Neuroendocrine mechanisms of grief and bereavement: A systematic review and implications for future interventions

Bereavement is associated with many negative behavioural, psychological and physiological consequences and leads to an increased risk of mortality and morbidity. However, studies specifically examining neuroendocrine mechanisms of grief and bereavement have yet to be reviewed. This systematic review is a synthesis of the latest evidence in this field and aims to draw conclusions about the implications of neurobiological findings on the development of new interventions. PRISMA guidelines for systematic reviews were used to search for articles assessing neuroendocrine correlates of grief. Findings were qualitatively summarised. The National Heart, Lung, and Blood Institute Study Assessment Tool was used to assess the quality of the included studies. Out of 460 papers, 20 met the inclusion criteria. However, most were of fair quality only. As a neuroendocrine marker, the majority of the studies reported cortisol as the outcome measure and found elevated mean cortisol levels, flattened diurnal cortisol slopes and higher morning cortisol in bereaved subjects. Cortisol alterations were moderated by individual differences such as emotional reaction to grief, depressive symptoms, grief severity, closeness to the deceased and age or gender. Research on neuroendocrine mechanisms of grief is still in its early stages regarding grief measures and the use and timing of neuroendocrine assessments. Most of the studies focus on cortisol as outcome, and only limited data exist on other biomarkers such as oxytocin. Future research might consider assessing a broader range of neuroendocrine markers and use longitudinal designs with a focus on the psychobiological reactions to loss. Based on this, individually tailored psychosocial interventions, possibly in the palliative care context, might be developed to prevent prolonged grief disorder.


| Social loss and its consequences
The loss of a loved person is one of the most devastating experiences in life and is associated with psychological, behavioural and physiological changes in the surviving close persons. The term loss is referred to the actual loss event, whereas grief entails the subjective reactions that are associated with loss. Although grief also occurs after a separation, in this review, we focus exclusively on grief after an actual loss of a loved one through death. Bereavement is defined as the state of having suffered the loss of a loved one and entails the time after a loss during which grief is experienced. 1 Physiological reactions to bereavement include neuroendocrine, immunological and somatic changes. 2 Psychological consequences include insecurity, anxiety, aggression and depressive and (psycho-) somatic symptoms, 3 which result in a greater vulnerability to somatic or psychiatric problems, such as cardiovascular diseases 4 or clinical depression. 5,6 Some studies even associate loss with increased mortality among the survivors, 7-10 highlighting the massive effects of this experience. More recently, for example, systematic research revealed that social loss triggers the development of Takotsubo cardiomyopathy, or "Broken Heart Syndrome". This syndrome is a reversible, stress-induced cardiomyopathy that mimics acute myocardial infarction and occurs after intense emotional or physical stress. 11 Above this, patients with Takotsubo cardiomyopathy have a higher prevalence of neurological or psychiatric disorders than those with an acute coronary syndrome. 12 The previously described non-pathological mourning process is an adaptive response and usually has no long-term negative effects. 13 If, however, grieving continues and symptoms occur, that are beyond typical grief, Prolonged Grief Disorder (PGD) or Persistent Complex Bereavement Disorder (PCBD) can be diagnosed. PGD is characterised by longing for and preoccupation with the deceased, along with emotional distress and significant functional impairments that persist beyond 6 months after the loss of a significant other. 14 Approximately 10%-20% of mourners develop PGD/PCBD. [15][16][17][18] The diagnosis has only recently been added to the latest versions of the International Classification of Diseases (ICD-XI, PGD) and the Diagnostic Manual for Psychiatric Disorders (DSM-5, PCBD), 19 and led to debate about the defining criteria and consequences. 7,16,20,21 The term Complicated Grief (CG), which was originally developed to distinguish grief from depression, 22 does not represent the official diagnosis but, instead, comprises a larger category with diagnostic disordered grief encompassing a smaller group. 23 This distinction has to be kept in mind when interpreting empirical studies on grief. In the following, we employ the original terms used in the studies in each case.

