Response was defined as a decrease in the Inventory of Complicated Grief
score of 20 points or more. CGT indicates complicated grief treatment; IPT,
Customize your JAMA Network experience by selecting one or more topics from the list below.
Shear K, Frank E, Houck PR, Reynolds CF. Treatment of Complicated GriefA Randomized Controlled Trial. JAMA. 2005;293(21):2601–2608. doi:10.1001/jama.293.21.2601
Context Complicated grief is a debilitating disorder associated with important
negative health consequences, but the results of existing treatments for it
have been disappointing.
Objective To compare the efficacy of a novel approach, complicated grief treatment,
with a standard psychotherapy (interpersonal psychotherapy).
Design Two-cell, prospective, randomized controlled clinical trial, stratified
by manner of death of loved one and treatment site.
Setting A university-based psychiatric research clinic as well as a satellite
clinic in a low-income African American community between April 2001 and April
Participants A total of 83 women and 12 men aged 18 to 85 years recruited through
professional referral, self-referral, and media announcements who met criteria
for complicated grief.
Interventions Participants were randomly assigned to receive interpersonal psychotherapy
(n = 46) or complicated grief treatment (n = 49); both
were administered in 16 sessions during an average interval of 19 weeks per
Main Outcome Measure Treatment response, defined either as independent evaluator-rated Clinical
Global Improvement score of 1 or 2 or as time to a 20-point or better improvement
in the self-reported Inventory of Complicated Grief.
Results Both treatments produced improvement in complicated grief symptoms.
The response rate was greater for complicated grief treatment (51%) than for
interpersonal psychotherapy (28%; P = .02)
and time to response was faster for complicated grief treatment (P = .02). The number needed to treat was 4.3.
Conclusion Complicated grief treatment is an improved treatment over interpersonal
psychotherapy, showing higher response rates and faster time to response.
Many physicians are uncertain about how to identify bereaved individuals
who need treatment, and what treatments work for bereavement-related mental
health problems.1 Bereavement-related major
depressive disorder is a well-recognized consequence of loss.2,3 Complicated
grief also occurs in the aftermath of loss but needs to be differentiated
from depression. Complicated grief can be reliably identified by administering
the Inventory of Complicated Grief (ICG)4 more
than 6 months after the death of a loved one. Key features of complicated
grief5,6 include (1) a sense of
disbelief regarding the death; (2) anger and bitterness over the death; (3)
recurrent pangs of painful emotions, with intense yearning and longing for
the deceased; and (4) preoccupation with thoughts of the loved one, often
including distressing intrusive thoughts related to the death.
Avoidance behavior is also frequent and entails a range of situations
and activities that serve as reminders of the painful loss. Studies indicate
that treatments for bereavement-related depression show minimal effects on
complicated grief symptoms.7,8 Complicated
grief bears some resemblance to posttraumatic stress disorder (PTSD), although
again, there are important differences.9 Factor
analysis shows that symptoms of complicated grief load separately from both
depression and anxiety.10,11 Comparisons
of complicated grief, major depression, and PTSD are listed in Table 1. Co-occurrence of complicated grief with major depressive
disorder and PTSD is also common. Prior studies indicate that rates of complicated
grief co-occurring with major depressive disorder range from 21%5 to
54%4 and co-occurring with PTSD range from
30%12 to 50%.13
Although it is not included in the Diagnostic and
Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV), complicated grief is a source of significant distress and
impairment and is associated with a range of negative health consequences.14-16 Prevalence rates
are estimated at approximately 10% to 20% of bereaved persons.17,18 Approximately
2.5 million people die yearly in the United States.19 Estimates
suggest each death leaves an average of 5 people bereaved, suggesting that
more than 1 million people per year are expected to develop complicated grief
in the United States.
