Analytic methods to bridging the gap between randomized controlled trials and observational studies

Article type
Authors
Kitio Dschassi B, Cecchin M, Dosquet P
Abstract
Background: Causal inferences can be drawn from both randomized experiments and observational studies. Unlike randomized controlled trials, where randomization theoretically balances confounding factors, observational studies pose the challenge of how to adjust appropriately for confounding. The potential for confounding bias may diminish the enthusiasm to use observational studies to assess intervention effect. However, the use of some analytic methods in observational studies narrow the gap between effects estimates of both designs. Objective: To review key analytic methods that seeks to estimate consistent causal effect in observational studies. Methods: A literature search was performed using MEDLINE to identify methodological literature on the development, application and comparison of analytic methods adjusting for confounding bias. Results: Four groups of alternative methods have been proposed, in addition to standard statistical techniques, to adjust for confounding in observational studies. The first group included propensity score methods which is an attempt to reconstruct situation similar to random assignment, albeit only with respect to observed prognostic variables. The second group used marginal structural model to estimate causal effect for time-dependent exposure in presence of time-dependent covariate that may be simultaneously confounder and intermediate variable. In these groups of methods, adjustment can be made only for observed confounders. A method that has the potential to adjust for all confounders, whether observed or not, is the method of instrumental variables.Sensitivity analyses are useful tools to assess how robust the finding is and how likely or unlikely it is that unknown confounders would change the effects estimated. Conclusions: The lack of randomization makes observational effect estimates vulnerable to confounding bias due to different prognosis of individuals between treatments groups. However, analytic methods are available to estimate consistent causal effect in observational studies. Systematic reviewers should be aware of these methods to include best evidence from observational studies on adverse effects in their review.