| Psychobiological models of pair bond formation and bond disruption
The death of a loved one goes along with several psychosocial consequences: loneliness, a disruption in daily routines, a substantial loss of coherence, impaired sleep, and, most centrally, being separated from the loved person. All of these factors individually have been associated with poor health outcomes. For example, loneliness enhances the risk of morbidity and mortality, 24 elevates cardiovascular activation, 25 leads to cortisol dysregulation [26][27][28] and is associated with a greater utilisation of health care institutions. 29 A lower level of sense of coherence is associated with increased burden in caregivers of patients with chronic illness. 30 Additionally, poor sleep quality is associated with blunted cortisol awakening responses. 31 As the above mentioned psychosocial consequences all come together in grieving survivors, it can be assumed that those neuroendocrine and psychological changes may be even more pronounced in those who suffer intensely from the loss.
In this context, attachment and attachment disruption theories give important indications towards a better understanding of grief and its role in physical and mental health. Sbarra and Hazan 32 postulated that understanding the functionality and cause of human adult attachment could give us deeper insights into human coregulation and biobehavioural reactions to loss. According to their model, 32 relationships function as interpersonal regulatory systems. Interpersonal regulation means that couples co-regulate their emotional and behavioural responses, which serves as an adaptive mechanism that is less effortful and more automatic than individually regulating them. 32,33 The disruption of a relationship ends these regulatory benefits and leads to stress-related grief responses (dysregulation). The main task in coping with loss would be to manage dysregulation by using behavioural, emotional or cognitive strategies (functional self-regulation), which then attenuate the physiological consequences. According to the model, the initial reaction to loss not only involves psychological, but also physiological changes accompanied by psychological reactions. 32 Therefore, it is important to know the associated biological mechanisms of grief to predict negative psychological changes and to prevent grieving persons from long-term negative effects such as PGD/PCBD.
On the neuroendocrine level, grief might be primarily associated with an unspecific neuroendocrine stress-reaction, especially hypothalamic-pituitary-adrenal (HPA) axis activity. HPA axis activation leads to the synthesis of corticotrophin-releasing hormones (CRH) and vasopressin (VP), stimulating the secretion of adrenocorticotrophic hormones (ACTH) into the peripheral circulation. 34 As a result, ACTH induces glucocorticoid (e.g., cortisol) release in the adrenal gland, leading to a negative-feedback inhibiting HPA axis activation in the brain. 34,35 Cortisol secretion normally reaches its peak 30-45 minutes after awakening (cortisol awakening response [CAR]), followed by a subsequent decline during the day and reaching its lowest point between midnight and 5.00 am 36,37 Besides its stress-dependence, a healthy HPA axis function shows strong diurnal patterns, and deviations from the typical decline throughout the day provide valuable information regarding the role of the axis in disease processes. Cortisol can be measured in several ways. Basal urinary free cortisol is often used to interpret aggregated cortisol levels. Hair or nail samples indicate hormone secretion over weeks or even months. 36 Recent studies have started to examine the circadian rhythm of cortisol by evaluating a strong CAR and daily pattern of pronounced cortisol decreases during the day as indicators of a highly functional feedback-sensitivity of the HPA axis. 37 As an additional neuromodulator, oxytocin (OT) is a hypothalamic neuropeptide that, after secretion from the paraventricular nucleus of the hypothalamus (PVN) and supraoptic nucleus (SON), is stored in the posterior pituitary lobe 38 and released into the peripheral blood circuit and into central-nervous brain areas, as parts of the pain network and the reward-system, 39 OT interacts with the HPA axis system by accompanying its response to a given stressor and exerting stress-reducing effects, for example heart rate, blood pressure and cortisol level decrease. [40][41][42] OT plays an important role in the formation and maintenance of social relationships. 43,44 In turn, the OT system is also altered after the disruption of a relationship. 45

