THE DIAGNOSIS OF COMPLICATED GRIEF AS A MENTAL DISORDER: A CRITICAL APPRAISAL

————— Birgit Wagner, University Clinic for Psychotherapy and Psychosomatic Medicine, University Hospital Leipzig; Andreas Maercker, Department of Psychopathology and Clinical Interventions, University of Zurich. Correspondence concerning this article should be addressed to Birgit Wagner, University Clinic for Psychotherapy and Psychosomatic Medicine, University Hospital Leipzig, Semmelweisstr. 10, 04103 Leipzig, Germany. E-mail: birgit.wagner@medizin.uni-leipzig.de THE DIAGNOSIS OF COMPLICATED GRIEF AS A MENTAL DISORDER: A CRITICAL APPRAISAL


Introduction
Grief reactions following the loss of a significant person often comprise a set of expected negative reactions involving functional impairment. This process can be considered normal, though depression and trauma-related symptoms might occur during this time of adjustment and working through loss. It is therefore difficult to find consensus with regard to the difference between pathological and normal grief reactions. Normal grief reactions include a combination of mostly negative symptoms, e.g., social retreat, crying, and intrusions, which persist for a certain length of time. Distressing moods and confusing thoughts are common during grief. The mourning process will, in most cases, lead to a restored equilibrium. However, in some cases, the grief can become extreme (Horowitz, Siegel, Holen, Bonanno, Milbrath, & Stinson, 2003). As many as 5-15% of bereaved people seem to develop severe long-term reactions to their loss. These severe reactions may lead to impairments on a physical and psychopathological level (Horowitz et al., 2003): the grief is then usually termed Complicated Grief (CG). Various factors, such as the circumstances of the death, relationship to the deceased, access to social support, and mental health state play a large role in impairment after a loss. Research has demonstrated the multidimensionality of grief reactions with different types of emotions, cognitive impairment, health problems, and impaired role functioning (Bonanno, Neria, Mancini, Coifman, Litz, & Insel, 2007). Studies have also shown that bereavement is associated with a higher risk of mortality, especially in the period immediately after the loss, and have shown a relation with the use of medical services (Stroebe, Schut, & Stroebe, 2007). Further, these ailments can include interpersonal problems, substance abuse, physical illness, and even death (Lichtenthal, Cruess, & Prigerson, 2004;Stroebe, Schut, & Finkenauer, 2001).
Yet, CG is still not in the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) or the International Classification of Diseases, 10th revision (ICD-10), though there is now one proposal (Prigerson, Horowitz, Jacobs, Parkes, Aslan, Goodkin et al., 2009) for CG (or alternatively 'prolonged grief disorder'), which has been developed out of two previously proposed sets of diagnostic criteria (Horowitz, Siegel, Holen, Bonanno, Milbrath, & Stinson, 1997;Prigerson, Shear, Jacobs, Reynolds, Maciejewski, Davidson et al., 1999), to be given official recognition in the fifth edition of the Diagnostic and Statistical Manual of Mental Disorders. However, there is still some scepticism and critical discussion concerning the validity of diagnostic criteria for CG. Some researchers claim that the current CG criteria are still not scientifically soundly proven (e.g., Hogan, Worden, & Schmidt, 2003). Diagnostic criteria are helpful for the identification of bereaved individuals who are suffering pathological grief reactions, but they should neither exclude too many cases of bereavement complications nor be over-inclusive.
A number of questions arise regarding the specific diagnostic criteria for CG: first of all, on which conceptualisation will the new diagnostic criteria be based? How can the interface of trauma and bereavement be addressed? Will the diagnostic criteria show a cohesive syndrome for all different groups of bereaved people (loss of a child, loss of a spouse, parental loss, loss of significant others, etc)? How could different pathways of complicated bereavement (delayed, intensive or absent) be assessed with the proposed criteria? Another dilemma is the ongoing relationship of the bereaved person to the deceased and the interpersonal aspects of the loss. The assessment of the cognitiveemotional organisation of the relationship to the deceased can give important insights into possible complications in the grief process (Rubin, Malkinson, & Witztum, 2008). And finally, what are the advantages and disadvantages of including CG in the future DSM-V and ICD 11? The multidimensional complexity of these questions poses a challenge for the diagnostic criteria of CG and grief processes.
A number of complex issues are raised by this proposed new diagnosis. In this article we will give an overview of the challenges associated with the evolution of a DSM diagnosis for CG. We will also introduce important conceptualisations of CG and their relevance to diagnostic criteria. Furthermore, we will present the historical development of two propositions of diagnostic criteria (Horowitz et al., 1997;Prigerson et al., 1999), out of which one diagnostic category has now been developed (Prigerson et al., 2009). Another important topic will be the relationship between trauma and bereavement and its implications for criteria of CG. Finally, in an effort to give a widely neglected aspect of complicated grief processes the attention we think it needs, we review the impact of the continuing relationship to the deceased.

Conceptualisation of complicated grief
The importance of identifying syndrome criteria for CG with acceptable reliability is evident and could facilitate research in the field of laboratory studies, family studies and treatment studies. Diagnostic criteria might serve to identify individuals in need of treatment. The establishment of diagnostic criteria for CG therefore asks for a 'gold standard', which is not only able to detect those who are not experiencing 'normal' grief, but also those who suffer from distinct disorders. Still, bereavement is a normal, non-pathological phenomenon, which occurs after the loss of a loved one. In order to define CG, one would need to know what exactly is understood by "normal grief". However, it appears nearly impossible to define 'normal' grief, as there are cultural and individual differences in bereavement reactions. There are common ways of grieving in one culture which do not fit the traditional idea of bereavement in another culture. Just as it is difficult to define normal grief, so is it also difficult to define CG (Dijkstra, 2000;Maercker, 2007).
