NeuroToxicology 57 (2016) 22–30

Contents lists available at ScienceDirect

NeuroToxicology

Full Length Article

Solvent exposure and cognitive function in automotive technicians

Michael N. Batesa,* , Bruce R. Reedb,1, Sa Liua , Ellen A. Eisena , S. Katharine Hammonda
a
Division of Environmental Health Sciences, School of Public Health, University of California, Berkeley, CA 94720, USA
b
Department of Neurology, University of California, Davis, CA, USA

A R T I C L E I N F O A B S T R A C T

Article history:
Received 6 May 2016 Automotive technicians are commonly exposed to organic and chlorinated solvents, particularly through
Received in revised form 15 August 2016 use of cleaning products. Occupational solvent exposures have been associated with deficits in cognitive
Accepted 16 August 2016 function but, to our knowledge, no previous studies have investigated automotive technicians. The
Available online 18 August 2016 purpose of the present study was to investigate whether previous exposures to n-hexane, in particular, or
general solvents posed a persistent neurotoxic hazard to automotive workers.
Keywords: Enrolled in the study were 830 San Francisco Bay Area automotive repair workers. Each participant
Automotive repair underwent a battery of cognitive function tests to investigate central nervous system impairment, with a
Automotive technician
primary focus on the domains of psychomotor speed, fine motor function, memory and mood. Cognitive
Cognitive function
test results regressed against estimated hexane and total solvent exposures showed little evidence of
Hexane
Occupational exposure associations. Exposures to both solvents and hexane were well below the occupational exposure limits.
Solvent Our results provide some reassurance about persistent neuropsychological effects in automotive
workers who use solvent-based products and those who previously used hexane-containing automotive
cleaning products, since this solvent is believed no longer to be used in automotive cleaning products. The
lack of observed effect in this study may be attributable to low exposures, or it may reflect improved
cognitive function since hexane use in automotive cleaning products was discontinued. However,
impacts on results of exposure misclassification and/or the healthy worker survivor effect cannot be
discounted. Irrespective of the outcome of this study, the main known neurologic effect of n-hexane is
peripheral neuropathy, and such an association in automotive technicians is not excluded by these
results.
ã 2016 Elsevier B.V. All rights reserved.

1. Introduction
Automotive technicians, who service and repair motor vehicles,
are commonly exposed to organic and chlorinated solvents,
particularly through use of spray cans to clean brakes and engine
parts, but also from solvent tanks into which engines and parts are
cleaned by dipping. Since the 1970s, evidence has accumulated
that occupational solvent exposure is associated with deficits in
cognitive function (Berr et al., 2010; Meyer-Baron et al., 2008;
Sabbath et al., 2012). Most studies have involved exposure to
solvent mixtures and most have been of painters of houses, cars
and ships, but, to our knowledge, none have involved automotive
technicians.
* Corresponding author.
E-mail address: m_bates@berkeley.edu (M.N. Bates).
1
Present address: Center for Scientific Review, National Institutes of Health,
Bethesda, MD 20892, USA.
The present study was prompted by the finding in 1998 of 3 San
Francisco Bay Area automotive technicians from one car repair
facility with evidence of mild sensory or motor nerve conduction
abnormalities (MMWR, 2001). Ages of these mechanics were 24,
45 and 52 years. Since 1989, automotive spray cans containing up
to 85% hexane, often with acetone, had been used in California and
other states. The technical grade of mixed hexanes used in industry
typically contains 20–80% n-hexane. N-hexane is well-established
as a peripheral neurotoxin. Its neurotoxic effects are enhanced in
animal models by concurrent exposure to acetone, although to our
knowledge the combined effects of the two solvents have not been
studied in humans (Noraberg and Arlien-Soborg, 2000). The levels
of hexane exposure associated with automotive spray can use have
been shown to be below the current occupational exposure limits
for exposure to n-hexane (Wilson et al., 2007).
The peripheral neuropathy findings (and color vision deficits in
the same persons) prompted the California Department of Health
Services in 2001 to issue a Health Hazard Advisory through HESIS,
its Hazard Evaluation System and Information Service. Because the
California market is large, manufacturers saw it as in their interest