| Neuroendocrine changes after social loss in animals
In the history of research on neurobiological changes after social loss in humans, researchers often relied on animal models of separation and loss. More specifically, they began to examine neurobiological factors of social loss in the prairie vole (Microtus ochraster), which serves as an animal model of human social loss. In these monogamous rodents, the loss of a companion is associated with the activation of the HPA axis with higher basal plasma corticosterone concentrations 46-48 and adrenal hypertrophy. 49 Vole mothers show significant increases in the corticotrophin-releasing factor (CRF) mRNA expression in the PVN, 46 when separated from their pups. Interestingly, the stress response to separation can be reduced through the peripheral, subcutaneous application of OT. 50,51 The separation from an adult attachment figure in voles leads to decreased OT mRNA expression in the PVN 49 and increased density of OT-immunoreactive cells in the PVN and the SON. The latter has been interpreted as a consequence of a decreased release and limited OT receptor activity in reaction to loss. 48 Furthermore, OT fibres signalling to the NAcc show decreased activation after loss in voles. 44 Translating these effects of OT to human attachment, one can assume that the OT system is also involved in social loss in human beings. Neuroendocrine mechanisms involving OT have already been discussed with relevance for different mental disorders. 52 Although they might only serve as one of many response domains after the death of a beloved person, they could be a key mediator in the relationship between grief and the development of psychiatric disorders such as PGD or PCBD. 32 Deviations from functional neuroendocrine stress responses have already shown to be involved in response to trauma [53][54][55] and could possibly serve as a prognostic indicator for the development of grief-related psychopathology. Furthermore, important implications could be derived regarding preventive psychosocial interventions before the death of the close person in order to enhance co-regulation, as well as the awareness of the upcoming relationship disruption.
To date, a number of articles exist reviewing literature on the neuroendocrine mechanisms of grief, although they either exclusively focus on animal studies 44,45 or on prolonged grief in the context of only one neuroendocrine marker. 56 Therefore, the aim of the current work is to extend the existing literature by systematically reviewing studies investigating neuroendocrine mechanisms in the early stage of grief with potential predictive value for long-term pathological reactions to loss. November 2019 by adding more specific neuroendocrine words (oxytocin OR OXT OR OT OR cort* OR insulin OR prolactin OR endorphin OR catecholamin*) to find all the relevant articles concerning specific neuroendocrine changes after bereavement. Additionally, reference lists of relevant reviews, primary studies, and theoretical frameworks were searched for potential articles. 6,17,32,[43][44][45][58][59][60][61][62] Two independent readers (DH and HM) screened the article abstracts and read the selected full-text articles in order to decide whether to include or exclude the articles according to predefined criteria. Non-consistent decisions were discussed until consensus was reached. The eligibility criteria for the studies were:

| Search strategy and eligibility criteria
Inclusion criteria: • Original study.
• Population: human adults (> 18 years) who lost a beloved person (partner, family member, close friend).
• Article available in English.
• Grief reactions did not occur as a result of death (eg, grief related to depression or post-traumatic stress disorder (PTSD); grief after divorce or break up).
• Article not available in English.

| Data extraction
Relevant data of the incorporated studies, including publication date, study design, sample characteristics, grief assessment tools, neuroendocrine measure and results, were extracted for qualitative data analyses. Study quality was assessed independently by three authors (DH, ME and CAR) using the National Heart, Lung, and Blood Institute (NHLBI) Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies. 63 This tool consists of 14 items (including 18 sub-items) assessing key issues of the study's internal validity; for example, population recruitment, statistical power considerations, assessment of exposure and outcome variables and consideration of confounding variables. The criteria can be met, not met, cannot be determined, are not applicable or not reported. The raters discussed their ratings to resolve discrepancies and come to a final decision.
Studies were rated as "good", "fair" or "poor" to describe the risk of bias. A "good" quality rating indicates the least risk of bias. We decided to rate studies as "good" if they met more than 2/3 of the criteria. A "fair" rating indicates that the study shows higher risk of bias but not enough to invalidate results. Studies were rated as "fair if they met at least half of the criteria. A "poor" rating indicates high risk of bias that could significantly compromise the accuracy of the results.
Studies were rated as "poor" if they met less than half of the criteria.

