(190g) Exposure Reconstruction of Multiple Chemicals from Human Biomonitoring Data Using Markov Chain and Differential Evolution Monte Carlo | AIChE

(190g) Exposure Reconstruction of Multiple Chemicals from Human Biomonitoring Data Using Markov Chain and Differential Evolution Monte Carlo


Karakitsios, S. - Presenter, Aristotle University of Thessaloniki
Sarigiannis, D., Aristotle University
Handakas, E., Aristotle University of Thessaloniki
Gotti, A., Aristotle University of Thessaloniki
Humans are exposed to a wide range of chemicals that can be found in multiple environmental and microenvironmental media such as dust, soil, food, air, daily health care products and through multiple routes namely through the skin, inhalation and orally. These chemicals are emitted from human activities and they can have a significant contribution to short and long term exposure events causing adverse effects to human. However, the external exposure and the intake can be evaluated and estimated using Physiologically Based Biokinetic models assimilating biomonitoring data. This approach is known as reverse dosimetry or exposure reconstruction and its major and defining benefit is the retrospective interpretation of biomonitoring data allowing the estimation of cumulative and aggregate exposure.

The current study aims at the estimation of external and target tissue exposure to 17 different chemicals, including both rapidly (BPA, DEHP, triclosan) and non-rapidly (PCBs, BDEs, HCB, DDT, metals) metabolized compounds, starting from human biomonitoring (HBM) data and focusing on perinatal and childhood exposure.

The simulations were carried out using the INTEGRA platform, a software that provides realistic exposure scenarios coupled with a generic physiologic based bio-kinetic (PBBK) model and numerical “reverse engineering” techniques for exposure reconstruction. The applied exposure reconstruction algorithm is based on the Markov chain Monte Carlo and Differential Evolution Monte Carlo techniques. The process starts from ancillary exposure-related data that are fed into the exposure model taking into account multiple exposure routes. The results are evaluated against the biomonitoring data distributions, aiming at the reduction of uncertainty in back-calculating doses, by minimizing the error between the predicted and the actual biomonitored data. Parameterization of the model to cover a large chemical space is feasible using quantitative structure–activity relationship (QSAR) models provide by INTEGRA. Ancillary exposure parameters were obtained from the INTEGRA database.

The methodological framework was based on reverse dosimetry modelling and creation of exposure probability distributions. The distributions were consistent with collected human biomonitoring (HBM) data. HBM data were obtained from cohort and biomonitoring studies from several Mediterranean countries, namely Spain, Italy, Croatia and Slovenia. Then, the reconstructed exposure estimates were used to feed the PBBK model that was executed in forward-mode using two different dietary scenarios to derive internal doses of chemicals in target tissues.

The first exposure scenario consisted of one exposure event for newborn until the 4th year of his/her life. The main assumption for the newborns until the first six months of age was that they were fed exclusively with breast milk. Then, it was assumed that from the 6th until the 18th month daily diet consisted of 6 different meals and from the age of 18 months to 4 years it consisted of differentiated meals every 3 hours. It was also assumed that the contribution of each meal to the daily intake dose is the same. This assumption was based on the fact that the modelled chemicals have long half-life time, and consequently they accumulate for years in the human body.

The second exposure scenario consisted of one exposure event lasting 30 years assuming a daily food consumption based on the actual dietary schedule of the generic European adult population. In this case the main assumption was that the generic population consumes 3 daily meals. The exposure scenario of the simulation was assumed to start at the age of 15 years.

The examined chemical substances were PCB and PBDE congeners (PCB28, PCB52, PCB99, PCB101, PCB153, PCB180, BDE28, BDE47, BDE153, BDE154), HBC, pp’DDT, mercury, arsenic, BPA, DEHP and triclosan.

The results of the simulations showed that the predicted intake dose is commensurate with intake estimates found in the literature for both short- and long-term exposure scenarios in the European population. For the cases of BPA, DEHP and triclosan external intake estimates (e.g. for BPA 0.4 μg/kg_bw/d) were significantly lower than the respective tolerable daily intake (TDI). The estimated internal dose and the respective concentration in breast milk of BPA, DEHP and triclosan was very low because of their rapid metabolism. Similar were the results on exposure of neonates and infants.

