|Home | About | Journals | Submit | Contact Us | Français|
A significant number of survivors of hematopoietic stem-cell transplantation (HSCT) report enduring adverse effects of treatment, including illness-related post-traumatic stress disorder (PTSD) symptoms and general distress. We report results of a randomized clinical trial that tested the effects of a 10-session, telephone-administered cognitive-behavioral therapy (CBT) intervention on PTSD, depression, and distress symptoms.
Survivors who had undergone HSCT 1 to 3 years earlier (N = 408) were assessed for study eligibility. Those who met study eligibility criteria (n = 89) completed a baseline assessment that included a clinical interview and self-report measures of PTSD symptoms (the primary outcome) and depression and general distress (the secondary outcomes). Next, they were randomly assigned to CBT or an assessment-only condition. Survivors in the CBT group completed 10 individual telephone-based CBT sessions (T-CBT) that included strategies to reduce PTSD symptoms, depression, and general distress. Follow-up assessments occurred at 6, 9, and 12 months after the baseline assessment.
Linear mixed-model analyses revealed that, compared with HSCT survivors in the assessment-only condition, survivors who completed T-CBT reported fewer illness-related PTSD symptoms, including less avoidance (P < .001) and fewer intrusive thoughts (P < .05) as well as less general distress and fewer depressive symptoms (P < .05) even after controlling for potential demographic and medical covariates. These results were consistent across the three follow-up assessments.
A brief, telephone-administered CBT intervention developed for HSCT survivors is an efficacious treatment for reducing illness-related PTSD symptoms and general distress.
Hematopoietic stem-cell transplantation (HSCT) is an increasingly common treatment for hematologic and lymphoid cancers and nonmalignant diseases.1 Many survivors of HSCT experience long-term medical and psychological adjustment problems, such as post-traumatic stress disorder (PTSD) symptoms.1–2 These symptoms include intrusive thoughts, avoiding reminders of illness experiences, and physiologic arousal on exposure to illness-related cues.3 Research suggests that as many as 41% of survivors of HSCT experience persistent PTSD symptoms up to 10 years post transplantation3–8 and that general distress and depressive symptoms affect up to 40% of survivors.2
Cognitive-behavioral therapy (CBT) has been found to reduce cancer-related adjustment problems.9–12 For example, a 10-session cognitive-behavioral intervention reduced depression among patients with early-stage breast cancer.13 With noncancer trauma populations, including military veterans and survivors of sexual assault, CBT has reduced PTSD symptoms and psychiatric comorbidities, such as depression.14–15
On the basis of these prior CBT trials11,13,15–16 and our clinical experience with patients who have cancer and who survived HSCT,17 we developed a telephone-administered CBT (T-CBT) intervention for survivors of HSCT that focused on the maladaptive emotional reactions to illness and transplantation. We hypothesized that this intervention would reduce symptoms of illness-related PTSD (the primary study outcome), general distress, and depression (the secondary study outcomes) in survivors of HSCT. We tested this hypothesis by comparing psychologic outcomes of a T-CBT group to those of an assessment-only control group. Treatment effects were evidenced either by a significant study group main effect or by a significant time-by–study group interaction. On the basis of unpublished pilot data for this study, a main effect for treatment group was anticipated for the continuous measure of PTSD, as the pilot data indicated that, even in the absence of treatment, a decline in PTSD symptoms occurred over time. Because no pilot data were obtained for the categoric measure of PTSD or for the measures of distress and depression, there were no clear a priori hypotheses with regard to a significant main effect for treatment condition or an interaction between time and treatment. Finally, we explored the influence of demographic and medical factors on the outcomes.
Potential participants were recruited from Memorial Sloan-Kettering Cancer Center, the Mount Sinai School of Medicine, and Hackensack University Medical Center. Inclusion requirements were HSCT performed 12 to 36 months before study enrollment, English fluency, 18 years of age or older, and significant distress as indicated by at least one of the following three criteria: probable illness-related PTSD on the PTSD Checklist-Civilian Version (PCL-C) by using the three- or four-symptom cluster criteria;8,18 subclinical PTSD symptoms as indicated by scores one or more standard deviations greater than the PCL-C mean; or general distress with some PTSD symptoms as indicated by scores exceeding the clinical cutoff on any two subscales of the Brief Symptom Inventory19 (BSI) or the BSI Global Severity Index and, according to either PCL-C scoring method, scores exceeding the cutoff for at least one PTSD symptom cluster.
