I was reading a new paper titled Reporting of Multi-Arm Parallel-Group Randomized Trials (https://jamanetwork.com/journals/jama/fullarticle/2731183) and thought this issue was very common in observational studies. For example, to relieve pain, we have different NSAIDs and opioids. We could get all the data from the registry database. Is there any method to simultaneously compare them and then rank them according to different outcomes. I know network meta-analysis could produce such results. But it only suitable for RCTs. The common causal inference tools (e.g., IPW and Instrumental variable) seems only work for two-arm design.
Thank you for your time and wish you have a great day!