Extension and validation of a physiologically based toxicokinetic model for risk assessment of aluminium exposure in humans

Abstract

The safety of aluminium (Al) exposure from sources such as food, parenteral nutrition or adjuvanted medicinal products is still a matter of uncertainty. Since toxicokinetic studies in humans are lacking, model predictions are warranted for risk assessment. Recently, we established a physiologically based toxicokinetic (PBTK) model for Al built on a comprehensive toxicokinetic $_{}^{26}\text{Al}$ database, which could describe Al biokinetics in rats and human adults after single oral and intravenous doses of soluble Al salts. Since then, we have substantially amended the model, rendering it applicable to accurately represent children and their dynamically changing physiology (including maturating renal function in neonates and increased bone turnover during puberty). Also, additional sources of exposure were implemented, including vaccinations, subcutaneous allergen immunotherapies, food, antacids and parenteral nutrition. The model predictions in plasma and tissues were then compared to own published data and literature Al measurements after exposure from food (human reference values), parenteral nutrition (toxic levels in children and adults), adjuvanted allergen products or vaccines in rats and humans, and whole-body retention data. Al levels were predicted remarkably well, in plasma and toxicologically important tissues like bone, liver and brain. To our knowledge, this is the first Al PBTK model in humans ready for use in regulatory risk assessment, allowing to simulate Al exposure in children and adults from various sources of Al exposure like food and drinking water, Al contaminations in parenteral nutrition solutions, or poorly soluble Al complexes in medicinal products including Al-adjuvanted immunotherapeutics and vaccines.

Keywords: Aluminium; PBTK; Toxicokinetics; Validation.

Fig. 1

Structure of the Al PBTK model. Each compartment represents an ordinary differential equation, straight arrows indicate an exchange of $\text{Al}$ between compartments and jagged arrows are dosing sites. All exchange processes were modelled as first-order, except that absorption of Al adjuvants was modelled as a zero-order process. Symbols written on top of arrows are model parameters, with (non-estimated) physiological parameters in parentheses. Modules/parametrizations which have changed or are newly added compared to Hethey et al. (2021) are highlighted in red. Abbreviations: AlCit, aluminium citrate; AlChl, aluminium chloride; PN, parenteral nutrition; i.v., intravenous; p.o., per oral; s.c., subcutaneous; i.m. intramuscular

Fig. 2

Reference Al concentrations in adults from enteral uptake via food according to simulations (left, adult range 20–50 years highlighted) and from literature data (right). (Calculated) median, interquartile range and/or range are displayed as reported in the respective literature source. Abbreviations: ww, wet weight; CI, confidence interval; IQR, interquartile range; Hellstr., Hellstroem; McLach., McLachlan; Rahil-K., Rahil-Khazen

Fig. 3

Fraction of injected Al amount remaining at the injection site (mean±standard deviation from n samples) over time from various animal studies after subcutaneous (s.c.) or intramuscular (i.m.) injection of products adjuvanted with different Al preparations. Dashed line (shaded area): predicted fraction remaining (90% confidence intervals) using the estimated adjuvant-specific zero-order absorption rates. Abbreviations: AH, Al hydroxide; AP, Al phosphate

Fig. 4

Model validation for Al exposure from adjuvants in rats (A, B) and humans (C). A/B: comparison of the model predictions to experimental data for two Al-containing adjuvants in rat plasma (A) and tissues (bone and brain, B). Simulated Al exposure is via adjuvants only, and data represent excess Al exposure compared to a control group, displayed as mean±SD (note that mean−SD or mean can become negative this way, and whiskers extend to the bottom plot range in this case). Simulations for other adjuvants can be found in the Supplementary Material (Fig. S6). C: comparison of the model predictions to measured urinary Al excretion rates in humans on subcutaneous immunotherapy vs. control (enteral uptake from food only). Abbreviations: pAP (i.m.), intramuscular administration of Al phosphate-containing Adjuphos; P2 (s.c.), subcutaneous administration of Al hydroxide-containing allergen product no. 2; SCIT, subcutaneous immunotherapy; CI, confidence interval

Fig. 5

Model validation using Al tissue data obtained in individuals on total parenteral nutrition (PN). A: one term newborn (Moreno et al. 1994), B: a population of children aged 18 to 34 months (Klein et al. 1984), C: a population of adults receiving PN for 1–6 years (Klein et al. 1982). Simulations were stopped at the timepoint of sampling. Abbreviations: ww, wet weight; CI, confidence interval

Fig. 6

Model validation using data from a combined p.o. and i.v. exposure scenario reported by Stoehr et al. (2006). An mixed male/female adult population was simulated and compared to data (mean±SD). Prolonged parenteral Al exposure led to the model predictions reaching steady-state. Subsequently, an oral antacid (sucralfate) was added, leading to increased exposure. The first observation is a pre-sucralfate baseline observation. Abbreviations: CI, confidence interval

Fig. 7

Model validation using $_{}^{26}\text{Al}$ retention data after single i.v. administration as citrate. A/B: Comparison of fraction of retained Al in the body reported in Talbot et al. (1995); Newton and Talbot (2012), on a timescale of days (A) and years (B). C: simulated distribution of retained Al in different tissues. Abbreviations: fid, fraction of injected dose; CI, confidence interval