Can observational studies replace or complement experiment?

The therapeutic effects of interventions in patients with rheumatoid arthritis (RA) are frequently modest. In the assessment of treatments effects, variability due to a variety of sources causes problems that are best controlled by randomized clinical trials. Currently most trials give only a short term picture of RA, a chronic disease whose outcome is multidimensional. Inclusion criteria of clinical trials are frequently very strict, raising concern about the external validity of the results. More longterm data is needed to guide clinical practice. Observational studies may contribute to the body of evidence, but have inherent shortcomings. They are liable to bias and supply weaker evidence. There must be creative development of trial designs suitable for evaluating longterm outcomes. Such trials may include many of the positive features of observational studies, but should not omit the principles of randomization and controlled comparison.