In DSM-IV, a mental disorder is defined as "a clinically significant behavioural or psychological syndrome or pattern that occurs in an individual and that is associated with present distress or disability or with a significantly increased risk of suffering death, pain, disability, or an important loss of freedom. In addition, this syndrome or pattern must not be merely an expectable and culturally sanctioned response to a particular event, for example, the death of a loved one" (American Psychiatric Association, 1994, xxi). The DSM-IV categorises the death of a significant person as a stressor with generally normative and predictable consequences. In the current issue of the coding system of the DSM-IV, bereavement is included among the "additional codes" (V codes) and it is conceptualised as a normal phenomenon, which is culturally varying among different groups (American Psychiatric Association, 1994, p. 684). The use of the V code also explicitly avoids the distinction between normal and complicated forms of grieving. Instead, the DSM-IV provides for the more pathological cases diagnosis in the form of existing categories, such as Major Depressive Disorder (MDD) or Posttraumatic Stress Disorder (PTSD).
Not only are the diagnostic criteria of CG still at a development and research stage, naming the "not normal" grief reactions has also gone through various changes and developments. In the last years, clinicians and researchers have described CG by using many terms and subtypes such as 'abnormal ', 'chronic', 'morbid', 'pathological', 'traumatic' and 'prolonged' grief. Recently, two research teams have focused on establishing specific diagnostic criteria for CG (Horowitz, Bonanno, & Holen, 1993;Horowitz et al., 1997;Prigerson, Maciejewski, Reynolds, Bierhals, Newsom, Fasiczka et al., 1995). Although CG was, from 1997 to 2002, referred to as 'traumatic grief' by Prigerson and colleagues in the literature, after the events of 9/11 they reverted to CG due to the upcoming misinterpretation of traumatic grief with PTSD (Lichtenthal et al., 2004). Later, Prigerson, Vanderwerker, and Maciejewski (2008) developed new diagnostic criteria named 'prolonged grief' (Prigerson et al., 2009;Prigerson et al., 2008). The authors explain this change of terminology by pointing out that complicated is defined as "difficult to analyze, understand and explain" and does not capture the nature of the bereavement syndrome. Instead, the term prolonged seemed to express the nature of the disorder more clearly. However, the authors state that duration is not the main factor of a dysfunctional bereavement (Prigerson et al., 2008). The frequent change of terminology which has taken place in the past years has not been useful for developing standard diagnostic criteria. The lack of consensus regarding specific terminology for the phenomenon of "not normal" grief reactions also mirrors the lack of consensus regarding the conceptualisation of CG. A conceptualisation of CG is needed to understand how bereavement is processed on an individual level. The future diagnostic criteria of CG in the DSM should be based on a clear conceptualisation of complicated grief, because this will have important implications for both diagnosis and treatment. Recently, a number of conceptualisations of CG have been suggested. In the following, we will describe the most relevant conceptualisations which have also partly influenced the current diagnostic criteria proposed by Prigerson et al. (2009). Horowitz and colleagues (1997) based their conceptualisation of CG on the stress response theory, which views bereavement as a stressful life event. Horowitz (2006) suggested a general change of the DSM categories, in the sense that PTSD should be removed from the category 'Anxiety Disorders' and a new category 'Stress Response Syndromes' should be created (Horowitz, 2006). These Stress Response Syndromes would include psychiatric disorders that are caused by the experience of stress: PTSD, Adjustment Disorder, Acute Stress Disorder, Stress Induced Psychosomatic Disorder, and Complicated Grief.
Another approach would be to conceptualise CG in terms of depression (Clayton, 1990). The depression model of bereavement defined CG as 'con-tinued depressive symptom', which needed to be treated as a depression (Clayton, 1990). However, a number of studies have shown that CG was distinct from depressive disorders: factor analytic studies have shown a significant distinction between grief and MDD or anxiety disorders (Boelen, van den Bout, & de Keijser, 2003;Bonanno et al., 2007). Furthermore, though research has shown that tricyclic antidepressants have proven to be effective in reducing depressive symptoms following the loss of a loved one, they did not prove to be particularly effective in ameliorating symptoms of CG (Reynolds, Miller, Pasternak, Frank, Perel, & Cornes, 1999). These findings are in line with another conceptualisation of CG which was based on the distinctiveness from other disorders (Stroebe, van Son, Stroebe, Kleber, Schut, & van den Bout, 2000). After the death of a significant person, most people experience a wide variety of symptoms, but also meet criteria for disorders such as MDD (Maercker, Forstmeier, Enzler, Krüsi, Hörler, Maier, & Ehlert, 2008;Zisook, Shuchter, Sledge, Paulus, & Judd, 1994), PTSD (Schut, de Keijser, Bout, & Dijkhuis, 1991) and other anxiety disorders (Jacobs, Hansen, Kasl, Ostfeld, Berkman, & Kim, 1990). Therefore, CG had often been subsumed under other disorders, even though bereavement, depression or trauma did not always overlap, and the unique symptoms of CG (i.e., strong yearning for the deceased) were not captured by the symptom criteria of these disorders (see Lichtenthal et al., 2004). The high comorbidity with these disorders added to the considerable confusion over the precise nature of symptoms that constituted a CG reaction.