Given observations regarding the specificity and clinical significance
of complicated grief symptoms, including the lack of response to standard
treatments for depression,20,21 we
developed a targeted complicated grief treatment (CGT). Since complicated
grief includes depressive symptoms such as sadness, guilt, and social withdrawal,
we used a framework for the treatment based on previous research with interpersonal
psychotherapy (IPT) for grief-related depression.22 In
view of the presence of PTSD symptoms of disbelief, intrusive images, and
avoidance behaviors, as well as unique symptoms related to the loss (eg, yearning
and longing for the deceased), we modified IPT techniques to include cognitive-behavioral
therapy–based techniques for addressing trauma. We used cognitive strategies
for working with loss-specific distress. As suggested by results of a comparison
of the 2 methods,23 we have previously found
that IPT and cognitive-behavioral therapy lend themselves to integration.24,25 Following completion of an open trial
of CGT,21 we report here the results of a randomized
controlled trial comparing CGT with standard IPT. We hypothesized that CGT
would be superior to IPT with respect to overall response rates and time to
response, with CGT producing a more rapid and greater resolution of complicated
grief symptoms than IPT.
Patients who met criteria for complicated grief, defined as score on
the ICG of at least 30, were recruited to a university-based clinic. To include
a broad range of participants, we also enrolled study participants at a clinic
attended by primarily low-income African American patients. Race was assessed
by self-report. We obtained this information as part of a concerted effort
to include low-income minorities in our study. The originally proposed sample
size was 60. Participants were randomly assigned to receive CGT or IPT in
a ratio of 1:1. Randomization was stratified by treatment site and, within
site, by violent (accident, homicide, or suicide) vs nonviolent death of a
loved one. A blinded randomization number was assigned using a computerized
random number generator without blocking. Decisions regarding eligibility
were based primarily on independent evaluator assessment and always made by
the study team prior to disclosure of the treatment assignment (Figure 1).
Treatment was provided in approximately 16 sessions over a 16- to 20-week
period. Three additional sessions could be added in the event of a second
death. Time could be extended if there was a serious life event (eg, hospitalization
for medical illness, severe stressor.) Patients whose treatment coincided
with the attacks of September 11, 2001, were offered an extra session to discuss
their reaction. Treatment could be shorter if both therapist and patient agreed
that the patient had successfully completed the course of treatment. Posttreatment
assessment was obtained on completion of treatment by evaluators blinded to
randomized treatment assignment. For early dropouts, therapists provided an
estimated global improvement score and a written paragraph justifying their
rating. The independent evaluator reviewed this information with all available
ratings prior to finalizing the response rating. Study nonresponders were
either treated openly or referred to geographically convenient or preferred
outside treatment. The study was approved by the University of Pittsburgh
Institutional Review Board. Participants were enrolled between April 2001
and April 2004.
Bereaved individuals recruited via professional referral, media advertisement,
and self-referral gave oral informed consent for a brief screening interview
by telephone (n = 405) or in person (n = 12). A subgroup
(n = 26) was recruited from the clinic with predominantly low-income
African American patients. Individuals who screened positive (n = 329)
on the ICG and signed written informed consent (n = 218) were assessed
for eligibility, initial symptom ratings, and drug stabilization for patients
taking antidepressant medication (n = 92). Patients were permitted
to take medication for depression during the study if (1) medication management
was transferred to the study pharmacotherapist and (2) medication use was
stable for a minimum of 3 months, with at least 6 weeks at the same dose.
The study pharmacotherapist made a judgment about adequacy of pharmacotherapy
and adjusted medications as necessary, prior to randomization.
Inclusion required a score of at least 30 on the ICG at least 6 months
after the death of a loved one and judgment by the independent evaluator that
complicated grief was the most important clinical problem. Individuals with
current substance abuse or dependence (past 3 months), history of psychotic
disorder or bipolar I disorder, suicidality requiring hospitalization, pending
lawsuit or disability claim related to the death, or concurrent psychotherapy
All therapists were master’s- or doctoral-level clinicians who
had at least 2 years of psychotherapy experience and who underwent extensive
training and certification in either IPT or CGT. Certification entailed completion
of 2 treatment cases in a manner judged competent by K.S. (for CGT) or E.F.
(for IPT). Therapists received ongoing group supervision, separately for IPT
and CGT, throughout the study period. Selected audiotapes or videotapes were
used in supervision sessions as a part of the discussion. Therapy sessions
were audiotaped for adherence and competence ratings, performed on a randomly
selected subset of sessions.
Interpersonal psychotherapy is a proven efficacious treatment, well
studied for the treatment of depression.26,27 Our
group has done extensive research using this treatment, and therapists in
this study had a strong allegiance to IPT. Interpersonal psychotherapy was
delivered as described in a published manual,28 using
an introductory, middle, and termination phase. During the introductory phase,
symptoms were reviewed and identified and an interpersonal inventory was completed.
