Table 1.

Xxx

Author	Year	Country	Study Design	Study Population	Comparison Subjects	Exposure Assessment	Exposure Levels	Outcome	Outcome Assessment	Confounders Considered	Statistical Approach	Estimate of Association (95% CI or CrI)	Funding Source
Neurodegenerative outcomes
Dhillon et al.	2008	U.S	Hospital-based case–control	100 Parkinson disease patients aged ≥ 50 years recruited from ~ 800 Parkinson disease patients seen at a single neurology practice in East Texas region, excluding patients with multiple sclerosis, schizophrenia, and "Parkinson's plus" diseases; years NR, participation NR (~ 12.5%)	84 controls without Parkinson disease selected from the same neurology practice; participation NR	Self-reported ever personal use/mixing or application of Roundup, Jury, or other glyphosate product and 54 other specific pesticides and pesticide products, as well as various occupations and other occupational exposures, assessed by telephone interviewer, not blinded to case–control status	Ever personally used/mixed or applied Roundup, Jury, or other glyphosate product (54 cases, 43 controls)	Parkinson disease	Parkinson disease diagnosed by a neurologist specializing in movement disorders, using standard clinical/laboratory diagnostic criteria (NR)	None	Chi-square test	Odds ratio = 1.1 (0.6, 2.0)	U.S. Centers for Disease Control and Prevention/National Institute of Occupational Safety and Health
Kamel et al.	2007	U.S	Prospective nested case–control	Agricultural Health Study cohort of 84,738 licensed pesticide applicators (n = 52,393; 84% participation) and their spouses ( n = 32,345; 74% participation), including 22,915 applicators (44% of cohort) who completed a supplemental questionnaire, recruited in 1993–1997 in Iowa and North Carolina, with 5-year follow-up of 57,251 cohort members (68% participation) 83 prevalent Parkinson disease cases (60 applicators, 23 spouses) at enrollment 78 incident Parkinson disease cases (56 applicators, 22 spouses) at follow-up	79,557 controls without Parkinson disease at enrollment, 55,931 controls at follow-up identified within same Agricultural Health Study cohort	Self-reported ever personal mixing or application of glyphosate and 49 other specific pesticides, assessed by written questionnaire at enrollment	Ever use of glyphosate (45 prevalent cases, 46,687 controls; 49 incident cases, 32,686 controls)	Parkinson disease	Prevalent Parkinson disease defined based on self-reported doctor's diagnosis per enrollment questionnaire, applicator questionnaire, or spouse questionnaire, excluding 5073 cohort members with missing data and 25 with conflicting data on enrollment and/or follow-up questionnaires Incident Parkinson disease defined based on self-reported doctor's diagnosis per follow-up telephone interview, excluding prevalent cases and 28,621 cohort members with missing data	Adjusted: age, state, applicator or spouse status, insecticides, herbicides, fungicides, fumigants, organophosphates, organochlorines, carbamates, phenoxyacetates, triazines/triazones	Two-stage hierarchical multivariable logistic regression, with covariates and indicators for specific pesticides with ≥ 4 cases in stage 1, and variables for functional pesticide groups and chemical groups with ≥ 3 pesticides in stage 2	Odds ratio, prevalent disease = 1.0 (0.6, 1.7) Odds ratio, incident disease = 1.1 (0.6, 2.0)	U.S. National Institute of Environmental Health Sciences, U.S. National Cancer Institute