http://dx.doi.org/10.1016/j.neuro.2016.08.009
0161-813X/ã 2016 Elsevier B.V. All rights reserved.
 M.N. Bates et al. / NeuroToxicology 57 (2016) 22–30
to remove hexane from all automotive spray cleaning products
used across the U.S. (Wilson, 2003) However, hexane remains a
widely used industrial solvent, for extraction of vegetable oils from
crops, as a solvent for glues, varnishes and inks, and as a cleaning
agent in the printing industry (http://www3.epa.gov/airtoxics/
hlthef/hexane.html) (last accessed July 22, 2016).
Our primary purpose in conducting this study was to
investigate whether previous exposures to hexane, previously
shown to be below the ACGIH Threshold Limit Value (TLV1)
(Wilson et al., 2007) posed a persistent neurotoxic hazard. Results
for color vision testing have already been published (Beckman
et al., 2016). Data on peripheral nerve function are being prepared
for separate publication. We also collected urine samples for
biomarker analysis.
Although the study focus was hexane, data were obtained on all
solvents used in the automotive repair facilities at which our
participants worked. As well as measures of peripheral nerve
function, a battery of cognitive function tests was used to
investigate central nervous system impairment. The results from
the cognitive test battery are the focus of this paper. We
hypothesized that we would find solvent exposure in general,
and hexane exposure in particular, to be associated with deficits in
cognitive function.
2. Methods
Institutional Review Board approvals for study procedures were
obtained at the University of California, Berkeley, and at the
University of California, Davis, under a University of California
multi-campus agreement. Written informed consent was obtained
from all participants before their participation. Participants were
each paid $75 to cover transport to and from the study clinic and as
some compensation for their time.
2.1. Participants
Participants were recruited from male members of 3 Northern
California locals of District 19 of the International Association of
Machinists and Aerospace Workers (IAMAW). Eligibility required
that participants be males 60 years old at time of participation,
have worked as an automotive technician for at least one year in
the period 1990–2000 (when hexane was in use), and currently
living in or near the San Francisco Bay Area. Women were not
included as there were few female members of the IAMAW during
the period of hexane use, and another part of the study
investigated possible testicular effects of hexane, manifesting in
infertility. Current work as an automotive technician was not a
requirement.
We began recruitment with members of the Oakland-San
Leandro local, moving to locals in San Mateo and Sacramento when
we were unable to further recruit eligible and willing members of
the first local. The first step in recruitment was obtaining the Local
1546 membership database, which contained records of former
and current members. From this database we identified all
members who fit the eligibility criteria, including their most up-
to-date contact details. Current addresses were confirmed using
other means, including Experian, California Department of Motor
Vehicles, and California voter registration records. Recruitment
was initiated by sending a letter to the current address, including a
brochure describing the study and a stamped and addressed
response envelope with a form to be returned indicating
willingness to participate. If necessary, a further letter was sent,
followed by telephone calls to the last recorded telephone number.
Where necessary, efforts were made to find new or alternative
telephone numbers. Recruitment efforts proceeded until we had
received participation acceptance or refusal, the invitee was found
23
to have moved out of the area, was found to be deceased, or we had
made no contact after multiple (up to 30) attempts to do so.
2.2. Data collection
Participants from Locals 1546 and 1414 attended a dedicated
clinic in San Leandro, California. A mobile clinical van was used for
participants from Local 2182 in Sacramento. At the clinic they
responded to questions in a questionnaire and underwent a series
of clinical tests, including the cognitive function test battery.
2.2.1. Questionnaire
The questionnaire was programmed using Casic BuilderTM
(West Portal Software Corporation), for direct data entry. Data
were downloaded as they were entered to a dedicated study server.
In addition to collecting data on demographics, medical history,
tobacco smoking, lifetime alcohol consumption, education and
income, the questionnaire obtained detailed automotive techni-
cian work histories, including names of employers, solvent-using
tasks and their frequency at each workplace, and names of solvent
products. We did not collect information on other possible
neurotoxic post-automotive work exposures, on the assumption
that they would be unlikely to be correlated with automotive
solvent exposures and, therefore, could not confound our results.
As a memory aid, a booklet containing 121 color pictures of
spray-can products was provided. The booklet was not complete,
as pictures of 17, mostly older, products could not be obtained.
The work history module of the questionnaire was developed
using a focus group of 14 experienced automotive technicians to
identify tasks with potential solvent exposures, work practices and
historical changes in those that would have affected solvent
exposure. A draft questionnaire was pilot-tested and revised,
repeating this procedure until no change was necessary, before the
final version was produced.
2.2.2. Cognitive testing
Tests assessed psychomotor speed, fine motor function,
memory, and mood–important cognitive functions that are
sensitive to adverse effects of a wide variety of injuries and
diseases (Lezak et al., 2004). These neurobehavioral domains have
been reported as affected by solvents or have been identified as
relevant to neurotoxic assessment (Anger et al., 1994; Anger et al.,
2000; Rohlman et al., 2003; White et al., 2003). Tests were
administered in a fixed order in a quiet, light-controlled room and,
for the Sacramento local, in a mobile clinic. Simple reaction time
and finger tapping were measured by computer-driven tests
administered using Presentation Software (http://www.neurobs.
com; Neurobehavioral Systems, Inc., Albany, CA) and run on a
personal computer with an LCD display and a high temporal
resolution mouse (Copperhead 2000 DPI Gaming Mouse; Razer
USA Ltd.).
Simple Reaction Time (SRT) Speed of simple motor response to
a visual cue is a basic measure of psychomotor speed. Reaction
time is sensitive to a wide variety of factors, including toxic
exposures (Anger et al., 2000; Rohlman et al., 2003). The primary
outcome variable was mean SRT across all trials after excluding
times <100 ms and >1000 ms, resulting in exclusion of 3
participants.
Finger Tapping (Cousins et al., 1998; Jobbagy et al., 2005) speed
for the index fingers, as a test of fine motor function, was measured
over 30 s intervals using Presentation software, as described in
Hubel et al. (Hubel et al., 2013a). The timing of each press and
release was recorded using the Windows programmable clock,
which has a temporal resolution of 0.1 ms, to provide a temporal
uncertainty measure for each response. Outcome variables were
taps per 30 s, separately for dominant and non-dominant hands.
24 M.N. Bates et al. / NeuroToxicology 57 (2016) 22–30
Digit Symbol requires multiple cognitive processes and cortical
systems, including the key functions of psychomotor speed,
attention and memory. It is thus sensitive measure to a wide
variety of types of brain damage, including effects of toxic
exposures (Anger et al., 2000; Joy et al., 2003; Joy et al., 2004).
The test was administered using standard materials and methods
(Wechsler, 1997). The primary outcome measure was the number
of correct responses in 90 s.
Memory was assessed with the Benton Visual Retention Test
(BVRT) (Sivan, 1992), using standard procedures. It was adminis-
tered using stimulus set C and a 5-s exposure duration. Two
measures quantified BVRT performance: “BVRT Correct” (number
of items correct; 0–10) and “BVRT Errors” (sum of item error
scores; 0–40).
Mood was assessed using the Profile of Mood States (POMS)–
part of the battery recommended by the WHO-NCTB (Anger et al.,
2000) for neurotoxicity assessment. The 30-item version was used
(Curran et al., 1995).
2.2.3. Solvent exposure assessment
Exposure to total solvents was calculated by combining
information obtained from air sampling data with year-specific
MSDSs (material safety data sheets) on the commercial products
used in each workplace, the solvent application method, duration
of task performance and quantity of solvent products used with
self-reported tasks performed by each participant. Each combina-
tion of task and application method (aerosol can, pump spray,
solvent tank, etc.) was assigned a concentration of solvent in the
breathing zone, which was multiplied by the reported estimate of
time spent performing that task per day. Task-based solvent
concentrations were based on measurements made in San
Francisco Bay Area automotive repair shops by Wilson et al.
(2007). The full shift average was calculated by combining
exposures from all of a participant’s daily tasks with estimates
of shop background levels. Total exposures were expressed in mg/
m3, as an 8 h time-weighted average, and estimated for each
month. Estimated dermal exposures were expressed as inhalation-
equivalent exposures in mg/m3. Cumulative exposure, in mg/m3-
years, was calculated by adding the monthly 8 h time-weighted
averages and dividing by twelve.
Hexane was present only in aerosol cans. Exposure to hexane
was calculated by applying the percent of hexane in a product,
obtained from Material Safety Data Sheets, to the total solvent
exposure from that product. Since aerosol products containing
hexane evaporate completely, the fraction of solvent as hexane in
the breathing zone was assumed to be the same as in the product,
as has previously been demonstrated to be the case (Wilson et al.,
2007). Acetone exposure was not quantified, but was classified as
either present or absent in any hexane formulation.
We were unable to identify the compositions of all aerosol
products used by all participants in all years, mainly for products
used before 1989, so the following procedure was employed:
Generally, for aerosol products with years for which we did not
have an MSDS, we assumed the average composition from MSDSs
of the same product before and after the missing period; for
products with missing MSDSs for the study period, we assumed
hexane concentration to be zero. For individual participants who
were unable to provide specific product identifications for a work
period, we assumed use of the same products as used by other
technicians who worked in the same automotive shop during the
same period; for individuals who worked in shops without product
composition information for the same period, averaged product
compositions for the same shop from before and after the missing
period were used; for individuals with no shop-specific informa-
tion for any time, the average composition of the three most
popular aerosol spray products during each year of work was
assumed.
2.3. Statistical analysis
Each neuropsychological test was evaluated separately follow-
ing a common analytic approach. First, participants whose test
scores may have been invalidated were excluded. Invalidating
factors included equipment failure, disruption during testing (e.g.,
a cell phone ringing), and sensory or motor problems that
compromised the test (e.g., presence of a finger splint would
invalidate Finger Tapping). The primary analyses were carried out
with multivariable linear regression, using quartiles of solvent
exposure metrics or trichotomous categories. Trichotomous
categories were used when a substantial number of participants
had estimated zero or close to zero exposure—for measures of
hexane exposure. They were created by setting all the participants
with zero exposure as the baseline group and then dichotomizing
those with exposure into equal sized groups.
Parallel analyses were run with each of the exposure metrics as
continuous variables. Covariates were selected based on known
associations with outcome variables. For simplicity and compre-
hensibility, we used a common set of covariates (age, union local,
race/ethnicity, education, income, alcohol) for all analyses. Some of
these covariates were selected because of well-established
associations with cognitive function (age, education, alcohol,
and race/ethnicity and education as indicators of socio-economic
status). Union local was included because for one local data were
collected in a clinic set up in a mobile van; we had concerns that
this might have impacted the data collection in some (unknown)
way. Age was categorized as shown in Table 1. Race was categorized
as Native American, Asian, Hawaiian/Pacific Islander, Black, White,
and other. Education was a categorical variable with 3 levels:
school only, some college, college degree. Income was an ordinal
variable with 5 levels. There were two variables to describe alcohol
use: consumption frequency (5 levels) and how often 6 or more
drinks were consumed on one occasion (5 levels). Categories are
shown in Table 1.
3. Results
Study participation took place during 2009–2012. Using
telephone and address information from electronic union records,
we attempted to contact 4186 potentially eligible automotive
technicians. Of the 2848 subjects contacted, 1765 were verified as
eligible to participate, according to our criteria. Of these, 908 (52%)
refused to participate. Respondents who declined to participate
were asked to give a reason for refusal. The largest number stated
their refusal was out of a lack of interest (n = 262), others
considered that they lived too far away to travel to the study
site (n = 132), and others were too busy to come in for an
appointment (n = 129). A few respondents were too ill to travel to
our clinic (n = 15), and several (n = 13) repeatedly failed to show up
after scheduling appointments. In 78 cases, a respondent (usually a
family member) informed the researcher that the subject was not
interested in participation and an additional 68 subjects across
Locals 1546 and 1414 declined by returning the postcard sent out
by our researchers. Seven respondents stated that the amount of
compensation offered for participation was inadequate for the
time and effort required.
We had data on age for both participants and refusals only for
Local 1546, which provided 80% of our participants. The mean age
for those agreeing to participate was 49.9 years and for those
refusing it was 49.1 years.
A total of 831 automotive technicians participated and provided
data for the study. We divided them into exposure quartiles for
 M.N. Bates et al. / NeuroToxicology 57 (2016) 22–30 25

