Table 3.
Parameter estimates (95% confidence intervals) for the best-fitting LCS model of weekly alcohol consumption and mental health symptoms in the Whitehall II study a (MCS → Δalcohol model)
| MCS → Δ alcohol model |
Age and sex adjusted |
Fully adjusted
b
|
||
|---|---|---|---|---|
| Alcohol | MCS | Alcohol | MCS | |
| Fixed effects |
|
|
|
|
| Intercept |
17.11 (16.69 to 17.53) |
51.54*** (51.28 to 51.79) |
17.58 (16.64 to18.52) |
53.41*** (52.81 to 54.00) |
| Slope (α) |
21.46** (8.50 to 34.43) |
4.96 (-8.62 to 18.54) |
23.31*** (11.00 to 35.62) |
7.20 (-5.55 to 19.96) |
| Autoproportional (β) |
-0.50*** (-0.61 to -0.40) |
-0.07 (-0.33 to 0.19) |
-0.50*** (-0.60 to -0.41) |
-0.11 (-0.35 to 0.12) |
| Coupling (γ) |
– |
-0.30* (-0.53 to -0.06) |
– |
-0.31** (-0.52 to -0.10) |
| Random effects |
|
|
|
|
| Residual variance |
35.77*** (34.23 to 37.3) |
35.02*** (33.51 to 36.54) |
35.77*** (34.25 to 37.29) |
34.94*** (33.45 to 36.42) |
| Intercept variance |
177.95*** (170.31 to 185.58) |
46.91*** (43.70 to 50.11) |
144.21*** (137.72 to 150.71) |
39.66*** (36.75 to 42.58) |
| Slope variance |
26.26*** (12.55 to 39.98) |
2.31*** (1.27 to 3.34) |
23.66*** (12.24 to 35.08) |
1.88** (0.81 to 2.95) |
| Intercept/slope correlation |
0.69*** |
-0.30 |
0.67*** |
-0.12 |
| Intercepts correlation |
-0.02 |
0.02 |
||
| Slopes correlation |
-0.11 |
-0.02 |
||
| Alcohol intercept, MCS slope correlation | -0.05 | -0.06 | ||
LCS, latent change score; MCS, mental health component score.
*** P < 0.001; ** P < 0.01; * P < 0.05.
an = 6,330.
bFully adjusted = age (centered around the sample mean), sex (male referent group), ethnicity (white (referent) versus non-white), socioeconomic status (defined by most recent recorded employment grade – entered as a linear term with high (referent), intermediate and low categories), marital status (married/cohabiting (referent) versus other), highest educational qualification (University (referent), post-secondary, secondary or no qualifications – entered as a continuous variable), economic activity (active (referent) versus inactive (merging retired and unemployed groups together)), social network (centered around the mean score), current smoking status (no (referent) versus yes), level of physical activity (active (referent), moderately active or low – entered as a linear term), CAGE caseness (no case (referent) versus case), use of anti-depressant medication was also controlled for (no (referent) versus current), self-reported long-standing physical illness (no (referent) versus yes), belonging to the lowest sex-specific SF-36 physical health component quartile (no (referent) versus yes), known diabetes (no (referent) versus yes), coronary heart disease (no (referent) versus yes), stroke (no (referent) versus yes), transient ischemic attack (no (referent) versus yes), total serum cholesterol (centered around the sample mean), systolic and diastolic blood pressure (centered around their mean values), a resting heart rate > 80 beats/minute (no (referent) versus yes) and body mass index (centered around the sample mean).