How to Conduct Systematic Reviews and Meta-Analyses to Improve Environmental Health Policy and Decision-Making
ADVANCED SESSION
UCSF Mission Bay Campus (in person only):
Beginner Workshop: Thursday, February 13, 2025; 10:30 am - 5 pm (Beginner course outline)
Advanced Workshop: Friday, February 14, 2025; 9 am - 4 pm
Advanced Workshop Course Outline
Part 1: Strengths and Limitations of Divergent Approaches for Assessing ‘Risk of Bias’ in human Environmental Epidemiology Studies
Systematic reviews are increasingly prevalent in environmental health due to their ability to synthesize evidence, while minimizing bias. An important step in conducting a systematic review is to evaluate internal validity, or risk of bias (ROB) of the studies. Different approaches have been developed and applied including by the U.S. Environmental Protection Agency (EPA)’s Integrated Risk Information System (IRIS), which uses a modified version of the Cochrane ROBINS-I tool (Risk of Bias In Non-randomized Studies of Interventions), and the University of California San Francisco’s Program on Reproductive Health and the Environment, Navigation Guide approach. Additionally, new approaches have recently been published for use, including the ROBINS-E tool (Risk of Bias In Non-randomized Studies of Exposures).
The objective of this session is to discuss the performance of these and other tools in assessing ROB. Specifically, we will discuss the impact the ROB ratings for each tool may have on evaluating the overall quality of the body of evidence. The session will also discuss results of a recent study that has applied several of these tools and what the possible public health implications are of using these tools when examining harms of environmental exposures in a systematic review.
Part 2: Evidence Synthesis When Meta-analysis is Not Possible
When meta-analysis is not possible due to too much heterogeneity or when data is not adequately reported, narrative reviews play an important role in summarizing the results. For example, studies may only report point estimates, estimates are reported on different scales, used different association metrics or the scales on figures are not fully reported preventing conversion of the results across studies into a single scale.
This session will discuss established methods by Cochrane for statistical synthesis when meta-analysis of effect estimates is not possible, including combining P values, estimating the proportion of effects favoring the intervention along with a confidence interval (e.g. using the Wilson interval methods) and the use of heat maps to highlight statistical significance.
The session will explore how to apply judicious consideration and detailed discussion on the studies characteristics, including the sample size (which would impact the weight of a study in a meta-analysis) and its risk of bias, along with any other characteristics that may strengthen or reduce the confidence in the results.