Table 1 Study Characteristics and results

Cerel et al., 1999 [34]

Ohio, United States

Cohort study

Follow-up: 1, 6, 13 and 25 months after the death

N = 26 children bereaved by suicide of parent, from 15 families

N = 322 children bereaved by non-suicide death of parent (reasons other than suicide or homicide)

BAMO, an unvalidated scale measuring somatisation disorder

Health/Sickness Questionnaire

School and Physician Rating Forms

No significant difference in scores of somatisation between suicide and non-suicide bereaved. Suicide-bereaved children visited doctor less frequently but missed significantly more days of school than non-suicide-bereaved children

Small sample size for suicide-bereaved children. No confounding factors were controlled for. Type 1 error is increased due to multiple testing of the data

Cleiren et al., 1994 [37]

Leiden, The Netherlands

Cohort study Follow-up: 4 and 14 months after death

N = 91 first-degree relatives and spouses bereaved by suicide

N = 93 first-degree relatives and spouses bereaved by traffic accident

N = 125 people bereaved by the illness of first-degree relative

General wellbeing questionnaire measuring physical health and somatic complaints (measure not described)

No differences were found for somatic complaints between the different modes of death groups (no p-value given). Mode of death was not significantly associated with physical health complaints

Some of the scales used are not validated. No confounding factors were controlled for. 10% loss to follow-up which may introduce attrition bias

Erlangsen et al., 2017 [35]

Denmark

Longitudinal cohort study Follow-up: 1980–2014

N = 15,607 suicide-bereaved spouses

N = 788,778 spouses bereaved by non-suicide death

Diagnosis of cancers, diabetes, sleep disorders, cardiovascular and chronic lower respiratory tract diseases, liver cirrhosis, and spinal disc herniation

Suicide-bereaved had lower risk of diagnoses of cancer, diabetes, cardiovascular and chronic lower respiratory tract disorders. They were less likely to take sick leave

Only people who were in a formal union or who were living together were included. While analyses was adjusted for some covariates, unmeasured confounders may be an issue

Fang et al., 2011 [44]

Sweden

Historic cohort study

Follow-up: 1990–2004

N = 102 parents bereaved by suicide of child

N = 124 parents bereaved by non-suicide death of child

N = 334 and n = 297 parents bereaved by non-cancer and cancer death of child

N = 46 and n = 251 parents bereaved by an infection-related or any other cause of death of child

A diagnosis of infection-related cancer using ICD codes

The risk ratio was higher for suicide-bereaved than for non-suicide bereaved but this didn’t reach statistical significance; confidence intervals greatly overlapped

Some potential confounders were not accounted for due to the registry-based nature of the study

Farberow et al., 1992 [22]

Three counties in California

Cohort study

Follow-up: 2, 6, 12 and 30 months after the death

N = 108 suicide-bereaved aged ≥55 years

N = 199 people aged ≥ 55 years bereaved by natural death

N = 144 people aged ≥55 years not experiencing any death or divorce of spouse

The somatisation subscale of the Brief Symptom Inventory (BSI)

Suicide-bereaved and naturally bereaved spouses did not differ significantly on the somatisation subscale. All of the mean scores of the scales, including somatisation, decreased over the 2.5 year period

There appears to be loss to follow-up in each group which may indicate the presence of attrition bias

Grad and Zavasnik, 1999 [39]

Slovenia

Cohort study

Follow-up: 2 and 12–14 months after the death

N = 30 suicide-bereaved spouses

N = 23 spouses bereaved by road traffic accident

N = 20 spouses bereaved by terminal illness

Slovenia Bereavement Scale (SBS) has 46 items, representing several categories, including physical health

There were no statistically significant differences (no data presented) between the bereaved groups on the physical health measures contained within the SBS

Small sample size and no confounding factors controlled for. First assessment was conducted 2 months post-death when acute grief is likely to be present

Kennedy et al., 2014 [45]

Sweden

Historical cohort study

Follow-up: 1961–2006

N = 19,535 offspring ≤18 years and n = 12,137 ≥ 18 years who lost a parent to suicide

N = 42,796 offspring ≤18 years and n = 136,786 ≥ 18 years bereaved by parental cancer death

N = 52,592 offspring ≤18 years and n = 178,393 ≥ 18 years bereaved by parental non-cancer death

N = 25,772 offspring ≤18 years and n = 18,566 ≥ 18 years bereaved by parental non-suicide death

