For every one point MCS increase, physical activity increased by 0.09 MET-hrs. (β = 0.09, 95% CI 0.04, 0.14), controlling for baseline physical activity and covariates. Fig. 1 shows the physical activity and mental health trajectories, of observed available data at each time-point. Fig. 1A shows the physical activity trajectory according to MCS caseness at baseline. Those with probable depression/dysthymia did less physical activity than those without. These differences persisted across follow-up, but narrowed over time. Fig. 1B shows the trajectory of MCS score according to whether participants met WHO recommendations for physical activity at baseline. Those who did Selleck PARP inhibitor had better mental

health at baseline and across follow-up, but differences also narrowed over time. Although those with good mental health decreased

activity over this website time and those with high levels of physical activity showed slower increases to mental health, differences persisted and both groups were always in a relatively better position from baseline to end of follow-up. These figures illustrate the expected change for each variable based only on the initial status of the predictor variable, ignoring information on repeated measures of the predictor. In contrast, the multivariate LGC model incorporates all three measures for both variables. Results from the multivariate LGC model are shown in Fig. 2. The model about had a good fit to the data (CFI = 0.99, TLI = 0.97, RMSEA = 0.03, SRMR = 0.01) (Hu and Bentler, 1999). In the model, both variables were treated as continuous to avoid loss of information and statistical power. Coefficients

are estimated for male participants aged 55 with intermediate employment grades. The intercept (estimated baseline value) for physical activity was 17.42 (95% CI 15.19, 19.64) which refers to the expected number of min/week at baseline for a participant with these covariate values. The slope (change over time) for physical activity was 3.69 (95% CI 1.25, 6.13) indicating a small increase per study wave. The intercept for mental health was 51.10 (95% CI 49.37, 52.82) which refers to the expected MCS score at baseline. The slope of 1.58 (95% CI 0.68, 2.53) indicated that MCS would be expected to increase by 1.58 points per study phase. The intercepts were positively correlated — higher levels of physical activity at baseline were associated with better mental health at baseline (β = 0.17, 95% CI 0.13, 0.21). The slopes were also positively correlated (β = 0.24, 95% CI 0.11, 0.37) indicating that over time as physical activity increased, so did mental health and at a similar rate. The variables ‘moved together’ over time. Higher mental health at baseline was associated with slightly slower increases in physical activity over follow-up (β = − 0.07, 95% CI − 0.11, − 0.03).