An additional conceptualisation of CG was the focus on relational aspects of bereavement (Rubin et al., 2008). The cognitive-emotional organisation of the relationship to the deceased plays an important role in bereavement dysfunction. Therefore, the evaluation of the nature of the continuing relationship might help to explain the grief process. Because of its high relevance to any treatment and diagnoses of CG, the influence of relational aspects will be described in a later, separate section.
In their overview of conceptualisations of CG, Stroebe and colleagues (2000) described how some experts have identified subtypologies of CG. These subtypologies included delayed, absent, unresolved, and chronic forms of grief. In the earlier proposed diagnostic criteria the research groups of Horowitz and Prigerson (Horowitz et al., 1997;Prigerson et al., 1999) mainly focused on grief processes for which the time spans were too long and too intensive. Aspects of delayed or absent complicated grief reactions were not considered at that time. Further, Stroebe and colleagues (2000) were right in saying that the concentration only on the high intensity of grief reactions as indicator for CG is problematic. Those bereaved individuals who avoid grieving or show delayed grief reactions may mask complications in their grief process. Further, if the complication lies in one specific symptom (e.g., yearning, guilt), a checklist might not reflect this problem (Stroebe et al., 2000).
A case of a patient illustrates the problem that complications may be masked: A 76-year-old woman lost her husband five years ago. The couple had a very good and close relationship together and enjoyed each other's company. The husband then died suddenly of a stroke. The patient was shocked, but organised the funeral and described herself during the immediate period after the loss as sad, though she could not cry at the funeral or afterwards. During the first year and a half, she managed very well, travelled a lot, and often visited her family abroad. On one of these trips, she suddenly became fully aware of the death of her husband and started crying. She felt a strong sense of yearning and physical pain with regard to her deceased husband. Subsequently, she experienced strong longing for her deceased husband, and cried every time uncontrollably when she saw a reminder of him in the house or when she talked about him. Even though she had a functioning social life, she felt deeply saddened as soon as she thought of her husband and started to cry. She did not, however, have any feelings of bitterness or anger, she did not feel numb, she could accept the loss, and she did not avoid reminders of the loss. Not until five years after the death did she start looking for therapy. This case illustrated very clearly two problems of the two current diagnostic criteria. First, the delayed/absent form of grief is not considered in the diagnoses, which might exclude a number of individuals who suffer from pathological grief (we will expand on this below). Second, if an individual suffers only of one or two specific symptoms (e.g., yearning) a diagnosis of CG cannot be made, even though the individual may suffer greatly.
The above example suggests that the subgroups of pathological grief processes (delayed, absent, chronic) would require multiple classifications if they were to be taken into account (Stroebe et al., 2000). However, there is so far very little scientific evidence to support the subtypologies of CG, because only a few studies have been conducted to investigate their relevance. In the most recent classifications of CG there has been a stronger focus on high intensity, chronic or prolonged grief. In conclusion, looking at the various conceptualisations, it becomes obvious that there is little agreement: instead, there is a great diversity of different models of CG. Even though there is growing evidence that CG is a distinct disorder, there is a lack of agreement between different researchers.

Diagnostic criteria for complicated grief
In recent years, a great deal of research has been carried out, mainly by Prigerson and colleagues (e.g., Prigerson & Jacobs, 2001) and Horowitz and colleagues (e.g., Horowitz et al., 1997), with the aim of defining diagnostic criteria for CG for the following edition of the DSM. In 2009 the research groups of Prigerson and Horowitz joined forces and developed a new diagnostic category, combining both criteria. Before we describe the new proposed criteria (Prigerson et al., 2009) in more detail, we would like to describe the original research work of both groups in order to achieve a better understanding of the historical development of the newly proposed criteria. Prigerson and colleagues (1995) began to evaluate symptoms of CG empirically after they found a distinct cluster of symptoms which forms a unified component of emotional distress that is clearly different from depression and anxiety (Prigerson, Maciejewski et al., 1995). These results were found in three independent samples of widows and widowers (Prigerson, Frank, Kasl, Reynolds, Anderson, Zubenko et al., 1995;Prigerson, Maciejewski et al., 1995). In 1997, a panel of experts led by the Prigerson research group, met to discuss the advantages and disadvantages of establishing diagnostic criteria, and agreed that there is evidence that CG is a symptom cluster which is distinct from depression and anxiety and which can predict mental and physical impairment. Consensus criteria proposed for CG by  were formerly defined in two categories: (a) symptoms of separation distress (i.e., preoccupation with thoughts of the deceased person, longing and searching for the deceased, loneliness after the loss); and (b) symptoms of traumatic distress, such as disbelief about the death; anger and feeling shocked, avoiding reminders of the deceased, feeling purposelessness and futility about the future, feeling that life is empty and unfulfilling without the deceased, having a fragmented sense of trust, security and control (Prigerson & Jacobs, 2001). In a preliminary test of the consensus criteria for CG analyses were conducted on data collected from the San Diego widowhood study (Zisook, Shuchter, & Lyons, 1987) with the Widowhood Questionnaire (Zisook et al., 1987), which covered nearly all the symptoms of the consensus criteria of CG. Receiver operator characteristics (ROC) analyses tested the performance of the proposed criteria on 306 widowed respondents at seven months post-loss with a mean age of 61 years (SD = 10.4). Each item was evaluated to determine its ability to identify individuals suffering from true cases of CG. Two items (avoidance and 'difficulty imagining a fulfilling life without the deceased') were deleted and the internal consistency coefficient improved after deletion (Prigerson & Jacobs, 2001). Originally, a set of symptoms which persist for more than 2 months were taken to give an appropriate marker for dysfunction; later the duration time of disturbance was increased to 6 months (Latham & Prigerson, 2004). The 6-month duration criterion was chosen because it might more easily distinguish bereaved individuals who are suffering more chronic stress from individuals with more temporary stress.
