How Toxicology Research Is Changing Chemical Safety

Toxicology research has moved well beyond the old idea that “the dose makes the poison.” Modern toxicologists study how chemicals interact with living systems at every scale, from single enzymes inside a cell to entire food webs in the ocean, and the tools and questions have shifted dramatically in the past two decades. The field now grapples with chemicals that behave unpredictably at low doses, contaminants so new that no regulatory framework covers them, and genetic differences that make one person far more vulnerable than the next to the same exposure.

Why Low Doses Sometimes Do More Harm Than High Ones

One of the foundational assumptions in traditional toxicology was that a higher dose always means a bigger effect. Plot the dose on one axis and the response on the other, and you expect a line that climbs steadily upward. For many poisons, that holds. But for a growing class of chemicals, especially those that mimic or interfere with hormones, the curve can dip, rise, and dip again. These non-monotonic dose-response relationships mean a moderate dose could cause an effect that a higher dose does not, because at the higher concentration the body’s feedback systems kick in or a different receptor takes over.

Endocrine-disrupting chemicals like bisphenol A have been a central case study. Non-monotonic dose-response curves have been demonstrated for natural hormones and endocrine disruptors across cultured cells, organ cultures, laboratory animals, and human populations.¹ A systematic review of such relationships found that more than half of the ones examined had a moderate to high level of plausibility, and several molecular mechanisms can explain them: multiple receptors with different binding strengths, receptor desensitization, negative feedback loops, or changes in how the chemical is metabolized as the dose shifts.² This matters for regulation because safety limits are usually set by finding a dose that causes no observable harm and then dividing by a safety factor. If the chemical’s danger does not follow a straight line upward, that approach can miss real risks at lower exposures.

How the Body Processes Foreign Chemicals

When a foreign substance enters the body, the liver’s primary job is to convert it into something water-soluble enough to be excreted. A large family of enzymes called cytochrome P450s handles the heavy lifting. These enzymes are responsible for metabolizing a wide variety of compounds, from prescription drugs to environmental pollutants, and they play a central role in detoxification.³

The catch is that detoxification sometimes backfires. An analysis of the reactions catalyzed by P450 enzyme families found that roughly 15% of those reactions were bioactivation reactions, meaning the enzyme converted a relatively harmless parent chemical into a toxic metabolite.⁴ In some of those cases the newly created metabolite reacts directly with DNA or proteins in the cell. In others, the metabolite feeds into downstream reactions whose final products are the real toxicants. This is why two chemicals with similar-looking structures can have wildly different safety profiles: the body’s own processing of one may generate a dangerous intermediate that the other never produces.

Once cells are exposed to a damaging chemical or metabolite, a common mechanism of injury is oxidative stress. Certain fungal toxins called mycotoxins illustrate this vividly. T-2 toxin, a mycotoxin found on contaminated grain, triggers a buildup of reactive oxygen species inside cells, which damages the membrane of the mitochondria and sets off a cascade of programmed cell death.⁵ That same oxidative-stress-to-mitochondrial-damage pathway has been confirmed in reproductive tissue as well, where T-2 toxin kills ovarian cells by the same route.⁶ In both cases, adding an antioxidant significantly reduced the damage, confirming that the oxidative burst was the trigger rather than a side effect.

Replacing Animal Tests With High-Throughput Screening and Organ Chips

Traditional toxicology tested one chemical at a time, usually by dosing laboratory animals over weeks or months. That approach is slow and expensive, and tens of thousands of industrial chemicals have never gone through it. The Tox21 program, a collaboration between several U.S. federal agencies, took a different path: automated high-throughput screening that runs thousands of chemicals against panels of cell-based assays simultaneously. Researchers have shown that activity profiles from these assays can serve as signatures of a compound’s mechanism and help prioritize which chemicals need deeper testing.⁷ In one screening campaign, the Tox21 collection of over 8,300 unique chemicals was tested for acetylcholinesterase inhibition, a key marker for nerve-agent-like toxicity, and about 2% showed substantial inhibitory activity, narrowing the pool to 111 compounds flagged for follow-up.⁸

Cell-based assays have their own limitations, though, because isolated cells in a dish do not behave exactly like cells embedded in a living organ. Organ-on-a-chip technology attempts to close that gap. A human liver organoid chip was recently shown to capture species-specific liver toxicity with greater clinical relevance than animal models.⁹ In a separate performance assessment, a liver chip correctly predicted toxicity for 12 out of 15 known toxic drugs, yielding about 80% sensitivity. That was nearly double the sensitivity of 3D hepatic spheroids, which identified only about 42% of the same toxic drugs.¹⁰