Kamel et al.	2012	U.S.	Prospective nested case–control	Agricultural Health Study cohort of 84,738 licensed pesticide applicators ( n = 52,393; 84% participation) and their spouses ( n = 32,345; 74% participation), including ~ 42% of applicators who completed a supplemental questionnaire, recruited in 1993–1997 in Iowa and North Carolina, linked to mortality data through February 7, 2010 41 ALS deaths	84,698 controls without ALS identified within same Agricultural Health Study cohort	Self-reported ever personal mixing or application of glyphosate and 49 other specific pesticides, assessed by written questionnaire at enrollment	Ever use of glyphosate (25 cases, 48,847 controls)	ALS death	Death from ALS as underlying or contributing cause of death identified from death certificates, including 7 deaths (of 41) with medical records available for review by study neurologist using published criteria; 5 diagnosed with ALS (1 definite, 2 probable, 2 possible), 1 with progressive bulbar palsy, 1 indeterminate	Adjusted: age, sex Considered: smoking, education, state, head injury	Multivariable logistic regression	Odds ratio = 1.2 (0.6, 2.5)	U.S. National Institute of Environmental Health Sciences, U.S. National Cancer Institute
Montgomery et al.	2017	U.S	Prospective nested case–control	Agricultural Health Study cohort of 84,739 licensed pesticide applicators ( n = 52,394; 84% participation) and their spouses ( n = 32,345; 74% participation), including a supplemental questionnaire completed by 53% of controls and 72% of cases, recruited in 1993–1997 in Iowa and North Carolina, with follow-up interviews in 1993–2003 and 2005–2010 (68% participation), restricted to 41,863 cohort members aged < 50 y on September 1, 2007, without retinal or macular degeneration at enrollment 161 incident AMD cases	39,108 controls without AMD or possible AMD identified within same Agricultural Health Study cohort	Self-reported ever personal mixing or application of glyphosate and 49 other specific pesticides, including duration in years and frequency in days per year, assessed by written questionnaire at enrollment	Ever use of glyphosate (103 cases, 23,493 controls), including cumulative days of use among applicators: 0: 15 cases, 5104 controls  > 0–10: 18 cases, 4929 controls  > 10–100: 33 cases, 7403 controls > 100: 28 cases, 4783 controls	AMD	Validated self-reported doctor's diagnosis of AMD at either follow-up telephone interview, with self-report affirmed by screening call and confirmed by treating physician with supporting pathology or by study ophthalmologist based on retinal photographs, excluding self-reported prevalent doctor's diagnosis of retinal or macular degeneration per enrollment questionnaire ( n = 324); further classified as early stage, late-stage, or unknown-stage	Adjusted: age, sex, smoking at enrollment; in some models: strongly correlated pesticide Considered: body mass index, education, state, sun exposure hours per day	Multivariable logistic regression	Odds ratio, ever use = 1.4 (0.99, 2.0) Odds ratio, ever use, me n = 1.8 (1.0, 3.1) Odds ratio, ever use, wome n = 1.2 (0.7, 2.0) Odds ratio, ever use, early AMD = 1.5 (0.8, 2.7) Odds ratio, ever use, late AMD = 1.3 (0.8, 2.3) Odds ratio, ever use, late vs. early AMD = 0.9 (0.4, 2.0) Odds ratio, ever use, adjusted for malathio n = 1.1 (not significant) Odds ratio, ever use, adjusted for carbaryl = 1.3 (not significant) Odds ratio, ever use, adjusted for 2,4-D = 1.3 (not significant) Odds ratio, > 0–10 days vs. 0 = 1.3 (0.6, 2.5) Odds ratio, > 10–100 days vs. 0 = 1.7 (0.9, 3.1) Odds ratio, > 100 days vs. 0 = 2.6 (1.4, 4.9) p-trend = 0.002	U.S. National Institute of Environmental Health Sciences, U.S. National Cancer Institute
Shrestha et al.	2020	U.S	Prospective cohort	Agricultural Health Study cohort of 38,274 male licensed pesticide applicators (84% initial participation; 44% participation in supplemental questionnaire) and 27,836 spouses (75% initial participation) recruited in 1993–1997 in Iowa and North Carolina, with completion of at least one follow-up interview in 1993–2003, 2005–2010, or 2013–2016 or a Parkinson disease validation screening questionnaire 491 incident Parkinson disease cases (373 applicators, 118 spouses), 65,619 non-cases 106 prevalent Parkinson disease cases	Cohort members without glyphosate exposure	Self-reported ever personal mixing or application of glyphosate and 49 other specific pesticides, including duration in years and frequency in days per year, assessed by written questionnaire