and medical records, and all bereaved parents in the cohort were correctly enrolled into the exposed group. One of our study’s limitations relates to the restricted information available on the entire JPS cohort from birth notifications. Although we were able to adjust for several potential sociodemographic variables, it was impossible to control for basic health status and lifestyle characteristics of the parents. Finally, the large number of participants and the long follow-up provided us with sufficient outcome events to obtain reasonably precise statistical assessments of HR estimates and to examine possible heterogeneity by various characteristics. Conclusion In conclusion, this study provides evidence for slightly increased mortality in parents after bereavement. Specifically, an association was identified for mothers’ deaths from CHD and for both parents who died from circulatory causes. The effect of psychological stress related to bereavement on cancer incidence and survival from cancer is likely to be negligible, if any. The considerable increased mortality risk during a period of a few years after child loss merits further investigation to support the finding and elucidate the mechanisms underlying this association. References [1] Mcewen BS. Protective and damaging effects of stress mediators. Library 1998;338(3):171e9. [2] Cohen S, Janicki-Deverts D, Miller GE. Psychological stress and disease. JAMA 2007;298(14):1685e7.
L. Schorr et al. / Annals of Epidemiology 26 (2016) 115e121 [3] Stroebe M, Schut H, Stroebe W. Health outcomes of bereavement. Lancet 2007;370(9603):1960e73. [4] Hart CL, Hole DJ, Lawlor DA, Smith GD, Lever TF. Effect of conjugal bereavement on mortality of the bereaved spouse in participants of the Renfrew/ Paisley Study. J Epidemiol Community Health 2007;61(5):455e60. [5] Davies R. New understandings of parental grief: literature review. J Adv Nurs 2004;46(5):506e13. [6] American Psychiatric Association. Diagnostic and statistical manual of mental disorders, 5th edition (DSM-5). Washington, DC: American Psychiatric Association; 2013. [7] Levav I, Kransnoff L, Dohrenwend B. Israeli PERI life event scale: ratings of events by community sample. Isr J Med Sci 1981;17:176e83. [8] Dyregrov A. Parental reactions to the loss of an infant child: a review. Scand J Psychol 1990;31(4):266e80. [9] Rubin SS. The death of a child is forever: the life course impact of child loss. In: Handbook of bereavement: theory, research, and intervention. US: Cambridge University Press; 1993. p. 285e99 http://dx.doi.org/10.1017/CBO9780 511664076.020. [10] McEwen BS. Protective and damaging effects of stress mediators: the good and bad sides of the response to stress. Metabolism 2002;51(6 Suppl 1):2e4. [11] Vassilas CA, Morgan HG. Suicide in Avon. Life stress, alcohol misuse and use of services. Br J Psychiatry 1997;170:453e5. [12] Anda RF, Williamson DF, Escobedo LG, Mast EE, Giovino GA, Remington PL. Depression and the dynamics of smoking. A national perspective. JAMA 1990;264(12):1541e5. [13] Rehm J, Gmel G, Sempos CT, Trevisan M. Alcohol-related morbidity and mortality. Alcohol Res Health 2003;27(1):39e51. Available from: http://www. ncbi.nlm.nih.gov/pubmed/15301399. [14] Ezzati M, Lopez AD. Estimates of global mortality attributable to smoking in 2000. Lancet 2003;362(9387):847e52. [15] Levav I, Friedlander Y, Kark JD, Peritz E. An epidemiologic study of mortality among bereaved parents. N Engl J Med 1988;319(8):457e61. [16] Levav I, Kohn R, Iscovich J, Abramson JH, Tsai WY, Vigdorovich D. Cancer incidence and survival following bereavement. Am J Public Health 2000;90(10):1601e7. [17] Li J, Precht DH, Mortensen PB, Olsen J. Mortality in parents after death of a child in Denmark: a nationwide follow-up study. Lancet 2003; 361(9355):363e7. [18] Centers for Disease Control and Prevention. ICD title and code cross reference file. Available from: http://www.cdc.gov/nchs/data/statab/gmwki_98.pdf. Accessed December 7, 2015. [19] International Statistical Classification of Diseases and Related Health Problems 10th Revision, Volume 2, Second Edition. Available from: http://www.who.int/ classifications/icd/ICD-10_2nd_ed_volume2.pdf. Accessed December 7, 2015. [20] Calderon-Margalit R, Friedlander Y, Yanetz R, Deutsch L, Perrin MC, Kleinhaus K, et al. Preeclampsia and subsequent risk of cancer: update from the Jerusalem Perinatal Study. Am J Obstet Gynecol 2009;200(1):63.e1e5. [21] Higgins J, Thompson S. Quantifying heterogeneity in a meta-analysis. Stat Med 2002;21:1539e58. [22] Abramson JH. WINPEPI (PEPI-for-Windows): computer programs for epidemiologists. Epidemiol Perspect Innov 2004;1:6. [23] Rostila M, Saarela J, Kawachi I. Mortality in parents following the death of a child: a nationwide follow-up study from Sweden. J Epidemiol Community Health 2012;66(10):927e33. [24] Vance JC, Boyle FM, Najman JM, Thearle MJ. Gender differences in parental psychological distress following perinatal death or sudden infant death syndrome. Br J Psychiatry 1995;167(6):806e11. [25] Moriarty HJ, Carroll R, Cotroneo M. Differences in bereavement reactions within couples following death of a child. Res Nurs Health 1996;19(6):461e9. [26] Vance JC, Foster WJ, Najman JM, Embelton G, Thearle MJ, Hodgen FM. Early parental responses to sudden infant death, stillbirth or neonatal death. Med J Aust 1991;155(5):292e7.
