To identify studies in the traditional medical literature as well as the complementary and alternative medicine literature, we searched the following databases: MEDLINE; PsychINFO; Cumulative Index to Nursing and Allied Health (CINAHL); BIOSIS Previews; ISI Science Citation Index Expanded and Social Sciences Index; Evidence Based Medicine Reviews (EBMR), including the Cochrane Database of Systematic Reviews (DSR), the Cochrane Controlled Trial Registry (CCTR), Database of Abstracts of Reviews of Effectiveness (DARE), and the American College of Physicians' (ACP) Journal Club Review; Sociological Abstracts; Alt HealthWatch; and Wilson Web from 1966 to 2003. We identified all relevant articles on bereavement care interventions by using the primary search terms of "bereaved or bereavement" and "grief", combined with secondary descriptors of "intervention or support or counselling or therapy" and "controlled or trial or design".

Our inclusion criteria specified that each study: 1) addressed the treatment of bereaved individuals, and 2) included an evaluation of a selected method of therapy aimed at reducing the grief reaction due to bereavement. We considered only articles written in the English language. We then reviewed the titles and abstracts of all articles we retrieved through our initial database search, and obtained the full texts of all applicable studies. We also reviewed the references in all applicable studies for additional pertinent studies.

The full articles of all studies that met inclusion criteria and passed subsequent title and abstract reviews were retrieved and examined independently by two of the authors. Each article was reviewed for measured outcomes, patient and decedent characteristics, and intervention characteristics. These measures included sample size, type of intervention, length of intervention, patient's relationship to the deceased, time since the bereaved death, and patient demographics. Data was extracted and any disagreements were resolved through discussion, clarification, and consensus within the research team.

The initial literature search generated 737 citations. Elimination of duplicate citations yielded 340 references. 2 studies, written in Chinese and Spanish, were excluded. Reviewing the titles culled the sample to 243 citations, and a review of the abstracts found 87 of these to be potentially relevant. Of these, 9 were dissertations, 2 were irretrievable, 2 were duplicate publications of the same study, and 15 were ineligible because they did not meet our inclusion criteria. The resulting set of 74 articles was subject to review for data extraction. A list of all citations found, including those excluded from this analysis, is available [see Additional file 1].

Of the 74 studies that met inclusion criteria, almost 6,000 participants within these studies experienced a multitude of losses – of parents, spouses, children, and other loved ones who had died from a wide range of causes, both sudden and protracted. The therapies utilized and outcomes evaluated varied widely. Heterogeneity among both the outcomes and the measures used to assess similar outcomes precluded an effort to summarize data across studies, even in the form of generic effect-size measures. Furthermore, for a significant portion of the studies, concerns regarding the internal or external validity of the reported results cautioned against making quantitative summary statements regarding treatment effects.

Eight studies feature pharmacotherapy, but only four compared active therapy to non-pharmacotherapy controls, and only one study clearly reported their random allocation method (Table 1) 678910111213. These studies targeted adults and seniors, ranged in sample size from 10–80 subjects, and used a variety of drugs, including tricyclic antidepressants (TCA), selective serotonin reuptake inhibitors (SSRI), buprioion, and benzodiazepines. Overall, these studies demonstrated a statistically significant beneficial effect of pharmacotherapy on ameliorating symptoms of depression and improving subjective sleep quality 6789101113. These benefits persisted only as long as the subjects continued to receive pharmacotherapy. Pharmacotherapy was found, however, to have a mixed effect on bereavement intensity as measured by symptoms of grief (i.e., Texas Revised Inventory of Grief, Inventory of Complicated Grief). For example, Warner and colleagues (2001) did not find evidence of an effect of benzodiazepines (diazepam) on bereavement-related grief intensity12. One study combined pharmacotherapy with psychotherapy in a 16-week double-blinded factorial design trial of nortriptyline (NT) and interpersonal psychotherapy 6. The 80 patients were randomly assigned to one of four treatment conditions: NT plus interpersonal psychotherapy, NT plus medication clinic (i.e., no interpersonal psychotherapy), placebo pill plus interpersonal psychotherapy, and placebo pill plus medication clinic (i.e., no interpersonal psychotherapy conditions). Details of the psychotherapy were not described. While the results displayed a statistically significant benefit of nortriptyline over placebo regarding remission of depression, none of the treatment conditions were associated with diminishment of grief.

Support groups or counselling constituted the intervention in 39 studies, of which 23 had control groups and 15 claimed random allocation, yet only three of these included clearly described allocation methods (Table 2) 141516171819202122232425262728293031323334353637383940414243444546474849505152. Ten of these were mutual/self-help, with the majority taking the form of informal group therapy. The remaining 29 studies were professionally led support groups targeting select subgroups including parentally bereaved children, college students, and seniors, as well as many specific adult populations. Program implementation across studies varied even further. This variation was found in terms of number of sessions (one to 25), whether the sessions proceeded with full-fledged patient-driven discussion or highly structured protocols, whether attendance was mandatory or individually motivated, as well as in the nature of the group leadership and the format (individual, group, or marital). Perhaps due to these or other differences in the interventions, some studies documented study treatment effects 2226293031333452 while other studies showed no effect 151727374651.