| Mean cortisol level
Five studies examined the association between bereavement and mean cortisol levels. Jacobs et al 84 compared 56 bereaved with nonbereaved spouses, both 1 and 2 months after hospitalisation of their spouse. It was hypothesised that adults with rising separation anxiety and distress during bereavement would show higher cortisol levels than those with lower anxiety. They collected 24-hour urinary free cortisol on three separate days in the week before the second interview and averaged the daily values of cortisol. Participants with high separation anxiety showed higher cortisol levels than those whose anxiety level fell from 1 to 2 months after hospitalisation. There was no difference in cortisol levels between the bereaved and the anticipatory bereaved. 84 Irwin et al 92  They report no differences in cortisol levels between the two groups.
One further study compared cortisol levels between 260 bereaved and 262 non-bereaved men and women at the same time as controlling for depressive symptoms. Levels were assessed at one time-point after the psychiatric interview. No significant differences between the two groups were found. 79 Minton et al 96 investigated changes in physiological stress 11, 12 and 13 months after loss in 47 widows. They compared mean morning and evening cortisol levels and hypothesised that during the first anniversary after their loss, physiological stress level would be the highest. However, no significant differences in cortisol levels were found. The authors suggest that the anniversary does not represent an immediate stressor, not being sufficiently salient to change neuroendocrine stress levels. 96 Andersen et al 97 investigated the psychological and physical health effects of repeated loss among university students after clustered peer deaths. Cortisol was measured via hair samples 3 months after the loss. A significant association of prior bereavement experiences with hair cortisol level was found, as well as a significant negative relationship between the number of bereavement experiences and cortisol levels. The latter finding is interpreted in the way that people with prior bereavement maintain average levels of cortisol across the extended period of loss, whereas those with no prior experience display dysregulated cortisol levels. non-bereaved men and women 2 weeks and 6 months post-loss by taking one sample each morning. They found significantly higher cortisol morning levels in the bereaved compared to the controls at both time-points. 80 The second study examined whether overnight basal urinary free cortisol 12 months after loss depended on gender, the emotional reaction to loss (emotional numbness) and circumstances of spousal bereavement (prolongation) which were assessed 6 months post-loss. It was hypothesised that longer forewarning of death ("How long before your spouse's death did you realise that s/he was going to die?") and higher emotional numbness would be associated with higher cortisol dysregulation (higher cortisol levels). 78 As expected, prolonged forewarning was significantly associated with elevated cortisol levels.
During 6-12 months, cortisol levels increased in widowers and decreased in widows. Bereaved men with emotional numbness at time 1 had higher cortisol levels at time 2 compared to men without emotional numbness. This association was not found in women. 78

QA rating
Age (mean/ range or SD) 62.6 (NR) SA 24-hour urinary free cortisol Assessment times: three separate days in the week before timepoint (2) Group, in which separation anxiety rose from (1) to (2) had sig. higher cortisol levels than group in which SA diminished or dropped sig. Higher cortisol levels were found both for the bereaved and the anticipatory bereaved subjects  (2) and (3) Fair (   Sig. lower levels of log-cortisol levels at wake and flatter diurnal slopes in group (3) compared to group (1) Not significant: differences between group (2) and (3), although descriptively flatter profile in group (3) compared to group (2) Bereavement independently of its strength is associated with dysregulated cortisol levels Subsidiary analysis: Those who most recently lost a spouse showed sig. greater cortisol dysregulation (higher log-levels at wake and flatter slope) than those who lost someone else than the partner Not significant: continuous PG did not predict log-cortisol Vitlic et al (2014) 85 Cross-sectional 2 x 2 design Young bereaved (1) vs young nonbereaved (2) vs old bereaved (3) vs old non-bereaved (4) Spouse (65 for (1) and 9.5 for (2)) Close relative (35 for (1) and 91.5 for (2))  (4) Sample (3) and (4) are used from earlier study  (1) and (2), although higher mean values in group (1) Fair