Table 1
Description of the study population, by quartiles (Q1–Q4) of cumulative inhaled solvent exposures and trichotomized (T1–T3) cumulative hexane exposures.

N (col%) Solvent exposure (mg/m3)-yrs Hexane exposure (mg/m3)-yrs

Q1 Q2 Q3 Q4 T1 T2 T3 Unknown
(<605) (605 to <1197) (1197 to <2069) (2069) (0) (>0 to <32) (32)
N 830 (100%) 208 209 207 207 386 210 217 18
Age-group (years)
<45 176 (21.2%) 33.7% 24.9% 16.4% 9.7% 19.5% 16.7% 30.4% 0
45 to <50 194 (23.4%) 20.2% 25.4% 26.1% 21.8% 20.5% 28.1% 24.0% 22.2%
50 to <55 237 (28.6%) 24.5% 29.7% 29.5% 30.6% 28.8% 31.4% 24.4% 38.9%
55+ 223 (26.9%) 21.6% 20.1% 28.0% 37.9% 31.2% 23.8% 21.2% 38.9%

Union local
Oakland-San Leandro 662 (79.8%) 80.3% 79.0% 80.2% 79.6% 77.7% 81.0% 83.9% 61.1%
Sacramento 60 (7.2%) 7.7% 5.7% 7.7% 7.8% 8.1% 7.1% 5.1% 16.7%
San Mateo 108 (13.0%) 12.0% 15.3% 12.1% 12.6% 14.3% 11.9% 11.1% 22.2%

