A large number of prognostic factors have been associated with recovery from an episode of back pain. The literature has placed much emphasis on psychosocial prognostic factors. The large number of prognostic factors and the lack of comparative analysis of different factors make their use difficult in clinical practice. The aim of this study was to evaluate the comparative usefulness of a range of factors to predict outcome using data from a randomized controlled trial. 312 patients with sub-acute to chronic back pain received a mechanical evaluation and were sub-grouped based on the presence or absence of directional preference (DP). Patients were then randomized to treatment that was matched or unmatched to that DP. Patients with a minimal reduction of 30% in RMDQ score were defined as the ‘good outcome’ group. 17 baseline variables were entered into a step-wise logistic regression analysis for their ability to predict a good outcome. 84 patients met the good outcome criteria and had a mean RMDQ decrease of 58.2% (9.8 points) in 4 visits. Leg pain, work status, depression, pain location, chronicity, and treatment assignment were significant predictors of outcome in univariate analysis. Only leg bothersomeness rating and treatment assignment survived multivariate analysis. Subjects with DP/centralization who received matched treatment had a 7.8 times greater likelihood of a good outcome. Matching patients to their DP is a stronger predictor of outcome than a range of other biopsychosocial factors.