121
[27] Rossi A, Rossi P. Of human bonding: parent-child relations across the life course. New York: Aldine de Gruyter; 1990. [28] Werthmann J, Smits LJM, Li J. Parental mortality rates in a western country after the death of a child: assessment of the role of the child’s sex. Gend Med 2010;7(1):39e46. [29] Raphael B. The anatomy of bereavement. New York: Basic Books; 1986. [30] Bolton JM, Au W, Walld R, Chateau D, Martens PJ, Leslie WD, et al. Parental bereavement after the death of an offspring in a motor vehicle collision: a population-based study. Am J Epidemiol 2014;179(2):177e85. [31] Russ TC, Stamatakis E, Hamer M, Starr JM, Kivimaki M, Batty GD. Association between psychological distress and mortality: individual participant pooled analysis of 10 prospective cohort studies. BMJ 2012;345:e4933. [32] Batty GD, Whitley E, Gale CR, Osborn D, Tynelius P, Rasmussen F. Impact of mental health problems on case fatality in male cancer patients. Br J Cancer 2012;106(11):1842e5. [33] Dior UP, Hochner H, Friedlander Y, Calderon-Margalit R, Jaffe D, Burger A, et al. Association between number of children and mortality of mothers: results of a 37-year follow-up study. Ann Epidemiol 2013;23(1): 13e8. [34] Martikainen P, Valkonen T. Do education and income buffer the effects of death of spouse on mortality? Epidemiology 1998;9(5):530e4. [35] Kaplan GA, Keil JE. Socioeconomic factors and cardiovascular disease: a review of the literature. Circulation 1993;88(4 Pt 1):1973e98. [36] Mittleman MA, Maclure M, Nachnani M, Sherwood JB, Muller JE. Educational attainment, anger, and the risk of triggering myocardial infarction onset. The Determinants of Myocardial Infarction Onset Study Investigators. Arch Intern Med 1997;157(7):769e75. [37] Li J. Myocardial infarction in parents who lost a child: a nationwide prospective cohort study in Denmark. Circulation 2002;106(13):1634e9. [38] Harlap S, Davies AM, Deutsch L, Calderon-Margalit R, Manor O, Paltiel O, et al. The Jerusalem Perinatal Study cohort, 1964-2005: methods and a review of the main results. Paediatr Perinat Epidemiol 2007;21(3):256e73. [39] Li J, Johansen C, Hansen D, Olsen J. Cancer incidence in parents who lost a child: a nationwide study in Denmark. Cancer 2002;95(10): 2237e42. [40] McGee R. Does stress cause cancer? BMJ 1999;319(7216):1015e6. [41] Kvikstad A, Vatten LJ. Cancer risk and prognosis in Norway: comparing women in their first marriage with women who have never married. J Epidemiol Community Health 1996;50(1):51e5. [42] Johansen C, Olsen JH. Psychological stress, cancer incidence and mortality from non-malignant diseases. Br J Cancer 1997;75(1):144e8. [43] Jacobs JR, Bovasso GB. Early and chronic stress and their relation to breast cancer. Psychol Med 2000;30(3):669e78. [44] Li J, Johansen C, Olsen J. Cancer survival in parents who lost a child: a nationwide study in Denmark. Br J Cancer 2003;88(11):1698e701. [45] Jamison RN, Burish TG, Wallston KA. Psychogenic factors in predicting survival of breast cancer patients. J Clin Oncol 1987;5(5):768e72. [46] Tross S, Herndon 2nd J, Korzun A, Kornblith AB, Cella DF, Holland JF, et al. Psychological symptoms and disease-free and overall survival in women with stage II breast cancer. J Natl Cancer Inst 1996;88(10):661e7. [47] Funch DP, Marshall J. The role of stress, social support and age in survival from breast cancer. J Psychosom Res 1983;27(1):77e83. [48] Ramirez AJ, Craig TK, Watson JP, Fentiman IS, North WR, Rubens RD. Stress and relapse of breast cancer. BMJ 1989;298(6669):291e3. [49] Forsen A. Psychosocial stress as a risk for breast cancer. Psychother Psychosom 1991;55(2-4):176e85. [50] Susser M. Widowhood: a situational life stress or stressful life event? Am J Public Health 1981;71:793e5. [51] Cohen S, Rabin BS. Psychologic stress, immunity, and cancer. J Natl Cancer Inst 1998;90(1):3e4. [52] Kiecolt-Glaser JK, Glaser R. Psychoneuroimmunology and cancer: fact or fiction? Eur J Cancer 1999;35(11):1603e7.