Humans are exposed to chemical compounds in everyday life, both from the environment and from endogenous processes. Some compounds constitute a risk for cancer development. One such compound is glycidol, which is genotoxic and an animal carcinogen. It is the model compound of this work, partly due to its presence in food. Glycidol, often together with 3-monochloropropane-1,2-diol (3-MCPD), occurs in the form of esters particularly in refined cooking oils, which are used in a variety of food products. The esters are hydrolyzed in the gastrointestinal tract to form glycidol (and 3-MCPD).

The aim of the thesis has been to evaluate an approach for cancer risk estimation of genotoxic carcinogens based on a multiplicative (relative) risk model and genotoxic potency. Further, the aim was to estimate the cancer risk for exposure to glycidol via food. Measurement of the internal doses (concentration × time) of glycidol in the studied biological systems, including humans, has been crucial. Glycidol is electrophilic and forms adducts with nucleophilic sites in proteins and DNA. The doses of glycidol were quantified by mass spectrometry: in vivo from adduct levels to hemoglobin (Hb); in vitro from adducts to cob(I)alamin.

The first part of the thesis concerns the genotoxic potency (genotoxic response per internal dose) of glycidol, measured in vitro by mutation studies and in vivo by micronuclei as a biomarker for genotoxicity (short-term studies in mice). The results were compared to that of ionizing radiation, used as a standard, to estimate the relative genotoxic potency of glycidol: 10 and 15 rad-equ./mMh from mutations and micronuclei, respectively. No induction of micronuclei was observed for the related compound 3-MCPD.

Tumor incidence from published carcinogenicity studies of glycidol in mice and rats, together with the measured in vivo doses, was evaluated with the relative cancer risk model. A good agreement between predicted and observed tumor incidence was shown, and no significant difference of the obtained cancer risk coefficients (risk per dose) between mice (5.1 % per mMh) and rats (5.4 % per mMh) was observed. The overall results support that the relative risk coefficient (β) is independent of sex, tumor site, and species, and indicated that it can be transferred also to humans. The doubling dose, expressed as 1/β, is the dose that is required to double the background tumor incidence. The mean of the doubling doses from mice and rats (19 mMh) was assumed valid for risk estimation for humans. Transfer of β of glycidol to rad-equ. via its relative genotoxic potency showed a risk coefficient in agreement with the relative cancer risk coefficient of ionizing radiation.

In the final work, the lifetime (70 years) in vivo doses of glycidol were calculated from measured Hb adduct levels in blood from 50 children and 12 adults, and compared to the doubling dose. A fivefold variation was observed in the in vivo doses. The estimated lifetime excess cancer risk from glycidol exceeds 1/1000. This is much higher than what is considered as an acceptable risk.

To conclude, the multiplicative (relative) risk model together with relative genotoxic potency is promising to use in an approach for cancer risk estimation and in line with 3R (reduce-refine-replace) initiatives.