Factors influencing occupational exposure to pyrethroids and glyphosate: An analysis of urinary biomarkers in Malaysia, Uganda and the United Kingdom

Background: Long-term exposure to pesticides is often assessed using semi-quantitative models. To improve these models, a better understanding of how occupational factors determine exposure (e.g., as estimated by biomonitoring) would be valuable. Methods: Urine samples were collected from pesticide applicators in Malaysia, Uganda, and the UK during mixing/application days (and also during non-application days in Uganda). Samples were collected pre- and post-activity on the same day and analysed for biomarkers of active ingredients (AIs), including synthetic pyrethroids (via the metabolite 3-phenoxybenzoic acid [3-PBA]) and glyphosate, as well as creatinine. We performed multilevel Tobit regression models for each study to assess the relationship between exposure modifying factors (e.g., mixing/application of AI, duration of activity, personal protective equipment [PPE]) and urinary biomarkers of exposure. Results: From the Malaysia, Uganda, and UK studies, 81, 84, and 106 study participants provided 162, 384 and 212 urine samples, respectively. Pyrethroid use on the sampling day was most common in Malaysia (n = 38; 47%), and glyphosate use was most prevalent in the UK (n = 93; 88%). Median pre- and post-activity 3-PBA concentrations were similar, with higher median concentrations post-compared to pre-activity for glyphosate samples in the UK (1.7 to 0.5 μg/L) and Uganda (7.6 to 0.8 μg/L) (glyphosate was not used in the Malaysia study). There was evidence from individual studies that higher urinary biomarker concentrations were associated with mixing/application of the AI on the day of urine sampling, longer duration of mixing/application, lower PPE protection, and less education/literacy, but no factor was consistently associated with exposure across biomarkers in the three studies. Conclusions: Our results suggest a need for AI-specific interpretation of exposure modifying factors as the relevance of exposure routes, levels of detection, and farming systems/practices may be very context and AI-specific.