at enrollment; in phase 2 (2–10 y after enrollment, mean 5 y), self-reported days of use of glyphosate in year prior to interview or in most recent year assessed by telephone interview; intensity-weighted lifetime days of use among pesticide applicators calculated as years of use × days per year × exposure intensity, with weights based on mixing practices, application methods, repair status, and personal protective equipment use	Ever use of glyphosate (291 cases, 35,406 non-cases) Through enrollment: Never use of glyphosate, applicators only (86 cases, 8,307 non-cases)  > 0– ≤ 677 intensity-weighted lifetime days of glyphosate use (106 cases, 8,996 non-cases)  > 677– ≤ 2,604 intensity-weighted lifetime days of glyphosate use (91 cases, 9,313 non-cases)  > 2,604 intensity-weighted lifetime days of glyphosate use (73 cases, 9,015 non-cases) Through phase 2: Never use of glyphosate, applicators only (62 cases, 5247 non-cases)  > 0– ≤ 970 intensity-weighted lifetime days of glyphosate use (132 cases, 9965 non-cases) > 970–≤ 3352 intensity-weighted lifetime days of glyphosate use (84 cases, 10,318 non-cases) > 3352 intensity-weighted lifetime days of glyphosate use (77 cases, 10,018 non-cases)	Parkinson disease	Incident Parkinson disease defined based on self-reported doctor's diagnosis per follow-up interview or linkage to National Death Index and state death registries; self-reported cases through phase 2 previously confirmed by movement disorder specialists via structured clinical examinations and medical records (confirmed in 84% of self-reported cases); all incident and prevalent self-reported cases through phase 4 re-contacted for validation through self- or proxy completion of detailed screening questionnaire on Parkinson disease diagnosis, symptoms, characteristics, and treatment, with adjudication by a movement disorder specialist, and review of medical records if consented ( n = 65; 91% confirmed, 9% questionable); excluded prevalent cases and those without information on age at diagnosis, without supporting symptoms or medications, or with inconsistent survey responses	Adjusted: applicator status, sex, state, smoking, alcohol, education, top four pesticides with Spearman correlation coefficient of ≥ 0.40	Multivariable Cox proportional hazards regression with attained age as time scale, left-truncated at enrollment, and stratification of baseline hazard by median age (63 y) when proportional hazards assumption failed; sensitivity analysis using inverse probability of censoring weights to adjust for loss to follow-up	Hazard ratio, ever use, applicators and spouses = 1.10 (0.87, 1.39) Hazard ratio, ever use, applicators = 1.02 (0.79, 1.30) Hazard ratio, ever use, spouses = 1.44 (0.92, 2.25) Hazard ratio, > 0– ≤ 677 intensity-weighted lifetime days through enrollment, applicators = 1.17 (0.88, 1.55) Hazard ratio, > 677– ≤ 2,604 days = 0.99 (0.73, 1.33) Hazard ratio, > 2,604 days = 0.85 (0.62, 1.17) p-trend = 0.09 Hazard ratio, > 0– ≤ 970 intensity-weighted lifetime days through phase 2, applicators = 1.21 (0.88, 1.65) Hazard ratio, > 970– ≤ 3,352 days = 0.92 (0.64, 1.34) Hazard ratio, > 3352 days = 0.88 (0.62, 1.25) p-trend = 0.10 No significant associations after stratification by ≤ 10 y vs. > 10 y follow-up, inclusion of prevalent cases, use of inverse probability weights, or classification of ever use through phase 2	U.S. National Institute of Environmental Health Sciences, U.S. National Cancer Institute
Neurobehavioral outcomes
