A Randomized Controlled Trial of 7-Day Intensive and Standard Weekly Cognitive Therapy for PTSD and Emotion-Focused Supportive Therapy

Individuals (N=121) were recruited between 2003 and 2008 from consecutive referrals to a National Health Service outpatient clinic for anxiety disorders in South London, U.K. (N=81) or a research clinic at the University of Oxford, U.K. (N=40). Patients were invited to participate if they met the following inclusion criteria: they were between 18 and 65 years old, met diagnostic criteria for chronic PTSD as determined by the Structured Clinical Interview for DSM-IV (14), their intrusive memories were linked to one or two discrete traumatic events in adulthood, and PTSD was the main problem. Exclusion criteria were history of psychosis, current substance dependence, borderline personality disorder, acute serious suicide risk, or if treatment could not be conducted without the aid of an interpreter. Figure 1 depicts the patient flow chart and Table 1 summarizes the details on trauma and the clinical, demographic, and treatment characteristics. No group differences were observed in any of the variables. Seventy-one patients (58.7%) were women, and 36 (29.8%) were from ethnic minorities. The most common index traumas were interpersonal violence (physical or sexual assault, 37.2%), accidents or disaster (38.0%), or traumatic death of others (7.4%). Most patients (71.9%) had a history of other traumas besides their index traumas. The majority (63.6%) had other comorbid axis I disorders (mainly mood and anxiety disorders or substance abuse), and 19.8% had axis II disorders (mainly obsessive-compulsive, depressive, paranoid, or avoidant disorders). Around one-third (36.7%) had had previous treatment for PTSD. Patients taking psychotropic medication (29.8%) were required to be on a stable dosage for 2 months before random assignment.

If the patients were suitable for the trial and willing to participate, they signed the informed consent form. The participants were then randomly allocated to one of the four trial conditions by an independent researcher who was not involved in assessing patients using the minimization procedure (15) to stratify for sex and severity of PTSD symptoms. The assessors determining the suitability of a patient for inclusion were not informed about the stratification variables and algorithm. The assessments of treatment outcome were conducted by independent evaluators without knowledge of the patient’s treatment condition. Patients were asked not to reveal their group assignment to the evaluators. Participants were not blind to the nature of the treatment, but care was taken to create similarly positive expectations in each treatment group by informing them that several psychological treatments were effective in PTSD and it was unknown which worked best, and by giving a detailed rationale for the treatment condition to which the patient was allocated. Patient ratings of treatment credibility (16) and therapeutic alliance scores (17) were high in all treatment conditions and did not differ (Table 1).

Patients in all treatment conditions received up to 20 hours of treatment by the 14-week assessment (posttreatment/wait). The sessions were spread evenly over 3 months for standard cognitive therapy and supportive therapy, whereas the main part of treatment occurred within the first 7–10 days for intensive cognitive therapy. The number of treatment or booster sessions received did not differ between the treatment groups (Table 1).

Data were collected from all participants, including dropouts. The primary assessment points were at baseline (pretreatment or assignment to waiting list), 6 weeks (self-reports only), and 14 weeks (posttreatment/wait). Follow-ups for treated patients were at 27 and 40 weeks after random treatment assignment. Figure 1 depicts the number of patients who provided data at each assessment point. In addition, patients receiving therapy also completed self-reports of PTSD symptoms, anxiety, and depression at 3 weeks.

The therapists were qualified clinicians who had completed a clinical psychology (AH, NG, JW, IA) or nurse therapist (SL, AD) degree and had received further training in all treatments used in this study. They had treated at least two individuals with each of the therapy protocols under supervision before treating trial patients. They received weekly supervision from a senior clinician (AE, AH, NG) trained in all treatment modalities for weekly cases, and daily supervision for intensive cases to ensure compliance with the treatment protocols.

