Calibration Runs of the Rat Model

Read Also

As indicated in Table 2-1, Kirman et al. (2003) chose to calibrate the key metabolic rates in their model based on information from four papers (Raymer et al. 1993; Miller et al., 1982; Ramsey et al. 1984; and Sumner et al., 1992).  As it happens these papers do not contain hemoglobin adduct information—which is our preferred direct measure of internal dose of acrylamide and glycidamide.[1]  Instead, calibration of model parameters that relate to the balance between different routes of metabolism is largely based on measures of urinary metabolite outputs [from Sumner et al. (1992).  The overall rate of processing of acrylamide in the model is also influenced by some direct measurements of blood and nervous tissue concentrations of acrylamide in the Raymer et al. (1993) and for total radioactivity in the Ramsey et al (1984) and Miller et al. (1982) papers]. 

The basic assumption in using ratios of urinary metabolites that are produced by different metabolic pathways (e.g. those via glycidamide vs. those involving direct reaction of acrylamide with GSH) is that the fraction of acrylamide urinary metabolites resulting from a particular pathway will reflect the overall fraction of acrylamide that is processed by that pathway.  Unfortunately, the fractional recovery of metabolites in urine in these experiments is typically far less than 100%.[2]  Therefore, some uncertainty remains about whether the fraction of original acrylamide that is disposed of as macromolecular reactions in tissues and other possible unmeasured pathways (e.g. complete metabolism to exhaled carbon dioxide) will be accurately reflected in the urinary metabolite ratios.  Moreover, although there are some useful measurements of parent acrylamide at a few time points in blood and other tissues from the work of Miller et al. (1982), the model as currently calibrated appears to under-predict the observed concentrations in blood and muscle by about 2-4 fold at time points greater than 2 hours, suggesting that the half-life of acrylamide might be somewhat longer than the Kirman  model would predict. 

A final difficulty is that the assembled data do not include direct measurements of the blood or tissue concentrations of glycidamide—the chemical of most direct interest for cancer risk projections.  The model contains one parameter that strongly influences the distribution of glycidamide between blood and tissues, but this parameter is particularly uncertain.  .  This is the partition coefficient multiplier, used to convert estimated tissue/blood partition coefficients for acrylamide into corresponding tissue/blood partition coefficients for glycidamide.  This is set at 3.2, by analogy with prior modeling by Kedderis et al (1996) for the epoxide of acrylonitrile in relation to acrylonitrile itself. 

In conclusion, the parameters that are the best candidates for adjustment in response to departures of model predictions from hemoglobin adduct observations are:

(1) The tissue/blood partition coefficient multiplier of 3.2 for glycidamide;

(2) The balance between P450 vs. GSH and other non-P450 metabolic routes for acrylamide, on the basis that the urinary metabolite profile observed at 24 hours may not fully represent the complete metabolic fate of acrylamide; i.e., the fraction that does not appear in urine but is irreversibly bound to tissues, or completely metabolized to building blocks that are incorporated into tissue constituents or exhaled; and

(3) The tissue/blood partition coefficients for acrylamide, which were based on statistical model projections (based on Poulin and Krishnan 1995; 1996a; 1996b) rather than direct measurements.  In fitting the human models in Section 3, we found it necessary to revise the acrylamide tissue/blood partition coefficients downward using our own model and database of human tissue/blood partition coefficients (previously used in modeling for the hydrophilic compounds caffeine and theophylline (Ginsberg et al. 2003.) and described in earlier work (Ginsberg et al., 1996, 1999).  The basic regression model used is patterned after methodology and reasoning first described by Patterson and Mackay (1989).  For completeness, we will add below (in Section 2.6) an alternative rat model based on a similar set of revised calculations of tissue/blood partition coefficients for acrylamide based on our own database of rat tissue/blood partition coefficients for other chemicals and the same fitting procedure patterned after Patterson and Mackay (1989).

---

[1] Hemoglobin adduct concentrations can be converted into integrated concentration X time estimates with the aid of simple rate constants for the bimolecular reaction between acrylamide or glycidamide and hemoglobin.  Formulas for this are provided in Section 2.4.

[2] The key paper of Sumner et al. (1992) reports recovery of about 50.4% of the administered dose in their 24 hour urine collection.  Of this fraction of the total dose recovered as urinary metabolites, 32.6% consisted of metabolites derived from the P450/glycidamide pathway at the 50 mg/kg dose used.  Thus the minimum total fraction of acrylamide dose that must definitely be attributed to metabolism by non-P450 routes at the 50 mg/kg dose used is (1-0.326)*0.504 = 0.34.