Beard et al.	2013	U.S	Prospective cohort	Agricultural Health Study cohort of 16,893 wives of licensed pesticide applicators who completed a take-home questionnaire (75% participation), recruited in 1993–1997 in Iowa and North Carolina, with 5-year follow-up (62% participation), excluding depression at enrollment 1,054 incident depression cases and 15,839 non-cases at follow-up	Cohort members without glyphosate exposure	Self-reported ever personal mixing or application of glyphosate (direct) or ever spousal mixing or application of glyphosate (indirect) if never personally used pesticides, including 49 other specific pesticides and 11 pesticide classes, assessed by written questionnaire at enrollment	Ever personal use of glyphosate (359 cases, 6,017 total) Husbands' ever use of glyphosate (if never personally used pesticides; 330 cases, 4935 total)	Depression	Self-reported doctor's diagnosis of depression per follow-up telephone interview, excluding self-reported prevalent doctor's diagnosis of depression per enrollment questionnaire ( n = 2252) and those with missing data or self-reported age at diagnosis before enrollment on follow-up interview	Adjusted: age, diabetes, education, state Considered: race/ethnicity, number of children in family, farm size, alcohol in past year, smoking, number of doctor visits in past year, heart disease, number of years lived or worked on a farm, job held off of farm, solvent (other than gasoline) exposure at longest held non-farm job, ever use of any pesticides, husbands' age, husbands' depression status, husbands' use of individual pesticides, most strongly correlated pesticide	Multivariable log-binomial regression, with various regression models to calculate stabilized weights accounting for confounding and selection bias from loss to follow-up, multiplied to obtain overall stabilized weights used for inverse probability weighting; sensitivity analysis using Cox proportional hazards regression with estimated date of depression diagnosis	Risk ratio, personal use = 0.95 (0.83, 1.08) Risk ratio, spousal use = 1.04 (0.84, 1.30)	U.S. National Institute of Environmental Health Sciences, U.S. National Cancer Institute
Beard et al.	2014	U.S	Prospective nested case–control	Agricultural Health Study cohort of 21,208 male licensed pesticide applicators (40% of cohort with 84% initial participation), including a supplemental questionnaire completed by 56% of analytic cohort, recruited in 1993–1997 in Iowa and North Carolina, with 5-year follow-up (68% participation) 1702 depression cases, including 474 at enrollment but not follow-up, 540 at enrollment and follow-up, and 688 at follow-up but not enrollment	19,506 controls without depression at enrollment and follow-up identified within same Agricultural Health Study cohort	Self-reported ever personal mixing or application of glyphosate and 49 other specific pesticides and 10 pesticide classes	Ever personal use of glyphosate (376 pre-enrollment only cases, 426 pre-enrollment and follow-up cases, 540 follow-up only cases, 15,053 controls)	Depression	Self-reported doctor's diagnosis of depression per enrollment questionnaire (including self-reported doctor's diagnosis of depression requiring medication or shock therapy per supplemental questionnaire) and/or follow-up telephone interview, excluding 1894 cohort members with missing depression data	Adjusted: age, diabetes, education, state Considered: marital status, number of children in family, alcohol in past year, smoking, farm size, use of chemical-resistant gloves when handling pesticides, number of doctor visits in past year, number of years lived or worked on a farm, job held off of farm, solvent (other than gasoline) exposure in longest-held non-farm job, most strongly correlated pesticide	Multivariable polytomous logistic regression, with various regression models to calculate stabilized weights accounting for confounding, missing covariate data, missing supplemental questionnaire (if applicable), and selection bias from loss to follow-up, multiplied to obtain overall stabilized weights used for inverse probability weighting	Odds ratio, pre-enrollment depression only = 1.2 (0.9, 1.6) Odds ratio, pre-enrollment and follow-up depressio n = 1.1 (0.9, 1.4) Odds ratio, follow-up depression only = 1.1 (0.9, 1.3) p = 0.80 for difference among odds ratios	U.S. National Institute of Environmental Health Sciences, U.S. National Cancer Institute, U.S. National Institute for Occupational Safety and Health
