OBJECTIVE: This study aimed to determine whether (1) initial and/or (2) changes in psychosocial functioning predict body mass index (BMI) z-score change over 4 years in overweight/mildly obese 5- to 9-year old children presenting to primary care. DESIGN AND METHODS: Eligible participants (n = 258) were overweight/mildly obese children (IOTF criteria) recruited into the LEAP2 trial (ISRCTN52511065) from 3,958 children visiting general practitioners in Melbourne, Australia from May 2005 to July 2006. Predictors were change scores calculated from repeated measures of parent- and child-reported child health-related quality of life (PedsQL) and self-esteem; child-reported desire to be thinner; and parent-reported child weight concern. Outcome was measured BMI z-score change from baseline to 4 years. RESULTS: The 189 respondents (61% female; 73% retention) showed little mean change in BMI z-score (-0.08) but wide variation (standard deviation 0.50, range -1.32 to 1.20). Only one baseline measure (better parent-reported PedsQL School Functioning) predicted improving BMI z-score. However, parents and children consistently reported that changes in psychosocial functioning (i.e., PedsQL Social and Global Self-esteem) were inversely related to BMI z-score change scores. The strongest predictors of decreases in BMI z-scores were changes in child-reported body-image variables, i.e., improvements in Physical Appearance Self-esteem (β =0.40, 95% CI -0.98 to -0.15, P < 0.01) and declines in Desire to be Thinner (β = 0.33, 95% CI 0.04 to 0.23, P < 0.01). CONCLUSIONS: At presentation to primary care, it seems unlikely that targeting the psychosocial factors measured in this study would influence BMI z-score change in overweight/mildly obese children. Subsequent change in psychosocial well-being covaries with BMI z-score change and may have important adolescent ramifications; the causal directions for these associations require further research.