Diagnosis of first malignant cancer before the age of 40 in the Cancer Register

The effect of suicide bereavement more than doubled the risk of human papillomavirus-related cancers before the age of 40, compared to those bereaved by non-suicide deaths. However, this finding was not statistically significant

No information on individual confounding factors including alcohol consumption and smoking

Momen et al., 2013 [26]

Denmark and Sweden

Population-based cohort study

Follow-up for Denmark/

Sweden: 1968–2007/1973–2006

N = <66 suicide-bereaved (exact number unavailable from authors)

N = 1217 children bereaved by unexpected deaths, other than accident, suicide or the violence death of their relative

A diagnosis of childhood cancers using ICD codes

The adjusted hazard ratio was higher for suicide-bereaved children than children bereaved by other causes of death. However this association did not reach statistical significance

Small numbers of suicide- bereaved may not make these findings generalisable to other suicide-bereaved family members

Séguin et al., 1995 [31]

Quebec City, Montreal

Cohort study

Follow-up: First interview M = 5.8 months after the death

Second interview

M = 9 months after the death

N = 30 parents who lost a son aged 18 to 35 years to suicide

N = 30 parents who lost a son aged 18 to 35 years by a car accident

Physical disorders were measured using items taken from Quebec’s 1987 Health Survey

Suicide-bereaved had more physical illnesses and consulted health professionals more frequently than accident survivors

Some important confounding factors including gender and age of the deceased not controlled for. High rate of attrition bias

Weinberg et al., 2013 [46]

United States

Prospective longitudinal controlled study

Follow-up: 5 years

N = 45 offspring bereaved by suicide of a parent

N = 27 offspring bereaved by accidental death of a parent

N = 51 offspring bereaved by sudden natural death of a parent

BMI was the outcome studied, by measuring the weight and height of offspring objectively

There were no differences in the BMI categories of offspring bereaved by suicide, accident and sudden natural death

Some participants recruited via advertising. Possible attrition bias as participants lost to follow-up more likely to be bereaved than those retained in the study

Wilcox et al., 2015 [47]

Sweden

Prospective cohort study

Follow-up: 3 years

N = 537 parents bereaved by suicide of an offspring

N = 716 parents bereaved by accidental death of offspring

N = 549 parents bereaved by natural death of offspring

Diagnosis-specific sickness absence exceeding 30 days due to somatic diagnoses

No statistically significant differences in the risk of somatic diagnosis between suicide-bereaved, accident-bereaved and naturally bereaved parents

Sickness absence due to specific somatic diagnoses were only included if they exceeded 30 days

Barrett and Scott, 1990 [36]

North Dakota and Minnesota, United States

Cross-sectional study

N = 14 suicide-bereaved spouses

N = 15 accident-bereaved spouses

N = 15 unanticipated natural death bereaved

N = 13 bereaved by theexpected natural death of spouse

Grief Experiences Questionnaire (GEQ): somatic reactions subscale

No significant differences in mean scores of somatic reactions for suicide-bereaved and non-suicide bereaved

Small sample size of suicide and non-suicide bereaved

De Groot et al., 2006 [30]

Northern Provinces in The Netherlands

Cross-sectional study

N = 153 first-degree suicide-bereaved relatives and spouses

N = 70 first-degree relatives and spouses bereaved by natural causes

RAND-36 used to assess general health, with nine subscales

Suicide-bereaved functioned less well in terms of pain and general health than naturally-bereaved

Possibility of selection bias due to difficulty in recruiting family members bereaved by natural death

Demi and Miles, 1988 [38]

United States

Cross-sectional study

N = 59 (15 fathers and 44 mothers) parents whose children died by suicide

N = 61 (13 fathers and 48 mothers) whose children died as a result of an accident or a chronic disease

Hopkins Symptom Checklist (HSCL)

Health problems measured using the Bereavement Health Assessment Scale (BHAS)

No difference on the 5 subscales of the HSCL (somatisation, obsessive-compulsive, interpersonal sensitivity, depression, anxiety) or across physical health outcomes between the 2 groups

Bereaved parents may not be representative as they were recruited from self-help groups

Dyregrov et al., 2003 [33]

Norway

Cross-sectional study

N = 128 suicide-bereaved parents

N = 68 accident-bereaved parents

N = 36 SIDS-bereaved parents

General Health Questionnaire (GHQ-28): somatic symptoms

SIDS-bereaved parents experienced significantly fewer problems on GHQ than suicide and accident-bereaved

