Recruitment of patients was performed at the University of Technology Dresden, Germany, from April 2002 to August 2004. Treatment and 3-month follow-ups on all participants were completed in February 2005. The protocol was approved by the local ethics board and all participants gave written informed consent. After the patients have been recruited and completed the initial assessment, they were matched into dyads based on the initial Clinician Administered PTSD Scale (CAPS) score, diagnosis (full vs. subsyndromal PTSD) and randomly assigned to one of two conditions: (1) a combination of cognitive behavioral treatment (CBT) procedures or (2) a Wait List Control (WLC) condition. The latter was crossed over after the post-assessment to the CBT condition.

Of 239 potential participants who were screened by telephone, 132 were offered an appointment for assessment and 110 completed the assessment. Eleven were excluded on the basis of co-morbid diagnoses (1 with bipolar disorder, 5 with current alcohol or drug abuse or dependence, and 2 with noticeable cognitive impairment secondary to the MVA – as clinically observed by assessors); 44 others were not included in the treatment study because of too few symptoms or too low CAPS score. Sixty-five patients were eligible for treatment and therefore offered treatment. Seventeen refused treatment due to geographical distance. Forty-eight attended at least one treatment session. Data from all of these were used in the intent-to-treat analysis. There were six dropouts (2 from CBT, 4 from Wait List). Thus, 42 participants completed the post-treatment assessment.

Our sample included 22 individuals who met DSM-IV criteria for PTSD and 20 individuals with severely symptomatic subsyndromal PTSD [meets criterion A, E and F for PTSD and two of criteria B, C, or D [see 8 or 25 for a description of the utility of this category] with a CAPS score of 30 or higher. Demographic variables for the participants in each of the treatment conditions, and for the treatment drop-outs combined, are presented in Table 1.

The CBT and WLC conditions did not differ significantly in any of the variables presented in Table 1 even though there were more men in the wait list control than in the CBT condition. The two dropouts from CBT condition did not differ significantly from treatment completers on any tabulated variable.

A reliability check on diagnosis of PTSD was made by having a trained graduate student assessor, blind to diagnosis, reassess half of the sample (n = 20) by listening to audio tapes of the initial interview. Kappa for diagnostic agreement was 0.78, p < 0.001. Correlation of the two total CAPS scores was r (n = 20) = 0.94, p < 0.001.

In order to increase the external validity of the study by working with a broad variety of therapists, six therapists (5 female, 1 male) were trained for the present CBT protocol. All were already licensed psychotherapists or in their last year of postgraduate psychotherapy education. They saw participants in the university practice offices.

The first author and three other licensed and experienced therapists supervised the therapists by regular meetings with video tapes or tele-conferences. All therapy sessions were video taped and scored for treatment adherence to our CBT manual by graduate students. The correspondence between raters and corresponding ideal protocols was kappa = 0.63 (s.d. = 0.13, p < .05).

Postgraduate students in Clinical Psychology conducted all assessments after the last author had extensively trained each student. To the extent possible, a participant was assessed by the same assessor at all assessments. By instruction of each study participant the assessors were kept blind to treatment condition.

One of the principal investigators (A.K.) matched participants into dyads, based on age, total CAPS score and diagnosis (either PTSD or subsyndromal), and number of comorbid diagnoses and then randomly assigned dyads to conditions and to therapists according to a list of random numbers previously assigned to each participant.

The wait-list control individuals were told that their treatment would be delayed for 2–3 months and that they would need to be reassessed after that interval before beginning their treatment. After the post-treatment assessment, they were all given the CBT if they were still interested in treatment.

In an effort to enhance the external validity, the treatment allowed the therapist a range of 8–12 weekly sessions with an expected mode of 10. Thus, the therapist could end treatment after 8 visits if in his/her judgment maximum benefit had been obtained; on the other hand treatment could be extended up to 12 visits if the therapist believed the extra sessions were necessary. Mean number of visits for CBT was 11.4 (s.d. = 3.2).

The CBT condition combined several behavioral and cognitive procedures which had been manualized 24. The initial visit consisted of a detailed description of PTSD and its symptoms and discussion of how the specific patient's symptoms fitted the description. Emphasis was placed on helping the patient to understand PTSD as a reaction to trauma, or "to normalize" the patient's view of his/her symptoms. A description of the treatment components was given. Lastly, the patients were asked to write a very detailed description of the MVA and its immediate aftermath, including their thoughts and sensory perceptions, and bring it to the next session.

Three kinds of exposure were emphasized. First was the reading aloud of his/her description of the MVA by the participant. This was done after the second session. Additionally, patients were asked to read it aloud repeatedly (up to three times) at home and record their 'subjective units of distress' (SUDs). This was tapered to once per day when the patient complained of boredom in the task. This procedure was designed to force the participant to learn to reduce cognitive avoidance.

The second form of exposure was in sensu exposure. It followed the standard procedure for prolonged imaginal exposure described by 16) and consisted of in-sensu imagination of worst moment(s) of the traumatic event. The therapists encouraged the patient to stay with the memory until habituation leads to reduction of subjectively experienced distress. Sequences of this imagination are repeated until the patient reaches stable levels of lowered or disappeared stress.

The third form of exposure was in vivo exposure to fear arousing cues related to traveling by automobile. An individual hierarchy of tasks was constructed with the patient and travel tasks assigned for each week beginning in Week 3. It could range from sitting behind the wheel of the car, starting it, and backing out of the driveway, in very avoidant cases, to driving near the MVA site or under conditions vaguely related to travel conditions at the time of the MVA. Gentle pressure was exerted to have the patient accomplish more each week. In addition, they were urged to use the newly learned relaxation skills to counter the arousal caused by working on travel hierarchy items. Involvement of spouse/significant other for these exercises was encouraged.

Cognitive procedures concerned four domains. First, basic cognitive restructuring 34 comprised instructions to the patients to monitor their thoughts and feelings associated with the accident-related negative emotional states. Patients were taught to identify cognitive fallacies and learned how to dispute them. In addition, they were taught to identify negative self-talk and how to correct it with positive coping self-talk.

Second, specific accident-related cognitions as described in the Ehlers and Clark model 35 of cognitive treatment of PTSD were identified and questioned, e.g., themes of safety ("I will never feel safe again"), over-generalisation of danger ("Traffic is never safe"), and memory disturbances ("I am loosing my mind because of what happened").

Third, special emphasis was given to identify subjective guilt and anger feelings. In case of predominating feelings of guilt or anger somewhat extended cognitive modules were applied to address and dispute these feelings 35. Advanced cognitive dispute procedures or behavioral experiments were introduced, e.g. "Anger chair" or "Guilt letter" assignments 24.

Fourth, attention was paid to existential issues such as the recurring thought that the patient could have died in the accident and if and how one could regard oneself as positively changed or personally grown by overcoming the traumatic experience and its aftermath 23. When patients mentioned positive changes as a result of coping with the trauma, the reported benefits were appreciated and attributed as the patients' personal successes. In addition, the repeated completion of a questionnaire on posttraumatic growth 36 provided further occasions for discussing this issue.

The primary analysis in this study is the comparison of CAPS scores for the two conditions from pre-treatment to post-treatment. The relevant mean scores are in Table 2 along with the means for the CBT condition available for the 3-month follow-up.

A repeated measures ANOVA (Group × Time) revealed a significant main effect of Time (F [1, 40] = 63.79, p < 0.001, Cohen's d = 2.53) and a significant interaction of Group × Time (F [1, 40] = 25.81, p < 0.001, d = 1.61) but no main effect of Group. A between group comparison at post-test by ANCOVA (controlled for pre-values) revealed that the CBT condition showed greater PTSD symptom reduction than WLC (F [1, 39] = 23.47, p < 0.001, d = 1.55).

The intent-to-treat analysis including data from drop-outs revealed similar results, a main significant effect of Time (F [1, 42] = 51.0, p < 0.001, d = 2.20) and a significant interaction of Group × Time (F [1, 42] = 18.9, p < 0.001, d = 1.34). For controlling the gender imbalance of the two groups a repeated measures ANOVA (Group × Time) was performed only for women (19 in CBT and 13 in WLC) and revealed an even higher significant main effect of Time (F [1, 30] = 83.02, p < 0.001, Cohen's d = 3.33) and a significant interaction of Group × Time (F [1, 30] = 33.69, p < 0.001, d = 2.12) but no main effect of Group.

A second way to view these data is in terms of the categorical variable of whether participants changed diagnostic category or improved from PTSD/subsyndromal PTSD to non-PTSD. This could also be seen as a measure of clinically significant change. The categorical diagnostic data from before to after treatment are shown in Table 3.

These data were analyzed with a 2 × 2 chi square test comparing status of full recovery (non-PTSD) in the CBT or control condition. It revealed that CBT was superior to Wait List (Chi² [1, N = 21] = 3.70, p < 0.05) post-treatment. Thus, in summary 67% of those with initial PTSD or sub-PTSD treated by CBT had improved, compared to 19% of those on the Wait List who were assessed twice.

Our assessments allowed us to examine possible changes in co-morbid major depressive disorder (MDD). It should be remembered that this co-morbid condition was not explicitly targeted by the treatment. Results for MDD were: Seven patients from CBT condition had MDD prior to treatment, all recovered from MDD at post-treatment. Three patients from WLC condition had MDD prior, 2 cases remained depressed by the post-treatment assessment. Chi squares test revealed, for those with MDD (49% of the sample), that CBT led to a significantly higher rate of recovery than Wait List (Chi² [1, N = 21] = 5.74, p < 0.05). Moreover, none of the participants in the CBT condition who were not depressed initially became depressed during treatment.

Table 4 lists the pre-treatment and post-treatment values for each of the psychological tests (in the interest of brevity the 3-month follow-up values for the CBT participants are also listed). Also in the table there are (a) results for the Group × Time interaction from the Group × Time repeated measures MANOVA and (b) whether the within group change is significant from pre-treatment to post-treatment (and from post-treatment to 3-month follow-up).

Examining the results in Table 4 one finds significant (all p < 0.01) change for those receiving CBT on measures of intrusive and avoidance symptoms, depressive symptoms (BDI), and posttraumatic cognitions (PTCI) from pre-treatment to post-treatment with no additional change in follow-up. Those in the Wait List condition did not significantly change on any measure. Between group comparisons at post-test by ANCOVAs (last column of table 4) revealed that the CBT condition showed greater change than WLC (all p < 0.01).

For the 3-month follow-up, data of only n = 17 participants were available due to logistic problems of the research project. A series of repeated measure MANOVAs showed stability of measured psychopathology (see superscripts in table 4).

Table 2 already showed the mean CAPS scores at follow-up. Because of the lack of untreated control condition only tests of stability of treatment success in the CBT condition are available (paired t-test_post/FU(16) = 0.7, p = 0.52). From post-treatment to follow-up, the recovery rate (i.e. rate of non-PTSD, see table 3) increased slightly from 67% to 76% of the initial CBT group.