Fuhrimann et al.	2021	Uganda	Cross-sectional	288 smallholder farmers aged ≥ 18 years, including approximately equal numbers of conventional farmers sampled from lists provided by local leaders, and organic farmers sampled from a list provided from a local non-governmental organization, using snowball recruiting (participation NR), Wakiso District, Uganda, 2017	Subjects without glyphosate exposure	Self-reported use of glyphosate and 13 other specific pesticides in 12 months before study; exposure intensity score calculated as (mixing + application) × overall average personal protective equipment use × time interval between pesticide application and change of clothes × time interval between application and shower; also multiplied by yearly number of application days	Glyphosate application (208 subjects) Glyphosate mixing (191 subjects) Glyphosate exposure intensity score (media n = 6.1, interquartile range = 3.0) Glyphosate yearly application days (media n = 9, interquartile range = 26) Glyphosate yearly exposure-intensity-score-weighted days (media n = 51.5, interquartile range = 179.5)	Language, memory, attention, executive function, and motor function	Neurobehavioral tests administered by trained psychometrician, including Semantic Verbal Fluency, Phonetic Verbal Fluency (language); Color Trail Part 2 (inhibition/flexibility); Digit Span backward (working memory); Trail Making Test A (processing speed); Digit Symbol Substitution Test, Digit Vigilance (sustained attention); Benton Visual Retention Test (recognition memory); Digit Span forward (short-term memory); Purdue pegboard (perceptual motor, fine motor ability, coordination); Finger Tapping test (hand motor speed)	Adjusted: pesticide applicator status, age, education, psychometrician, language of assessment, sex, literacy, alcohol use, history of head injury, HIV status	Bayesian model averaging to compute inclusion probability for each predictor by summing posterior model probabilities over models including that predictor, with Jeffreys‐Zellner‐Siow prior for regression coefficients and beta‐binomial prior for model space, accounting for multiple testing across different models	Benton Visual Retention: marginal inclusion probability = 0.665, slope per interquartile increase in exposure intensity score = -0.103 (-0.236, 0) Finger Tapping, dominant hand: marginal inclusion probability = 0.483, slope = -0.217 (-0.712, 0) Trail Making A, log10: marginal inclusion probability = 0.176, slope = 0.002 (0, 0.013) Finger Tapping, non-dominant hand: marginal inclusion probability = 0.235, slope = -0.087 (0.531, 1e-04) Digit Symbol: marginal inclusion probability = 0.176, slope = -0.068 (-0.522, 0) Semantic Verbal Fluency: marginal inclusion probability = 0.078, slope = 0.007 (0, 0.09) Other neurobehavioral measures: marginal inclusion probability of empty model ≥ 0.5	Swiss Network for International Studies, Swiss National Science Foundation
Neurodevelopmental outcomes
Juntarawijit et al.	2020	Thailand	Hospital-based case–control	442 children aged < 5 years with suspected developmental delay identified from National Child Developmental Screening Program in 15 of 21 randomly selected hospitals in one rural area (Bang Rakam district) and 10 of 30 randomly selected hospitals in one urban area (Muang district), Phitsanulok Province, Thailand, 2019; 87% participation	413 controls with normal development identified from National Child Developmental Screening Program at same hospital, matched on age, sex, area of residence; 81% participation	Maternal self-reported ever prenatal or postnatal use of glyphosate and 13 other pesticides, reported in questionnaire administered by trained village health volunteers	Ever maternal use of glyphosate (33 cases, 34 controls) Ever prenatal maternal use of glyphosate (29 cases, 32 controls) Ever postnatal maternal use of glyphosate (16 cases, 11 controls)	Suspected developmental delay	Suspected developmental delay identified based