br Further support for a common
Further support for a common risk coefficient β is given from pre-vious evaluations of the applicability of the model for other simple genotoxic epoxides/epoxide metabolites; 4u8C oxide (Granath et al., 1999), 1,3-butadiene (Fred et al., 2008), and acrylamide (Törnqvist et al., 2008), where approximately the same estimates of β per internal dose were obtained for different tumor sites, sexes and species (mice and/or rats) for respective compound. The evaluations throughout have shown solid results. Furthermore, in the evaluations of the carcino-genicity studies of ethylene and 1,3-butadiene, the risk coefficient was in agreement with relative genotoxic potency per AUC (compared with ionizing radiation) from in vitro genotoxicity studies. For glycidol, the relative genotoxic potency (from in vitro and in vivo experiments) has been compared with the relative cancer risk coefficient in a preliminary evaluation, which indicated a good agreement (Aasa, 2018).
A common β per internal dose (AUC) in mice and rats supports the transfer of the risk coefficient to other species, as humans. Therefore, the mean doubling dose of glycidol (19.1 mMh, c.f. Table 2) for mice and rats is assumed to be the best estimate of the cancer risk coefficient of glycidol also for humans. Translation of the doubling dose to the risk coefficient per exposure (intake) dose for humans requires that the relation between AUC and intake of glycidol in humans is known. In a 4-week exposure study, recently published by Abraham et al. (2018), 11 human individuals received a daily portion of palm fat equivalent to a mean daily dose of glycidol of 4.2 μg/kg body weight. The Hb adduct levels (diHOPrVal) were monitored from blood samples using the FIRE procedure. The relation between the internal and administered dose of glycidol was calculated to 4.3 μMh per mg/kg using the rate constant for diHOPrVal adduct formation (kval) from glycidol with human Hb, from our studies (Aasa et al., 2019). In Supplementary Table S3 the relation between the internal and administered dose of glycidol for several species, obtained in published studies, are given, showing a difference between the studied species up to ca. 8-fold. As this total variation also includes differences between the methods and labora-tories (and without inter-calibrations), the data indicate that the in-terspecies variation in the detoxification of glycidol and the resulting ratio between AUC and administered dose is not very large.
Thus, the doubling dose of 19.1 mMh (1/β) of glycidol is corre-sponding to a lifetime intake of ca. 4400 mg/kg in humans, according to the relation between internal and administered dose of glycidol. This means that a mean daily intake of glycidol of ca. 0.17 mg/kg throughout a lifetime of 70 years is estimated to double the background cancer incidence in the population. The calculations are illustrated in Scheme 1 below. A lifetime cancer risk of 1/105 is often used as a figure for acceptable risk in the population. The daily intake associated with a lifetime risk level of 1/105 was further calculated to 0.40 μg/day (70 kg) using Equation (5) (Section 2.5). In this calculation the total background cancer incidence in humans, 30%, is used as a conservative approximate of P0 (P0 for the Swedish population below 75 years; Cancerfonden, 2017). At low doses and at low background tumor in-cidence Equation (5) may be approximated with Equation (6), which gives a simplified illustration of the risk model. The approximation using Equation (6) in this case (ΔP = 1/105 and P0 = 30%) would give a deviation of only 20% of corresponding daily intake of glycidol compared to the exact calculation.
There is only one other quantitative cancer risk estimation of gly-cidol found in the literature which allows a comparison. The California Environmental Protection Agency (C. EPA) has performed a risk
estimation for glycidol and arrived at a no significant risk level (NSRL) for humans of 0.54 μg glycidol per day considering a lifetime cancer risk of 1/105 (OEHHA, 2010). The risk estimation by C. EPA is based on the same carcinogenicity studies as in the present evaluation study but estimated by using an additive model. The cancer risk due to exposure to glycidol measured in humans (children) is further discussed in a recently published paper by Aasa et al. (2019).
4.4. Comparison of relative and additive risk models and transfer across species
The risk coefficients for glycidol estimated by the approach used by C. EPA and by the relative risk model used by us, respectively, would not be expected to yield so similar values considering the large differences in the applied models. Fundamental differences are that the additive risk model used by C. EPA uses linear extrapolation of the dose-response data for the most sensitive species (rat) and without considering background tumor incidence or internal doses. Potential species differences were considered through allometric scaling