Survivors were excluded from study participation if they were currently awaiting another transplantation or receiving treatment for disease relapse; had severe cognitive impairment assessed with the six-item Mini-Mental State Exam;20 experienced active psychosis assessed with six items from the Psychotic Symptoms module of the Structured Clinical Interview for the Diagnostic and Statistical Manual of Mental Disorders fourth edition text revision (DSM-IV-TR);21 reported suicidal ideation assessed with one item from the Beck Depression Inventory22 and one from the BSI;19,23 or had substance dependence assessed with the four-item Rapid Alcohol Problems Screen-424 and the two-item Conjoint Screen25 for alcohol and other drug problems.
After approval by the institutional review boards, oncologists or their representatives identified potentially eligible survivors of HSCT. Each survivor was mailed a letter and brochure describing the study, a consent form, and psychosocial questionnaires. Trained research assistants obtained verbal consent from survivors via telephone. Participants first completed a telephone screening interview that assessed sociodemographics and symptoms of PTSD, general distress, and depression. Next, those who met study eligibility criteria completed a baseline telephone interview that included a PTSD diagnostic interview and measures of PTSD, general distress, and depressive symptoms. Computerized random assignment occurred approximately 1 week postbaseline. Interviewers were blinded to participant group assignment.
The 10-session manualized T-CBT intervention was delivered during a period of 10 to 16 weeks. The first session was approximately 90 minutes in duration, and sessions 2 to 10 were approximately 60 minutes. Interventionists were postdoctoral psychology research fellows who completed 12 hours of training with an expert in CBT for PTSD and study team members. They were supervised individually and as a group by senior CBT clinicians throughout the study. The intervention manual from a previous study17 was revised, and participants received a workbook to accompany the telephone sessions. The intervention included education regarding illness-related PTSD symptoms and CBT (eg, characterizing PTSD symptoms in cancer and other trauma populations and ways in which CBT can reduce these symptoms), self monitoring and alteration of maladaptive beliefs,26 guided exposure to cues associated with PTSD symptoms,27 enhancement of social support through training in communication skills,28–30 and relaxation training31 (study manual available from K.D.). Each component focused on distress related to the participant's disease and transplantation. The percentage of participants who completed each session ranged from 92% for sessions 1 and 2 to 79% for sessions 7 to 10. The mean number of sessions attended by the participants was 8.36 (standard deviation [SD], 3.27).
Baseline measures were readministered during three telephone follow-up interviews. Follow-ups occurred approximately 6 months (mean, 5.71 months; SD, 1.97), 9 months (mean, 8.57 months; SD, 2.21), and 12 months after the baseline assessment (mean, 12.16 months; SD, 1.86).
The PCL-C18 is a measure of illness-related PTSD symptoms previously used with survivors of HSCT.3 The PCL-C yields a total score and subscale scores for intrusive thoughts, avoidance, numbing, and hyperarousal.8
The BSI19 is a questionnaire that assesses general psychological distress. Because general distress and depressive symptoms often co-occur with PTSD,7 the Global Severity Index (GSI) and the depression subscale were used in this study.
The Clinician-Administered PTSD Scale for DSM-IV (CAPS)32 is a structured diagnostic interview that is based on DSM-IV criteria for PTSD. The CAPS is sensitive to short-term changes in PTSD symptoms.32–33
Participants reported their sociodemographic data, and medical information, such as the type of transplantation and graft-versus-host disease (GVHD), was obtained from chart review at screening and approximately 6 to 12 months postbaseline.
Before undertaking the main analyses, all continuous explanatory and outcome variables were assessed for normality and the presence of outliers. The primary analyses focused on PTSD symptoms by using a linear mixed-model repeated measures approach for the continuous PCL-C scores (SAS Proc Mixed, version 9.2; SAS Institute, Cary, NC) and a Generalized Estimating Equations (GEE) model for the binary PTSD diagnosis that is based on the CAPS (SAS Proc Genmod, version 9.2; SAS Institute). Because both the linear mixed-model and GEE procedures utilize all the data each participant provided (even if the participant was a study dropout), an intent-to-treat (ITT) framework was implemented. For each outcome, the model included the main effects for time, the distress criteria for study entry, study group, and the time-by–study group interaction. Two-tailed tests of significance were employed. Treatment effects were evidenced either by a significant study group main effect or by a significant interaction between time and study group.