on failure of one or more skills (gross motor, fine motor, receptive language, expressive language, and personal and social) assessed at ages 9, 18, 30, and 42 months using Developmental and Surveillance Promotion Manual, modified from Denver Development Screening Test II, conducted by trained nurse or health personnel as part of National Child Developmental Screening Program	Adjusted: maternal age at pregnancy, education, occupation, income, chronic disease, alcohol consumption, gestational age, birth order, delivery method, birth weight, and breastfeeding	Multivariable logistic regression	Odds ratio, glyphosate ever, adjusted = 0.93 (0.46, 1.90) Odds ratio, glyphosate ever, unadjusted = 0.90 (0.54, 1.47) Odds ratio, glyphosate prenatal, adjusted = 0.92 (0.45, 1.91) Odds ratio, glyphosate prenatal, unadjusted = 0.83 (0.49, 1.40) Odds ratio, glyphosate postnatal, adjusted = 1.32 (0.49, 3.55) Odds ratio, glyphosate postnatal, unadjusted = 1.37 (0.63, 2.98)	Faculty of Nursing, Naresuan University, Thailand
Other neurological outcomes
Fuhrimann et al.	2022	Uganda	Cross-sectional	253 smallholder farmers aged ≥ 18 years, including approximately equal numbers of conventional farmers chosen using random clustered convenience sampling from lists provided by local leaders, and organic farmers sampled from a list provided from a local non-governmental organization in Wakiso District, Uganda, using snowball recruiting in 2017 (participation NR), with follow-up interview of initial study subjects in 2019 (84% follow-up participation)	Subjects without glyphosate exposure	Self-reported use of glyphosate and 29 other specific pesticides on crops, livestock, or household in past year or 7 days prior to interview	Glyphosate use in past week (31 subjects) Glyphosate use in past year but more than 1 week ago (120 subjects)	Sleep problems	Self-reported sleep problems during past 1 week assessed by modified 12-item Medical Outcomes Study Sleep Scale questionnaire, used to derive measures of overall sleep problems (6 items or 9 items), sleep disturbance (4 items), sleep inadequacy (2 items), daytime somnolence (3 items), snoring (1 item), awakening short of breath or with a headache (1 item), and non-optimal sleep quantity (1 item); proportionally transformed to 100-point scale and dichotomized at 30 points. Modified from standard sleep scale by changing usual 1-month measurement period to 1 week, and changing 6-point Likert scale from "not at all" to "all the time" to 8-point Likert scale from 0 to 7 days	Adjusted: age, sex, current alcohol consumption, body mass index, sleep disruption during past week (yes or no) by mosquitoes, bedbugs, noise, infectious disease, wearing actimeter, or any other reason	Multivariable logistic regression	Odds ratio for 6-item sleep problem index, glyphosate use in past year but more than 1 week ago = 1.29 (0.64, 2.59) Odds ratio for 6-item sleep problem index, glyphosate use in past week = 3.75 (1.24, 11.8) 6-item sleep problem index and frequency of glyphosate use in past week (0, 1–2, or > 2 days): odds ratio NR, p < 0.05 Results NR for 9-item sleep problem index or any of 6 sleep dimensions measured; not significantly associated with frequency of glyphosate use in past week (0, 1–2, or > 2 days)	Swiss National Science Foundation; Swiss Network for International Studies; CropLife Europe
Shrestha et al.	2018	U.S	Prospective nested case–control	Agricultural Health Study cohort of 20,591 male licensed farmers, recruited in 1993–1997 in Iowa and North Carolina (84% participation at enrollment), with follow-up interview in 2013–2015 (46% participation) 1623 dream-enacting behavior cases	16,441 controls without dream-enacting behaviors identified within same Agricultural Health Study cohort	Self-reported ever personal mixing or application of glyphosate and 49 other specific pesticides	Ever personal use of glyphosate (1143 cases)	Dream-enacting behaviors	Self-reported dream-enacting behaviors based on question "Have you ever been told, or suspected yourself, that you seem to 'act out dream' while sleeping? For example, punching or flailing arms in the air, shouting, or screaming while asleep." If yes, further classified by frequency of symptoms (< 3 times in life, < 1/month, 1–3 month, 1/week, or > 1/week), with sensitivity analyses excluding 179 subjects diagnosed with Parkinson disease or restricting to cases with ≥ 3 lifetime episodes	Adjusted: age, smoking, alcohol, marital status, education, state, head injury; in some models: other pesticides and functional/chemical classes of pesticides with statistically significant associations before mutual adjustment	Multivariable logistic regression, with logistic regression to calculate stabilized weights accounting separately for loss of participants and for missing covariates, multiplied to obtain overall stabilized weights used for inverse probability weighting	Odds ratio, unadjusted for other pesticides = 1.3 (1.1, 1.5) Odds ratio, unadjusted, excluding Parkinson disease = 1.3 (1.1, 1.6) Odds ratio, unadjusted, ≥ 3 lifetime episodes = 1.3 (1.1., 1.5) Odds ratio, adjusted for other pesticides = 1.2 (1.0, 1.4) Odds ratio, adjusted, excluding Parkinson disease = 1.2 (1.0, 1.5) Odds ratio, adjusted, ≥ 3 lifetime episodes = 1.2 (1.0, 1.4)	U.S. National Institute of Environmental Health Sciences, U.S. National Cancer Institute, Michigan State University
Shrestha et al.	2021	U.S	Prospective nested case–control	Agricultural Health Study cohort of 20,409 licensed pesticide applicators recruited in 1993–1997 in Iowa and North Carolina (84% participation at enrollment, 44% with supplemental questionnaire), with completed follow-up interview in 2013–2016 (participatio n ~ 40%) 2,069 cases of olfactory impairment	18,340 controls without olfactory impairment identified within same Agricultural Health Study cohort	Self-reported ever personal mixing or application of glyphosate and 49 other specific pesticides, including duration in years and frequency in days per year, assessed by written questionnaire at enrollment; in phase 2, self-reported days of use of glyphosate in year prior to interview or in most recent year assessed by telephone interview; intensity-weighted lifetime days of use among pesticide applicators at enrollment or through phase 2 calculated as years of use × days per year × exposure intensity, with weights based on mixing practices, application methods, repair status, and personal protective equipment use	Ever personal use of glyphosate (1,678 cases with no time restriction, 991 cases ≤ 10 y before phase 4 [third follow-up], 14,086 controls) Never use of glyphosate (381 unrestricted cases, 226 cases ≤ 10 y before phase 4, 4,163 controls)  > 0–672 intensity-weighted lifetime days of glyphosate use (573 unrestricted cases, 340 cases ≤ 10 y before phase 4, 4590 controls) > 672–2610 intensity-weighted lifetime days of glyphosate use (523 unrestricted cases, 302 cases ≤ 10 y before phase 4, 4748 controls) > 2610 intensity-weighted lifetime days of glyphosate use (567 unrestricted cases, 342 cases ≤ 10 y before phase 4, 4572 controls)	Olfactory impairment	Self-reported olfactory impairment based on question "Do you suffer from a loss of sense of smell or significantly decreased sense of smell?" If yes, further classified by timing of when loss of sense of smell began (≤ 1, 1–5, 5–10, or > 10 y before phase 4 interview [third follow-up])	Adjusted: age, sex, state, education, smoking, other farming tasks, correlated pesticides with Spearman ρ ≥ 0.40	Multivariable logistic regression; sensitivity analyses using inverse probability of censoring weights to adjust for loss to follow-up	Odds ratio, ever use = 1.33 (1.18, 1.50) Odds ratio, ever use, onset ≤ 10 y prior = 1.31 (1.13, 1.53) Odds ratio, > 0–672 intensity-weighted lifetime days = 1.38 (1.21, 1.59) Odds ratio, > 672–2,610 days = 1.22 (1.06, 1.41) Odds ratio, > 2,610 days = 1.41 (1.22, 1.62) p-trend < 0.01 Odds ratio, > 0–672 