Based on these diagnostic criteria, a widely utilised assessment tool was developed called the 'Inventory of Complicated Grief' (ICG), later renamed the 'Inventory of Traumatic Grief' (ICG), and after these changes again to the 'Inventory of Complicated Grief' and recently into the 'Inventory of Prolonged Grief' -a questionnaire which provides a self-report symptom severity score. The study used to test the consensus criteria (Prigerson & Jacobs, 2001) has a number of limitations, which are important to acknowledge. First, the group of elderly widows and widowers were not entirely random or unbiased: only 34% responded to the initial assessment; people who did not participate might have been more distressed than those who participated. Low mean levels of the proposed symptoms of CG support this case. Second, two thirds of the spouses lost their partner after prolonged illness, which so far has not been identified as a risk factor for CG. Therefore the sample used to base the consensus criteria on might report biased results. The sample is typically associated with comparatively low complications; therefore it probably shows low levels of CG and might thus not be the ideal sample to validate the criteria on.
The diagnostic criteria of Prigerson & Jacobs (2001) caused much discussion. Hogan et al. (2003) conducted a study with bereaved parents using the diagnostic criteria of CG (Prigerson & Jacobs, 2001). Bereaved parents (N = 166), who lost their child through automobile accidents (65%), suicide (15%), homicide (9%), illness (7%) or other (4%) participated in this study (Hogan et al., 2003). The results showed that the basic criteria 'separation distress' and 'traumatic distress' could not be isolated as distinct constructs in confirmatory factor analyses. The conceptualisation as a diagnostic entity did not provide an adequate fit of the model. Hogan and colleagues (2003) suggested that the separation distress and the traumatic distress criteria needed further investigation and refinement with other samples, which also include subtypes of bereavement (e.g., traumatically bereaved, suicide survivors). This argument is important, because both samples on which the algorithms for complicated grief, respectively prolonged grief, are based on were conducted with elderly widowers and widows. However, a rejoinder was published by Prigerson and Maciejewski (2006), where they claim that the examination of the data by Hogan et al. (2003) had not been carried out in accordance with a fair and unbiased standard.
A second group of researchers (Horowitz et al., 1997) have published parallel criteria for CG disorder (Horowitz et al., 1997), based on stress response theory. According to Horowitz, CG disorder has a generic relationship to PTSD and acute stress disorder (ASD), resulting from exposure to a stressful event. In a 1984 study (Horowitz, Marmar, Weiss, DeWitt, & Rosenbaum, 1984), they first showed that symptoms of prolonged grieving resulted in the following symptom cluster: intrusion, avoidance, and failure to adapt to the loss. Later, Horowitz and colleagues (1997) published criteria for CG disorder and constructed operational definitions of these symptoms. This CG module (Horowitz et al., 1997) comprised of 30 items includes the following symptoms: intrusions (e.g., unbidden memories, frequent reminiscences of life with the deceased), avoidance (e.g., avoiding places that evoke the deceased) as well as maladaptive behaviour (feeling alone or empty, trouble sleeping). The symptoms should persist longer than 14 months after the loss. In the event of CG, intrusions manifest themselves in the recurrent realisation of the absence of the lost relationship, leaving a painful reminder of the empty space left by the deceased. An additional form of intrusion has also been identified: the bereaved person may wilfully indulge in positive memories and images of the deceased to the extent that it becomes a problem in the process of re-orientation in the present. Therefore, the positive and negative memories and images of the deceased may interfere with assessment of the grief process. Avoidance manifests itself in various ways: staying away from places or people; avoiding talking about the deceased in the family. Finally, the third criterion, namely failure to adapt, may be observed in feelings of being far too much alone or unusual levels of sleep disturbance. Recent research has found evidence for the stress response operationalization of CG (Langner & Maercker, 2005). Using the stress response model of CG (Horowitz et al., 1997), Langner and Maercker (2005) examined a sample of 75 participants who had lost their children, parents or spouses. The authors could confirm the classification of the symptoms into intrusion, avoidance, and failureto-adapt categories. ROC analyses indicated high diagnostic accuracy and showed predictive validation to standard measures of distinct disorders and normal grief reactions.
In a joint effort to integrate the previous two approaches, Prigerson and colleagues (2009) determined a new specific algorithm for criteria for Prolonged Grief Disorder (PGD). The main difference to the previous approach by Prigerson et al. (2009) is to include avoidance symptoms in the definition of the disorder. The new criteria involve the experience of yearning and at least 5 of the following 9 symptoms experienced at least daily or to a disturbing degree: avoidance of reality of the loss; emotional numbness; feeling stunned; feeling that life is meaningless; confusion of identity; mistrust; difficulty accepting the loss; bitterness over the loss; and difficulty moving on with life. Symptom disturbance must last at least 6 months from the death and be associated with functional impairment.
