The classical dogma regarding the design is that Randomized Controlled Trials (RCTs) have strengths and observational studies have limitations. Randomized trials are considered as the gold standard to evaluate interventions because randomization controls for known and unknown confounding makes study groups comparable. In comparison, non randomized or observational studies (cohorts and case controls) are considered of lower methodological quality. Non randomization predisposes to potentially important imbalances in baseline characteristics). Actually, both designs have strengths and weaknesses that must be recognized. The different types of observational data (issued from Cohorts, Registries, or Administrative Databases) have potentially different methodological limits.
Furthermore, it is an unrealistic expectation that we will have randomized trials for every intervention and its combinations in every patient subgroup (for example if for a disease such as Rheumatoid Arthritis we have 15 different treatment options and 3 different subgroups of patients, we need theoretically at least 315 head to head trials). Therefore, 85% of the Comparative Effectiveness Research evidence will come from non-experimental data according to the American Institute of Medicine.
Disclosure of Interest None Declared