DST/CRH stimulation test
Core Bereavement Items (CBI) IES Venous blood samples, Cortisol, DHEAS, cortisol:DHEAS-ratio Sig. lower DHEAS, higher cortisol and higher cortisol:DHEAS ratio in (3) compared to (4) No significant differences in these outcomes between (1) and (2) Those with higher CBI -scores showed higher cortisol:DHEAS ratios Those with higher social support reported lower cortsiol:DHEAS ratios  (2) and (4) No significant differences in ACTH-levels Sig. smaller ACTH responses to CRH in group (1) compared to group (2) and (4) Sig. greater cortisol responses to CRH in group (1) compared to groups (2) -(4) No significant differences in ACTH responses to CRH between groups (1) and (2) Fair 44.7 (14.1) None Blood cortisol Assessment times: 4.00 pm and 11.00 pm 1 day after dexamethasone application 11.00 pm day before) Among the affective disorder patients of the early loss group, younger age at first loss significantly a correlated with higher 4.00 pm cortisol levels First loss as strongest predictor for HPA axis functioning Late loss predicts higher cortisol levels at 11.00 pm Blood cortisol blood ACTH DST Assessment of blood samples: between 9.00 and 11.00 pm at times (1), (2) and (3) Sig. higher cortisol plasma levels after DST in time (1) compared to time (2) and (3) (2) Spouse   (2) Intervention study (activation program) Anticipatory bereaved and bereaved men and women who are about to lose/who lost a close relative (1) Activation programme Group sig.,significant. Overall plasma levels of OT, measured through one simple blood collection

TA B L E 1 (Continued)
Sig. higher plasma OT levels for group (1) compared to group (2) No significant OT differences between (1) and (3) ICG symptom severity explained only 2% of the variation in plasma OT levels Secondary analysis: a primary or probable CG diagnosis is positively associated with plasma OT levels 58 At time 2, significant higher cortisol morning responses and overall cortisol responses were found in the CG group compared to the same group at time 1. Furthermore, higher scores in grief severity were associated with lower morning cortisol levels. 77

| Cortisol:DHEAS ratio
Two studies investigated the association between bereavement and the cortisol:dehydroepiandrostheron-sulphate (DHEAS) ratio.
DHEAS is a sulfated steroid-hormone that is associated with HPA axis activity. By contrast to cortisol, which has immunosuppressive effects, DHEAS enhances the immune response. Studies have shown that DHEAS can buffer the suppressive effects of cortisol on neutrophil function. 89 Additionally, an increased cortisol:DHEAS ratio, which represents an imbalance between those biomarkers, appears to be a contributing factor to the process of age-related immunosenescence. Khanfer et al 89 hypothesised that ageing and stress had an additive and deleterious effect on immunity and that bereaved older adults should have higher cortisol:DHEAS ratios than non-bereaved older adults. They used the cortisol:DHEAS ratio as an indicator of neutrophil function and assessed cortisol levels in bereaved and non-bereaved older adults. Although cortisol levels were slightly higher in the bereaved group, a higher cortisol:DHEAS ratio was found in the bereaved compared to the non-bereaved subjects. 89 Vitlic et al 85 compared cortisol:DHEAS ratios between younger and older bereaved vs non-bereaved adults and found significant lower DHEAS, higher cortisol and higher cortisol:DHEAS ratios in the older bereaved compared to the older non-bereaved.
These differences were not shown in the young groups. 85 Although the younger bereaved showed higher psychological effects of loss than the older subjects, these changes were not reflected in neuroendocrine outcomes. Finally, those with stronger grief symptoms showed higher cortisol:DHEAS ratios, whereas those with higher levels of social support showed lower ratios. 85

| Dexamethasone suppression test (DST)/CRH stimulation test
The DST is applied to assess HPA axis feedback sensitivity. 103  and women showed lower cortisol levels than the non-depressed.
Among the affective disorder patients of the early loss group, younger age at first loss significantly correlated with higher afternoon cortisol levels. Furthermore, in the afternoon, men in the early loss group showed significantly higher cortisol levels than women.
Late loss significantly predicted higher cortisol levels in the morning. 102

