There was a good meta analysis in 2016 on this, you lose the mortality benefits of moderate drinking once you correct for study design, which can basically dictate results here. One big confound, a lot of current strict abstainers are former heavy drinkers with serious health problems:
Even without being former heavy drinkers, the fact that a fair percentage of teetotalers abstain due to serious health issues is enough to sway the results
https://www.ncbi.nlm.nih.gov/m/pubmed/26997174/