The main advantage of the randomized controlled trial (RCT) is the random assignment of treatment that prevents selection by prognosis. Nevertheless, only few RCTs can be performed given their high cost and the difficulties in conducting such studies. Therefore, several analytical methods for removing the effects of selection bias in observational studies have been proposed. The first aim of this paper is to compare three of those methods: the multivariable risk adjustment method, the propensity score risk adjustment method, and the instrumental variable method. The second aim is to compare the results from observational studies using the instrumental variable method with those from RCTs aiming to answer the same study question.
Keywords: epidemiology; nephrology; patient selection; statistics; study design.