SRA Annual Meeting 2024 – Austin TX, December 8-12

The Center for Truth in Science will be active at this year’s Society for Risk Analysis (SRA) Annual Meeting taking place in early December in Austin, Texas. This long-running meeting will feature a diverse mix of researchers, risk management professionals, policy makers, and industry representatives focused on the use of risk analysis in decision making. The Center will once again be promoting our research and funding opportunities as an exhibitor – we hope you will come see us!

Additionally, we are excited to announce an oral presentation at the meeting that will showcase the Center’s work applying advanced analytical research tools and methods to questions of causality and claims of harm. The session is titled: Objective causal analysis using interventional probability of causation (IPOC), and it will be led by Dr. Tony Cox and Bill Thompson, MS.

This session was initiated in part as a response to major public health and medical research journals recently suggesting that determining if something causes harm from observational data (i.e., from surveys, case control studies, etc. rather than randomized clinical trials or laboratory studies) must rely on untestable assumptions that ultimately can only be evaluated using informed judgments by experts. Said more simply, the experts are telling us that when things get complicated, we just have to trust the experts because it isn’t possible to do it more objectively.

This subjective approach to determine if something in our lives can cause harm undermines scientific rigor and trustworthiness by failing to provide empirically testable (and potentially falsifiable) causal claims.  As a result, it is likely public policies to protect our health and wellbeing are driven by unverified judgements of selected experts, without the possibility of empirical refutation, since the key assumptions are said to be untestable. This deprives decision-makers and the public of key benefits of traditional objective science that invites scrutiny and independent verification. We think we can do better!

At last year’s SRA annual meeting, Dr. Cox and a team of researchers presented a case study using IPoC to demonstrate the carcinogenicity of benzene exposure – a well-known cause of cancer. IPoC was then applied to the question of whether inhaled formaldehyde causes cancer. This work has since been published in Critical Reviews in Toxicology, and was completed as part of a multi-year collaborative effort to demonstrate an alternative, objective (independently verifiable) approach to causal analysis of exposure-response relationships in observational data that uses empirically testable models in place of untestable potential outcome models.

The IPOC method emphasizes that interventional causal hypotheses and claims can indeed be tested empirically using well-defined statistical properties such as Invariant Causal Prediction. By defining causal models in terms of testable predictive generalizations about effects of interventions, IPOC ensures that causal claims about how changing exposure would change risk can be independently tested and verified using data.

This year’s presentation goes more deeply into the IPoC method, demonstrating that IPOC is empirically testable, allowing for the validation of causal claims through Invariant Causal Prediction (ICP) tests across multiple studies. This approach addresses the problem of hidden confounders and ensures the reliability of causal inferences.

To demonstrate how it can be used in the real world of risk analysis, the presenters will contrast this objective approach to causality with the subjective judgement approach that has recently been used in risk assessments of gas stove emissions, formaldehyde, particulate matter, and other high-profile public health applications.

Top Takeaway: The IPOC approach introduces a paradigm shift in causal analysis by establishing a scientifically rigorous, empirically testable approach based on replacing untestable claims about potential outcomes with empirically verifiable predictions across multiple studies and conditions.  This could potentially revolutionize how causal inferences are drawn from observational data, impacting fields from epidemiology to public policy.