Meta-analysis of diverse study methodologies is very difficult and admittedly dubious, and multivariate analysis of each retrospective or observational study is unavoidably fraught with selection bias; but the alternative gold standard randomized double blind placebo controlled trial design can never be ethically done, especially in an emotive issue as pregnancy (which is a non-disease state) and lactation (during which most mothers will rather forego treatment than accept the slightest of risk).
So our decisions and recommendations are at best informed by these "poor stats". And, by the Delphi consensus of a panel of experts drawn from the relevant disciplines, the weight of majority wisdom carries the day. This is "Eminence-based medicine" at its best, when "Evidence-based medicine" will be found forever wanting in a dystopic world.
All said, Table 4 sums it all, and should be printed out and kept handy on every rheumatologist's and obstetrician's desk.