The new criteria proposal for PGD still lacks comprehensive validation since it has so far only been tested in a field trial of the Yale Bereavement Study, a study which only included elderly widowed persons. However, any revised version of the diagnostic criteria requires renewed testing with new data and valid inferences across subgroups (e.g., bereaved parents, widows and widowers, the traumatically bereaved, the elderly), and it is therefore still too early to say if these new and integrated concepts can address the requirements for psychodiagnostic criteria.
Comparison of the two diagnostic criteria Prigerson and Jacobs (2001) and Horowitz et al. (1997) Both Prigerson and Jacobs (2001) and Horowitz et al. (1997) have emphasised the impairment of social functioning, and severe symptoms of separation distress, which Horowitz et al. characterises as intrusive symptoms. The Horowitz group and the consensus panel of experts have proposed a similar set of criteria independently. However, there are differences which are important to acknowledge, such as the symptom of avoidance, sleep disturbances, functional disturbance, and the duration of bereavement. While avoidance is one of the core symptoms of the criteria set of Horowitz et al. (1997), Prigerson and Jacobs (2001) omitted the symptom of avoidance due to low specificity and item-total correlation in order to increase the diagnostic accuracy of the traumatic stress set. Other studies (Raphael & Martinek, 1997) observed that bereaved people wish to avoid reminders of the absence of the deceased person, while people with PTSD avert reminders of the trauma. While some researchers (Kaminer & Lavie, 1993) suggest that avoidance can be an adaptive way of coping with the loss, some theorists (e.g., Horowitz et al., 1993) claim that grief reactions persist if the emotional processing and adjustment is prevented due to avoidance behaviour. The symptom of avoidance can interfere with the integration of the death of a close person into existing schemas and the development of new schemas (Horowitz et al., 1993). Research shows that there is increasing evidence of avoidance processes in CG (Boelen, van den Bout, & van den Hout, 2003;Langner & Maercker, 2005;Schut et al., 1991). Boelen, van den Bout, and de Keijser (2003) evaluated bereaved individuals who had been confronted with the death of a close relative, looking at the role of negative interpretations of grief reactions in emotional problems after bereavement. They reported that avoidance behaviour was significantly related to the severity of CG and depression. The degree to which mourners experienced their grief reactions as distressing influences the degree to which they engage in avoidance strategies that are likely to impede recovery and prolong grief reactions. Therefore, taking the relevance of avoidance behaviour into account, the avoidance item has been added to the algorithm of the new Prolonged Grief Disorder (Prigerson et al., 2009).
Another unique symptom of the diagnostic criteria of Horowitz et al. (1997) is interference with sleep, while sleep disturbances, presumably reflecting hyperarousal, was also omitted from the original consensus criteria (Prigerson et al., 1999). The latter decision was based on a sleep study with 65 recently bereaved people over 60 with varying levels of CG symptoms (McDermott, Prigerson, & Reynolds, 1997). The CG symptoms were associated with mild subjective sleep disturbance but there were no main effects on electroencephalographic (EEG) sleep measures. However, newer studies show that people suffering CG actually show poor sleep quality (Forstmeier & Maercker, 2007;Germain, Caroff, Buysse, & Shear, 2005). In addition to these differences, the two sets of criteria also differ with regard to the time relationship to the death.
Further differences of the two diagnostic criteria could be shown regarding prevalence of CG. In a treatment study for CG (Wagner & Maercker, 2008), in which both diagnostic criteria systems were used (Horowitz et al., 1997;Prigerson & Jacobs, 2001), comparably more patients were diagnosed with CG at pretreatment and posttreatment assessment when the Horowitz Grief Module Scale was applied. Similar findings were found in a study of bereaved family members, who lost a significant person through assisted suicide (Wagner & Maercker, submitted). In this study, 13.4% fulfilled the complicated grief criteria of Horowitz et al. (1997) and 5% met criteria of Prigerson and Jacobs (2001). These findings are in line with results of Forstmeier and Maercker (2007), who assessed the two diagnostic systems in a sample of 570 elderly people using data from the Zurich Older Age Study, and compared the respective findings with regard to prevalence. The prevalence of CG as measured by the two diagnostic systems differed widely: 4.2% (n = 24) when the Horowitz et al. (1997) criteria were applied, and 0.9% (n = 5) when the Prigerson and Jacobs (2001) criteria were applied. There was little overlap: only 0.3% of participants were diagnosed with CG according to both systems. The likelihood of developing CG after experiencing bereavement was 22.2% when the Horowitz et al. (1997) criteria were applied, and 4.6% when the Prigerson and Jacobs (2001) criteria were applied. These findings imply that the Horowitz et al. (1997) criteria are less strict and more inclusive than the Prigerson and Jacobs (2001) criteria. It is therefore vital that clinicians and researchers indicate which diagnostic system they have used. But it also shows clearly that the two criteria sets do not measure the same syndrome criteria. These findings demonstrate the importance that further research and studies are needed to investigate the validity of CG disorder as a diagnosis, and verification of the validity of the diagnostic criteria is warranted before declaring this phenomenon appropriate for inclusion in standardised psychiatric taxonomies.