| Catecholamines
Two studies examined the association between bereavement and catecholamines as outcomes of sympathetic adrenal medullary function (SAM). Jacobs et al 83 investigated 24-hour urinary free epinephrine and norepinephrine on three successive days in 59 bereaved and anticipatory-bereaved subjects and found higher catecholamine outputs in the bereaved compared to the anticipatory bereaved; however, these differences were not significant.
Norepinephrine was inversely correlated with depression scores and positively correlated with age. The latter finding is in line with past research showing that the SAM system in older adults adapts more slowly to stress. 83

| Oxytocin
Bui et al 99 investigated peripheral plasma OT levels in men and women with CG. They compared a single assessment of OT levels of participants with a primary CG diagnosis to participants suffering from depression as primary diagnosis and bereaved control participants with no comorbid diagnosis. They found significantly higher OT levels in the CG group compared to the depressed group.
There were no significant differences between the CG group and the group of non-pathological grief. 99 Secondary analyses revealed that a primary or probable CG diagnosis was positively associated with plasma OT levels.

| Prolactin
Lane et al 104 investigated sex differences in prolactin (PRL) changes during mourning in 26 spouses. Amongst others, PRL plays a role in the stimulation of maternal care, acts as an endogenous anxiolytic agent and regulates oxytocin neurones. 104 They assessed serum PRL before and after a semi-structured interview. The aim was to examine sex differences in the association between the developmental levels of the survivors' object representation (DLOR). The DLOR represents the verbal description of a person and the level of cognitive complexity of that description. 100 The results show a significant larger mean PRL change in women compared to men. A negative correlation between PRL change and DLOR was found in women, whereas a positive correlation was found in men. 100

| Effects of bereavement interventions on neuroendocrine stress markers
Three studies examined the effects of bereavement interventions on stress-related neuroendocrine markers. In the first study, the effect of an activation programme on plasma cortisol and prolactin levels was examined in 72 close female relatives of cancer patients. 101 Plasma cortisol and prolactin, as well as anxiety, depression and mental exhaustion, were assessed during the intervention, right before the death of the relative and 1 and 2 months after loss. The results show that an increasing degree of mental exhaustion during the treatment period is significantly associated with increasing cortisol levels and decreasing prolactin levels. Furthermore, significantly higher cortisol levels were found 1 month after death compared to the last assessment before death. Also, lower prolactin levels during treatment were found in the activation group compared to the control group. 101 In the second study, the effects of a short-term bereavement support group intervention with 119 widowed men infected with HIV on immune variables and cortisol levels were assessed. 94 Recently bereaved HIV seropositive (HIV+) and HIV seronegative (HIV-) men were randomly assigned to either a bereavement support group intervention or a standard care group. Plasma cortisol was assessed pre, post and at 6-month follow-up. Significantly lower cortisol levels were found in the intervention group compared to the control group 6 months after the intervention. HIV + men in the intervention group showed significant decreases in cortisol levels from pre-assessment to follow-up, whereas HIV-men in the intervention group showed increased levels of cortisol within the same time-period. 94 The third study assessed predictive effects of catecholamines as moderators of a bereavement intervention and CG treatment outcomes after bereavement. 95 Sixteen bereaved individuals provided information on the Inventory of Complicated Grief (ICG) pre-and post-psychotherapy and blood epinephrine, norepinephrine and dopamine were assessed 4 weeks before the intervention. The posttreatment ICG-score was significantly predicted by pre-treatment epinephrine levels. SAM activity and autonomous function in the participants showed impaired CG outcomes after therapy. 95

| Summarized results of good-quality studies
In summary, the results of "good quality" studies suggest the following neuroendocrine changes after the loss of a loved one: • The more deaths of loved ones someone experiences, the higher his/her the cortisol levels. 77 • Morning cortisol levels are significantly higher in bereaved compared to non-bereaved 2 weeks and 6 months after bereavement. 80 • The longer the forewarning of someone's death, the higher the cortisol levels after bereavement. 80 • Bereaved men suffering from emotional numbness 6 months after loss show higher cortisol levels 12 months after death compared to bereaved men who do not suffer from emotional numbness. 78 • Compared to non-pathologically grieving subjects, people with CG show flattened diurnal cortisol slopes, suggesting that HPA axis dysregulation is more pronounced in prolonged grief. 79 People with CG show significantly lower morning and overall cortisol levels compared to non-pathological grievers 2 years after loss. 77 • People with CG show significant higher cortisol morning responses 5 years after loss compared to two years after loss. 77 • Higher scores in grief severity 5 years after loss are associated with lower morning cortisol levels. 77

