Population-Based Observational Studies in Oncology: Proceed With Caution

A principle goal of research in Oncology is to determine the optimal treatment for our patients. This often takes the form of comparing 2 existing therapies to one another to determine which is superior, or to introduce a new therapy and determine if it is superior or noninferior to the existing standard of care. This type of research is termed comparative effectiveness research (CER), and the gold-standard is through the conduct of randomized trials. This is the preferred approach, and the only true methodologic approach that can assign a cause-and-effect relationship between a treatment effect and the observed outcome. An alternative approach that is gaining popularity is the use of population-based registry analysis given that it is quicker, cheaper, and easier to perform. However, there are unavoidable forms of bias and confounding that exist when using observational research to perform CER, and recent evidence suggests that population-based CER often results in erroneous results, and that statistical methods to minimize bias are ineffective to overcome the numerous limitations of these databases. In this article, the strengths and weaknesses of both randomized and observational research will be discussed.