Two secondary outcomes (BSI Global Distress and BSI Depression) were also evaluated by using the same ITT approach. Probabilities for these four outcomes were Sidak adjusted for correlated multiple outcomes (average correlation, 0.56; adjusted probability for significance, 0.0280). In additional secondary analyses, covariates were examined on the basis of correlations of the sociodemographic and medical variables with each outcome.
Of the 1,434 survivors approached regarding the study, 452 consented to participate, and 408 of these individuals completed the screening assessment to determine eligibility (Fig 1). Common reasons for inability to complete screening were inability to contact the participant (n = 25) and participant refusal (n = 6). Ten participants were not screened because the study ended before they could be scheduled. None of the variables that had a sufficient amount of nonmissing responses (eg, age, ethnicity, sex, time elapsed since treatment) were significant predictors of screening completion by using logistic regression.
Of the 408 participants screened, 89 (21.8%) met study distress criteria and were randomly assigned. Of these 89 participants, two were lost to follow-up, and six withdrew because of serious adverse events and/or disease relapse. This resulted in a final sample of 81 participants.
Table 1 presents descriptive statistics by group assignment. At study entry, 15 (19%) of the 81 participants met criteria for a three-symptom cluster diagnosis of PTSD, and 58 (72%) met criteria for a four-symptom cluster diagnosis of PTSD. Almost all participants (n = 78 [96%]) met criteria for general distress with some PTSD symptoms.
A main effect for treatment group was anticipated and, indeed, the ITT analysis of post-traumatic stress symptoms (ie, PCL-C) yielded a significant (t(80) = 2.37; P = .0201) main effect for study group in favor of T-CBT (MControl = 33.03; 95% CI, 29.87 to 36.18; MT-CBT = 28.34; 95% CI, 25.04 to 31.62). Mean scores for this primary outcome can be found in Table 2.
Because the two groups differed significantly on the total PCL-C score and because there is evidence that the PCL-C scale consists of four factors (intrusive thoughts, avoidance, hyperarousal, numbing),8 follow-up analyses were conducted to examine intervention effects on these four symptom clusters.
The ITT analysis of intrusive thoughts yielded a significant (t(80) = 2.60; P = .011) main effect for study group in favor of T-CBT (MControl = 10.36; 95% CI, 9.10 to 11.62; MT-CBT = 8.62; 95% CI, 7.77 to 9.48). The ITT analysis of avoidance also yielded a significant (t(80) = 3.95; P < .001) main effect for study group again in favor of T-CBT (MControl = 4.13; 95% CI, 3.60 to 4.65; MT-CBT = 2.98; 95% CI, 2.59 to 3.37). There were, however, no treatment effects for the numbing or hyperarousal subscales.
Next, we conducted an ITT GEE analysis of the PTSD diagnosis that was based on the CAPS. There was no a priori reason to expect either a significant study group main effect or a time-by–study group interaction. There was a significant (z = −2.22; P = .027) time-by–study group effect. To understand this interaction, group differences were examined at the beginning and end of the assessment period (ie, at baseline and 12 months postbaseline). As expected, there were no significant (z = 1.10; P = .27) study group differences at baseline. However, at the 12-month follow-up, there was a significant (z = −2.06; P = .04) difference in the log odds of not being diagnosed with PTSD (odds ratio ]OR], 0.07; 95% CI, 0.006 to 0.88) in favor of the T-CBT group. Mean log odds scores for each group across time are listed in Table 2.
The ITT analysis of BSI Global Distress yielded a significant (F(4,80) = 4.05; P = .005) time-by–study group interaction (Table 3). The interaction means displayed in Figure 2 show a steady decline in general distress for the T-CBT group, whereas the mean scores for the control group show some decline from screening to baseline but remain relatively stable over subsequent assessments.
The ITT analysis yielded a significant (F(4,208) = 2.89; P = .023) time by study group interaction on BSI Depression (Table 3). The means displayed in Figure 3 follow the same pattern as that found for general distress. Notably, the CIs for mean depressive symptoms at the three follow-up assessments for the T-CBT group contained zero, which indicated that they did not differ significantly from zero.
