Predictors and moderators of outcome: A systematic review

OCD is a disabling mental health condition. Despite effective psychological therapies for OCD, around 50% of treated patients fail to experience significant and lasting benefit. Factors underlying variable treatment response are poorly understood. Moderators of outcome can help understand ?for whom? and ?under what circumstances? an intervention works best and thus improve service effectiveness. This review identifies and synthesises the current evidence on predictors and moderators and assesses the quality of reporting of related analyses in psychological therapies for adults with OCD.Relevant trials of any language, reporting on psychological treatments in adults with OCD were identified through electronic database searches (CENTRAL, MEDLINE, PsycINFO and EMBASE), key author and reference list searches of relevant systematic reviews. The quality of predictor analyses was assessed using published criteria. The quality appraisal and study extractions were conducted independently by two researchers. Fifty five per cent (38/69) of eligible trials reported baseline factors associated with outcome; these were categorised as clinical, demographic, interpersonal, OCD-symptom specific, psychological/psychosocial, and treatment-specific variables. Predictors were commonly assessed via a validated pre-randomisation measure, though few trials adopted existing guidelines by stating a-priori hypotheses or conducting a test of interaction. None of the common predictors showed a consistent association with outcome. Commonly reported variables showing a consistent lack of association comprised medication use, age of onset, OCD related beliefs, and educational level. Whilst analyses of predictors and moderators are common, their prospective utility is currently limited by important methodological weaknesses. Advances in this field require the use of gold standard statistical procedures in the assessment of moderators of outcome, a priori hypotheses stating anticipated predictor effects, and increased consideration of statistical power in the planning and conduct of predictor analyses.