Interpersonal psychotherapists used a grief focus, sometimes accompanied by
a secondary focus on role transition or interpersonal disputes. The relationship
between symptoms and grief and other interpersonal problems was discussed.
The middle phase was used to address grief and other interpersonal problems,
as indicated. The IPT therapist helped patients to arrive at a more realistic
assessment of the relationship with the deceased, addressing both its positive
and negative aspects, and encouraged the pursuit of satisfying relationships
and activities. In the termination phase, treatment gains were reviewed, plans
were made for the future, and feelings about ending treatment were discussed.
Complicated grief treatment, delivered according to a manual protocol,
also included an introductory, middle, and termination phase. In the introductory
phase, the therapist provided information about normal and complicated grief
and described the dual-process model of adaptive coping, entailing both restoration
of a satisfying life and adjustment to the loss.29 This
model posits that grief proceeds optimally when attention to loss and restoration
alternate, while coping with both processes proceeds more or less in concert.
Thus, in addition to discussion of the loss, the introductory phase of CGT
included a focus on personal life goals. In the middle phase, the therapist
addressed both processes in tandem. Similar to IPT, the termination phase
focused on review of progress, plans for the future, and feelings about ending
In contradistinction to IPT, however, traumalike symptoms were addressed
using procedures for retelling the story of the death and exercises entailing
confrontation with avoided situations, modified from imaginal and in vivo
exposure used for PTSD.30,31 We
called the retelling procedure “revisiting.” To conduct a revisiting
exercise, the therapist asked patients to close their eyes and tell the story
of the death. The therapist tape-recorded the story, and periodically asked
the patient to report distress levels. The patient was given the tape to listen
to at home during the week. Distress related to the loss (eg, yearning and
longing, reveries, fears of losing the deceased forever) was targeted using
techniques to promote a sense of connection to the deceased. These included
an imaginal conversation with the deceased and completion of a set of memories
questionnaires, primarily focused on positive memories, though also inviting
reminiscence that was negative. The imaginal conversation was conducted with
the patient’s eyes closed. The patient was asked to imagine that he/she
could speak to the person who died and that the person could hear and respond.
The patient was invited to talk with the loved one and then to take the role
of the deceased and answer. The therapist guided this “conversation”
for 10 to 20 minutes. For the restoration focus, patients defined personal
life goals using a technique derived from motivational enhancement therapy.32 Patients were encouraged to consider what they would
like for themselves if their grief was not so intense. The therapist then
helped patients identify ways to know that they were working toward their
identified goals. Concrete plans were discussed and the therapist encouraged
the patient to put these into action. Standard IPT procedures targeting role
transition and/or interpersonal disputes were also used, as needed, to encourage
patients to reengage in meaningful relationships. More detailed information
describing the treatment is available from the authors.
Independent evaluators were experienced master’s- or doctoral-level
clinicians trained for reliability on rating instruments and monitored throughout
the study. Evaluators were blinded to treatment assignment, and study staff
closely monitored procedures to maintain the blinding. Independent evaluators
conducted assessments prior to as well as after treatment. Additionally, for
randomized participants who dropped out after at least 1 treatment session,
a Clinical Global Improvement (CGI) Scale score was generated. To do this,
therapists provided a global improvement rating and a brief narrative justifying
their rating, without including information related to the treatment. The
independent evaluator reviewed the rating and narrative as well as available
participant self-report assessments from the final session to finalize the
CGI score. The CGI Scale33 is a single Likert-type
rating from 1 to 7 where 1 through 3 indicate very much, much, and minimally
improved, respectively; 4 indicates no change; and 5 through 7 indicate minimally,
much, and very much worse, respectively.
Pretreatment assessment included the Structured Clinical Interview for
the DSM-IV,34 Hamilton
Rating Scale for Depression,35 Hamilton Rating
Scale for Anxiety,36 structured clinical interviews
for complicated grief and for suicidality, and screening medical evaluation.
