We conducted neurological testing on 701 participants in the Agricultural Health Study (AHS), a large prospective study of licensed pesticide applicators from Iowa and North Carolina (Alavanja et al. 1996
). Male AHS participants were invited to participate in the present study based on their lifetime use of OPs, completion of earlier parts of the study, and proximity to the testing sites.
The AHS enrolled private pesticide applicators between 1993 and 1997 when they completed a self-administered “enrollment” questionnaire at the time of pesticide licensing and recertification; 44% of private pesticide applicators also completed a take-home questionnaire. These two questionnaires comprised phase 1 data collection. After enrollment, two 5-year follow-up phone interviews were administered (phase 2 and 3 data collection). The present study sample is limited to those who completed all AHS questionnaires. The questionnaires are available online (AHS 2011).
To enrich the sample with applicators with higher lifetime use of OP pesticides, we oversampled the high end of the OP lifetime use distribution based on the lifetime days of use of 10 OPs assessed in detail in phase 1. Among eligible participants, a stratified sample was selected based on equal random sampling from the upper and lower portions of the OP lifetime days distribution. In Iowa, a cut point of approximately 75% was used to separate individuals; in North Carolina the cut point was lower (66%) because the North Carolina cohort is more geographically dispersed and has fewer members. Although the cut point was shifted for selection, all analyses were based on lifetime use of pesticides. Thus, although the sampling frame allowed for a sample enriched for OP use, it was not used as an analytical variable.
For the present study, AHS participants with amyotrophic lateral sclerosis, diabetes, multiple sclerosis, Parkinson’s disease, retinal or macular degeneration, hypothyroidism, and stroke were excluded. In addition, participants who, during phase 3 AHS interviews, reported drinking ≥ 42 alcoholic beverages/week or reported being diagnosed with pesticide poisoning were excluded. After exclusions, a total of 1,807 AHS participants were initially eligible to participate. The overall participation rate was 39%.
In Iowa, testing was conducted in Iowa City and Dubuque between November 2006 and March 2007. In North Carolina, testing was conducted in Greenville and Wilmington between January and March 2008. Institutional review boards approved the study protocol, and all participants provided written informed consent.
Exposure assessment. For each participant, pesticide use information from the three AHS phases and a questionnaire administered on the day of neurological testing was used to create ever use and lifetime days of use variables for the 16 OP pesticides used by at least 50 neurological study participants. Ever use was based on any report of use at any interview, and lifetime days of use was based on the sum of lifetime days of use calculated for each interview period.
Pesticide use information was assessed in slightly different ways in each AHS phase. In phase 1 questionnaires, participants were queried in detail about 50 pesticides and asked to provide information on ever use, frequency of use, and years of use. Additionally, on the phase 1 take-home questionnaire, participants were asked to complete a checklist indicating ever use of specific chemicals but were not asked about frequency or duration of use. In phases 2 and 3, participants provided open-ended responses about their pesticide use since the last interview, and their responses were used to create lifetime use information for those time periods. The questionnaire administered on the day of neurological testing collected pesticide use information for the past 12 months. Of the 16 pesticides included in the present study, 9 were reported in detail on the AHS phase 1 questionnaires (chlorpyrifos, coumaphos, diazinon, dichlorvos, fonofos, malathion, parathion, phorate, and terbufos); 6 were initially queried on the phase 1 take-home questionnaire checklist (acephate, dimethoate, disulfoton, ethoprop, phosmet, and tetrachlorvinphos); and a new chemical introduced in 1995 (tebupirimfos) was reported initially on the phase 2 questionnaire.
For phase 1 lifetime days of use, we multiplied the number of days used per year by the numbers of years used to create the summary measure. Because this information was not available for the chemicals listed on the take-home checklist, we assumed that the days per year applied was equal to the median number of days of insecticide use per year for that individual and that the number of years used was equal to the median number of years that individual had applied insecticides using the categories in the phase 1 questionnaire. These values were used to create an estimate of lifetime days of use in phase 1 for the chemicals listed in the take-home checklist. For phases 2 and 3 as well as the neurological study questionnaire, we multiplied the number of days used per year by the number of years since the last interview to create the lifetime days for that period. We then summed the lifetime days for phase 1–3 questionnaires and the neurological study questionnaire to create cumulative lifetime days of use for each pesticide. All pesticide use occurred prior to neurological testing. A summary measure of OP pesticide use (cumulative lifetime days of all OP pesticides) was also created.
We created similar variables for the four carbamate pesticides (aldicarb, benomyl, carbaryl, and carbofuran). In addition, an overall measure of cumulative lifetime pesticide use (all chemical classes) was estimated from all AHS interviews for the 50 pesticides assessed on the AHS enrollment questionnaire. Finally, information on high pesticide exposure events (HPEEs) was obtained from three AHS interviews.
In summary, for every participant, pesticide exposures were characterized as a) ever use of 16 OP and 4 carbamate pesticides, b) cumulative lifetime days of use for each of these pesticides, c) cumulative lifetime days of any OP pesticide use, d) cumulative lifetime days of any pesticide use, and e) ever having an HPEE.
Neurological outcomes. The purpose of the neurological evaluation was to assess neurological function rather than to diagnose clinically apparent peripheral neuropathy.
Neurological physical examination.
Standard neurological physical examinations (NPxs) (Campbell 2005
) were performed by a physician (F.G.) blinded to pesticide exposure status. Assessments of vibration perception (128 Hz) and proprioception were performed on the great toes, bilaterally. Achilles deep tendon reflexes were examined bilaterally, and Romberg test performance, tandem gait, and postural tremor were assessed.
