1- the differences between various types of ETOH is one major confounding variable. The benefits of red wine vs. moonshine, both in "moderate" amount, are likely to be widely different.

2- If you control for wealth, you also likely inject all the other variables associated with wealth, such as likely lower stress level, better health-care, etc., thus mainly demonstrating, again, that being poor is "not so good" for you.

3- Although not in the range of the study, the subjective element, as usual, is missing. How to quantify the temporary relief, the feeling of happiness, control or whatever, offered by drugs/alcohol, even in the face of long-term lack of benefit or even negative consequences? Would moments of happiness count in a life of misery? Could such moments postpone the onset of health issues?

4- I believe that the intent of a statistical study needs to be initially defined. I believe there needs to be a broad "idea" prior to formulating a hypothesis and churning numbers. If it is not stated, it becomes implicit and open the results to endless discussion, thereby, in effect, negating the scientific (i.e. quantitative) aspect.