Race
Native American 24 (2.9%) 1.9% 3.8% 2.9% 2.9% 3.1% 3.8% 1.8% 0
Asian 88 (10.6%) 12.5% 8.6% 13.0% 8.3% 9.9% 7.6% 15.2% 5.6%
Hawaian/Pacific Islander 23 (2.8%) 0 3.8% 3.9% 3.4% 1.3% 1.9% 6.0% 5.6%
Black 29 (3.5%) 3.9% 3.8% 2.9% 3.4% 4.2% 3.3% 2.3% 5.6%
White 672 (81.0%) 83.7% 79.0% 76.3% 85.0% 81.6% 86.7% 73.7% 88.9%
Other 29 (3.5%) 1.9% 4.8% 4.8% 2.4% 3.9% 1.9% 4.6% 0

Ethnicity
Not Hispanic or Latino 697 (84.0%) 85.1% 84.2% 83.6% 83.0% 83.4% 84.8% 84.3% 83.3%
Hispanic or Latino 132 (15.9%) 14.4% 15.8% 16.4% 17.0% 16.4% 15.2% 15.7% 16.7%
Unknown 1 (0.1%) 0.5% 0 0 0 0.3% 0 0 0

Education
High school only 269 (32.4%) 26.4% 30.6% 36.2% 36.4% 33.0% 34.3% 31.8% 5.6%
Some college, no degree 364 (43.9%) 48.6% 41.6% 41.1% 44.2% 44.2% 40.5% 44.7% 66.7%
College degree 197 (23.7%) 25.0% 27.8% 22.7% 19.4% 22.9% 25.2% 23.5% 27.8%

Household income
$40,000 84 (10.1%) 12.5% 7.7% 7.7% 12.6% 10.9% 7.6% 11.1% 11.1%
$40,001–$60,000 125 (15.1%) 16.8% 10.1% 17.9% 15.5% 13.3% 15.2% 18.4% 11.1%
$60,001–$80,000 189 (22.8%) 16.8% 27.8% 21.3% 25.2% 20.2% 26.2% 24.0% 22.2%
$80,00–$100,000 156 (18.8%) 17.3% 21.1% 18.8% 18.0% 19.2% 17.6% 18.0% 33.3%
>$100,000 268 (32.3%) 34.6% 33.0% 33.3% 28.2% 35.1% 32.4% 28.1% 22.2%
Missing 8 (1.0%) 1.9% 0.5% 1.0% 0.5% 1.3% 1.0% 0.5% 0

Alcohol consumption frequency

Non-drinker 189 (22.8%) 27.4% 19.1% 20.3% 24.2% 23.9% 16.7% 28.1% 5.6%
.1 time per month 69 (8.3%) 8.7% 5.3% 9.7% 9.7% 7.8% 8.1% 9.7% 5.6%
4 times per month 142 (17.1%) 15.4% 20.1% 16.4% 16.5% 17.4% 18.6% 14.3% 27.8%
3 times per week 148 (17.8%) 17.3% 20.6% 18.4% 15.1% 15.8% 24.3% 15.2% 16.7%
4 times per week 282 (34.0%) 31.3% 34.9% 35.3% 34.5% 35.1% 32.4% 32.7% 44.4%

Frequency of consuming  6 alcoholic drinks

Never 356 (42.9%) 46.6% 41.6% 39.6% 43.7% 44.9% 40.5% 43.8% 16.7%
Less than monthly 217 (26.1%) 23.1% 24.9% 30.4% 26.2% 24.4% 30.0% 24.9% 33.3%
Monthly 101 (12.2%) 14.4% 14.4% 9.2% 10.7% 9.6% 15.7% 13.4% 11.1%
Weekly 102 (12.3%) 10.6% 14.4% 12.1% 12.1% 13.3% 8.6% 12.9% 27.8%
Daily or almost daily 54 (6.5%) 5.3% 4.8% 8.7% 7.3% 7.8% 5.2% 5.1% 11.1%

total solvents, including hexane, and trichotomized categories for
hexane-based metrics. Table 1 shows the distribution of demo-
graphic factors according to quartiles of estimated cumulative
inhaled solvent exposure and hexane exposure (with and without
acetone). Expectedly, older ages were more frequent in the higher
solvent exposure quartiles, both because they have more years of
workplace exposure and because the use of solvents in these jobs
has been decreasing over the past two decades; however, this
pattern was not evident in the hexane exposure groups, because
eligibility required working as a technician during the years
hexane was in use. No other major differences were evident across
the solvent and hexane categories.
Because of excluded scores, the number of participants with
acceptable cognitive battery test scores was smaller than for the
total number of participants. Results, by test battery component,
for all solvent exposure, by estimated total inhaled and dermal
exposure and by estimated inhalation exposure only are shown in
Table 2. The figures are shown for quartile 1 (lowest exposure) and
the mean score for that quartile and the standard error of the
mean; for quartiles 2–3, the differences from the quartile 1 mean
are shown, with their 95% confidence intervals. The column labeled
D indicates the direction of a score for better performance on the
particular test. In almost every case, the confidence intervals for
quartiles 2–4 include the null value (zero), indicating no evidence
of a difference from quartile 1. The two exceptions, for simple
 26
Table 2
Cognitive function and mood–test score differences from the first quartile (Q1) of all solvent exposure concentration in (mg/m3)-yrs, after controlling for covariates. Results where the 95% confidence interval excludes the null value
are in bold type face.

