Residential exposure to agricultural pesticides and acute and recurring health outcomes: protocol for a retrospective dynamic cohort study using primary care data in the Netherlands.

INTRODUCTION: Agricultural pesticide use leads to exposure in residents and raises concerns for possible health effects. Although international studies report associations between pesticide exposure and adverse effects, most studies focus on occupational settings and chronic effects. This protocol describes an approach to assess the relation between modelled pesticide exposure near homes of residents living in the vicinity of agricultural plots and the incidence of acute and recurring health symptoms and conditions. The study will also assess exacerbations in patients with asthma and chronic obstructive pulmonary disease, as well as the healthcare utilisation, using Dutch primary care registrations. The findings of this study will inform policymakers and address societal concerns of pesticide exposure in residents.

METHOD AND ANALYSIS: A retrospective dynamic cohort study design will be used spanning from 2011 to 2025. Data will be extracted from pseudonymised electronic health records of general practices in the Netherlands, using Nivel Primary Care Database (Nivel-PCD). This includes approximately 900 000 individuals residing in moderately urban to non-urban areas. Exposure data for the analytical period will be provided by the Institute for Risk Assessment Sciences (IRAS) of Utrecht University. Two formats of individual exposure will be merged with the health outcome data: amount (kg) of the active substance (AS) within defined buffers (50, 100, 250 and 500 m) and modelled concentrations at the individual homes. ASs have been pre-selected based on epidemiological and toxicological evidence of health effects, societal and scientific concerns and reported usage patterns in the Netherlands. For the primary analyses, potential associations between pesticide exposure and incidence or prevalence of health endpoints will be investigated by means of longitudinal multilevel Poisson regression models. For the analysis examining the relationship between pesticide exposure and exacerbations or total healthcare utilisation, zero-inflated Poisson and negative binomial regression models will be considered. All models will be adjusted for a number of relevant covariates at individual and neighbourhood level. Power calculations indicated overall feasibility for the selected outcomes, but detection of smaller effects for low-prevalent outcomes may require larger population sizes.

ETHICS AND DISSEMINATION: The study does not fall within the scope of the Medical Research Involving Human Subjects Act and therefore does not require ethical approval. Approval of applicable governance bodies of Nivel-PCD is requested. No informed consent is needed; individual opt-out of individuals is possible via their general practitioner.