Glyphosate in food, and how measured exposure compares with the limit

1. 01

   Detection is near-universal and the concentrations are low

   In the first nationally representative US data, glyphosate was quantified in 2,310 urine samples with a weighted detection frequency of 81.2 percent and a median of 0.392 micrograms per litre, interquartile range 0.263 to 0.656, or 0.450 micrograms per gram of creatinine, leading the authors to conclude that over four-fifths of the US general population aged 6 and over experienced recent exposure (Ospina et al. 2022). Detection at that sensitivity is not the same as a dose that matters.

2. 02

   Exposure tracks recent eating

   Children, adolescents and adults who had fasted more than 8 hours had significantly lower model-adjusted geometric means than those fasting 8 hours or less, and the likelihood of being above the 95th percentile was 1.94 times higher, 95 percent confidence interval 1.06 to 3.54, in those fasting 8 hours or less (Ospina et al. 2022). That pattern is consistent with diet as a real route and with rapid clearance rather than accumulation in tissue.

3. 03

   The reference value and the arithmetic against it

   A biomonitoring study of children in an intensive agricultural area computed geometric mean estimated daily intakes of 4.38 micrograms per kilogram bodyweight per day for glyphosate and 2.26 for its metabolite, and calculated hazard quotients against the 0.5 milligrams per kilogram bodyweight per day acceptable daily intake established by the European Food Safety Authority as a health-based reference value; the quotients were below 1 and low health risk was expected (Berni et al. 2023). That is a margin of roughly a hundredfold.

4. 04

   The cohort was selected for high exposure

   That study covered 48 children in an intensive agricultural area, detecting glyphosate in 73 percent and its metabolite in 75 percent of samples at mean concentrations of 0.97 and 0.79 micrograms per litre, and found children living within 50 metres of spraying fields had 14.91 micrograms per litre more glyphosate than children in urban counties, 95 percent confidence interval 8.14 to 20.91 (Berni et al. 2023). The margin above holds in a population chosen to be worse than average.

5. 05

   The hazard classification: IARC Group 2A

   An International Agency for Research on Cancer monograph working group classified glyphosate as probably carcinogenic to humans, published as a summary in 2015 (Guyton et al. 2015), and the classification has been described in subsequent peer-reviewed work in exactly those terms (Berni et al. 2023). A hazard classification addresses whether a substance can cause cancer under some conditions, not whether dietary exposure levels do.

6. 06

   The regulatory risk assessments reached different conclusions

   The European Food Safety Authority's 2023 peer review of the risk assessment of glyphosate did not identify any critical areas of concern, while reporting data gaps including impurities, consumer dietary risk, aquatic plant risks and toxicity information for formulation components, and the European Chemicals Agency's 2022 hazard assessment concluded glyphosate did not meet the scientific criteria to be classified as carcinogenic, mutagenic or reprotoxic (EFSA 2023).

7. 07

   Why the two sides talk past each other

   Hazard identification asks whether a substance can cause harm under any conditions. Risk assessment asks whether it does at realistic exposure. The IARC classification is the first kind of statement and the EFSA conclusion is the second, so both can stand without contradiction. Reporting either one as the settled answer to the other question is the single most common error in coverage of this topic.

8. 08

   The observational signal that warrants watching

   Using NHANES 2013 to 2016, a study of non-persistent pesticide exposures and liver fibrosis found significant and persistent associations for glyphosate with both continuous and dichotomous outcomes, with a FIB-4 coefficient of 0.09, 95 percent confidence interval 0.06 to 0.12, among the highest effect estimates in the panel, and a posterior inclusion probability of 1.000 in the mixture analysis (Shu et al. 2025). Cross-sectional associations within a correlated mixture do not establish causation.

What people ask us next.

Q01

How much glyphosate do people actually take in?

Very little, and almost everyone takes in some. The first nationally representative US biomonitoring data found a weighted detection frequency of 81.2 percent across 2,310 urine samples with a median of 0.392 micrograms per litre (Ospina et al. 2022), and a study in a high-exposure agricultural setting calculated geometric mean estimated daily intakes of 4.38 micrograms per kilogram bodyweight per day against a European reference value of 0.5 milligrams per kilogram bodyweight per day (Berni et al. 2023).

Q02

Is glyphosate a carcinogen?

The classifications disagree, and the disagreement is about different questions. An International Agency for Research on Cancer working group classified it as probably carcinogenic to humans in 2015 (Guyton et al. 2015), which is a hazard statement. The European Food Safety Authority's 2023 risk assessment peer review did not identify any critical areas of concern, and the European Chemicals Agency's 2022 hazard assessment concluded it did not meet the criteria for classification as carcinogenic, mutagenic or reprotoxic (EFSA 2023).

Q03

Does washing produce reduce glyphosate exposure?

It is the wrong lever for this particular residue. Biomonitoring showed concentrations tracking recent eating, with those fasting 8 hours or less 1.94 times more likely to be above the 95th percentile (Ospina et al. 2022), and the residue burden is concentrated in the commodity crops where the herbicide is applied rather than on the surface of fresh produce. Diet composition is the actionable variable; surface washing is not.

Q04

Are children at higher risk?

Their measured concentrations are higher and the modelled risk was still below the reference. Urinary concentrations decreased from ages 6 to 11 through ages 20 to 59 (Ospina et al. 2022), and in an agricultural cohort children under 5 had higher concentrations than those aged 6 to 12, with children living within 50 metres of spraying fields showing 14.91 micrograms per litre more glyphosate than urban children. Hazard quotients against the European reference value were still below 1 (Berni et al. 2023).

Q05

What about the studies linking glyphosate to liver problems?

They are real signals with real limitations. In NHANES 2013 to 2016, urinary glyphosate showed significant and persistent associations with liver fibrosis indices, with a FIB-4 coefficient of 0.09, 95 percent confidence interval 0.06 to 0.12, and a posterior inclusion probability of 1.000 in mixture analysis (Shu et al. 2025). These are cross-sectional associations within a correlated mixture of pesticide exposures, and the authors framed the work as providing insight into hepatotoxic potential rather than demonstrating it.

Q06

Does buying organic eliminate the exposure?

It reduces one route rather than eliminating exposure, and the size of the reduction is not what the detection headlines imply. Biomonitoring found 81.2 percent detection across the general population at a median of 0.392 micrograms per litre (Ospina et al. 2022), which reflects a widely distributed low-level exposure. The defensible framing is the same one applied to the arsenic and heavy metal analyses on this site: compare measured intake with the reference value before restructuring a diet.