Computational approaches add another layer. Machine learning models trained on large toxicity databases can predict a chemical’s acute toxicity from its molecular structure alone. These quantitative structure-activity relationship models work well when the training data are clean but remain limited by the quality of the underlying datasets.¹¹ For narrower chemical families, the accuracy can be quite good. Models developed specifically for polycyclic aromatic hydrocarbon derivatives, trained on oral toxicity data from nearly 800 substances, achieved strong predictive performance using ensemble machine learning algorithms.¹²

The Emerging Contaminants Nobody Planned For

Some of the most active areas in toxicology research involve chemicals that were not on anyone’s watch list a generation ago. Per- and polyfluoroalkyl substances, commonly called PFAS or “forever chemicals,” are one prominent example. Cross-sectional analyses using U.S. national health survey data found a significant association between PFAS exposure and fatty liver disease, and in silico analyses pointed to a specific molecular pathway involving a fat-metabolism receptor as the mechanism.¹³ Studies in zebrafish larvae reinforced this, showing that PFHxS, one member of the PFAS family, caused measurable liver damage through the same signaling pathway, and blocking that pathway reduced the damage.¹⁴

Nano- and microplastics are another emerging concern. Animal studies have shown that nanoplastics, because of their tiny size, can penetrate deeply into organs, distributing to the liver, spleen, heart, lungs, kidneys, reproductive organs, and even crossing the blood-brain barrier.¹⁵ Chronic exposure experiments in mice found that ingested polystyrene nanoplastics caused structural damage to the intestine, including erosion of the finger-like villi that absorb nutrients, along with a surge in inflammatory signaling molecules.¹⁶

Engineered nanomaterials such as carbon nanotubes present a different kind of hazard. Long, rigid multi-walled carbon nanotubes trigger a reaction in immune cells called frustrated phagocytosis: the cell tries to engulf the fiber but cannot wrap around it, and it spews out inflammatory signals instead. Laboratory comparisons found that industrially produced long carbon nanotubes were routinely more damaging to macrophages than asbestos fibers at the same dose.¹⁷ Follow-up work established that rigidity, not just length, matters: only the stiffest nanotubes and asbestos fibers actually pierced immune cells and triggered the inflammation cascade.¹⁸

Pollutants That Climb the Food Chain

Not all toxicants stay at the concentrations they are released into the environment. Persistent organic pollutants like PCBs and DDT-related compounds accumulate in living tissue and concentrate as they move up the food chain, a process called trophic magnification. A study of a Mediterranean coastal food web found that while some flame retardants stayed concentrated in plankton and did not magnify much, PCBs preferentially built up in predatory fish, reaching median concentrations of several thousand nanograms per gram of fat in species like Atlantic bonito.¹⁹ In a subtropical river system in South Africa, DDT-related compounds showed unexpectedly high magnification factors, exceeding values reported for colder climates where these chemicals had traditionally been studied.²⁰ This finding matters because risk assessments often borrow magnification data from temperate ecosystems, and applying those numbers to tropical or subtropical regions could underestimate the dose reaching top predators, including people who eat the fish.

Rethinking How Safe Limits Are Set

For decades, regulators identified a safe exposure level for a chemical by finding the highest dose in an animal study that produced no observable adverse effect and then dividing by uncertainty factors. This “no-observed-adverse-effect level” approach is straightforward but has well-known weaknesses: it depends heavily on what doses the experimenters happened to test, and it ignores much of the data from the rest of the dose-response curve. The benchmark dose approach uses all the data points from a study to model the full dose-response relationship, and both the European Food Safety Authority and the UK Committee on Toxicity have endorsed it as a more data-driven and transparent alternative.²¹²² A practical evaluation across a range of chemical classes found the benchmark dose particularly useful when the traditional approach could identify only a lowest-observed-adverse-effect level rather than a true no-effect level, or when the tested doses were spread far apart.²³

Mixture toxicity adds another complication. People are never exposed to just one chemical at a time. When two chemicals act on the same molecular target, their effects tend to be additive. In a different scenario, one chemical may boost the concentration of a second chemical at its site of action, making the second one more harmful than it would be alone.²⁴ Whether the sheer number of low-dose exposures people face daily could collectively cause harm even when each chemical is individually at a safe level remains one of the field’s most debated questions.

