Aim: To determine whether baseline levels or changes in physical activity (PA) and screen based behaviours (SBB) predict changes in BMI-z among Fijian adolescents; and test a potential reverse causal pathway of baseline BMI-z predicting changes in behaviours.

Method: Demographic, anthropometric, self-reported PA and screen-based behavioural data were collected from adolescents (13-18 years) from 18 secondary schools in peri-urban areas in Viti Levu, Fiji in 2005/2006 and followed up in 2008 (n=2781). Multiple linear regression analyses were used to assess the longitudinal relationships between PA and SBB (independent variables) and BMI-z (dependent variable) and reverse relationships.

Results: Overall, about 40-80% of adolescents had changed PA and SBB behaviours. However, a mixed pattern of behaviour changes was seen with some increasing and some decreasing in the frequency of healthy behaviours. For example, more adolescents increased rather than decreased the following behaviours: e-gaming (33% vs 23%), TV viewing (28% vs 38%), active travel (41% vs 33%), activity after school (38% vs 35%). All longitudinal analyses (direct and reverse) showed no significant relationships between behaviours and BMI-z.

 Conclusion: The baseline or changes in behaviours did not predict changes in BMI-z (direct pathway hypothesis); nor did baseline BMI-z predict changes in behaviours (reverse pathway hypothesis). Potential explanations for these null findings include: 1) PA/SBB is a weak determinant of energy balance, therefore body size; 2) changes in measured PA are compensated for other non-measured PA and/or dietary intake.  Nevertheless, increasing PA has many proven health benefits and decreasing SSB may be beneficial (for TV but uncertain for e-games). Therefore, expectations of BMI-z impact of PA programs should be low and programs need to be evaluated.