Longitudinal studies of populations who are healthy at baseline are less likely to suffer from these biases as reporting of alcohol consumption is made prior to subsequent diagnosis. This problem with investigating causation in medicine is illustrated nicely by a study published in the Journal of Neurology by a Spanish team. They looked at 31 studies in the field and broke down the analysis into different types of study as well as looking seperately at men and women and different ways of classifying alcohol consumption.
When looking at the case-control studies, as in previous studies, they found higher numbers of 'never-drinkers' in the group with Parkinson's. However, when looking at longitudinal studies, there was no difference in having ever drunk alcohol between people who went on to develop Parkinson's and people who didn't. However, there were far more case-control studies in the article, giving more power to detect these kind of effects, as you can see in the diagram below.
|Pooled odds ratios comparing "never drinking" to "ever drinking"alcohol in a)case control and b)prospective longitudinal studies, showing higher likelihood of Parkinson's disease in those who never drank in case control but not longitudinal studies. From https://link.springer.com/article/10.1007/s00415-018-9032-3|