Trauma and bereavement
A finer-grained consideration of the distinction versus overlap between trauma and bereavement is also essential when considering potential criteria for complicated grief. The loss of a significant person is by nature a shocking and very often traumatic event. One major concern is how to address the overlap between trauma and bereavement. There is bereavement with and without a traumatic experience and trauma with and without a bereavement experience. Additionally, a bereaved individual can suffer after the loss of a significant person from PTSD and CG, or only from CG or only PTSD.
In DSM-IV, an event can be considered traumatic if it includes the experiencing, or confrontation with actual or threatened death (DSM-IV, APA, 1994, p. 424). Further, the stressor criterion is defined as "learning about unexpected or violent death … experienced by a family member or close associate". Therefore, witnessing or learning of the death permits the diagnosis of PTSD because bereavement could be considered a traumatic event. But there are also arguments that bereavements, which occur under normal non-traumatic circumstances, are not necessarily traumatic events (Stroebe & Schut, 2006). For example, Stroebe and Schut stated that an expected and peaceful death of an elderly person should generally not be included in the definition of a traumatic event, though there might be individual cases in which the same situation could be traumatic, for example, for a grandchild.
The interaction between trauma and bereavement has been described in different ways, representing different research interests, in the scientific community (Stroebe et al., 2001). One position focuses on the phenomenology of the bereavement reaction rather than on the type of stressor, and proposes that trauma and bereavement are distinct (Pynoos, Nader, Frederick, Gonda, & Stuber, 1987;Raphael & Martinek, 1997). Other groups suggest that bereavement should be considered a traumatic event, and that therefore CG can be subsumed under PTSD (Figley, Bride, & Mazza, 1997). Yet another research group proposes that the diagnostic criteria 'traumatic grief' should be based on the nature of the death event (Green, 2000;Rando, 2000). The final position is the focus of recent research on CG, which posits that CG can occur as a consequence of both traumatic and nontraumatic bereavement (Horowitz et al., 1997;Prigerson & Jacobs, 2001).
Two case examples will illustrate the complexity of the problem of overlapping diagnoses: A 47-year old married woman lost her 13-year-old only son, who committed suicide by lying down in front of a train. The son's death was completely unexpected and came as a great shock to her and her husband. The bereaved mother suffered positive and negative intrusions of her son: on the one hand, she had visual images and nightmares of her son lying down in front of the train, but on the other hand, she also had positive intrusions of the last birthday of her son and how happy he had been getting his new computer. She had strong feelings of guilt and yearned and pined for her lost son. The patient had high PTSD and CG scores. In the second case, a 35-year old bereaved mother grieved the death of her 7-year-old daughter, who had suffered from cancer for three years and who died at home, surrounded by her family. Even though the death of her daughter was expected, the final hours and the actual death of her daughter were experienced by her mother in a traumatic way. When the undertaker came to carry out the body of her daughter, she started to scream and refused to let the men take her daughter's body out of the house. Four years after the death of her daughter, she still suffered intrusions relating to the moment of dying and the circumstances which surrounded the death. The patient showed high PTSD and CG scores when beginning with psychotherapy. These two examples clearly illustrate the complexity of CG as a distinct disorder and the difficulties in differentiating between the three categories of bereavement: non-traumatic bereavement, traumatic bereavement and trauma. These categories can overlap and can be distinct from each other. The diagnosis of PTSD in traumatic bereavement has been shown in a number of studies. In a study on conjugally bereaved individuals who have lost their spouse in a natural way, 10% met diagnostic criteria for PTSD (Zisook, Chentsova-Dutton, & Shuchter, 1998). By contrast, more than one-third of study participants who lost their spouses of suicide or accidents suffered of PTSD. These studies suggest that loss due to violent death may greatly influence an eventual grieving process.
Nevertheless, in recent years, a number of studies have shown that PTSD and CG are distinct disorders (Momartin, Silove, Manicavasagar, & Steel, 2004;Silverman, Jacobs, Kasl, Shear, Maciejewski, Noaghiul et al., 2000). For example, Silverman and colleagues (2000) found in their study of bereaved spouses (N = 85) that 18% met the criteria of CG and 7% met the criteria of PTSD at 4 months after their loss. Momartin et al. (2004) conducted a study with Bosnian refugees (N = 126) who had experienced particularly tragic losses. Most of the losses occurred either in public, or in concentration camps in the form of murder and torture, witnessed by relatives and friends. More than half of the participants showed PTSD, whereas only widowhood, but not gender or age, was associated with CG. Nevertheless, PTSD was unrelated to CG, a finding that might support evidence that the syndromes are for the most part distinct.
The question arises as to what the differences are between a diagnosis of PTSD or CG. Research conducted in the past few years showed some important features of different symptomatology between the two disorders. Bereavement reactions involve distinct anxiety reactions: traumatised individuals are typically anxious about the threat experienced with the traumatic event, whereas bereaved individuals experience separation anxiety (Stroebe et al., 2001). A sense of safety is often diminished after experiencing a trauma, while this does not typically occur with bereaved people after a nontraumatic bereavement. The core symptoms of yearning, pining, feeling sad and lonely are also not necessarily experienced after trauma without bereave-ment. Similar symptoms of the two disorders are e.g., intrusive thoughts, even though they appear to be qualitatively different. Intrusions of PTSD involve negative and distressing memories of the traumatic event and related memories (Horowitz et al., 1993), whereas CG intrusions after a non-traumatic bereavement are typically of the deceased person and can be also experienced as positive and comforting. These positive and treasured memories can be permitted to such a degree that they have maladaptive qualities and prevent the bereaved person from re-orientation (Horowitz et al., 1993). Bereaved individuals can also experience the symptom of avoidance in different ways than traumatised individuals. Traumatised individuals typically avoid reminders of the traumatic event, whereas bereaved people might avoid places, people and conversations related to the deceased person, but also specifically seek out reminders (Stroebe et al., 2001). However, if a bereaved individual has experienced the loss in a traumatic way, avoidance and intrusions might similarly be focusing around the traumatic circumstances, and it might be extremely difficult for the bereaved person experiencing positive memories of the deceased. One important distinction is the separation distress component. Yearning and searching for the deceased, intrusive thoughts about and longing for the deceased and the loss of the person are often the source of distress.
