(PS11-C70) The Impact of Study Characteristics on Post-treatment Outcomes in Randomized Clinical Trials of Psychotherapies for Post-traumatic Stress Disorder

Individual RCTs and meta-analyses have demonstrated that psychotherapy for PTSD is generally efficacious, with the largest evidence base for CBTs for PTSD. However, RCTs often rely on idiosyncratic samples and methods that may bias effect sizes.  The primary aim of this meta-analysis is to examine whether the characteristics of clinical trials of psychotherapies for PTSD influence meta-analytic estimates of effect sizes, both in comparisons between active treatments and to conditions intended as active or inactive controls. The primary characteristics we are interested in are methodological quality of the trial, design elements (eg type of control), and inclusion/exclusion criteria. Systematic searches were conducted using prior meta-analyses as well as searches in databases such as PsychInfo and PubMed. Inclusion criteria for the meta-analysis were RCTs of evidence-based psychotherapy for PTSD vs control conditions. Patients included non-geriatric adults (aged 18-65) diagnosed with PTSD using a reliable diagnostic interview. Outcome measures included reliable and valid PTSD symptom measures and secondary outcome measures include depression symptom measures, measures of dissociative symptoms and psychosocial functioning.  We examined whether effect sizes were, in part, a function of study characteristics, including number and frequency of sessions, risk of bias scores (using RCT-PQRS), exclusion criteria for other mental health diagnoses, exclusions for medical conditions, memory difficulties, psychotropic medication, individual demographic characteristic, number of trauma exposures, and veteran vs non-veteran populations. The R package "robumeta" will be employed to both estimate an overall effect size for the above comparisons, and to explore predictors of effect size using meta-regression. This package conducts robust variance estimation (RVE) meta-regression using both large and small sample RVE estimators under various weighting schemes. These methods are distribution free and provide valid point estimates, standard errors, and hypothesis tests even when the degree and structure of dependence between effect sizes is unknown. This allows for using all effect estimates available from a given study. In total, 85 studies were identified that met inclusion/exclusion criteria and were coded for study characteristics, effect sizes, dropout, and study quality.  Minimum criteria were met for numbers of studies providing data on prior trauma, childhood trauma, personality disorder.  All 85 studies could be coded for number and frequency of sessions, type of active treatment, and control group.  Inclusion and exclusion criteria were for nearly all studies thus providing viable power to examine study characteristics and inclusion and exclusion criteria as predictors of ES (outcome).  This data provides evidence for how study design and methodology impact standardized effect sizes of treatments and whether effect sizes may be inflated for studies with stricter inclusion or exclusion criteria. It provides further recommendations for designing RCTS for optimizing generalizability of ES.