Does anyone know why these medicinal studies do not seem to use more modern statistical methods of observational causal inference? I mean yes you can not RCT a

Does anyone know why these medicinal studies do not seem to use more modern statistical methods of observational causal inference? I mean yes you can not RCT and yes a linear model should be the baseline, but then why not go ahead and do some synth control or double ML, or heck, take the survey twice and do some matched diff and diff? Surely more can be done or not?