In covariate analyses, we examined the effects of the sociodemographic and medical measures reported in Table 1. None of these covariates predicted total PCL-C score. However, in the GEE analysis of the PTSD diagnosis that was based on the CAPS, the time that had elapsed since HSCT was significant (z = −2.36; P = .019). The shorter the time since transplantation completion was, the greater the likelihood that the survivor was diagnosed with PTSD (OR, 1.9; 95% CI, 1.02 to 1.17). Survivors declared disease free were marginally (z = 1.80; P = .07) more likely to be diagnosed with PTSD (OR, 4.58; 95% CI, 0.88 to 23.69). After analysis was adjusted for these covariates, the time-by–study group interaction continued to be significant (z = −2.46; P = .014). This interaction was examined through analysis of treatment effects at baseline and the last assessment. Once more, there was no significant baseline difference (z = 1.02; P = .308) between the two groups in PTSD diagnosis. By study end, however, those in T-CBT were significantly (z = −2.69; P = .007) less likely to be diagnosed with PTSD (OR, 0.046; 95% CI, 0.004 to 0.43).
Two significant covariates predicted general distress on the BSI Global Distress scale. Married survivors reported significantly lower general distress (t(78) = −3.04; P = .003), and less well-educated survivors reported significantly greater general distress (t(78) = −2.44; P = .017). The time-by–study group interaction continued to be significant (F(4,78) = 4.34; P = .003). Table 4 contains the covariate-adjusted means. Moreover, the pattern of change for those in T-CBT continued to show a temporal decline, whereas average scores for control condition survivors remained relatively constant.
Marital status was a significant (t(63) = −2.44; P = .018) predictor of BSI depression. Married survivors reported significantly fewer depressive symptoms. Number of hospitalizations was marginally significant (t(63) = 1.93; P = .058). As the number of hospitalizations increased, depressive symptoms also increased. The time-by–study group interaction continued to be significant (F(4,177) = 2.43; P = .049) after analysis was controlled for these variables. Table 4 summarizes the adjusted means and shows a decline in depressive symptoms for the T-CBT group. In contrast, survivors in the assessment-only condition demonstrated an initial decline (a nonsignificant difference from screening to baseline) and a subsequent increase in depressive symptoms at the third, fourth, and final assessments.
Survivors who participated in the 10-session T-CBT intervention experienced fewer PTSD symptoms, including fewer illness-related intrusive thoughts and less avoidance across all follow-up assessments. T-CBT participants also were less likely to meet diagnostic criteria for PTSD at the final assessment and showed lower general distress and depressive symptoms at all follow-up assessments. Previous in-person CBT interventions have reduced general distress during cancer treatment,12,34 whereas T-CBT is the first intervention to reduce illness-specific PTSD during the early phase of HSCT survivorship. Several components of T-CBT (ie, education regarding PTSD and CBT, relaxation training, guided exposure to cues associated with PTSD responses, and changing maladaptive beliefs) targeted PTSD symptoms. The goal of T-CBT was to reduce illness-related fears of the survivors and to facilitate cognitive processing of their transplantation experiences (ie, incorporation of the experiences into their worldviews). As depressive symptoms and general distress are often comorbid with PTSD symptoms and were part of the study entry criteria, these symptoms were addressed through training in methods of accessing social support and challenging negative thoughts.
Although the T-CBT intervention reduced total PTSD symptom scores, more detailed analyses indicated that T-CBT was effective for intrusive thoughts and avoidance but not for numbing and hyperarousal. This finding may be because, although avoidance and intrusive thoughts were targeted via exposure exercises, feelings of numbness or emotional detachment from others were not directly addressed. In addition, relaxation and challenging maladaptive beliefs did not reduce hyperarousal associated with life-threatening illness. Additional intervention development is needed to address these symptoms.
