Yes! The higher the risk, the greater the reduction with steroids, exactly to be expected with regression to the mean. Gill Lancet 1985 showed the higher the initial BP, the greater the reduction with treatment, whatever you give! It's just a numerical phenomenon, with no greater meaning than that.

I've read several SR recently, where the included articles are wrongly referenced, and it's impossible to find the original, if it ever existed. Some of these are being written by AI, and, I think, some of the original trials are also written by AI, without the trial ever having really existed.

I think that editors and reviewers should require that authors supply copies of articles they include in an SR. Automated translation works reasonably well for the languages I know, hopefully it's OK for Chinese also.