To further evaluate treatment integrity, a randomly selected recording from each patient was reviewed by a trained assessor for compliance with the treatment protocol, using a detailed checklist of procedures used. Only one minor deviation was discovered: one of the supportive therapy patients worked on spotting memory triggers for a few minutes. Another randomly selected session from each patient was rated for therapist competency. Cognitive therapy sessions were rated by a psychologist experienced in cognitive therapy using an adapted version of the Cognitive Therapy Scale (29), on a scale from 0 to 6. A score of 3 is considered satisfactory, and scores ≥4 indicate good-to-excellent competency. The mean score was 4.7 (SD=0.41) for standard cognitive therapy and 4.8 (SD=0.35) for intensive cognitive therapy (p>0.18). Supportive therapy sessions were evaluated for therapist competency by a counseling psychologist experienced in supportive therapy (on a scale from 0 to 6 with anchors as above, informed by ratings of dimensions of good nondirective therapy such an empathic understanding) (30). The mean rating was 4.7 (SD=0.49).

All analyses were intention-to-treat using all 121 randomly assigned participants. Dichotomous outcomes were compared with chi-square tests. Continuous outcomes were analyzed with hierarchical linear modeling (31). This analysis models random slopes and intercepts for participants and tests the fixed effects of treatment condition and repeated assessments over time, using data from all participants. Differential treatment efficacy shows in significant interactions between treatment condition and time. Significant overall effects were followed up with contrasts between conditions. All variables were centered for the analysis (32). Significance levels were set at p<0.05 (two-tailed). To test whether the three treatment conditions led to better outcomes than the waiting list, linear trends for symptom change over assessments points from baseline to 6 weeks and 14 weeks were compared between the four trial conditions. To compare the efficacy of the three treatment conditions, hierarchical linear modeling compared symptom scores from baseline to the 40-week follow-up, fitting linear and quadratic trends for symptom change over the five assessments (baseline and 6, 14, 27, and 40 weeks). Interactions of site, sex, medication status, and trauma type with condition and time were explored in additional analyses, but as effects were far from significant, these were omitted from the final models.

For comparison with meta-analyses, we report Cohen’s d effect sizes (33) for adjusted between-group differences (controlling for pretreatment scores) and confidence intervals at posttreatment. Effect sizes ≥0.5 are considered medium effects and ≥0.8 are considered large effects. To compare the speed of recovery between the treated groups, a further analysis compared symptom scores on the Posttraumatic Diagnostic Scale, the Beck Anxiety Inventory, and the BDI at 3 weeks for the treatment groups, controlling for initial symptom severity. Effect sizes for within-group changes in symptom scores between the pretreatment and posttreatment/wait assessments were calculated as Cohen’s d statistic (33), using the pooled standard deviation as reference, which is more conservative in estimating improvement than using pretreatment standard deviations.

Recovery from PTSD diagnosis according to the CAPS was coded if the patient no longer met the minimum number of symptoms in each symptom cluster required by DSM-IV, with a score of at least 1 for both frequency and intensity and a global severity score of at least 2 (9–11). Recovery was determined for all randomly assigned participants. The status of a few participants with missing CAPS observations was based on the Posttraumatic Diagnostic Scale (if available for this time point) or the last available value on the CAPS. In addition, for comparisons with other research (21), we calculated the percentages of patients who were totally remitted according to assessor ratings and self-report, using cutoffs recommended in the respective manual: a CAPS score of below 20 (“asymptomatic”) and a Posttraumatic Diagnostic Scale score below 11. PTSD symptom deterioration was defined using established cutoffs for statistically reliable change, i.e., symptom increases greater than 6.15 on the Posttraumatic Diagnostic Scale (34) and greater than 10 on the CAPS (21).

Sample size was determined by power analysis on the basis of effect sizes for cognitive therapy observed in previous trials. A group size of N=30 per condition yields 85% power for an effect size of 0.8.

No adverse effects (i.e., negative reactions to treatment procedures such as significant increases in dissociation, suicidal intent, or hyperarousal) were reported in any of the groups. Dropouts were defined as attending fewer than eight sessions (35), unless earlier completion was agreed with the therapist. Dropout rates were low and did not differ between conditions (Table 2). Only one patient in the supportive therapy group reported symptom deterioration on the Posttraumatic Diagnostic Scale (Table 2). On the CAPS, fewer patients treated with intensive and cognitive therapy were rated as having symptom deterioration than those in the waiting list condition. The supportive therapy group did not statistically differ from the other groups.

Table 2 summarizes the recovery rates for the treatment and waiting list conditions. All treatment conditions were more likely to lead to recovery from PTSD diagnosis than the waiting list. Intensive and standard cognitive therapy had excellent number-needed-to-treat statistics of 1.50 (95% confidence interval [CI]=1.18–2.06) and 1.41 (95% CI=1.14–1.87), respectively. For supportive therapy, the number needed to treat was 2.73 (95% CI=1.77–5.95). Similar results were obtained for assessor-rated and self-reported total remission.