Test Na Db All solvents cumulative exposure– All solvents cumulative inhalation exposure (mg/m3-yrs)
Inhalation and dermal (mg/m3-yrs)

Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4
(<676) (676 to <1381) (1381 to <2491) (2491) (< 605) (605 to <1197) (1197 to <2069) (2069)
Mean (SE)c Difference from Q1 Mean (SE)c Difference from Q1
d d
(95% confidence interval) (95% confidence interval)
Psychomotor Speed
Simple reaction time (ms) 707 <0 246 2.41 5.82 2.46 246 2.00 9.09 0.21
(1.86) ( 3.98, 8.80) ( 0.79, 12.4) ( 4.14, 9.06) (1.82) ( 4.43, 8.43) (2.63, 15.56) ( 6.78, 6.37)
Digit symbol # correct 756 >0 45.9 1.43 0.45 0.66 45.9 1.87 0.26 0.62
(0.74) ( 3.28, 0.43) ( 1.46, 2.36) ( 2.57, 1.26) (0.75) ( 3.74, 0.01) ( 1.62, 2.14) ( 2.53, 1.28)

Memory
BVRTe correct 758 >0 5.92 0.16 0.03 0.38 5.84 0.06 0.03 0.20
(0.14) ( 0.54, 0.22) ( 0.43, 0.36) ( 0.77, 0.02) (0.14) ( 0.45, 0.32) ( 0.42, 0.36) ( 0.59, 0.19)
BVRTe errors 758 <0 5.59 0.20 0.04 0.57 5.64 0.20 0.06 0.34
(0.23) ( 0.47, 0.88) ( 0.73, 0.66) ( 0.12, 1.26) (0.23) ( 0.48, 0.88) ( 0.63, 0.75) ( 0.35, 1.03)

M.N. Bates et al. / NeuroToxicology 57 (2016) 22–30

Fine Motor Function
Tapping, dominant hand, # taps 702 >0 54.2 0.19 0.17 0.24 53.9 0.44 0.07 0.29
(0.59) ( 1.89, 1.51) ( 1.95, 1.60) ( 1.52, 2.00) (0.60) ( 1.28, 2.16) ( 1.80, 1.67) ( 1.46, 2.05)
Tapping, non dominant, # taps 696 >0 48.6 0.46 0.41 1.22 48.1 0.33 0.38 0.60
(0.55) ( 2.15, 1.23) ( 2.18, 1.35) ( 2.97, 0.53) (0.55) ( 1.38, 2.04) ( 2.11, 1.34) ( 2.34, 1.15)

Mood
POMSf -Tension 795 <0 4.02 0.01 0.55 0.26 3.98 0.06 0.37 0.03
(0.25) ( 0.67, 0.70) ( 1.25, 0.15) ( 0.44, 0.96) (0.25) ( 0.63, 0.75) ( 1.06, 0.32) ( 0.67, 0.72)
f
POMS -Depression 795 <0 3.01 0.24 0.43 0.30 3.00 0.25 0.40 0.14
(0.23) ( 0.44, 0.92) ( 1.12, 0.27) ( 0.39, 0.99) (0.24) ( 0.43, 0.93) ( 1.09, 0.28) ( 0.55, 0.82)
POMSf -Anger 795 <0 4.39 0.13 0.25 0.21 4.28 0.38 0.26 0.14
(0.28) ( 0.64, 0.89) ( 1.04, 0.53) ( 0.57, 1.00) (0.28) ( 0.39, 1.15) ( 1.03, 0.52) ( 0.64, 0.92)
POMSf -Fatigue 795 <0 6.22 0.51 0.37 0.72 6.05 0.83 0.46 0.82
(0.31) ( 0.41, 1.42) ( 0.57, 1.30) ( 0.21, 1.65) (0.31) ( 0.09, 1.75) ( 0.47, 1.38) ( 0.11, 1.75)
f
POMS -Confusion 795 <0 4.47 0.32 0.19 0.07 4.47 (0.20) 0.21 0.15 0.06
(0.20) ( 0.27, 0.91) ( 0.79, 0.41) ( 0.53, 0.68) ( 0.39, 0.80) ( 0.75, 0.45) ( 0.65, 0.54)
f
POMS -Vigor 795 >0 9.43 0.09 0.25 0.15 9.34 0.03 0.44 0.31
(0.29) ( 0.94, 0.76) ( 0.62, 1.12) ( 0.71, 1.02) (0.30) ( 0.83, 0.88) ( 0.42, 1.30) ( 0.56, 1.17)
a
N, number of participants in the model.
b
D, direction of better performance.
c
Unadjusted values.
d
Tabled values are the mean (95% CI) of the difference between the estimated outcome mean for exposure quartile 1 and quartiles 2-4 in models adjusted for age, union local, ethnicity, education, income, and alcohol consumption.
e
BVRT Benton Visual Retention Test.
f
POMS, Profile of Mood States.
Table 3
Cognitive function and mood–test score differences from the zero exposure group (T1) of hexane exposure concentration in mg/m3-yrs, after controlling for covariates.

Test Na Db Any hexane exposure (mg/m3)-yrs Hexane exposure with acetone (mg/m3)-yrs

T1 T2 T3 T1 T2 T3
(0) (>0 to <32) (32) (0) (>0 to <3.5) (3.5)
Mean (SE)c Difference from T1 (95% confidence interval)d Mean (SE)c Difference from T1 (95% confidence interval)d

Psychomotor Speed
Simple reaction time (ms) 693 <0 248 5.57 0.66 248 3.20 3.03
(1.49) ( 0.04, 11.19) ( 4.95, 6.26) (1.31) ( 2.59, 8.99) ( 2.84, 8.90)
Digit symbol # correct 740 >0 43.6 1.21 0.78 44.1 0.75 0.54
(0.53) ( 0.39, 2.82) ( 0.85, 2.40) (0.47) ( 0.92, 2.41) ( 1.16, 2.24)

Memory
BVRTe correct 742 >0 5.61 0.12 0.13 5.71 0.05 0.25
(0.10) ( 0.21, 0.46) ( 0.47, 0.21) (0.09) ( 0.29, 0.40) ( 0.61, 0.10)
BVRTe errors 742 <0 6.12 0.29 0.29 5.94 0.12 0.48
(0.19) ( 0.88, 0.30) ( 0.30, 0.89) (0.17) ( 0.73, 0.49) ( 0.14, 1.11)

