Exploring the relationship between blood toxic metal(oid)s and serum insulin levels through benchmark modelling of human data: Possible role of arsenic as a metabolic disruptor.

The major goal of this study was to estimate the correlations and dose-response pattern between the measured blood toxic metals (cadmium (Cd), mercury (Hg), chromium (Cr), nickel (Ni))/metalloid (arsenic (As)) and serum insulin level by conducting Benchmark dose (BMD) analysis of human data. The study involved 435 non-occupationally exposed individuals (217 men and 218 women). The samples were collected at health care institutions in Belgrade, Serbia, from January 2019 to May 2021. Blood sample preparation was conducted by microwave digestion. Cd was measured by graphite furnace atomic absorption spectrophotometry (GF-AAS), while inductively coupled plasma-mass spectrometry (ICP-MS) was used to measure Hg, Ni, Cr and As. BMD analysis of insulin levels represented as quantal data was done using the PROAST software version 70.1 (model averaging methodology, BMD response: 10%). In the male population, there was no correlation between toxic metal/metalloid concentrations and insulin level. However, in the female population/whole population, a high positive correlation for As and Hg, and a strong negative correlation for Ni and measured serum insulin level was established. BMD modelling revealed quantitative associations between blood toxic metal/metalloid concentrations and serum insulin levels. All the estimated BMD intervals were wide except the one for As, reflecting a high degree of confidence in the estimations and possible role of As as a metabolic disruptor. These results indicate that, in the case of As blood concentrations, even values higher than BMD (BMDL): 3.27 (1.26) (male population), 2.79 (0.771) (female population), or 1.18 (2.96) μg/L (whole population) might contribute to a 10% higher risk of insulin level alterations, meaning 10% higher risk of blood insulin increasing from within reference range to above reference range. The obtained results contribute to the current body of knowledge on the use of BMD modelling for analysing human data.