Use of epidemiology studies in Integrated Risk Information System (IRIS) health assessments
On this page:
Epidemiology studies can be highly influential when assessing human health effects of chemical exposures. They are a core component of the hazard identification step of risk assessment and, when sufficient data exist, are preferred over animal toxicology studies for dose-response analysis and subsequent toxicity value derivation. However, due to inherent limitations of observational studies (i.e., lack of randomization and inability to assign exposure levels/timing), it is critical to evaluate the potential impact of bias in epidemiology studies when interpreting the evidence base. Characterizing the potential presence, direction, and magnitude of bias enhances transparency and clear communication of sources of uncertainty in the conclusions made during evidence synthesis and integration.
A number of approaches have been developed to evaluate individual studies for risk of bias. A greater challenge, and a key element of synthesizing epidemiology evidence, is determining how risk of bias impacts certainty in the overall evidence for a health effect of interest, which requires considering the risk of bias across the available studies. This is not as simple as counting up the number of studies at high vs low risk of bias, as a sufficient number of high quality studies may result in low risk of bias overall even if the majority of studies are at high risk of bias. Additionally, sometimes identified concerns can be ameliorated by further analysis (such as sensitivity analysis or bias analysis). One method for evaluating bias during evidence synthesis is triangulation, wherein causal inferences are strengthened by integrating results obtained from different approaches, where each approach has different sources of potential bias.
When the evidence synthesis (considering bias as well as factors such as consistency, coherence, effect magnitude, and exposure-response gradient) demonstrates that bias is not anticipated to explain the observed effects and the evidence is sufficient to support a hazard conclusion of a likely adverse effect, epidemiology studies can often be used for toxicity value derivation. This presentation will provide examples of the use of epidemiology studies for both hazard identification and dose-response in IRIS assessments, including application of methods for evidence synthesis such as triangulation to evaluate risk of bias across studies.