Fine Motor Function

Tapping, dominant hand, # taps 689 >0 53.6 0.86 0.29 53.9 1.25 0.29

M.N. Bates et al. / NeuroToxicology 57 (2016) 22–30

(0.43) ( 2.35, 0.63) ( 1.77, 1.19) (0.41) ( 2.77, 0.28) ( 1.84, 1.25)
Tapping, non-dominant, # taps 684 >0 47.9 0.73 0.83 47.9 1.19 0.34
(0.44) ( 2.20, 0.74) ( 2.29, 0.63) (0.41) ( 2.70, 0.32) ( 1.86, 1.19)

Mood
POMSf -Tension 778 <0 3.81 0.03 0.04 3.75 0.16 0.25
(0.19) ( 0.57, 0.62) ( 0.63, 0.56) (0.17) ( 0.45, 0.77) ( 0.37, 0.88)
f
POMS -Depression 778 <0 2.90 0.02 0.04 2.94 0.26 0.07
(0.18) ( 0.57, 0.61) ( 0.64, 0.55) (0.17) ( 0.86, 0.35) ( 0.55, 0.69)
POMSf -Anger 778 <0 4.08 0.12 0.04 4.16 0.40 0.29
(0.20) ( 0.54, 0.78) ( 0.71, 0.62) (0.18) ( 1.08, 0.29) ( 0.41, 0.99)
POMSf -Fatigue 778 <0 6.09 0.31 0.41 6.18 0.26 0.49
(0.24) ( 0.48, 1.10) ( 0.38, 1.21) (0.21) ( 0.56, 1.07) ( 0.34, 1.33)
f
POMS -Confusion 778 <0 4.24 0.02 0.28 4.26 0.16 0.48
(0.16) ( 0.53, 0.49) ( 0.23, 0.79) (0.14) ( 0.68, 0.37) ( 0.06, 1.01)
f
POMS -Vigor 778 >0 9.51 0.00 0.03 9.47 0.21 0.35
(0.21) ( 0.73, 0.74) ( 0.76, 0.70) (0.19) ( 0.97, 0.54) ( 0.42, 1.11)
a
N, number of participants in the model.
b
D, direction of better performance.
c
Unadjusted values.
d
Tabled values are the mean (95% CI) of the difference between the estimated outcome mean for exposure tertile 1 (zero exposure) and tertiles 2–4 in models adjusted for age, union local, ethnicity, education, income, and alcohol
consumption.
e
BVRT Benton Visual Retention Test.
f
POMS, Profile of Mood States.

