Problems in identifying predictors and correlates of weight loss and maintenance: implications for weight control therapies based on behaviour change.

Weight management is a dynamic process, with a pre-treatment phase, a treatment (including process) phase and post-treatment maintenance, and where relapse is possible during both the treatment and maintenance. Variability in the statistical power of the studies concerned, heterogeneity in the definitions, the complexity of obesity and treatment success, the constructs and measures used to predict weight loss and maintenance, and an appreciation of who and how many people achieve it, make prediction difficult. In models of weight loss or maintenance: (i) predictors explain up to 20-30% of the variance; (ii) many predictors are the sum of several small constituent variables, each accounting for a smaller proportion of the variance; (iii) correlational or predictive relationships differ across study populations; (iv) inter-individual variability in predictors and correlates of outcomes is high and (v) most of the variance remains unexplained. Greater standardization of predictive constructs and outcome measures, in more clearly defined study populations, tracked longitudinally, is needed to better predict who sustains weight loss. Treatments need to develop a more individualized approach that is sensitive to patients' needs and individual differences, which requires measuring and predicting patterns of intra-individual behaviour variations associated weight loss and its maintenance. This information will help people shape behaviour change solutions to their own lifestyle needs.