• rheumatoid arthritis
• outcomes research
• epidemiology
• DMARDs (biologic)
• DMARDs (synthetic)

Bergstra et al conducted a sophisticated study of a pertinent clinical question, ‘How to treat patients with rheumatoid arthritis when methotrexate has failed?’1 In particular, I welcome the introduction of the ‘multiple’ propensity score (PS) (better known by its more established name: generalised propensity score2 3) to observational studies in rheumatology. The authors provided a clear step-by-step tutorial which I am sure will be frequently used and cited. Since their article also aimed to teach, I would like to highlight some important points to readers who consider replicating these methods.

Use of PS methods has grown exponentially in clinical research. Suboptimal practice has unfortunately led to a level of scepticism for these versatile techniques.4 The theory and assumptions are often more involved, but some points can be easily conceptualised using …