1
Introduction
Human health risk assessments1 provide the basis for public health decision-making and chemical regulation in the United States. Three evidence streams generally support the development of human health risk assessments—epidemiology, toxicology, and mechanistic information. These evidence streams differ in generalizability and design and have various strengths and weaknesses for hazard identification and dose-response analyses. Toxicology studies have traditionally served as a primary or sole evidence stream for risk assessments. They are available for many agents, and dose-response relationships are typically well characterized. However, extrapolation of toxicology results to assess hazards and risks in humans can be challenging because of toxicokinetic or toxicodynamic factors. In addition, in vivo toxicology studies are declining in number. Mechanistic investigations, in contrast, are increasing in number as well as complexity. They have less frequently served as the sole basis for hazard and dose-response analyses, but recent developments offer promise through systematic approaches to identify, organize, and interpret the diverse and complex mechanistic information for application in human health risk assessments. Epidemiologic studies are generally the preferred evidence stream for assessing causal relationships during hazard identification. However, the available studies may be limited in scope, subject to bias, or otherwise inadequate to inform causal inferences. In addition, there are challenges in assessing coherence, validity, and reliability during synthesis of individual epidemiological studies with different designs, which in turn affects conclusions on causation.
Triangulation aims to address the challenge of synthesizing evidence from diverse studies with distinct sources of bias. Bias is a systematic error that leads to inaccurate study results. Tools for assessing risk of bias provide a structured list of questions for systematic consideration of different domains (such as confounding, selective reporting, and conflict of interest). These tools also provide a structured framework for identifying potential sources of bias and informing judgments on individual studies. Examples include those used by the Navigation Guide,2 the National Toxicology Program’s (NTP’s) Office of Health Assessment and Translation3 tool, the NTP Report on Carcinogens handbook,4 and the U.S. Environmental Protection Agency’s (EPA’s) Integrated Risk Information System (IRIS) Program.5 Best practices call for use of a tool that evaluates individual domains rather than creates overall quality scores (Eick et al., 2020). Internal validity and risk of bias may vary across the individual studies, all of which are included in the evidence synthesis according to standard practice (NASEM, 2021). Triangulation is an approach to evidence synthesis that acknowledges and accounts for different sources of bias in studies that address a research question from different angles (see Box 1-1).
Although Lawlor et al. (2016) specifically discussed triangulation in epidemiologic studies, this approach is sometimes broadened to integrate epidemiology with other streams of evidence (i.e., toxicology and mechanistic evidence) (Munafo and Davey Smith, 2018).
___________________
1 For further background information about human health risk assessment, see https://www.epa.gov/risk/humanhealth-risk-assessment#tab-1 and the National Research Council report Science and Decisions: Advancing Risk Assessment (https://nap.nationalacademies.org/catalog/12209).
2 See https://prhe.ucsf.edu/navigation-guide.
3 See https://ntp.niehs.nih.gov/whatwestudy/assessments/noncancer/riskbias/index.html.
4 See https://ntp.niehs.nih.gov/ntp/roc/handbook/roc_handbook_508.pdf.
These seminal publications and ongoing discussions about best practices in evidence synthesis led to convening of the Triangulation Forum in Bristol, England, in October 2019. The forum included presentations and discussions by academics, U.S. federal government scientists, U.S. federal contractors, and representatives from international health authorities. The attendees expressed particular interest in applying triangulation in environmental epidemiology, because of the challenges posed by causal inference evaluation of environmental exposures that affect population health (Pearce et al., 2019).
Building off the Bristol Triangulation Forum and the 2019 workshop Evidence Integration in Chemical Assessments: Challenges Faced in Developing and Communicating Human Health Effect Conclusions (NASEM, 2019), the National Academies of Sciences, Engineering, and Medicine (the National Academies) developed the Workshops to Support EPA’s Development of Human Health Assessments: Triangulation of Evidence in Environmental Epidemiology.6 This EPA-sponsored workshop was held virtually on May 9 and 11, 2022. EPA defined the workshop aims as “(1) to characterize and summarize existing examples of triangulation used as a tool in risk assessment for evidence synthesis, focusing on use for synthesis within and across epidemiologic studies; and (2) to solicit input on how triangulation can be implemented, transparently documented, and clearly communicated in science assessments overall.” To address these aims, the National Academies planning committee organized a workshop on triangulation as a framework for strengthening causal inference and evidence integration. With this context in mind, the planning committee7 developed three sets of questions as the foundation of the workshop (see Box 1-2).
___________________
6 See https://www.nationalacademies.org/event/05-09-2022/workshops-to-support-epas-development-of-humanhealth-assessments-triangulation-of-evidence-in-environmental-epidemiology.
7 This proceedings was prepared by the workshop rapporteurs as a factual summary of what occurred at the workshop. The planning committee’s role was limited to planning and convening the workshop. The views contained in the proceedings are those of individual workshop participants and do not necessarily represent the views of all workshop participants, the planning committee, or the National Academies of Sciences, Engineering, and Medicine.
The workshop agenda, including speaker names and presentation titles, is in Appendix A. The workshop began with an introduction to triangulation as a concept and a discussion of methodologic considerations for its application. The first day also included presentations and discussions by scientists from state, national, and international health authorities on their approaches to evidence integration and synthesis. The second day covered case studies of triangulation and evidence synthesis in epidemiology, future opportunities for applying triangulation, and a summary session led by the planning committee. Both days of the workshop included poster sessions. The poster sessions will not be summarized in this proceedings, but copies of the posters are available on the event webpage.8
These proceedings are organized in order of presentation with additional synthesis across sessions when relevant. Chapter 2 covers the stage-setting introductory presentations, including a presentation by Deborah Lawlor from the University of Bristol. Chapter 3 discusses the health authority presentations, which began with Rebecca Nachman’s presentation on EPA’s approach to triangulation within the IRIS Program. Chapter 4 summarizes the case studies presented during the workshop, including topics such as per- and polyfluoroalkyl substances (PFAS), air pollution, radiation, and health equity. Chapter 5 provides examples of next steps and opportunities for applying triangulation. Finally, Chapter 6 presents the planning committee’s session summaries as well as ideas about the potential future use of triangulation for human health assessment.
___________________
8 See https://www.nationalacademies.org/event/05-09-2022/workshops-to-support-epas-development-of-humanhealth-assessments-triangulation-of-evidence-in-environmental-epidemiology#sectionEventMaterials.