Several studies documented substantial spontaneous improvements in bereavement symptomology in the control groups. Kay and others (1993) report a bereavement intervention for Mexican-American widows 33. They found that all widows improved on all depression scales, state anxiety, life satisfaction, and emotional and somatic symptom scales over the course of two years. However, those widows in the experimental support group exhibit significantly improved changes in these scores. Tudiver and colleagues (1992) conducted a mutual-help support group for recently bereaved widowers 17 that can be compared to Vachon and colleagues' (1980) and Barrett's (1978) widow studies 1439. Tudiver and others found significant improvement over time (baseline to eight months) for all widowers, but found no significant differences between those who received treatment and a comparison group of windowers who were on the wait list to receive treatment but had not.

Psychotherapy-based treatments, another form of psychological interventions, can be done in different formats (family, group, or individual), and via different approaches. Of the 25 studies that use psychotherapy as an intervention, approaches included cognitive-behavioral, psychodynamic, psychoanalytical, and interpersonal approaches, as well as combinations of these and modality and social support (Table 3)6192235385354555657585960616263646566676869707172. Seventeen of these studies utilized control groups, only 13 claimed randomization, and only five of these clearly stated their method of allocation.

Cognitive-behavioral therapy was employed in nine trials, four of which used individual sessions while five studies used group sessions. Murphy and colleagues (1998) studied an intervention for parents bereaved by the violent death of their children 57. The results show no treatment effect between intervention and control groups over the five main tested outcome variables. The authors then proceeded with a post-hoc subgroup analysis, which identified mothers with high Global Severity Index scores and grief at baseline as potentially benefiting from intervention during the period, while fathers who received the intervention appeared to have more posttraumatic stress disorder (PTSD) symptoms at six-month follow-up.

Kleber and Brom (1987) conducted a comparative outcome study of three forms of short-term psychotherapy 69. They compared the results of 83 patients suffering from a major loss who had been randomized into hypnotherapy (behavioral), trauma desensitization (behavioral), psychodynamic therapy, and a delayed-treatment control group. They found all three therapies successful in improving patients' conditions, but did not find any particular therapy to be significantly more effective than another. While the control group showed slight recovery, over time the three therapies were more effective in reducing symptoms of the bereavement response.

Studies of psychodynamic therapy, which strives for the patient to understand and cope better with feelings by re-experiencing them and talking them through with the aid of the therapist, was found to be quite prevalent in the bereavement care literature. Overall, the results are mixed, with more support found in the group format of psychodynamic therapy than in individual therapy. Of the studies we evaluated as psychodynamic therapy, six were individual in format, seven had a group format, and eight employed control groups; five of these claimed random allocation (one additional study randomly assigned subjects to two experimental conditions but lacked a control group).

Psychoanalysis, as exemplified by Freud, proceeds with an inward investigation of unconscious mental processes and childhood experiences as the principal therapeutic procedure. Problematic measurement methodology beset the one study that utilized a group format to provide a psychoanalytic-based intervention (with no details regarding the tasks assigned to the patients)68. This study focused primarily on the relationship between the patient's personal affect (measured by an unvalidated affect assessment scale) and a favorable treatment outcome (measured again by an ad-hoc unvalidated measure).

Behavioral therapy uses learning principles (such as behavior modification, systematic desensitization, and aversion) to eliminate or reduce unwanted reactions to either external situations, one's thoughts and feelings, and bodily sensations or functions. Behavioral therapy was used in only one study, which compared traumatic desensitization to hypnotherapy and psychodynamic therapies 69. As described in the section on cognitive-behavioral therapies above, all three therapies resulted in significant improvements from pre- to post-treatment as compared to controls, and no one therapy was found to be more effective than the others in treating bereavement-related symptoms.

Interpersonal therapy aims to improve communication skills and increase self-esteem during a short time period by focusing on a patient's behaviors and social interactions with family and friends, directly teaching how better to relate to others. Only one study used interpersonal therapy as a bereavement care intervention, and this study found no effect on grief as the only measured outcome 6.

Seven studies featured interventions that altered the manner in which the healthcare system interacted with patients, family, and friends prior to death, guided by an underlying (yet not fully explicated) notion that interactions experienced by loved ones prior to death can influence the subsequent bereavement process (Table 4) 73747576777879. Six of the seven interventions provided enhanced or augmented care, in the form of palliative care, hospice care, or care coordination. One intervention gave family members the option of witnessing resuscitation efforts 79. Overall, the studies that reported systems-oriented interventions produced mixed results of efficacy, with only three of the seven studies showing any treatment effect, mostly in long-term follow-up ranging from 60–365 days post-death. In fact, no study found significant treatment effects when measured during the intervention.

Ringdal and colleagues (2001) found no significant differences between those family members whose relative received palliative care and those who received traditional care 76. This intervention, however, was not directed to the bereaved relatives, but rather to their terminally ill relatives. The bereaved relatives did show an overall significant decline in TRIG grief scores over one year post-bereavement for both palliative and traditional care groups.