Relations of sociodemographic and medical factors to study outcomes also were explored. Shorter time since transplantation was associated with a greater likelihood of being diagnosed with PTSD (on the basis of the CAPS) during this early phase of survivorship, whereas total self-reported PTSD symptoms were not associated with any sociodemographic or medical variables. In addition, being married was associated with less general distress and fewer depressive symptoms, and more education was associated with less general distress, consistent with prior research.35–37
Study limitations include the under representation of ethnic minorities and exclusion of those who lacked English fluency. None of the variables tested (age, ethnicity, sex, time elapsed since treatment) were predictors of screening completion by using logistic regression. It is possible that level of distress may have been associated with screening completion, which could not be tested in this study. However, levels of distress among those who completed the screening interview did not differ from those found in prior research with HSCT survivors.38 In addition, the study entry criteria involved a combination of sex-keyed, nonpatient, BSI clinical cutoffs19 and PTSD symptoms for which normative data were unavailable. Finally, use of an assessment-only control condition did not control for attention provided to survivors. Future research should compare T-CBT for HSCT survivors to alternatives such as supportive therapy, which can control for time and attention and which is frequently used in CBT trials with other populations.39 T-CBT also needs to be modified for culturally diverse populations and its cost effectiveness should be examined.
This study informs clinical care by showing that a novel, telephone-administered CBT intervention can reduce illness-related PTSD, general distress, and depressive symptoms among HSCT survivors. The rigorous methodology (eg, random assignment, blind follow-ups, manualized treatment, a well-defined population who met criteria for distress) enhances confidence in the findings. Finally, this telephone intervention may be readily disseminated to geographically dispersed survivors and those with physical impairments that limit their abilities to take advantage of in-person treatments. In addition, telephone-based counseling may be preferable for some survivors with PTSD who avoid contact with the hospital because of the anxiety it provokes. The next step is to disseminate this intervention and to increase awareness of PTSD symptoms and their treatment among HSCT survivors, their families, and their health care team.
We thank the survivors of hematopoietic stem-cell transplantation; and Julian Silva, Maria Nenova, and Jacob Stebel for their assistance.
Supported by Grants No. R01 CA093609, K07CA109236 (K.H.), and F32CA130600 (C.M.) from the National Cancer Institute and by Grant No. PF-01-018-01-CPPB from the American Cancer Society (A.R.).
Presented in part at the National Institutes of Health 4th Biennial Cancer Survivorship Research Conference, June 18-20, 2008, Atlanta, GA; at the European Health Psychology Conference, September 23-26, 2009, Pisa, Italy; at the 43rd Annual Meeting of the Association for Behavioral and Cognitive Therapies, November 19-22, 2009, New York, NY; and at Cognitive Behavioral Therapy training for clinicians providing supportive care to cancer survivors, March 18-20, 2010, Duarte, CA.
Authors' disclosures of potential conflicts of interest and author contributions are found at the end of this article.
The author(s) indicated no potential conflicts of interest.
Conception and design: Katherine N. DuHamel, Christine Rini, Jane Austin, Paul B. Greene, Esperanza B. Papadopoulos, Jack E. Burkhalter, Karen E. Hurley, Andreas R. Bollinger, William H. Redd
Administrative support: Paul B. Greene, Karen E. Hurley, William H. Redd
Provision of study materials or patients: Larissa E. Labay, Yeraz Markarian Meschian, Jane Austin, Anna Rusiewicz, Celia L. Grosskreutz, Craig H. Moskowitz, Esperanza B. Papadopoulos, Scott Rowley, Eileen Scigliano, Andreas R. Bollinger
Collection and assembly of data: Larissa E. Labay, Christine Rini, Yeraz Markarian Meschian, Jane Austin, Paul B. Greene, Catalina R. Lawsin, Anna Rusiewicz, Jack E. Burkhalter
Data analysis and interpretation: Katherine N. DuHamel, Catherine E. Mosher, Gary Winkel, Christine Rini, Yeraz Markarian Meschian, Paul B. Greene, Catalina R. Lawsin, Luis Isola
Manuscript writing: Katherine N. DuHamel, Catherine E. Mosher, Gary Winkel, Yeraz Markarian Meschian, Paul B. Greene, Luis Isola
Final approval of manuscript: Katherine N. DuHamel, Catherine E. Mosher, Gary Winkel, Larissa E. Labay, Christine Rini, Yeraz Markarian Meschian, Jane Austin, Paul B. Greene, Catalina R. Lawsin, Anna Rusiewicz, Celia L. Grosskreutz, Luis Isola, Craig H. Moskowitz, Esperanza B. Papadopoulos, Scott Rowley, Eileen Scigliano, Jack E. Burkhalter, Karen E. Hurley, Andreas R. Bollinger, William H. Redd