Genetic Variation and Personal Susceptibility

Not everyone metabolizes toxicants the same way, and inherited differences in detoxification enzymes help explain why one worker in a chemical plant develops liver disease while the colleague at the next station does not. Glutathione S-transferases are a family of enzymes that neutralize reactive metabolites before they damage cells. Certain gene variants make these enzymes more or less active. Among petrochemical workers, those carrying a particular variant of the GSTP1 gene (the Val/Val form) were significantly less likely to develop toxic liver damage, suggesting the variant acts as a protective factor.²⁵

Cancer susceptibility follows similar logic. A meta-analysis of 12 case-control studies found that people with a complete deletion of the GSTM1 gene had about 40% higher odds of developing lung cancer compared to those who retained it.²⁶ When multiple unfavorable variants line up in the same person, risks multiply: individuals who lacked the GSTM1 gene and also carried the less active GSTP1 variant faced roughly seven times the odds of developing small cell lung cancer compared with people whose detoxification enzymes were fully functional.²⁷ These findings feed into what some researchers call “precision toxicology,” the idea that individual genetic profiles should eventually inform personalized exposure limits and workplace protections.

When Exposure Happens During Development

Timing matters as much as dose when it comes to the developing embryo. A recent study mapping the critical windows for structural birth defects in animal models found that most organs are most vulnerable to toxic insult far earlier in pregnancy than many people realize. Central nervous system malformations had the earliest window, craniofacial defects overlapped and extended slightly later, and limb malformations clustered just after that. Translated to human development, the critical window for most chemically-induced structural malformations falls approximately between the third and sixth week of pregnancy, a period when many people do not yet know they are pregnant.²⁸

Lead is one of the best-studied developmental neurotoxicants. Exposure during brain development alters a specific type of receptor crucial for learning and memory. Animal studies have shown that lead selectively changes which subunits of this receptor are produced in the hippocampus, the brain region most involved in forming new memories. Some subunits are overproduced, others underproduced, and the resulting receptor behaves abnormally.²⁹³⁰ The consequences can persist long after the lead exposure ends, because the abnormal receptor composition alters the signaling cascades that wire neurons together during critical developmental periods.³¹

Toxicity That Echoes Across Generations

One of the more unsettling findings in modern toxicology is that some chemical exposures leave marks not just on the person exposed but on their offspring’s reproductive cells. Cadmium, a heavy metal found in cigarette smoke and certain industrial emissions, provides a striking example. When pregnant mice were exposed to cadmium, their offspring’s egg cells showed selective changes in epigenetic markers: certain histone modifications were reduced and DNA methylation patterns at a specific gene were altered. These epigenetic shifts were not caused by direct cadmium exposure of the offspring themselves but were inherited through the germ line.³² This is still early-stage research, mostly in rodents, but it raises the possibility that a chemical exposure in one generation could influence fertility or development in the next.

Measuring What Is Actually Inside Us

To link chemical exposures to health effects, researchers need to know which chemicals are actually present in a person’s blood, and at what concentrations. The field of chemical exposomics aims to capture as broad a snapshot as possible. Recent advances in analytical chemistry have improved the picture significantly. A method combining lipid removal from blood plasma with large-volume injection gas chromatography and high-resolution mass spectrometry was shown to detect more target chemicals and identify more unknown molecules than prior approaches.³³ The choice of how to prepare a blood sample matters more than it might seem: different preparation methods extract different subsets of chemicals, so what researchers find depends in part on how they look.³⁴

Neonicotinoids and Honeybee Navigation

Toxicology research extends to the ecological consequences of chemicals released into the environment. Neonicotinoid insecticides, widely used in agriculture, illustrate how sublethal doses can cause serious harm without outright killing the organism. When honeybees were exposed to non-lethal doses of three different neonicotinoids, their ability to navigate home was significantly impaired: they made more wrong turns at landmarks and flew less directed routes. The researchers concluded that the pesticides either blocked retrieval of navigational memories or degraded the memories themselves.³⁵ A separate study showed that imidacloprid, one of the most common neonicotinoids, impaired bees’ ability to learn to avoid threats, reducing aversive learning by about 87% and memory retention by about 85% compared with untreated bees.³⁶ A bee that cannot navigate home or learn which flowers are dangerous is functionally compromised even if it is technically alive, which is precisely the kind of harm traditional acute-toxicity testing was not designed to detect.

Reading Drug Levels After Death

Forensic toxicology occupies a very different corner of the field, but it shares the same fundamental challenge of interpreting chemical concentrations in a biological system. When a person dies, drug levels measured from the body do not necessarily reflect what was circulating in the blood at the time of death. After death, drugs stored in tissue reservoirs leak back into the blood, concentrations shift between body compartments, and microbial activity can generate or destroy certain compounds. This phenomenon, called postmortem redistribution, complicates efforts to determine whether a drug caused or contributed to death.³⁷ Forensic toxicologists account for it by sampling from multiple sites, preferring blood drawn from the leg veins (which are farther from drug-rich organs) and comparing concentrations across sites. Even with these precautions, interpreting postmortem drug levels remains one of the trickier tasks in death investigation, and compilations of postmortem-to-antemortem drug ratios continue to be refined as data accumulate.³⁸