Fitted autoregressive linear mixed-effects model of BMI (kg/m2) as a function of baseline age, sex, and centred time: The intercept term in the following model corresponds to the expected BMI value for a female patient, aged 18-24 y, measured in the middle of the study (approximately January 1, 2014). Random intercept variance was 31.35, residual variance was 4.12, and the AR1 working correlation parameter was 0.41.
| VARIABLES | β | SE | P VALUE |
|---|---|---|---|
| Intercept (constant) | 25.68 | 0.03 | < .0001 |
| Male sex | 1.02 | 0.02 | < .0001 |
| Age* groups, y | |||
| • 18-24 | Reference | - | - |
| • 25-34 | 1.28 | 0.03 | < .0001 |
| • 35-44 | 2.12 | 0.03 | < .0001 |
| • 45-54 | 2.60 | 0.03 | < .0001 |
| • 55-64 | 2.72 | 0.03 | < .0001 |
| • 65-74 | 2.59 | 0.03 | < .0001 |
| • ≥ 75 | 2.06 | 0.04 | < .0001 |
| Centred time, y | 0.06 | 0.002 | < .0001 |
AR1—first-order autoregression, β—regression coefficient, BMI—body mass index, SE—standard error.
↵* Age calculated from birth date to January 1, 2011 (study baseline).