Table 3.

GEP Criteria for Pesticide Exposure Assessment Using Questionnaires, Interviews, and Expert Record Review.

Category	Criteria for Causal Inference and Quantitative Risk Assessment	Comments
Study Protocol	An a priori study protocol must include standard and detailed specifications for developing questionnaires and conducting interviews; training personnel; extracting, coding, and processing data; keeping records and storing data; and quality assurance (QA) and quality control (QC) procedures that minimize the potential for bias and human error.
Exposure Window	1.The exposure metric should be assessed in the time period during which the exposure could have effects relevant to the outcome of interest. The selection of the exposure window should consider induction and latency and should be based on biological and clinical pathways.	▪Not applicable to case-control studies. For a developmental endpoint (e.g., congenital heart defects), the exposure metric should capture the exposure that occurred during the period of fetal development associated with that effect (e.g., fetal cardiac development). For cancer, the exposure must precede the diagnosis by a sufficient period of time (typically, at least 10 years).
Study Population	The study population should be representative of the target population with regard to the exposure distribution (i.e., it should capture the entire exposure range).
Design of Self-administered and Interview Questionnaires	The questions should be clear and simple so as to avoid ambiguity and enhance recall. Ideally, questionnaires should be designed with multiple ways of obtaining a specific piece of information, to allow for an internal cross-check of the validity of the provided answers.	▪For example, when assessing occupational exposures, workers should be asked about factors that are easily recalled, such as tasks, raw materials, equipment, and processes [27]. Avoid leading questions, such as, “Did you ever feel ill after using the pesticide product?”
	3.The questions should be sufficiently comprehensive and detailed to address specific research questions. For quantitative risk assessment, the questions should capture ordinal or semi-quantitative exposure information.	▪For example, the questions should capture multiple exposure metrics (e.g., exposure frequency, exposure time, exposure intensity, time since first exposure, time since last exposure) rather than a dichotomous exposure status (i.e., yes/no, ever/never).
Blinding	Blinding of study participants and researchers should be implemented, if possible, to reduce the potential for information bias. Particularly for retrospective studies, exposure must be assessed independent of outcome.	▪The study participants should be blinded to the specific research question (i.e., the exposure of interest). ▪For retrospective studies, the person(s) conducting the interviews or reviewing records/questionnaires should be blinded to the outcome status of the study participants, if possible.
Validation of Exposure Assessment Methods	Self-reported exposures from questionnaires and interviews should be validated by objective records or measurements. For quantitative risk assessment, self-reported exposure information needs to validated against biomonitoring or environmental sampling data, ideally in a subset of the study population. For record review, the expert needs to be sufficiently qualified and rely on environmental measurements to determine the relative rankings of the study participants with regard to the exposure. For quantitative risk assessment of occupational exposures, a valid job-exposure matrix, based on environmental sampling data, should be constructed. The environmental sampling should capture the temporal changes and variations in the exposures by task.	▪For example, self-reported use of pesticides can be validated with purchasing/inventory records, biomonitoring data, or personal/environmental exposure monitoring. ▪Self-reported use of pesticides can be combined with Geographic Information Systems (GIS) methods to develop surrogate exposure estimates, but due to inability to assess fine spatial variations, GIS can introduce uncertainty or error into exposure estimates, and approaches should be validated on multiple datasets [28,29,30,31]. ▪For assessing occupational exposures by record review and/or job-exposure matrix, the expert should have relevant experience and credentials (i.e., certified industrial hygienist).
Reporting Requirements	The study protocol should be made publicly available. Details regarding personnel training and credentials, questionnaire development, computer software employed, data extraction and processing, and methods used to estimate exposure should be reported. Any deviations from the study protocol and justifications for such deviations should be reported.

GEP = Good Epidemiology Practice; Underlining indicates criteria that are needed for quantitative risk assessment.