intensity-weighted lifetime days, onset ≤ 10 y prior = 1.39 (1.16, 1.65) Odds ratio, > 672–2610 days, onset ≤ 10 y prior = 1.17 (0.98, 1.41) Odds ratio, > 2610 days, onset ≤ 10 y prior = 1.40 (1.17, 1.68) p-trend = 0.02 No appreciable changes in sensitivity analyses excluding those with positive or unknown history of head injury, excluding those self-reporting Parkinson disease, using unweighted lifetime days of use, using average days/year of use, using intensity-weighted lifetime days through phase 2, or using inverse probability of censoring weights	U.S. National Institute of Environmental Health Sciences, U.S. National Cancer Institute, Michigan State University, Parkinson's Foundation, Office of the Assistant Secretary of Defense for Health Affairs through the Parkinson's Research Program
Zhang et al.	2018	China	Prospective cohort	218 farmers identified as main pesticide users in 20–25 farm households randomly selected from two villages per county, two counties per province in Guangdong, Jiangxi, and Hebei Provinces, China, followed from beginning to end of 2012 growing season; 89% participation	Cohort members without glyphosate exposure	Self-reported amount of glyphosate and other specific agricultural pesticides used in 2012, including chemical name, active ingredient percentage, amount used in kg, and date and duration of spray, recorded after each spray application; pesticide application records checked every other week, pesticide containers saved and checked twice per month	Glyphosate applied per farmer in 2012: mea n = 0.62 kg (52% of total herbicides applied)	Peripheral nerve conduction velocity	Conventional peripheral nerve conduction studies implemented at beginning of planting season (March 2012) and prior to but close to end of crop harvest (March 2012 in Jiangxi and Hebei; December 2012 in Guangdong), including 22 parameters of peripheral nerve conduction examined using surface electrodes with standard placement; classified as nerve conduction velocity, motor conduction velocity, sensory conduction velocity, distal motor latency, amplitude of action potential, amplitude of compound muscle action potential, and amplitude of sensory nerve action potential, categorized as normal vs. abnormal or aggregated into counts of abnormal parameters	Adjusted: age, sex, smoking, alcohol consumption, use of personal protective measures (e.g., wearing masks, gloves, or clothes with long sleeves), diabetes mellitus, body mass index, baseline peripheral nerve conduction, and other classes of pesticides (non-glyphosate herbicides or organophosphorus, organonitrogen, organosulfur, pyrethroid, and other insecticides and fungicides)	Multivariable logistic and negative binomial regression	Odds ratio for abnormal overall nerve conduction velocity = 0.70 (0.38, 1.30) Incidence rate ratio for # abnormal parameters of overall nerve conduction velocity = 0.86 (0.67, 1.10) Odds ratio for abnormal motor nerve conduction velocity = 1.34 (0.30, 6.03) Incidence rate ratio for # abnormal parameters of motor nerve conduction velocity = 1.11 (0.81, 1.53) Odds ratio for abnormal sensory nerve conduction velocity = 0.64 (0.35, 1.18) Incidence rate ratio for # abnormal parameters of sensory nerve conduction velocity = 0.74 (0.52, 1.06) Odds ratio for abnormal distal motor latency = 1.05 (0.81, 1.37) Incidence rate ratio for # abnormal parameters of distal motor latency = 1.02 (0.85, 1.22) Odds ratio for abnormal overall amplitude = 1.21 (0.75, 1.97) Incidence rate ratio for # abnormal parameters of overall amplitude = 0.96 (0.65, 1.43) Incidence rate ratio for # abnormal parameters of motor amplitude = 1.25 (0.67, 2.34) Incidence rate ratio for # abnormal parameters of sensory amplitude = 1.04 (0.49, 2.19)	National Natural Science Foundation of China

Statistically significant associations are shown in bold font

ALS amyotrophic lateral sclerosis, AMD age-related macular degeneration, CI confidence interval, CrI credible interval, NR not reported