Control group was heterogenous (violent and non-violent deaths)

Kitson, 2000 [41]

Two Midwestern counties in United States

Cross-sectional study

N = 85 suicide- bereaved widows

N = 56 homicide-bereaved widows

N = 135 accident-bereaved widows

N = 167 sudden death bereaved widows

N = 106 long-term illness bereaved widows

The somatisation subscale of the Brief Symptom Inventory (BSI)

No differences between the 5 bereaved groups on somaticism

Control group contained both bereavement from violent and non-violent deaths which may have introduced selection bias

McNiel et al., 1988 [42]

United States

Cross-sectional study

N = 13 widows bereaved by the suicide death of their husband

N = 13 widows bereaved by the accidental death of their husband

General Health Questionnaire (GHQ): somatic complaints subscale

No significant differences in the mean scores of suicide and accident-bereaved

Very small sample size and no confounding factors were adjusted for

Miyabayashi and Yasuda, 2007 [32]

Japan

Cross-sectional study

N = 21 suicide-bereaved adults

N = 23 accident-bereaved adults

N = 9 adults bereaved by acute illness

N = 74 adults bereaved by shorter illness

N = 88 adults bereaved by longer illness

General Health Questionnaire (GHQ), including somatic symptoms

No group differences were found for somatic symptoms. Multiple comparison tests indicated that those bereaved by suicide had poorer general heath than those bereaved by a longer illness (p < 0.05)

Selection bias may be present as participants recruited from self-help group. Response bias may be present due to the small sample of suicide-bereaved and those bereaved by acute illness. Some important confounders were not controlled for

Pfeffer et al., 2000 [24]

United States

Cross-sectional study

N = 11 families (made up of 16 children) where a parent died by suicide

N = 57 families (made up of 64 children) where a parent died from cancer

Child Behaviour Checklist (CBCL) has a subscale for somatic complaints

Mean scores of somatic complaints did not significantly differ between children bereaved by the cancer death of a parent and the suicide death of a parent

Some participants recruited via advertising which could lead to response bias. Very small sample of suicide-bereaved which will not be generalisable

Reed and Greenwald, 1991 [43]

United States

Cross-sectional study

N = 85 suicide-bereaved relatives and spouses

N = 96 accident-bereaved relatives and spouses

Measure for somatic complaints with 6-items

No significant differences in somatic complaints between the two groups

Use of unvalidated measures throughout the study

Xu and Li, 2014 [23]

China

Cross-sectional study

N = 92 immediate family members (parents, siblings, children, spouses) bereaved by suicide

N = 64 immediate family members (parents, siblings, children, spouses) bereaved by accidental death

The Symptom Checklist-90-Revised (SCL-90-R) has nine subscales, including somatisation

No significant differences were found on the score of somatisation between the suicide-bereaved group and the accidental death group

May not be generalisable to wider bereaved group as findings may be culturally specific

Bolton et al., 2013 [19]

Manitoba, Canada

Longitudinal case-control study

N = 1415 parents of children that died by suicide

N = 1132 parents of children who died in an motor vehicle crash

N = 1415 non-bereaved parents

Physical health disorders based on ICD 9 and 10 codes

Outpatient physician visits for physical health and hospitalisation for physical illnesses

Two years pre and post-death, suicide-bereaved parents had significantly higher rates of CVD COPD, hypertension, diabetes, and outpatient physician visits for physical illnesses compared to motor-vehicle bereaved parents

Prevalence of physical disorders were examined two years pre-death and two years post-death. This time may not be sufficient for the development of certain physical health problems

Harwood et al., 2002 [40]

England

Case-control study

N = 46 adults bereaved by the suicide of an older adult

N = 46 adults bereaved by the natural death of an older adult

Grief Experiences Questionnaire (GEQ): somatic reactions subscale

No significant difference on somatic reactions for suicide-bereaved and naturally-bereaved

Small sample size may have increased the risk of type II error

Huang et al., 2013 [20]

Sweden

Nested case-control study

N = 792 parents bereaved by the suicide of a child

N = 1451 bereaved by non-self-inflicted death of child

N = 1066 bereaved by cancer of child

N = 2814 bereaved by non-cancer death of child

Pancreatic cancer, identified by the Swedish Cancer Register

It appears that suicide-bereaved have a higher risk of cancer but this finding is not statistically significant when compared with non-suicide bereaved

Unmeasured potential confounders for pancreatic cancer, including smoking and BMI could not be controlled for.