It's amazing that people take hopelessly confounded studies like this and infer anything from them.

Yes the study says they tried to remove confounders. But if people could effectively remove confounders, there'd be no need for randomized trials (which are enormously expensive). Despite doing everything possible, scientists' best efforts to remove confounders still don't effectively substitute for actual randomization.

This study is from 1990-1992. Even with all the sophisticated software today, we still cannot remove confounders, so how bad was it in 1992?