Session Chair: Ian Shrier, McGill University
This workshop will introduce participants to directed acyclic graphs (DAGs). We will review the basic principles and show how they can be used to determine appropriate sets of variables for estimating total causal effects of exposure (treatment). Participants will work through concrete examples of increasing complexity. We will also introduce how DAGs can be used in more advanced applications, including natural and controlled direct and indirect effects and study design.
Session Chair: Jennifer Lund, University of North Carolina at Chapel Hill
Randomized clinical trials (RCTs) are considered the gold standard for assessing efficacy of new therapies and are required for regulatory approval. However, patients enrolled on trials are often not representative of patients in whom treatment will ultimately be delivered in clinical practice. When response to therapy varies across subgroups, differences between trial and clinical populations can contribute to the “efficacy-effectiveness gap” – where a treatment’s efficacy in a trial differs from its effectiveness in clinical practice. Methods for generalizability and transportability can help bridge this gap. These methods combine RCT and clinical practice data to generate evidence that directly addresses therapy effectiveness in target populations. Such approaches leverage the internal validity of RCTs with the external validity of clinical practice data to better inform real-world decision-making.
In this workshop, we will provide an overview of methods for generalizing and transporting treatment effects from RCTs to defined target populations. Participants will receive SAS and R code to combine publicly available RCT and real-world data. Participants will gain an understanding of the theory underlying external validity. Using graphics and quantitative metrics, participants will evaluate the suitability of and compare effect estimates transported to various target populations.
This workshop requires an introductory level of epidemiology training and is relevant for all interested in expanding their epidemiological toolkit. This workshop may be of particular interest to those focused on causal inference methods, pharmacoepidemiology, and comparative effectiveness research.
Session Chair: Mike Jackson, Kaiser Permanente
Epidemiologists can use statistical graphics to understand our data and to guide us toward correct inferences. Well-designed graphics can also be powerful tools for communicating our study findings. However, while statistical software makes it easy to produce certain types of figures, the default options leave much to be desired. Too often, the result is figures that distract, confuse, or even distort data. In this workshop, participants will first learn the fundamentals of effective data visualization. This includes selecting appropriate chart types, drawing attention to the relevant data, using effective visual cues, and providing helpful context. We will discuss how to put these principles into practice, leading viewers to make comparisons, identify trends, and find meaningful correlations. Finally, we will walk through techniques for going beyond the default settings of various software packages to produce well-designed figures.
Session Chair: Chuck Huber, Stata
Meta-analysis is a statistical technique for combining the results from multiple similar studies. The talk will provide a brief introduction to meta-analysis and will demonstrate how to perform meta-analysis in Stata 16. The -meta- command offers full support for meta-analysis, from computing various effect sizes and producing basic meta-analytic summaries and forest plots to accounting for between-study heterogeneity and potential publication bias. Examples demonstrating how to conduct meta-analysis within Stata will be provided. These examples will focus on the interpretation of meta-analysis under various models, meta-regression, subgroup analysis, small-study effects and publication bias, and various types of forest, funnel, and other plots.