How to evaluate the robustness of causal findings in observational studies: Application of E-values and Bounding Factors

Background Observational studies are often susceptible to unmeasured confounding. Traditional sensitivity techniques frequently require strong and untestable assumptions, making a sensitivity analysis challenging in causal analysis. Frameworks such as GRADE (Grading of Recommendations, Assessment, Development and Evaluation) and the TARGET (Transparent Reporting of Observational Studies Emulating a Target Trial) encourage robustness and transparency in causal inference. GRADE recommends assessing the impact of residual confounding, whereas TARGET encourages routine reporting of sensitivity techniques. Objectives This article provides an overview of E-values and the bounding factor, a relatively new metric to assess robustness to unmeasured confounding. It discusses their practical application, advantages, limitations, and best practices for interpretation in epidemiologic research. Sources Relevant peer-reviewed literature, methodological and applied papers, and tools for E-value calculation were reviewed until October 2025. Content E-values quantify the minimum strength an unmeasured confounder would need to have with both the exposure and the outcome to explain away an observed effect, conditional on the measured covariates. E-values require no additional assumptions, easy to compute and enhance transparency when routinely reported. Limitations include the absence of a universal cut-off for interpretation, inability to account for other biases, and potential for selective reporting. E-values should be reported within the context of the study and the measured confounders. Researchers are encouraged to identify plausible unmeasured confounders to guide future investigations. Bounding factors quantify how much an observed effect estimate would be altered if an unmeasured confounder of a specific strength was associated with both the treatment and outcome. This approach enables the investigators to evaluate realistic bias scenarios and assess their potential impact on the observed effect estimate. Implications E-values and bounding factors should be considered as complementary approaches to other sensitivity techniques. Routine application of such techniques can help researchers to explore the plausibility of unmeasured confounders in causal studies.