Background:
Toxicological assessment of chemical mixtures is becoming more important as humans are chronically exposed to multiple agents rather than single compounds. Metal mixtures are especially concerning due to their dual role as essential trace elements and toxic pollutants, with further relevance to occupational exposures and smoking. However, many in vivo studies lack clear relevance to humans and have significant methodological flaws.

Methods:
This critical review systematically examined published in vivo studies of metal mixtures, focusing on common pitfalls. Shortcomings include: (i) inadequate mixture definition without justification of human relevance; (ii) inappropriate biomarker selection, failing to capture both exposure and effects; (iii) excessive numbers of mixture components, making factorial designs unmanageable; (iv) dose selection based on uncertain reference levels and irrelevant exposure routes; (v) neglect of chemical speciation, despite organic and inorganic metal compounds exhibiting markedly different toxicokinetics and toxicodynamics. While dissociation produces the free metal ion, which binds sulfhydryl groups, disrupts metalloenzymes, redox balance, and ion exchange, the target organ determines divergent clinical manifestations (e.g., neurotoxicity vs. nephrotoxicity). Additional flaws include (vi) improper euthanasia and tissue preparation, such as the use of anaesthetics or omission of organ perfusion; (vii) neglect of species, strain, sex, and age in animal models; and (viii) flawed comparisons between studies that are not comparable.

Results:
These flaws explain the variability and poor reproducibility reported in the literature. They hinder interpretation of mixture effects (synergism, additivity, antagonism), distort biomarker outcomes, and impair extrapolation to humans. Moreover, standard analytical methods such as atomic absorption spectrometry cannot differentiate whether a measured metal originates from an organic or inorganic compound, further limiting mechanistic understanding. Flawed results are often incorporated into toxicological databases, creating misleading inputs for in silico models.

Conclusion:
Future studies should: (i) limit mixtures to no more than three metals relevant to humans; (ii) apply mechanistically based dose selection; (iii) use realistic exposure routes; (iv) ensure proper euthanasia and mandatory organ perfusion before tissue sampling; (v) select biomarkers that reflect both exposure and effects; and (vi) consider species, strain, sex, and age. Only with these improvements can in vivo toxicology of metal mixtures yield meaningful and translatable insights. Poorly designed studies that propagate flawed data into databases ultimately weaken the predictive power and regulatory reliability of in silico toxicology.