Table 3 summarizes the results for the continuous outcome measures. We observed significant condition-by-time interactions (p<0.002 in all cases) for all primary and secondary outcome measures: PTSD symptoms as measured by CAPS (F=21.50, df=3, 135.35) and the Posttraumatic Diagnostic Scale (F=21.16, df=3, 106.56) (see also Figure 2); disability (F=14.01, df=3, 109.86); anxiety (F=13.57, df=3, 106.85); depression (F=5.16, df=3, 122.20); and quality of life (F=6.96, df=3, 106.85). All contrasts between treatment conditions and the waiting list were significant (except for quality of life between supportive therapy and waiting list), indicating greater improvement for intensive and standard cognitive therapy and supportive therapy compared with waiting list. As summarized in Table 4, pre-post effect sizes (Cohen’s d) for both intensive and standard cognitive therapy revealed a very large improvement in PTSD symptoms and disability and large improvements in anxiety, depression, and quality of life.

At the posttreatment and follow-up assessments, more patients receiving intensive and standard cognitive therapy had recovered from a PTSD diagnosis than patients receiving supportive therapy (Table 2). Similar results were obtained for assessor-rated and self-reported total remission. For all primary and secondary continuous outcomes except depression (Table 3), hierarchical linear modeling revealed significant interactions between condition and linear time effects: PTSD symptoms as measured by CAPS (F=7.83, df=2, 154.13, p=0.001) and the Posttraumatic Diagnostic Scale (F=4.42, df=2, 215.14, p=0.01); disability (F=7.45, df=2, 220.14, p=0.001); anxiety (F=5.40, df=2, 176.80, p=0.005); depression (F=0.79, df=2, 213.98, p>0.23); and quality of life (F=3.27, df=2, 231.98, p=0.04). Contrasts revealed that both intensive and standard cognitive therapy led to greater improvement than supportive therapy on the primary outcome measures (CAPS and Posttraumatic Diagnostic Scale scores), disability, and anxiety. For quality of life, standard cognitive therapy was superior to supportive therapy, and we observed a trend for intensive cognitive therapy to be superior (p<0.10). Baseline-adjusted mean group differences at posttreatment and effect sizes are listed in Table 4.

Comparison of the treatment groups at 3 weeks, controlling for initial severity, revealed significant differences on Posttraumatic Diagnostic Scale scores (F=10.35, df=2, 87, p<0.001) and measures of anxiety (F=4.23, df=2, 87, p=0.018) and depression (F=5.27, df=2, 87, p=0.007). The intensive cognitive therapy group scored lower on PTSD symptoms than the standard cognitive therapy and supportive therapy groups (baseline-adjusted means, 16.65 [95% CI=13.19–20.12], 24.05 [95% CI=20.64–27.46], and 27.65 [95% CI=24.18–31.12], respectively). The intensive therapy group also had lower depression scores at 3 weeks than both other treatment groups and lower anxiety scores than patients receiving supportive therapy.

To further test the comparability of outcomes between the intensive and standard cognitive therapy groups, waiting list patients who still had PTSD at the post-waiting period assessment and still wished treatment were randomly assigned to either standard (N=13) or intensive (N=11) cognitive therapy. The comparison of all patients treated with intensive (N=41) and standard cognitive therapy (N=44) had 80% power in detecting a difference of 4.4 points on the Posttraumatic Diagnostic Scale. We found no interactions between treatment condition and time on any measure, indicating comparable outcomes. Baseline-adjusted differences at 14 weeks between all standard weekly and intensive cognitive therapy patients were as follows: CAPS, −2.19 (95% CI=−12.97 to 8.60), d=0.08; Posttraumatic Diagnostic Scale, −1.48 (95% CI=−5.35 to 2.39), d=0.15; disability, 0.51 (95% CI −2.74 to 3.75), d=0.06; anxiety, −2.59 (95% CI=−6.79 to 1.63), d=0.24; depression, 0.27 (95% CI=−3.59 to 4.13), d=0.03; and quality of life, 4.8 (95% CI=−3.18 to 12.72), d=0.23.