Another different feature is the impact intensity of the traumatic event. The extremity of a traumatic event has a strong impact on the PTSD symptoms. The greater the level of traumatic exposure, the greater impact has this on the individual who experiences the trauma (Green, 2000). However, in traumatic bereavements, the enormity of the impact is strongly combined with aspects of the relationship, which has a strong influence on the grief process (Stroebe & Schut, 2006). In traumatic bereavements, the kind of relationship to the deceased and the closeness to the deceased have an additional impact.
Concluding, it appears that -even though there seems to be evidence that PTSD and CG are distinct disorders which can also overlap -one question regarding future diagnostic criteria which remains is whether the two different types of bereavement (non-traumatic and traumatic bereavement) should not be taken into account to a greater extent than is currently the case. Would we not obtain more valid diagnoses if we would suggest CG as one diagnostic criterion with two separate subcategories, namely traumatic bereavement and non-traumatic bereavement? The current diagnostic criteria were developed on the basis of a sample of elderly widows and widowers who experienced losses, which were not specifically traumatic or violent, and therefore they might very well identify normal or non-traumatic bereaved individuals. However, these samples might not necessarily mirror the algorithm of traumatically bereaved individuals. If the consensus criteria would have been based on traumatically bereaved individuals (i.e., suicide survivors), they would very likely find a different CG algorithm that would reflect more as-pects of (negative) intrusions, relational aspects, feelings of guilt, and/or trauma-related avoidance behaviour. The question therefore remains whether the traumatically bereaved can in fact be satisfactorily identified with the present proposed criteria. The inclusion of the mode of death (non-traumatic or traumatic) into the criteria would provide a better conceptual and empirical background for future research. There are, however, also other important aspects of bereavement which could give us more diagnostic validity about the grade of complication of a grief process. In the following, we will describe the factor of the relation to the deceased as an indication of the level of suffering.
Relationship to the deceased Another feature relating closely to complicated grief that has frequently been omitted from consideration in defining and categorising complicated grief has to do with the bereaved person's past and present relationship to the deceased. For the surviving person, the death of a significant person often leaves an empty space, which -especially shortly after the death -cannot be easily comprehended. However, after the immediate shock period, the bereaved individual often experiences that the deceased person has not just left an 'empty space', but that the relationship to the deceased is continuing, and in fact the relationship to the deceased person is a central factor for the measurement of the grief process over time (Rubin et al., 2008). The importance of the interpersonal relationship to the deceased has been the focus predominantly of the psychodynamic and the attachment theories (Rubin, Malkinson, & Witztum, 2000). But clinicians, too, experience that the emotional attachment and the re-organising of the relationship to the deceased are important aspects of the grief process. Other features of the relationship, such as the type of the relationship or closeness, also have an influence on grief reactions (Bowlby, 1980;Rubin, 1999). Rubin (1999) developed the Two-Track Model showing that the bereavement process occurs along two main tracks. The first track refers to the bereaved's functioning or is living his or her life affected by the loss, and the second addresses how the bereaved individual is maintaining the inner relationship with the deceased. Rubin Malkinson, and Witztum (2003) described the death of a significant person as an attack on the relationship and the internal representation of that person needs therefore to be reorganised (Rubin et al., 2003). They describe this process as a shift from a psychological relationship with a living person to a person who is now deceased, a potentially traumatic experience. According to the authors, the coherence and association of the deceased person in the mind of the bereaved plays a prominent role in the bereavement process. The internal representation of the deceased person can be very complex. The memories of the deceased person can be very emotional, intrusive, and can give discomfort as well as a sense of well-being, and the nature of the relationship to the deceased gives important information of the perception and the relation to the deceased. Therefore, Rubin and colleagues (2008) emphasised the necessity to include assessment of the cognitive-emotional organisation of the relationship with the deceased in the future DSM. They stated that the nature of the post-loss relationship can give an important understanding of the complications of grief. In the proposed criteria the items 'yearning' and 'strong emotion' mainly represent the post-loss relationship to the deceased. Aspects of how the psychological relationship with the deceased continues should be included in conceptualisation of any classification of bereavement dysfunction (Rubin et al., 2008). Interventions could especially profit from the inclusion of the relational assessment. The continuing bond and how this can be actively involved in the life without the deceased is an important focus of most grief interventions. Therefore, not only should the symptoms and dysfunction criteria be taken into account, but the relational aspects should also be considered.

