Promoting Education and
   Research in Clinical and Translational
Science across the Career Pipeline

Comparative Effectiveness Research Program

Home > Comparative Effectiveness Research (CER) > Methodologies

Comparative Effectiveness Research (CER): Methodologies

The ICRE specializes on the following CER methodologies:

Meta-analysis and Systematic Reviews

Meta-analysis is the most common method for summarizing data from multiple clinical trials. It can be used to evaluate differences in clinical characteristics of the participants in the various trials and to make inferences regarding the effect of these differences on outcomes.

Clinical and Administrative Database Analysis

One of the problems with the use of clinical trials for CER is that many of the trials are conducted in populations that are chosen specifically to enhance the ability of the study to find a difference between therapeutic methods. In such cases, the results may not give a true indication of the effectiveness of a particular therapy when used in standard practice and under less-controlled conditions. Therefore, to evaluate the relative effectiveness of treatments in "actual-use" situations, many analysts advocate the use of administrative claims databases and clinical registries to conduct CER. However, observational studies are complicated by selection bias, indication bias, and the potential for unknown factors that differentiate patients exposed to one treatment versus another. Current epidemiologic and statistical methods have made substantial progress in these areas, and the use of administrative databases and registries for CER is increasing.

"Throughout fellowship, I had 5 first author publications, 1 second author, 1 book chapter, and two oral first author presentations. Currently, I have 5 studies (1 as first author) submitted for publication, and also received two grants (thanks to my grant writing class!). I could not have done these things without my additional education."

—Former CER Trainee

Modeling and Computational Methods

Mathematical and decision analytic models are used to represent diseases and treatments and to estimate the optimal effects of different diagnostic or therapeutic interventions under diverse circumstances. These models have been increasingly used as tools to conduct CER. The models vary in complexity, with examples including simple branch-and-node decision trees, time-varying Markov models, individual microsimulations, and mechanistically detailed agent-based models. At lower levels of complexity, decision models simply represent a method for summarizing knowledge from many sources. As models become more complex and biologically or mechanistically based, they have the potential to substantially enhance knowledge and allow CER through the use of virtual clinical trials that can help address questions that are either impossible or unlikely to be addressed by randomized clinical trials (RCTs).

Clinical Trials

The most straightforward CER methodology is to conduct an RCT that represents the specific comparison required. However, conducting multiple head-to-head RCTs is expensive and time-consuming. One solution to this problem is for RCTs to be conducted in more thoughtful ways that allow for future evaluations and post hoc analyses. For example, if two RCTs, each comparing a specific drug to a placebo, are conducted with sufficient attention to baseline data, inclusion and exclusion criteria, and appropriate follow-up, some inferences can be made regarding the direct comparison between the two drugs.