Table 1 Overview of probabilistic risk methods for non-cancer risk assessment
Approach | Steps/Result | Benefits | Limitations |
|---|---|---|---|
I. Replacing Default Uncertainty Factors with Probability Distributions1,2,3 | 1. Derive POD for certain response magnitude 2. Apply uncertainty/variability distributions to estimate risk of that response at any dose 3. Define risk-specific dose (e.g., dose at 1-in-1000 risk for 5% magnitude effect with 95% confidence) | - Retains familiar steps of RfD/RfC derivation - Discards default of fixed population variability - Can be used to estimate level of risk associated with RfD in addition to exposures above and below the RfD | - Currently available uncertainty/ variability distributions do not represent sensitive subpopulations - Requires risk managers to define acceptable risk - May be difficult to monetize in economic analyses if the endpoint is a non-clinical health effect |
II. Using Clinical Vulnerability Distributions to Determine Probability of Impairment or Disease4,5 | 1. Identify biomarker shared by aging or disease process and chemical effect and the dose-response relationship for this biomarker 2. Identify population distribution of shared biomarker 3. Estimate additional risk of functional impairment or clinical disease due to shift in biomarker distribution from chemical exposure | - Captures effect-specific population variability based upon distribution of risk biomarker - Use of clinical outcomes means results are likely adaptable for economic benefits analysis | - Data-intensive due to requirement for shared biomarker between disease and chemical - Disease risk may be more complex than predicted by a single biomarker |
IIIa. Continuous Dose-Response Models with Human Epidemiological Data6 | 1. Meta-analysis or single high-quality studied based on common biomarkers of exposure and effect across studies 2. Fit risk function to resulting epidemiological dose-response data | - Captures as much variability as is represented in underlying epidemiological studies | - May require advanced statistical methods if combining across epidemiological datasets -Relies on human epidemiology which can be difficult to obtain and means adverse effects are already occurring in human populations |
IIIb. Quantifying Risk of Non-Cancer Effects from Continuous Dose-Response Models with Animal Data7 | 1. Use benchmark dose model to estimate POD 2. extrapolate below POD to estimate probability of effect at RfD or other doses | - Relies on existing methodologies - Expands use of BMD model | - Assumes same mechanism and dose-response at higher and lower doses -Only applies to reference values that are based on benchmark dose as POD - Does not capture human variability or any other uncertainty not reflected in the experimental model |