27
28 M.N. Bates et al. / NeuroToxicology 57 (2016) 22–30
reaction time and digit symbol, are in bold typeface. There is no
evidence of a trend across quartiles.
Table 3 shows results of an analysis similar to that in Table 2, but
based on hexane exposure—with and without acetone. All of the
confidence intervals for the two higher exposure categories
include the null value and there is no evidence of an association
with hexane exposure, either with or without acetone included in
the product.
4. Discussion
The results of this analysis are reassuring. They provide no
evidence that San Francisco Bay Area automotive technicians
experienced persistent cognitive dysfunction as a result of their
cumulative exposure, either to all solvents or to hexane, with or
without acetone. They cannot, however, exclude the possibility
that cognitive defects could have been apparent in this population
closer to the time of exposure to hexane.
A recent meta-analysis of studies of neurobehavioral effects in
solvent-exposed occupational groups provided evidence of deficits
in neurobehavioral parameters, particularly to attentional perfor-
mance (Meyer-Baron et al., 2008). Weaker associations were
frequently present for longer duration exposures. This was
interpreted by the authors as possibly due to the healthy worker
effect, although, without actual evidence to confirm the healthy
worker effect, this remains a supposition and weakens the
evidence for associations between solvents and cognitive effects.
Assuming, for the sake of this discussion, that the associations
observed by Meyer-Baron et al. (2008) are real, there are a number
of possible reasons why such effects were not observed in our
study. The occupational groups studied in the Meyer-Baron
investigation were mostly painters, although there were some
printers, nail technicians, petrochemical workers and workers in
adhesive manufacturing. None were automotive technicians. To
the best of our knowledge, ours is the first study of solvent
exposure in that occupational group. This distinction raises the
possibility of differences in solvent type and in levels of exposure.
In terms of solvent types, this study is complicated by the fact
that many different solvents were used in the various automotive
cleaning products. Some were used in spray cans, but others were
used in solvent tanks (although solvent tanks had generally
changed to aqueous formulations by about 1997). The wide variety
of solvents used and the fact that there were significant gaps in the
data reported, meant that it was impractical for us to attempt to
carry out an analysis based on individual solvent types. The one
exception was hexane, which was the primary focus and
hypothesis of our study and for which much effort was expended
in obtaining as complete a record of use as possible.
There are occupational exposure limits for n-hexane exposure:
500 ppm for the 8-hr time-weighted average (TWA) for the OSHA
personal exposure limit (PEL) (equivalent to 1760 mg/m3), 50 ppm
(equivalent to 176 mg/m3) for the TLV1 and Cal/OSHA, and 180 mg/
m3 for the the German MAK (Maximale Arbeitsplatz Konzentra-
tion). The estimated mean 8-h TWA hexane (n-hexane plus other
hexanes) exposure in our study participants was 14 mg/m3 (SD:
21 mg/m3), range: 0.1–201 mg/m3), with a median of 7.7 mg/m3
(range: 0.1–201 mg/m3). This indicates that most participants in
our study were exposed to n-hexane at concentrations well below
commonly applied occupational exposure limits.
Meyer-Baron et al. present measures of an exposure index (EI)
for the 17 of the 46 publications that they included in their meta-
analysis that reported solvent concentrations. The EI was
calculated for each workplace as S(conci/limiti) across the range
of co-exposed solvents, where i represented an individual solvent.
Conci was the workplace solvent concentration and limiti was the
occupational exposure limit, being the German MAK concentration
for the particular solvent (Deutsche_Forschungsgemeinschaft,
2012). Values of EI for other occupations, including painters, floor
layers, adhesive factory workers, printers and shoe manufacturing
workers, ranged from 0.07 to 14.86, with a median value of 0.66.
The highest EI value was for cleaners. The estimated EI for mean n-
hexane concentration in our study was 0.08—at the lower end of
the EIs in Meyer-Baron. However, their study provided no data
specifically for hexane. This supports the idea that solvent
concentrations to which automotive workers were exposed may
have been lower than concentrations experienced by many other
occupational groups. However, it does not take into account the
other solvents to which our participants were probably co-exposed
at the same time.
In terms of tests used, the tests we applied were for the domains
of psychomotor speed, fine motor function, memory and mood.
The tests applied by the studies summarized by Meyer-Baron et al.
(2008) were for the domains of attention, memory, motor
performance, construction, reasoning and concept formation.
Tests which were found in the meta-analysis to produce
statistically significant (p  0.05) associations included several
which we used in our study – digit symbol, simple reaction time,
BVRT, and tapping with the non-dominant hand – some of which
have also been shown elsewhere to be sensitive to demographic
and clinical differences (Hubel et al., 2013b; Woods et al., 2015).
Thus a difference in cognitive tests used does not easily explain the
difference in results between our study and the Meyer-Baron
meta-analysis. The solvent exposures in our study may have been
too low to produce discernible effects in those tests. Another
possibility is that many of the studies in the meta-analysis were
confounded by lead exposure, since the majority of the studies
were of painters. The authors stated that “in an unknown number
of studies co-exposures to lead as a result of lead-containing paints
cannot be excluded”.
Other possible explanations for why we found no positive
associations involve confounding, selection bias and information
bias. Negative confounding might explain the lack of evidence for
an effect in our study, if there were an exposure that was correlated
with lower solvent exposures and also associated with higher
cognitive risk. It is not clear what that could be. One potential
confounder of solvent-cognitive effect relationships is alcohol
consumption. Our data, however, do not clearly show a clear
relationship between alcohol consumption and solvent exposure
(Table 1) and all regression analyses (Tables 2 and 3) were adjusted
for alcohol consumption.
Since we attempted to recruit actively working, as well as
former or retired, automotive technicians, this would have reduced
some of the potential for the healthy worker effect to impact our
results. However, even if follow-up extends beyond termination,
the healthy worker survivor effect may operate if workers with
cognitive impairment were more likely to terminate work–thereby
reducing exposure. The healthy worker survivor effect, a problem
of time varying confounding and selection bias, causes downward
bias toward the null and beyond (Eisen et al., 2012). Moreover, the
low participation rate introduces another potential source of
selection bias. If people who were disabled were more likely to
participate, then this would possibly lead to an exaggerated
measure of effect, if disabilities were associated with exposure.
However, since we found no evidence of effect, our data do not
support this. Conversely, if people disabled by solvent exposure
were less likely to participate, since they needed to be able to travel
to our study clinic, then this could have dampened detection of any
effect of solvents. However, few reported poor health as a reason
for not participating. Nonetheless, the results may to some extent
have been attenuated by selection bias due to the healthy worker
survivor effect or informative participation. Mitigating any impact
of the healthy worker effect on our results would have been our
 M.N. Bates et al. / NeuroToxicology 57 (2016) 22–30
recruitment from the union membership: unions commonly
continue to maintain contact with their membership and we
were able to take advantage of this to recruit participants no longer
in the automotive repair workforce.
Information bias can affect both health outcomes and
exposures. Since our neurocognitive tests were fairly standard,
test administrators were well-trained, and test discrepancies
