Longitudinal associations between TV viewing and BMI not explained by the ‘mindless eating’ or ‘physical activity displacement’ hypotheses among adults

Table 5 Association between change in television viewing and BMI (weighted associations¹)

Daily TV viewing (hrs/day)	BMI at CDAH2 (adjusted for BMI at CDAH1) B (95% CI)
Model 1 (M1)	Age, sex, education
Stable	(ref)
> 1 h increase	0.38 (− 0.17, 0.93)
> 1 h decrease	0.01 (− 0.48, 0.49)
Model 2 (M2)	M1 + Overall food & beverage consumption
Stable	(ref)
> 1 h increase	0.37 (− 0.18, 0.91)
> 1 h decrease	0.02 (− 0.47, 0.50)
Model 3 (M3)	M1 + LTPA
Stable	(ref)
> 1 h increase	0.38 (− 0.17, 0.93)
> 1 h decrease	0.01 (− 0.47, 0.50)
Model 4 (M4)	M1 + M2 + M3
Stable	(ref)
> 1 h increase	0.36 (− 0.19, 0.91)
> 1 h decrease	0.02 (− 0.46, 0.51)

¹Inverse probability weights were created from the comprehensive data collected over three time points. Observations that were missing were imputed using multiple imputation by chained equations so that a complete set of weights for the analysis sample was available; we imputed 50 different datasets, and for each a set of weights were derived and applied to the analysis model. The average of these model estimates was used to derive a set of weights for the analysis model
CDAH1 Childhood Determinants of Adult Health study (2004–06); CDAH2 Childhood Determinants of Adult Health study (2009–10); TV television