Background Relatively little is known about how clustering of energy-balance related behaviours (EBRBs) in young children influences weight status in later childhood. We assessed clustering of six ERBs in 5-year-old British children (on primary school entry), and explored whether this predicted weight status at 11 years (upon transition to high school).

Methods We used complete-case data from 9,726 children who participated in the Millennium Cohort Study, a nationally representative prospective cohort of children born in the UK. Mothers reported their 5-year-old child’s EBRBs (fruit consumption; sweetened drink consumption; active transport to school; frequency of sport/exercise; TV viewing; computer use). Body Mass Index (BMI) was objectively measured at 5 and 11 years. Latent Class Analysis was used to derive clusters of EBRBs (using the Latent Gold package), accounting for child’s sex, ethnicity, BMI at age 5, and maternal educational attainment. Multivariable linear regression was used to estimate whether EBRB clusters at age 5 predicted BMI z-score in 11-year-olds (using STATA13/SE).

Results Four EBRBs clusters were identified (in order of prevalence): ‘low screen-time, sweetened drinks’ (32%); ‘healthy’ (28%); ‘unhealthy’ (26%); and ‘high screen-time’ (14%). Children in the ‘healthy’ cluster had low screen-time, frequent sport participation, and a good diet (high fruit, low sweetened drinks consumption); the opposite EBRBs were observed in 5-year-olds assigned to the ‘unhealthy’ cluster. Those in the ‘low screen-time, sweetened drinks’ and ‘high screen-time’ clusters exhibited a mixture of EBRBs, with moderate sports participation and fruit intake.

Children in the ‘healthy’ cluster were more likely to be girls, have mothers with higher educational attainment, and were less likely to be obese at age 5; those in the ‘unhealthy’ cluster were more likely to have lower educated mothers. ‘High screen-time’ children were more likely to be boys and from middle-income homes.

In regression analyses, compared to children in the ‘healthy’ cluster, being in any other cluster was associated with higher BMI z-scores at age 11 [‘low screen-time, sweetened drinks’: z BMI 0.16 [0.10, 0.23] higher; ‘unhealthy’: 0.19 (0.12, 0.26); and ‘high screen-time’: 0.24 (0.16, 0.33)].

Conclusion Four clusters of energy-balance related behaviours were identified in 5-year-old UK children, and clusters appeared to be socially graded. Even after accounting for socio-demographic factors, as early as age 5, clusters of less healthy EBRBs predicted higher BMI z-scores in middle childhood, and how these EBRBs cluster may also be important. Potential pathways and mediating factors will be described.