We diagnosed major depression without making an effort to discriminate grief
from depression. Self-reported measures included the ICG4 and
the Work and Social Adjustment Scale.37 The
Beck Depression38 and Anxiety39 Inventory
scales as well as the ICG and the Work and Social Adjustment Scale were completed
at treatment sessions. Responder status was determined in 2 different ways:
independent evaluator score of 2 or lower on the CGI and self-reported improvement
of at least 20 points (2 SDs above baseline mean) on the ICG.
The study was designed to address the question of whether CGT produced
better results than standard IPT for the treatment of complicated grief. To
answer this question, we examined rate of response, defined using either an
interviewer (CGI) or a self-reported (ICG) measure for all randomized patients
who attended at least 1 treatment session (modified intention-to-treat study
group, n = 95).
Data were first descriptively analyzed to check range and distribution
of all variables. We further checked to ensure equivalent distribution of
scores across study groups. Baseline comparisons included all demographic
and clinical variables.
Cochran-Mantel-Haenszel general association analyses were used to compare
CGI responder rates for IPT and CGT. Statistical significance was defined
as P<.05 with a 2-tailed test. We used a survival
analytic strategy to compare time to response using the ICG criterion. Kaplan-Meier
curves were used to investigate time to response and proportion surviving
by treatment groups. Wilcoxon χ2 tests were used to assess
differences in survival curves. We further calculated number needed to treat
as 1 divided by the proportion responding in CGT-IPT as an estimate of the
number of patients who would need to be given CGT for 1 of them to achieve
a response outcome who would not have achieved it with IPT. For most efficacious
treatments, the number needed to treat falls between 2 and 4.40
Continuous measures were evaluated using end-point analysis with baseline
score as covariates in both modified intention-to-treat and completer analyses
for the self-reported measures, ICG, Work and Social Adjustment Scale, and
Beck Depression and Anxiety Inventory scales. Interview-rated Hamilton Depression
and Anxiety scores were obtained only at baseline and posttreatment assessment
points and so are available only for completers.
To examine the possible difference in response by baseline measures,
a Cochran-Mantel-Haenszel test, stratified by treatment, was used. To examine
differential treatment response in different subgroups, a Breslow-Day test41 was used, stratifying by group. A significant result
indicates that a differential treatment × group interaction exists.
SAS software, version 8.2 (SAS Institute Inc, Cary, NC) was used for all analyses.
There were no significant differences in demographic measures or baseline
ICG scores between the 2 randomized groups (Table
2). Because no significant stratum effect for site or type of death
was observed, we aggregated data across strata.
Treatment completion rates (73% for CGT and 74% for IPT) did not differ
across groups. Mean number of CGT sessions for completers was 16 (range, 7-19).
Mean number of IPT sessions was 16 (range, 15-16). Mean time to completion
of CGT was 19.4 weeks. Mean time to completion of IPT was 18.4 weeks. Mean
number of sessions prior to dropout for CGT was 5.9 (SD, 3.7; range, 1-12)
and for IPT was 4.3 (SD, 2.6; range, 1-8). Three patients in CGT each had
3 additional sessions to deal with a second death and 2 had 1 additional session.
Two had extra sessions to address an intercurrent medical problem (kidney
stone and blepharospasm) and 1 had an extra session to discuss the September
11 attacks. Three patients ended treatment early with the agreement of their
therapists. A total of 6 CGT and 3 IPT patients had treatment lasting more
than 20 weeks. Twenty IPT (43%) and 23 CGT (47%) patients continued to take
antidepressant medication begun prior to randomization.
Using the independent evaluator criterion of a CGI score of 2 (much
improved) or 1 (very much improved), rate of response in the modified intention-to-treat
sample was greater for CGT than for IPT among all randomized participants;
51% (95% confidence interval [CI], 37%-65%) treated with CGT responded compared
with 28% (95% CI, 15%-41%) treated with IPT (χ21 = 5.07; P = .02; cohort relative risk [RR], 1.69 [95%
CI, 1.03-2.77]). Among completers, 66% (95% CI, 50%-82%) vs 32% (95% CI, 16%-48%)
responded (χ21 = 7.56; P = .006; cohort RR, 2.03 [95% CI, 1.16-3.49]). The number
needed to treat was 4.3 for modified intention to treat and 2.9 for completers.
Median time to response using the self-report (ICG) criterion was shorter
for CGT than for IPT (Figure 2) (Wilcoxon
χ21 = 5.65; P = .02).