| D ISCUSS I ON
The loss of a loved one can be associated with neuroendocrine alter- longer forewarning before death lead to higher cortisol levels than experiencing an unexpected loss. 78 Sudden, unexpected losses, as well as a rising number of losses, are associated with lower insulin levels, showing that those context variables influence health-reducing neuroendocrine alterations after bereavement. 98 A positive affect was inversely correlated with cortisol levels, 90 whereas rising emotional numbness in men during the course of bereavement enhanced cortisol levels, 78 suggesting again that psychological variables are important when examining neuroendocrine changes after loss. Regarding gender differences, one study revealed that men showed decreasing cortisol levels, whereas women showed increasing cortisol levels during the course of bereavement. 78 Older men and women showed stronger alterations in their neuroendocrine stress responses than younger cohorts, 85,89 indicating that high age may have an additive effect on loss consequences. Furthermore, changes in stress-related alterations were shown, especially in the early stage of bereavement, although there is inter-individual variability. 88 In the latter study, however, almost no direct correlations between psychological and biochemical reactions were found. Neuroendocrine alterations were not only found directly after bereavement, but also months after loss experience. 97 Interesting results evolved with regard to psychiatric diseases: especially depressive symptoms were associated with higher cortisol levels 86,102 and higher cortisol nonsuppression 88 in bereaved subjects. Furthermore, individuals suffering from PTSD after a traumatic loss showed higher cortisol levels than those with no trauma-related psychiatric diagnosis. 80 The same study suggests that trauma-related psychopathology may foster a prolonged neuroendocrine response to social loss up to 8 years after the event. One study investigated OT as a biomarker of grief and found higher OT levels in people suffering from CG. 99 Regarding prolactin changes, women have higher prolactin levels than men after having been interviewed about the deceased partner. Interestingly, women who have a more complex insight into the deceased person also show higher prolactin levels, whereas the opposite association is observed in men. 100 In summary, these studies suggest that not only bereavement by itself, but also bereavement-associated psychopathology in particular is associated with stress-related neuroendocrine alterations.
This is in line with research on trauma, 53 The latter finding is consistent with the hypothesis that grief is activated by an intervention and that the active mourning may have a prophylactic value to the relative's grief reaction. 101 The results indicate that, even years after loss, bereavement might be associated with neuroendocrine changes. These changes are moderated by grief severity, psychiatric state and psychological reactions to loss, as well as age and gender. On a psychobiological level, neuroendocrine responses may serve as moderators between the loss-event and long-term psychological outcomes ( Figure 2). However, because of the methodological difficulties and contradictory results of the studies, these conclusions must be treated with caution.
For example, Ong et al 90 found that prolonged forewarning of death was associated with higher cortisol levels. They argue that a longer duration of care is associated with more stressful experiences, and thus leads to stronger physiological stress reactions. 90 Research on the development of PGD/PCBD shows that suddenness of death is a risk factor, 107 which initially appears to contradict the findings of Ong et al 90  with CG and non-CG. Additionally, two studies found significantly elevated cortisol levels in bereaved and anticipatory bereaved, 86,92 whereas two other studies did not. 79,93 One reason for the conflict-

| Limitations
The studies included in this systematic review reveal some limita- OT mechanisms in the central nervous system cannot be measured in the human living brain so far, which limits the possibilities to test for direct involvement of OT in the grieving process. Therefore, 44 animal models can be helpful to better understand those mechanisms.
Moreover, human and animal studies can complement each other in a meaningful way because their methods lead to context-dependent results: experimental settings are artificial and may lead to different reactions than real-life events. 45 Above this, animal models cannot give us sufficient insights into the psychological reactions to loss.
On the other hand, so far, the highly individual human grief-reaction cannot be investigated in a standard procedure or related to specific neuroendocrine changes in the living brain.

