For the binary outcome measure of harmful use, estimates indicate the odds-ratio for a one-SD change in depression intercept or slope. For males there is weak evidence (p ~ 0.02) of an association between depressive symptoms at baseline and harmful use at 18 years 8 months. This association is weak with only a 17 % increase in odds per one SD change in symptom level. Adjusting for confounders has little effect until the final adjustment in which the already-moderate estimates are attenuated considerably. The primary source of this attenuation was the measure of conduct problems, with baseline depressive symptoms having little effect once this was included in the multivariable model. In contrast to the males’ results, there is evidence of a stronger association in females for both depression intercept and slope. A one SD difference in baseline depression represents almost a 30 % increased odds of harmful drinking, with a slightly more moderate increase of 22 % associated with a one SD difference in symptom slope, i.e. both baseline and change appear to confer a risk of harmful drinking. Estimates