Table 3 shows results for the
ICG, Beck Depression Inventory, Beck Anxiety Inventory, and Work and Social
Adjustment Scale. In the modified intention-to-treat analysis, outcome was
marginally better for CGT than for IPT. Results for completers showed significantly
better outcome for CGT with medium effect size differences on the ICG, Beck
Depression Inventory, and Work and Social Adjustment Scale.
Early treatment discontinuation occurred for 13 (27%) of 49 CGT and
12 (26%) of 46 IPT participants. Reasons for discontinuation differed; 6 CGT
patients (12%) considered the treatment too difficult and/or did not believe
that telling the highly painful story of the death could help them. An additional
7 participants (14%) discontinued CGT for serious medical illness (n = 3;
after sessions 4, 10, and 12), insurmountable child care conflicts (n = 2;
after sessions 3 and 5), a death in the family (n = 1; after session
6), and sufficient improvement (n = 1; after session 12). Also of
note, 5 CGT patients who completed the treatment refused participation in
the imaginal exposure exercise because they considered it too difficult.
For IPT, 7 (15%) of 46 left treatment dissatisfied because of perceived
lack of effectiveness. Five additional IPT patients (11%) discontinued treatment.
Reasons included scheduling problems (n = 1; after session 8), hospitalization
for active suicidal ideation (n = 1; after session 5), beginning
antidepressant medication (n = 2; after sessions 10 and 12), and
withdrawal because of serious protocol violation on the part of the therapist
(n = 1; after session 9) related to insertion of CGT into the IPT
We found no statistically significant differences in response based
on race, age, sex, time since the loss, or relationship to the deceased. Patients
taking antidepressant medication had marginally better response rates: for
CGT, 13 of 22 (59% [95% CI, 38%-80%]) vs 11 of 26 (42% [95% CI, 23%-61%])
not taking antidepressant medication and for IPT, 8 of 20 (40% [95% CI, 19%-61%])
vs 5 of 26 (19% [95% CI, 4%-34%]) not taking antidepressant medication. Patients
who lost a loved one through violent death (suicide, homicide, or accident)
had a 56% (95% CI, 32%-80%) response rate with CGT and 13% (95% CI, 0%-30%)
response rate with IPT, while for natural, nonaccidental death, there was
a 47% (95% CI, 30%-64%) response to CGT and 35% (95% CI, 18%-52%) response
to IPT. Parents who lost a child had a low response rate to CGT (17% [95%
CI, 0%-52%]) compared with those who lost a spouse, parent, or other friend
or relative (average, 60%), while this was not true for IPT, for which the
response rate (28%) did not differ by type of loss. While provocative, none
of these comparisons was statistically significant.
This randomized controlled trial showed better response to CGT than
to IPT, with a number needed to treat of 4.3. Since this is the first such
study in this chronically ill population, this result is encouraging. Nevertheless,
only 51% responded to CGT, and it is clear that more work is needed. In other
studies,20 antidepressant medication alone
has shown small changes in complicated grief symptoms. However, patients taking
antidepressant medication prior to starting this study did have a marginally
better outcome than those not taking medication. Systematic study of combined
medication and psychotherapy is needed.
Participants in our study spanned the adult age range and included individuals
who lost parents, spouses, children, other relatives, or close friends through
violent (33%) or natural (66%) deaths; 22% of participants were African American
and 40% were older than 50 years. The heterogeneity of the sample provides
further evidence that complicated grief, like most DSM-IV disorders, can be identified in different adult populations and in
different psychosocial contexts.
This study has several important limitations. Forty-five percent of
study participants were taking psychotropic medications. We considered it
necessary to permit continued use of medication for co-occurring DSM-IV Axis I disorders for which CGT, and sometimes IPT, had not been
studied. We believed we would unnecessarily limit the generalizability of
our findings if we excluded such patients. There was no difference in the
rate of medication use in CGT vs IPT. There was a marginally significant effect
of medication on outcome, which was more pronounced for IPT (2.1 times the
response rate of those not taking medication) than CGT (1.4 times the response
of those not taking medication.) A similar proportion of patients taking concurrent
antidepressant medication responded to IPT (40%) as those who responded to
CGT without medication (42%).
