E-64d br coefficient obtained from these species was express
coefficient obtained from these species was expressed as risk per life-time dose (per mg/kg or per mMh) (Table 2). The evaluation showed that a common relative risk coefficient (β) could be derived across all responding tumor sites, doses, and sexes in mice and rats, respectively.
A significant difference between the species was observed when the risk coefficient was calculated per administered dose, compared to when it was calculated per internal dose (Table 2). The mean doubling doses (1/β), expressed as lifetime AUC's were not statistically significant different for the species: 19.6 mMh for mice and 18.6 mMh for rats. The applicability of the relative risk model to tumor data for glycidol is illustrated in Fig. 2, showing that a good agreement between the ob-served and predicted number of tumors in responding sites in glycidol-treated mice and rats was obtained.
We have evaluated a relative (multiplicative) cancer risk model for application to the genotoxic compound glycidol, using data from pub-lished carcinogenicity studies where background tumor incidence in responding sites and the internal doses (from short-term exposure stu-dies) from glycidol exposure in the studied species (mice and rats) were included in the evaluation.
4.1. Selection of tumor sites
As discussed in our first evaluation on the model (Granath et al., 1999), we limit the study to responding sites, even though neoplasms are observed in other sites in control animals. The spectrum of re-sponding tumor sites are rather similar for the studied genotoxic com-pounds within one strain of animal (present study; Granath et al., 1999; Fred et al., 2008, Törnqvist et al., 2008). In the glycidol carcinogenicity studies most of the neoplasms that occur in non-responding sites have low incidence and would likely not affect the relative risk coefficient (β) if included in the calculations. There are also non-responding sites with high background incidence. In rats, these correspond to e.g. neoplasms in the testes and leukemia (males) and in the pituitary gland (females) (NTP, 1990). In mice the sites with high incidence are mainly the pi-tuitary gland and lymphoma, both in females (NTP, 1990). These strains have been developed to be sensitive, and likely have a genetic predisposition for the development of tumors of this kind and therefore no treatment-related effect is observed. The high background tumor incidence in Leydig E-64d tumors and mononuclear cell leukemia is for instance one reason for the switch from F344 rats to Sprague Dawley rats in the carcinogenicity studies, decided in 2006 (Maronpot et al., 2016).
4.2. Internal doses as a basis for the relative cancer risk coefficient
The ultimate, although difficult, dose measurements would be in target organs for the derivation of cancer risk estimates in those tissues. A low-molecular mass compound like glycidol can be assumed to be distributed rather evenly throughout the body, and the internal dose in blood inferred from Hb adduct levels can be used as an estimate for
Scheme 1. Calculation of the daily intake of glycidol in humans, corresponding to the doubling dose.
doses of the reactive compound in the target tissues. This has been demonstrated for the structurally similar epoxide ethylene oxide, which gave about equivalent doses (AUC) in several organs and in the blood in mice, as inferred from levels of adducts to DNA and Hb (Segerbäck, 1985). A quantitative risk estimate for glycidol, as the relative risk coefficient (β), for all responding tumor sites is therefore estimated based on internal dose measured in blood.
In the present evaluation of the applicability of the relative risk model to data from published carcinogenicity studies of glycidol, a common β was derived across all responding tumor sites and both sexes for mice and rats administered different doses of glycidol. An ap-proximately two-fold higher lifetime relative risk coefficient per ad-ministered dose was indicated for rats compared to mice. This variation between the species was to a large extent explained by the observed differences (statistically significant) of the mean of internal doses per administered doses of glycidol (both sexes): mice ca. 1.6 μMh per mg/ kg per day; and rats ca. 2.9 μMh per mg/kg per day (p < 0.01). These internal doses reflect the pharmacokinetics of the species, with slower detoxification of glycidol in rats compared to mice.
Recalculation of the relative risk coefficients per internal dose (mMh) instead of administered dose resulted in a non-significant dif-ference between the species; 5.1% per mMh (mice) and 5.4% per mMh (rats) (Table 2), illustrating the importance of in vivo dosimetry for a more accurate estimation of β. A larger number of animals and more dose groups in an in vivo dosimetry study could give a further im-provement of the precision in the estimation of internal doses and β, but could only give a small adjustment of the risk estimate and not change the overall conclusion. Altogether, the obtained relative risk coeffi-cients per internal dose indicate that the probability for tumor devel-opment due to glycidol exposure is approximately the same for mice and rats, and the adequate agreement between predicted and observed tumor incidence, for the responding sites demonstrated in Fig. 2, sup-ports the assumption of a common β per internal dose.