Robinson (1998) examined the psychological effect of witnessing resuscitation efforts of patients in the emergency room on bereaved relatives 79. They found no psychological differences between the control group who did not witness the resuscitation attempt and the experimental group who had the option of viewing the resuscitation effort. In fact, at the three- and nine-month follow up, the experimental group exhibited median scores lower (that is, better) than the controls on five of the eight measured scales. At nine months, the authors found the difference in TRIG2 scores approaching the 5% significance level with a reported p = 0.08. These findings provide no evidence to support the popular belief that relatives should be excluded from the resuscitation room, and provide only weak evidence of possible psychological benefit of witnessed resuscitation; they do not, however, suggest that having witnessed an unsuccessful resuscitation attempt alleviates the grief reaction of the bereaved.

As the history of science and medicine suggests, successful scientific inquiry into a topic is typically a cumulative process undertaken by a community of investigators working within a shared scientific paradigm 8081. The field of bereavement care intervention studies does not appear to be organized in such a manner, but instead consists of distinct groups of investigators working within disparate theoretical frameworks: pharmacologic, psychodynamic, psychoanalytic, behavioral, cognitive-behavioral, interpersonal, and social supportive theories each vie for attention. Indeed, although specification of an underlying treatment-theory conceptual model may improve causal inference 82, the bereavement care literature may be too invested in and reliant on theoretical justifications of treatments. Consequently, the compiled published reports demonstrate a cumulative 'Tower of Babel' phenomenon, with the different theory-dominated perspectives failing to engage each other meaningfully: the sum is no greater than the parts, and perhaps less.

Treatments featured in published studies vary almost as much as the authors who tested them. One can observe substantial variation across studies regarding the type of intervention generally or regarding the specific implementation of a specific type of intervention (such as different doses of pharmaceuticals); regarding characteristics of targeted patient populations; regarding outcome measurements and study design methodology. Scrutinizing the key outcome measures listed in the accompanying tables illustrates this remarkable heterogeneity. Although these differences have been due in part to diverse treatment-theory paradigms, even studies conducted within the same theoretical paradigm often differed markedly in terms of what potential benefit was being tested, and how it was being measured. Such substantial variation between studies stymies comparison or confirmation of treatment effects.

Aside from the pharmacological studies, which reported the dosing of the intervention medication, very few reported studies describe the intervention procedures in sufficient detail for readers to envision clearly what tasks or activities intervention subjects were asked to perform. This under-specification prevents sensible analysis, within a class of treatments (such as cognitive-behavioral therapy), of observed differences in treatment effects (since the implementation of cognitive-behavioral therapy, for instance, may have been quite different in seemingly similar intervention studies).

Inadequate specification of intervention procedures, combined with other factors at work within the community of bereavement care investigators, may have resulted in the dearth of published replication studies. This lack of replication prevents the accumulation of a body of evidence that would confirm, refute, or refine prior estimates of treatment effects.

A final factor inhibiting research progress in the realm of bereavement care interventions encompasses a number of recurring methodological flaws that greatly limit inferences regarding treatment effects. First and foremost is the omission of control groups. Control groups are essential for the valid evaluation of a bereavement intervention, particularly because of the typically self-limited course of grief: even absent any treatment, most bereaved people show "diminished pathological symptoms and fewer signs of disturbance within two years of the loss"65. Purported beneficial treatment effects observed in an intervention group without a suitable control group therefore may in fact be simply the natural grief remission process. A second common study design feature is the non-random assignment of study subjects into treatment and control groups, which again limits the strength of inference regarding observed 'treatment' effects, as these differences between treatment and control groups may be due to selection or assignment bias. Third, many studies measured subject outcomes using untried assessment tools that had been created on an ad hoc basis, and which may therefore have compromised measurement accuracy and inference validity. Lastly, studies that failed to demonstrate a statistically-significant difference for the main outcome measure often performed numerous post hoc subgroup analyses, a practice that negates the rigor of statistical hypothesis testing.

If these five factors are indeed hampering progress towards improving bereavement care interventions and quality of care for bereaved individuals, then concrete actions could facilitate progress within the field of bereavement care, specifically: 1) Convening a consensus-building conference among key stakeholders and investigators to define a specific research agenda that would draw on a limited number of theoretical paradigms and delineate key elements of treatment theory 82; 2) Focusing on interventions to improve key outcomes that are valued by bereaved individuals; 3) Targeting well-defined patient populations at well-defined phases of bereavement; 4) Conducting high-quality randomized controlled trial research designs, employing rigorous tests of hypotheses defined prior to the conduct of the study, and eschewing unplanned subgroup analyses; 5) Weighing the ethical arguments for and against the use of randomized control subjects in such research; 6) Increasing incentive to conduct and publish highly-comparable replication studies; and 7) Enforcing the adoption of uniform standards regarding clinical trial study reporting (such as outlined in the CONSORT statement 83) by journal editors and the bereavement research community.