Conclusions
During the past years, researchers and clinicians have emphasised the need to recognise CG as a separate diagnostic category with the status of a psychiatric disorder in the forthcoming DSM-V. A large number of studies has been conducted, notably by Prigerson and colleagues (e.g., Prigerson & Jacobs, 2001, Prigerson et al., 2009) and Horowitz and colleagues (e.g., Horowitz et al., 1997). Studies provided evidence that a small but important minority of bereaved individuals show a more extreme and enduring grief process. Therefore, the attempt of researchers to identify these individuals with a diagnostic approach is understandable. But do we really need a separate, multifaceted taxonomy of grief reactions for this small minority? Or, is the current approach of the DSM-IV at this stage of research possibly the best solution? This chapter has reviewed the different theoretical frameworks and different perspectives and important features of CG from various researchers with different conceptual backgrounds. However, many important questions remain concerning the conceptualisation, validity of the syndromes and the distinction between traumatic bereavement and trauma, as well as between normal versus pathological forms of grieving.
A main criticism of our review is the absence of a clear conceptualisation of grief processes and the problem of different conceptualisations which exist in the literature. Sometimes concepts of CG overlap, but in general, there is little agreement between the research groups and the symptoms they focus on when they conduct their empirical studies. Some conceptualisations concentrate on the distinctiveness of CG from other psychiatric disorders, others focus more on relational aspects, or subtypologies of grief. Unfortunately, very little consideration has been given to cultural or relational aspects of normal and complicated grieving (Stroebe et al., 2001). Bonanno and Kaltman (2001) suggest there is a pressing need for cultural comparative grief assessment in a longitudinal design to identify what is human bereavement in general. It is to be assumed that different cultures show different grieving patterns, but so far we know little about normal or pathological grief processes in different cultures. This is even more important since patients often come from different cultural backgrounds, due to migration. The grieving process of a Swedish bereaved individual might be very different from a bereaved person with a Bosnian or Turkish cultural background, even if they are born in the same country.
We also described the historical development and differences of proposed diagnostic criteria (Horowitz et al., 1997, Prigerson et al., 19992009) and our review showed that there are a number of nosological and conceptual concerns for a criteria set for complicated grief. Over the past years, the research group of Prigerson and colleagues undertook the most efforts at finding a diagnostic criterion for CG. Various studies, the development of a consensus criterion, and the recently published psychometric validation of criteria proposed for DSM (Prigerson et al., 2009) have been conducted. While Prigerson and colleagues (2008) claim that there is consensus about the syndrome of CG, we are in line with Stroebe and Schut (2006) that there is still a large diversity and considerable disagreement. One main concern is the approach of traumatic and non-traumatic bereavements in CG in a diagnostic taxonomy. As stated, we have concerns about the algorithm of CG and Prolonged Grief (Prigerson & Jacobs, 2001;Prigerson et al., 2009), which are both based on elderly widows and widowers, a specific subgroup, and only a very small number of whom have experienced traumatic bereavement. There is therefore a need for a systematic comparison of grief severity for specific subgroups of bereavement, such as bereaved parents, elderly widows and widowers, the normally bereaved, the traumatically bereaved, bereaved adolescents, and prenatal or neonatal losses.
As already stated above, the most recent concept of Prolonged Grief Disorder (Prigerson et al., 2009), where the two categories 'separation distress' and 'traumatic distress' were substituted by the categories 'Separation Distress' and 'Cognitive, Emotional and Behavioural Symptoms', remains hampered by a lack of robust empirical evidence. It would therefore be sensible to investigate and validate this new group of criteria again especially in relationship to traumatic and non-traumatic bereavement. These investigations are even more urgent since the proposed algorithm for diagnostic criteria of Prolonged Grief needs more research-based evidence which is also valid for different subgroups (Hogan et al., 2003).
The question remains as to what the advantages or disadvantages of CG with its own separate diagnostic criteria would be. Ideally, a consensus about the diagnostic criteria for CG and its inclusion in the nosology would lead to more research and controlled studies. More research would result in suitable new treatment possibilities for individuals with CG, which would reduce the psychological strain on these people. In addition, more research would help clarify the phenomenology of CG and its risk factors. It would be worthwhile exploring whether convergent evidence from other post-stress conditions (e.g., PTSD, adjustment disorders) could enrich the body of knowledge on CG (Maercker, Einsle, & Kollner, 2007).
However, including CG in the nosology could also result in disadvantages. One concern could be that a CG diagnosis may pathologize or stigmatise a normal grief reaction (Prigerson & Jacobs, 2001;Stroebe et al., 2000;Stroebe et al., 2001;Lichtenthal et al., 2004).
In conclusion, the inclusion of CG in one of the leading guides for researchers and practitioners such as the DSM would have a far-reaching impact (Stroebe et al., 2000) and there are good arguments for a new diagnostic category. However, this review has come to the conclusion that at this point of research the inclusion of CG as an entity may still be too early. Too many important issues concerning the nature of CG remain unanswered, and preliminary acceptance of the proposed criteria might lead to the converse of the original aim. So far research in the field of diagnostic criteria has focused on non-traumatic samples. However, traumatic and non-traumatic bereavement might differ widely regarding intervention, and especially clinicians who work with specific groups of bereaved patients might need to focus on different conceptualisations. For example, practitioners who have mainly traumatically bereaved patients might focus more on the stress-response theory concept. A diagnostic criterion should therefore first identify those who actually suffer from traumatic or non-traumatic bereavement. In a next step the appropriate diagnoses and, at a later date, intervention should be offered.