Look out for treatment comparisons where the comparison groups were not alike.
For a comparison of treatments to be fair, the people in the comparison groups should be as similar as possible at the start of a study. When they were not similar, it is hard to know how much a difference in what happened to the two groups was because of that, and how much it was because of the treatments that were compared.
For example, if people decided themselves whether to use a treatment or not, the people who chose to use the treatment may start out healthier or sicker than those who chose not to use it. That could make the treatment look either better or worse than it is.
The best way to make sure that the people in the comparison groups are alike is to decide who gets which treatment by chance (randomly) – something like pulling out names from a hat or flipping a coin.