Hazard characterisation imputation¶
In some cases it may be that there are substances that are known to cause (or may possibly cause) the effect of interest, but for which there are no data available for obtaining hazard characterisations. I.e., for these substances, there are no points of departure or dose response models. Instead of excluding these substances in quantitative analyses, it is also possible to impute hazard characterisations for these substances based on hazard characterisations of other (similar) substances, and use these for calculating, e.g., relative potency factors or for risk assessment.
Munro P5 (TTC approach)¶
The Threshold of Toxicological Concern (TTC) is an example of a tier for extrapolation of hazard characterisations from other substances that is already in common use (see [Munro et al., 1996]). The Munro collection of NOELs/LOAELs is a collection of NOELs/LOAELs for chemicals for the critical (i.e., first occurring) effect. In the TTC approach, the toxicity of an unknown substance is, depending on its Cramer class (see [Cramer et al., 1976]), imputed by the 5th percentile NOAEL of the sub-collection of the corresponding Cramer class.
Two variations of this approach are to use the empirical NOAEL distribution itself (just sample from the NOAEL data), or to fit a distribution (e.g. lognormal) to the empirical data and sample from the parametric distribution. MCRA provides an implementation of the TTC approach that uses the empirical distribution. In the nominal run, this implementation imputes the hazard characterisations with a value equivalent to the TTC. In the uncertainty runs, NOAELs are sampled from the empirical distribution.
The TTC is a conservative estimate of the NOAEL for at least two reasons:
TTCs are calculated from a collection of NOELs for the critical (i.e., first occurring) effect within each study and often the effect of interest will not be the critical effect, and therefore higher NOAELs are expected.
The TTC is a low percentile and therefore a conservative estimate for a random class member with unknown NOAEL.
Munro central value¶
To avoid the conservatism of taking the 5th percentile in the Munro P5 approach, alternatively, a nominal (or central) value could be taken from the Munro collection for each Cramer class. For a nominal run without uncertainty, the expected contribution of a substance with missing hazard characterisation to the risk as quantified in the hazard index is obtained from
Here SF are all combined safety factors. It follows from this equation that an unbiased estimate for the contribution from a substance with missing hazard characterisations is obtained by taking the harmonic mean from the available NOAELs:
This is the value to use in a nominal run without uncertainty for the Munro central value approach. For the uncertainty runs, this approach also uses random sampling from the empirical distribution of the corresponding Cramer class.
Available hazard characterisations distribution P5¶
Another conservative aspect of the TTC approach is the fact that the Munro set lists NOELs/LOAELs for critical effects, not for the specific effect under study. Therefore an alternative is to use the effect-specific hazard characterisations of the substances for which these are available. This collection will have on average higher NOAELs than those of the Munro NOEL collection, because for many substances, the effect of interest will not be the critical effect.
Available hazard characterisations distribution central value¶
Similar to the Munro central value approach, a central value could also be obtained from the set of effect-specific hazard characterisations distribution for imputation of hazard characterisations. This approach may yield the most realistic, or unbiased imputation value for missing hazard characterisations.