Clinical examination results were recorded as normal, equivocal, or abnormal. For all tests performed bilaterally (ankle reflex, toe proprioception, and toe vibration), examination results were classified as “abnormal” if ratings were abnormal or equivocal bilaterally, abnormal unilaterally and equivocal on the contralateral side, or abnormal unilaterally and missing (because of injury/amputation) on the contralateral side. For tests without laterality (postural tremor, Romberg test, and tandem gait), we combined abnormal and equivocal results to create a dichotomized variable (normal vs. not normal) for each outcome.
Standard noninvasive electrophysiological measures of the dominant peroneal motor nerve were performed with a factory-calibrated TECA Sapphire II electromyograph (TECA Corp., Pleasantville, NY) by one examiner (F.G.) as described by Kimura (2001)
. Foot temperature was maintained above 32°C during testing. Distal motor amplitude (millivolts), distal and proximal motor latency (milliseconds), and short F-wave latency (milliseconds) were obtained and nerve conduction velocity (meters per second) was calculated.
Gross grip strength and key and palmar pinch strength were obtained bilaterally using digital grip and pinch dynamometers (JTech Medical, Salt Lake City, UT) (Mathiowetz et al. 1984
). A mean z
-score for all six hand strength tests (three tests performed bilaterally) combined was calculated to create one summary measure for hand strength.
Standing sway speed was measured with a CATSYS 2000 Force Plate (Danish Product Development, Snekkersten, Denmark) (Despres et al. 2000
). Four 80-sec trials were administered, two with eyes open and two with eyes closed. Average sway speed (millimeters per second) is reported separately for eyes open and eyes closed.
The Vibratron II (Sensortek, Inc., Clifton, NJ) was used to measure bilateral great toe 120-Hz vibrotactile threshold with a standard protocol (Gerr et al. 1990
). A single bilateral mean vibrotactile threshold is reported in log micrometers.
Assessment of potential confounders. Age, height, state (Iowa or North Carolina), smoking status (never, current, or past), education (≤ high school or > high school), alcohol use (0, 1–7, or > 7 drinks/week), ear infection within the past 12 months, prior inner ear surgery, exposure to other potentially neurotoxic substances such as solvents, soldering, and welding fumes (ever used ≥ 8 hr/week), and body mass index (BMI; kilograms per meter squared) were assessed for potential confounding of associations between pesticide use and PNS outcomes.
Study participant exclusions. Participants were excluded from all PNS analyses because of past polio (n = 6), cancer chemotherapy (n = 4), alcohol consumption on the day of testing (n = 3), physician diagnosis of alcoholism (n = 3), reporting drinking ≥ 42 alcoholic beverages/week during the past year (n = 3), diabetes (n = 1), dialysis (n = 1), severe dementia (n = 1), and being struck by lightning (n = 1).
Results of all tests except the electrophysiological tests were excluded for 5 participants with a history of brain tumor. For tests of postural stability, we excluded 2 participants who reported current use of the drug meclizine and 2 with Ménière’s disease. Electrophysiological results of 2 participants were excluded after linear regression diagnostics showed studentized residuals with absolute values > 6.0. In addition, a small number of participants were unable to perform certain tests because of recent surgery, amputation, or injury.
Statistical methods. Logistic regression was used to estimate odds ratios (ORs) of association between pesticide use and dichotomized NPx results. A base model with no pesticide exposure variables was developed using backward elimination. Adjusted models were run separately for each individual pesticide parameterized as ever use versus never use. Exposure response was examined by creating a three-level variable for individual pesticides with the distribution of lifetime days of use split at the median among the pesticide users to create two exposure categories (≤ median, and > median), with never use as the referent category. The distributions of the pesticide summary variables lifetime days of all OP pesticides and lifetime days of all pesticides were split in quartiles with the lowest category as the referent group. Chi-square tests for trend were used for all exposure–response models with exposure levels assigned 0, 1, and 2 for nonexposed, ≤ the median, and > the median. Analyses were restricted to pesticides with at least five exposed cases per category.
Linear regression was used to examine associations between pesticide use and the continuous outcomes. A base model for each neurological outcome was created with an outcome-specific set of covariates using backward elimination. The final multivariate base model for each outcome included only those covariates with p-values < 0.20. Each pesticide was examined both as a dichotomized variable (ever/never use) and as a continuous variable. The cumulative lifetime days of pesticide use variables were log10 transformed to normalize the distribution of residuals. Adjusted associations between neurological outcomes and pesticide exposures were estimated with linear regression models in which the neurological outcome was regressed on the pesticide exposure variable while controlling for base model covariates. For greater ease of interpretation, parameter estimates for peroneal nerve distal motor latency and short F-wave latency, sway speed, and vibrotactile threshold were multiplied by –1 so that lower scores indicated poorer performance for all continuous outcomes.
We examined potential confounding of the association between neurological outcomes and each pesticide by other pesticides in both linear and logistic regression models. Specifically, pesticide pairs with r ≥ 0.30 were added simultaneously to final base models, and the pesticide variable parameter estimates were compared with models with one pesticide.
To evaluate whether associations between pesticide use and neurological outcomes were influenced by previous pesticide poisoning, the 8 participants who reported ever being physician diagnosed with pesticide poisoning at the time of AHS enrollment were excluded from the analyses, and the results were compared with models that included them. The results did not change.
All analyses were performed using SAS software (version 9.2; SAS Institute Inc., Cary, NC).