| Future research and implications for psychosocial interventions
To help establish a comprehensive model of the neuroendocrine factors underlying the psychobiological reactions to social loss, in addition to the neuroendocrine stress response, future research can benefit from a focus on further and interacting neuroendocrine systems. Animal research on social loss suggests that, for example, the OT system interacts with the HPA axis and might be involved during grief reactions. Both CRH and OT have been shown to interact with the dopamine) system, which regulates reward and is involved in depressive disorders and addiction. In both animals and humans, dopamine appears to play a role in the formation of a romantic relationship; for example, reward-associated brain regions are highly activated in association with positive attachment interactions, [114][115][116] and it is assumed that this system could also be affected after loss by remaining under-stimulated. 111 Indeed, human studies already indicate activations in brain-regions with high OT and dopamine receptor density. 91,92 In line with this, withdrawal from drug abuse has been associated with similar activation patterns compared to separation from a partner. 46,111 In addition, longitudinal studies assessing subjective and neuroendocrine markers before and after loss could minimise confounding inter-individual variations and thereby improve statistical power and long-term predictive power. Although necessarily in such studies, loss would always be predicted by a lethal illness, thereby limiting the range of different possible events to trigger grief.
Initial studies investigating neuroendocrine alterations after a bereavement intervention show promising effects and suggest that, beside subjective measures, neuroendocrine and stress-related outcomes can serve as meaningful indicators of treatment success. 94,95,101 In this context, it is important to keep in mind that subjective and objective measures often diverge in research on stress, and thus it is important to reveal the differences between these measurements. This leads to the next step of exploring the reasons why these differences occur and what they mean for treatment success. However, to better interpret the meaning and importance of neuroendocrine measures for therapeutic success, more research is necessary. Furthermore, the assessment of neuroendocrine measures is associated with some hurdles. For example, the assessment of blood, urine or saliva samples is time-consuming and may be a reason for grieving participants not to take part in a study. One possibility to address this issue and to enable data on aggregated cortisol levels to be collected over an extended time period of weeks to months is the use of hair samples to quan- factors of an adaptive grief coping process. In this context, interventions that help to strengthen the bond, and which make unresolved issues a subject of discussion, might foster a healthy coping process and therefore affect neuroendocrine as well as psychological health changes after the loss. Although there is no study investigating the effects of pre-death interventions on neuroendocrine reactions such as OT signalling, initial studies show that psychosocial interventions before loss are able to improve the well-being of the participants. 121,122 However, this hypothesis needs further investigation and additional research is necessary to understand whether mechanisms such as OT signalling contribute to the efficacy of such treatments. Furthermore, it is important to consider inter-individual differences when deciding on whether to implement an intervention or not.
In summary, neuroendocrine correlates of anticipatory grief and grief after social loss could help us identify individual needs and serve as tools to evaluate not only impairment, but also treatment success. In the long run, this knowledge might allow the development of specific interventions that improve stress-related responses in the survivors and thereby their health.

ACK N OWLED G EM ENTS
We thank the German FAZIT-Stiftung and the Marsilius Kolleg at Heidelberg University for making this review possible by financially supporting DH and BD. CAR is supported from the Olympia Morata Program at Heidelberg University. Furthermore, we thank Hannah Melles (HM), who helped us find relevant articles and Star Dubber for proofreading the manuscript submitted for publication. Open access funding enabled and organized by Projekt DEAL.

CO N FLI C T O F I NTE R E S T S
The authors declare that they have no conflicts of interest.

AUTH O R CO NTR I B UTI O N S
DH, ME, CAR and BD defined the literature search criteria. DH conducted the literature search and summarised the findings. DH, ME and CAR rated the internal validity of the studies by the National Heart, Lung, and Blood Institute Study Assessment Tool. DH, CAR, ME and MW wrote the paper. BD reviewed the manuscript and gave critical advice.

PE E R R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1111/jne.12887.