Regarding non-rapidly metabolized chemicals, the BDE-47 simulation for Spain population showed that in steady state, at the age of 30, in utero concentration in the fetus of a pregnant mother had a median (min – max) value of 0.2 (0.1 - 0.8) ug/L. In Spain for the PCBs the results showed that in steady state and at the age of 4 years, PCB-153 concentration in the liver and in the brain are similar with a median value respectively of 1.3 (0.4 – 4.2) ug/L and 1.5 (0.6 – 3.8) ug/L while the concentration levels in gastrointestinal tract were one order of magnitude lower (median 0.1 (0.01 – 0.7) ug/L). Additionally, the simulation results for PCB28 and PCB101 showed brain concentrations with median value 1.9(0.9 – 3.0) and 0.9(0.0 – 4.0) ug/L. Liver concentration for PCB52 and PCB180 was 0.5(0.8 – 3.5) and 2.1(0 – 14) ug/L respectively. Moreover, PBBK simulations for pp’-DDT were carried out for the Menorca and Valencia population. Results revealed that in steady state condition and at the age of 30 years, the concentration of pp’-DDT in brain and in utero are about 1.5 - 2 times higher for the population of Menorca than in Valencia. Results showed that in Menorca the concentration in the brain and in utero was 3.5(0.2 – 12.1) and 2.5(0 – 2.2) ug/L. Last but not least, the HCB concentration in the adult brain of Valencia population had median value 2.1(0.6 – 7.7) ug/L.

Examining the concentration of arsenic in Slovenian and Italian populations in skin tissue were 5.5 (1.0 – 29.7) and 18.0(6.8 - 47.3) ug/L respectively. Additionally, arsenic concentration in the lung tissue for the Slovenian and Italian population had median values of 1.7(0.4 – 7.1) and 4.1(1.1 – 15.2) ug/L. Moreover, the concentration of mercury in Slovenian and Croatian populations in the brain was 46.5(12.1 – 174.8) and 9.2(2.8 – 29.0) ug/L respectively. Also, for the above mentioned populations the gastrointestinal concentration was calculated 53.8(9.6 – 58.4) and 134.9(56.0 – 321.1) ug/L.

From the methodological point of view, using an integrated exposure framework for linking mechanistically external and internal exposure, provides a comprehensive overview on how realistic exposure scenarios are translated into internal dose to humans, accounting for the age-dependent and route specific bioavailability differences. To this aim a generic PBTK modelling framework that captures lifetime internal exposure is a valuable tool with many applications in chemical risk assessment, by exploiting the continuously growing wealth of in vitro testing and biomonitoring data. It is worth mentioning that the advantage of using an approach based on PBBK modeling is that it allows the estimation of internal doses of xenobiotics that exceed levels associated with biological pathway alterations and, eventually, induction of biological perturbations that may lead to health risk. This represents a more appropriate metric to be linked with adverse health outcomes providing a more mechanistic and biology-based approach to human health risk assessment.

Assessing exposure at multiple scales across the source-to-dose continuum, needs to take into account the actual complexity of the environmental and biological/physiological processes that are critical to the proper description of the phenomena involved. This results in targeted interventions and consequently more cost efficient risk management. In addition, a comprehensive integrated exposure framework estimating tissue dosimetry for the various relevant exposure scenarios, could be of great use in exploiting the in vitro HTS (High-Throughput Screening) results rapidly produced by ToxCast21, advancing thus both exposure science and toxicology towards serving the needs of risk assessment in the 21st century. It is remarked that through PBBK models it can be effectively predicted and estimated the uptake dose and the deposition in tissues although the uncertainty that exists to the design and develop of the model.

Overall, our comprehensive modelling framework supports the association of a variety of environmental, exposure and biomonitoring data, as well as the incorporation of recent advances of in vitro toxicology using high-throughput systems in the risk assessment process enhancing thus significantly the artillery of environmental health science and chemical safety assessors and managers.