Heterogeneity is another potential limitation. It is possible that subgroups
might respond differently to different treatment approaches. We had no prior
hypotheses regarding these variables; however, we had insufficient power to
detect differences. For example, we observed that patients experiencing violent
loss had a very low response to IPT (13%). On the other hand, parents who
lost a child showed a much lower rate of response to CGT than patients with
other losses (17% vs 60%). Our study was not large enough to have confidence
in these observations; thus, they should be considered preliminary. Our conclusions
are also limited by the 26% dropout rate from both treatments and the additional
10% who refused to undergo key CGT procedures.
Intervention studies for bereaved individuals often recruited participants
without regard to symptom status and used supportive interventions.46,47 A recent meta-analysis of bereavement
support interventions showed an effect size of 0.15.48 However,
2 earlier studies49,50 examined
efficacy of an exposure-based treatment for individuals considered to have
pathological grief and showed significant treatment effects on measures of
anxiety and depression. There was no measure of complicated grief in these
Our treatment is the first to target complicated grief symptoms directly.
The dual-process model of coping of Stroebe and Schut29 forms
the framework for our approach. Complicated grief treatment is implemented
using loss-focused cognitive-behavioral therapy techniques and restoration-focused
IPT strategies. Cognitive-behavioral therapy techniques include repeated retelling
of the story of the death and work on confronting avoided situations. Cognitive
techniques include an imaginal conversation with the deceased and work on
memories. Interpersonal psychotherapy techniques enhance rapport building,
assistance in restoring effective interpersonal functioning, and guided treatment
In summary, we conducted the first randomized controlled trial of therapy
targeting symptoms of complicated grief. We found better response to CGT compared
with IPT, which is a proven efficacious psychotherapy for depression. Similarity
of ICG scores across age, cultural, and death-related variables supports the
diagnostic validity of the syndrome. Our treatment findings suggest that complicated
grief is a specific condition in need of a specific treatment. More research
is needed to confirm our findings, to test potential moderators of treatment
response, and to improve treatment acceptance.
Corresponding Author: Katherine Shear, MD,
Department of Psychiatry, University of Pittsburgh School of Medicine, 3811
O’Hara St, Room E-1116, Pittsburgh, PA 15213 (email@example.com).
Author Contributions: Dr Shear and Ms Houck
had full access to all of the data in the study and take responsibility for
the integrity of the data and the accuracy of the data analysis.
Study concept and design: Shear, Frank.
Acquisition of data: Shear, Frank.
Analysis and interpretation of data: Shear,
Frank, Houck, Reynolds.
Drafting of the manuscript: Shear, Houck.
Critical revision of the manuscript for important
intellectual content: Shear, Frank, Reynolds.
Statistical analysis: Shear, Frank, Houck,
Obtained funding: Shear.
Administrative, technical, or material support:
Study supervision: Shear, Frank.
Financial Disclosures: Dr Shear has received
financial support from Pfizer and Forest Pharmaceuticals. Dr Frank has received
financial support from Pfizer, Pfizer Italia, Eli Lilly, Forest Research Institute,
and the Pittsburgh Foundation.
Funding/Support: This work was supported by
grants R01MH60783, P30MH30915, and P30MH52247 from the National Institute
of Mental Health (NIMH).
Role of the Sponsor: The NIMH had no direct
input into the design or conduct of the study; collection, management, analysis,
or interpretation of the data; or preparation, review, or approval of the
Acknowledgment: We acknowledge the contributions
of the following individuals, without whose assistance this project would
not have been possible: Krissa Caroff, BS (study coordinator); Jacqueline
Fury, BS (study research associate); Russell Silowash, BS (data manger); Mary
Herschk (study administrative assistant); Rose Zingrone, LCSW, and Randi Taylor,
PhD (independent evaluators); Andrea Fagiolini, MD (study pharmacotherapist);
Bonnie Gorscak, PhD (CGT backup supervisor); Allan Zuckoff, PhD (study psychotherapist
and trainer in motivational enhancement therapy); Daniel Ford, MD, Wayne Katon,
MD, and Sidney Zisook, MD (data and safety monitoring board consultants);
and David J. Kupfer, MD, Edna Foa, PhD, Holly Prigerson, PhD, and Camille
Wortman, PhD (consultants).
Create a personal account or sign in to: