This systematic review is based on two health technology assessments (HTAs) of early psychosocial interventions following traumatic events [6, 7]. The HTAs were commissioned by the Norwegian Directorate for Health and Social Affairs to obtain an overview of all kinds of early psychosocial interventions following all types of traumatic events, and to use the evidence in the development of clinical guidelines. As the objective of this review is narrower than that of the HTAs, we have performed a new literature search and applied a refined set of study eligibility criteria.
We searched MEDLINE, Embase, PsycINFO, CINAHL, Cochrane Central Register of Controlled Trials (CENTRAL), ISI Web of Science and PILOTS, with each database being searched from inception to June or July 2007. We used subject headings and text words for PTSD symptoms and cognitive-behavioural therapy combined with Ovid's optimised search strategy for randomised trials developed and validated by the Health Information Research Unit at McMaster University . The search was restricted to adult populations.
We included studies that met the following criteria:
randomised controlled trial (RCT) published in peerreviewed scientific journal
a study population of adults with symptoms of acute stress disorder (ASD) or symptoms of post-traumatic stress disorder (PTSD)
individual TFCBT initiated within three months post trauma
a non-pharmacological comparison intervention
outcomes measured as symptoms and/or diagnosis of PTSD (primary outcome), anxiety and/or depression (secondary outcomes) at follow up (minimum one month after treatment completion).
Individual TFCBT was defined as an intervention with at least four planned sessions, regardless of the number of sessions actually completed. At least one of the following techniques should be included in the intervention: exposure, systematic desensitization, stress inoculation training, cognitive processing therapy, cognitive therapy, assertiveness training, biofeedback, relaxation training. The minimum of four sessions was chosen to differentiate TFCBT from debriefing techniques, which may resemble TFCBT in some respects, but are given over one or two sessions only. We did not include studies that compared the effectiveness of TFCBT to conditions with no interventions, such as waiting list controls. There is, at least in Norway, a public expectation to offer some kind of mental health care to traumatised people, and many would perceive no intervention or delayed interventions as unethical. Therefore, our objective was to evaluate whether or not TFCBT was more effective than other interventions. We excluded studies in other languages than English, the Scandinavian languages, French and Italian. Independent pairs of reviewers selected studies for inclusion. A third reviewer was consulted to resolve any disagreement regarding inclusion decisions.
We assessed the methodological quality of included studies on the basis of randomisation, adequate concealment of randomisation, level of blinding, use of intention-to-treat (ITT) analysis and description of loss to follow up. Two reviewers assessed study quality independently using a checklist developed at the Norwegian Knowledge Centre for the Health Services, based on "User's Guides to the Medical Literature" .
One reviewer extracted data from the studies into pre-designed data forms including study, patient and intervention characteristics, relevant outcome measures and study results. At least one other reviewer checked extracted data and any disagreement were resolved by discussion. We reported data from studies with multiple publications as a single study.
We used the Review Manager 4.2 software (Nordic Cochrane Centre, 2003) for meta-analysis where patients, interventions and outcomes were consistent enough across studies to justify pooling. Effect estimates were risk ratio (RR) and standardised mean difference (SMD) with 95% confidence intervals (CI) for dichotomous and continuous outcomes, respectively. We used SMD even when studies used the same assessment instrument for an outcome to accommodate the evaluations of clinical meaningfulness (see below). We used a fixed-effects model to calculate effect estimates when the I2-test for heterogeneity was less than 30%. Otherwise we used the random-effects model. We adhered to Bisson and coworkers' threshold criteria for clinically meaningful effect estimates when two active treatments are compared: SMD ≤ -0.5 or ≥ 0.5, and RR ≤ 0.80 or ≥ 1.25 . Further, 95% confidence intervals for clinically meaningful effect estimates should not cross the thresholds.