were excluded from the database, we think misclassification of
the outcome is a lesser likelihood than misclassification of
exposure. Exposure misclassification seems much more likely,
since we were relying on participants’ memories of products used
and frequencies of their use, going back two or more decades.
When participants did not remember products used during
particular periods of time, we needed to infer their exposures
from other information, including the experiences of other
workers in the same automotive shop or automotive spray
products in widespread use at the time of the data gap.
Additionally, we were not able to get a complete record of
MSDSs for automotive spray cleaning products over the study
period. This resulted in further assumptions, based on the MSDSs
for the same product in different years or the compositions of the
most commonly used comparable products at the time. The
overall impact of this misclassification would likely have been a
bias towards the null. Such assumptions would mostly have
affected estimates of hexane exposure and have had little impact
on estimates of overall solvent exposure.
If our results have been influenced by a bias, we think the most
likely explanations are misclassification of hexane exposure and
selection bias. This will almost certainly have caused a bias towards
the null of any true hexane-related effect, but we cannot say with
any certainty whether such an effect actually occurred in our study
population.
5. Conclusions
In conclusion, our results provide some reassurance about
persistent neuropsychological effects to automotive workers who
use solvent-based products and those who previously used
hexane-containing automotive cleaning products. We recruited
only unionized workers and so our results are confidently
applicable only to that group. We have no reliable information
on the extent of unionization in this profession, although we
believe it is high in the San Francisco Bay Area, but may be lower
elsewhere.
Possibly the lack of effect observed in this study is attributable
to low exposures in this occupational group. However, the
possibilities that it might be attributable to exposure misclassifi-
cation or the healthy worker survivor effect cannot be discounted.
There is also the possibility that the results reflect the benefits of
past exposure reduction in the industry, although it is difficult to
conclude this without results from a comparable study closer to
the time that hexane was used in automotive cleaning products.
Much effort was expended to create detailed longitudinal
solvent exposure profiles for each participant and it seems unlikely
that exposure misclassification would have obscured a strong and
persistent effect on cognitive function, but a weakly positive effect
cannot be ruled out. Overcoming this exposure misclassification
would probably necessitate a prospective study, although the
likelihood of this ever taking place is now slight, since hexane has
not been used in automotive spray products for over a decade and
cleaning tanks are now aqueous, rather than solvent-based.
Irrespective of the outcome of this study, the main known
neurologic effect of n-hexane is peripheral neuropathy, and such
an association in automotive technicians is not excluded by these
results.
29
Conflict of interest
None.
Disclaimer
The opinions expressed in this article are the author's own and
do not reflect the view of the National Institutes of Health, the
Department of Health and Human Services, or the United States
government.
Acknowledgments
This work was supported by the National Institute of
Environmental Health Sciences of the National Institutes of Health
(grant number R01 ES014049).
The authors thank the members of District 19 of the IAMAW for
their participation in the study. Thanks also to the following people
for their assistance with the study, including data collection: Frank
Alvarez, Kimen Balhotra, Lina Borgo, Susan Burns, David Cano, Yu-
teh Cheng, Don Crosatto, Justin Girard, Jackie Hayes, Roger
Hofmann, Sophie Horiuchi, John Kaufman, Meagan Loftin, Ryan
Loomba, Kathy Peixoto, Bill Phillips, Bill Schecter, Genie Tang,
David Woods, Zakia Young.
References
Anger, W.K., Letz, R., Chrislip, D.W., Frumkin, H., Hudnell, K., Russo, J.M., et al., 1994.
Neurobehavioral test methods for environmental health studies of adults.
Neurotoxicol. Teratol. 16, 489–497.
Anger, W.K., Liang, Y.X., Nell, V., Kang, S.K., Cole, D., Bazylewicz-Walczak, B., et al.,
2000. Lessons learned—15 years of the WHO-NCTB: a review. Neurotoxicology
21, 837–846.
Beckman, S., Eisen, E.A., Bates, M.N., Liu, S., Haegerstrom-Portnoy, G., Hammond, S.
K., 2016. Acquired color vision defects and hexane exposure: a study of San
Francisco Bay area automotive mechanics. Am. J. Epidemiol. 183, 969–976.
Berr, C., Vercambre, M.N., Bonenfant, S., Manoux, A.S., Zins, M., Goldberg, M., 2010.
Occupational exposure to solvents and cognitive performance in the GAZEL
cohort: preliminary results. Dement. Geriatr. Cogn. Disord. 30, 12–19.
Cousins, M.S., Corrow, C., Finn, M., Salamone, J.D., 1998. Temporal measures of
human finger tapping: effects of age. Pharmacol. Biochem. Behav. 59, 445–449.
Curran, S.L., Andrykowski, M.A., Studts, J.L., 1995. Short form of the profile of mood
states (POMS-SF): psychometric information. Psychol. Assess. 7, 80–83.
Deutsche_Forschungsgemeinschaft. List of MAK and BAT Values. Commission for
the Investigation of Health Hazards of Chemical Compounds in the Work Area,
2012.
Eisen, E.A., Robbins, J.M., Picciotto, S., 2012. Healthy worker effect. In: El-Shaarawi,
A.H., Piegorsch, W. (Eds.), Encyclopedia of Environmetrics. John Wiley & Sons
Ltd., Chichester, UK, pp. 1269–1272.
Hubel, J.A., Reed, B., Yund, E.W., Herron, T.J., Woods, D.L. Computerized measures of
finger tapping: The effects of hand dominance, aging, gender and fatigue.
Perceptual and Motor Skills, 2013a, in press.
Hubel, K.A., Reed, B., Yund, E.W., Herron, T.J., Woods, D.L., 2013b. Computerized
measures of finger tapping: effects of hand dominance, age, and sex. Percept.
Mot. Skills 116, 929–952.
Jobbagy, A., Harcos, P., Karoly, R., Fazekas, G., 2005. Analysis of finger-tapping
movement. J. Neurosci. Methods 141, 29–39.
Joy, S., Fein, D., Kaplan, E., 2003. Decoding digit symbol: speed, memory, and visual
scanning. Assessment 10, 56–65.
Joy, S., Kaplan, E., Fein, D., 2004. Speed and memory in the WAIS-III digit symbol—
coding subtest across the adult lifespan. Arch. Clin. Neuropsychol. 19, 759–767.
Lezak, M.D., Howison, D.B., Loring, D.W., 2004. Neuropsychological Assessment, 4th
ed. Oxford University Press, New York, NY.
MMWR, 2001. n-Hexane-related peripheral neuropathy among automotive
technicians—California, 1999–2000. Morb. Mortal. Wkly. Rep. 50, 1011–1013.
Meyer-Baron, M., Blaszkewicz, M., Henke, H., Knapp, G., Muttray, A., Schaper, M., et
al., 2008. The impact of solvent mixtures on neurobehavioral performance:
conclusions from epidemiological data. Neurotoxicology 29, 349–360.
Noraberg, J., Arlien-Soborg, P., 2000. Neurotoxic interactions of industrially used
ketones. Neurotoxicology 21, 409–418.
Rohlman, D.S., Gimenes, L.S., Eckerman, D.A., Kang, S.K., Farahat, F.M., Anger, W.K.,
2003. Development of the Behavioral Assessment and Research System (BARS)
to detect and characterize neurotoxicity in humans. Neurotoxicology 24,
523–531.
Sabbath, E.L., Glymour, M.M., Berr, C., Singh-Manoux, A., Zins, M., Goldberg, M., et al.,
2012. Occupational solvent exposure and cognition: does the association vary
by level of education. Neurology 78, 1754–1760.
30 M.N. Bates et al. / NeuroToxicology 57 (2016) 22–30

Sivan, A.B., 1992. Benton Visual Retention Test, 5th ed. The Psychological
Corporation, San Antonio, TX.
Wechsler, D., 1997. WAIS-III Administration and Scoring Manual. The Psychological
Corporation, San Antonio.
White, R.F., James, K.E., Vasterling, J.J., Letz, R., Marans, K., Delaney, R., et al., 2003.
Neuropsychological screening for cognitive impairment using computer-
assisted tasks. Assessment 10, 86–101.
Wilson, M.P., Hammond, S.K., Nicas, M., Hubbard, A.E., 2007. Worker exposure to
volatile organic compounds in the vehicle repair industry. J. Occup. Environ.
Hyg. 4, 301–310.
Wilson, M.P., 2003. n-Hexane Exposure in the California Vehicle Repair Industry:
Risk Assessment and Policy analysis. University of California, Berkeley.
Woods, D.L., Wyma, J.M., Yund, E.W., Herron, T.J., Reed, B., 2015. Factors influencing
the latency of simple reaction time. Front. Hum. Neurosci. 9, 131.