FUNDAMENTAL AND APPLIED TOXICOLOGY 36, 3 0 - 3 8 (1997)
ARTICLE NO. FA972287
Concentration-Time Relationships for the Effects of Inhaled
Trichloroethylene on Signal Detection Behavior in Rats12
PHILIP J. BUSHNELL
Neurotoxicology Division, U.S. Environmental Protection Agency, Research Triangle Park, North Carolina 27711
Received July 5, 1996; accepted January 9, 1997
TCE associated with changes in signal detection and thus underestimates the risk of behavior change from short-term exposures to
TCE. On the other hand, the fact that SIo., and RT100 did increase
with shorter exposure times indicates that the converse assumption, that the toxicity of inhaled TCE is independent of the duration of exposure, yields an overly conservative estimate of risk.
Concentration-Time Relationships for the Effects of Inhaled
Trichloroethylene on Signal Detection Behavior in Rats. BUSHNELL, P. J. (1997). Fundam. Appl. Toxicol. 36, 30-38.
The risk from inhaled volatile organic compounds (VOCs) is
presently assessed on the basis of lifetime exposure to average
concentrations of the vapor. This strategy yields rational predictions of risk if the product of concentration (C) and the duration
of exposure (0 yields constant effects on health (Haber's Rule).
The validity of this assumption was evaluated by assessing the
acute behavioral effects of inhaled trichloroethylene (TCE) vapor
at various values of C and t. Adult male Long-Evans rats (n =
11) were trained to perform a signal detection task in which a
press on one lever produced food on trials containing a signal (a
brief, unpredictable lightflash);a press on a second lever produced
food on trials lacking a signal. Response time (RT) and indices of
sensitivity (SI) and bias (RT) derived from the theory of signal
detection were calculated at three times during repeated daily 60min tests conducted in air containing 0,400, 800,1200,1600, 2000,
or 2400 ppm TCE. Behavior remained stable during tests in air.
In TCE, SI declined and RT increased as functions of both C and
t. RI was not affected by TCE. Effects on SI and RT were not
predictable from the C X ( product both endpoints were more
affected by C than by r. To quantify the change in the effect of
TCE across exposure times, concentration-effect relationships for
inhaled TCE on SI and RT were modeled with cubic polynomial
functions at each of the three exposure durations. Concentrations
of inhaled TCE associated with preselected changes in SI and RT
were then estimated for each animal from these functions. Criterion concentrations, SIo., and RTIOo, were denned as the concentration of TCE associated with a 0.1-unit decrease in SI or a 100msec increase in RT, respectively. Both SIo., and RT100 increased
as exposure duration decreased, but did so more slowly than would
be predicted by Haber's Rule. This pattern indicates that application of Haber's Rule overestimates the concentration of inhaled
O 1997 Society of Toikoksgy.
Volatile organic compounds (VOCs), including organic
solvents, comprise a large proportion of the chemicals
released into the atmosphere, and acute exposure to these
VOCs has been associated with neurotoxicity. For example, 18 of the highest-volume 25 chemicals released into
the air have been reported to be acutely neurotoxic VOCs
(OTA, 1990). Because the nervous system responds rapidly to inhaled VOCs, it is a likely target organ for shortterm releases of these chemicals, and evaluation of risk
from short-term exposure to VOCs should include assessment of neurotoxicity.
In response to concerns regarding episodic VOC releases,
the Clean Air Act of 1990 requires evaluation of "residual
risk" arising from short-term excursions of the concentration
of a VOC above regulated ambient levels (EPA, 1994). Ideally, assessment of the residual risk from a VOC is based
upon knowledge of the relationships among exposure, target
organ dose, and the mechanism of action of the chemical at
the target organ (Jarabek, 1995). Here, exposure refers to
the temporally variable concentration of the VOC in the
inhaled air, which depends upon the characteristics (frequency, concentration and duration) of release events,
whereas dose refers to a measure of the amount of the agent
reaching the target organ.
Typically, neither the dosimetry nor mechanism of toxicity of a VOC is known; thus, current approaches to noncancer risk assessment rely upon default assumptions regarding the relationship between exposure and effect. One such
assumption is that both the internal dose and the effect of
an airborne pollutant are directly related to the product of
the concentration (C) of the chemical in the air times the
duration (f) of exposure to it (Jarabek, 1995). While providing a convenient metric of exposure, this assumption yields
' This article has been reviewed by the National Health and Environmental Effects Research Laboratory, U.S. Environmental Protection Agency,
and approved for publication. Mention of trade names or commercial products does not constitute endorsement or recommendation for use. Portions
of these data were presented at the Society of Toxicology annual meeting,
Anaheim CA, March 11, 1996.
2
Correspondence should be addressed to the author at Neurotoxicology
Division, MD 74B, U.S. Environmental Protection Agency, Research
Triangle Park, NC 27711. Fax: 919/541^849. E-mail: bushnell.philip®
epamail.epa.gov.
0272-0590/97 $25.00
Copyright O 1997 by the Society of Toxicology.
AU rights of reproduction in any form reserved.
30
SIGNAL DETECTION, TCE, CONCENTRATION, AND TIME
a risk assessment strategy which ignores potential risk from
short-term, high-level exposure episodes (i.e., residual risk).
The assumption that a constant effect arises from a given
C X ( product, regardless of the particular values of C and
t, was first formulated by Warren (1900) but has become
known as "Haber's Rule" after Haber's studies of nerve
gases used in World War I (Hayes, 1975). This relationship
underlies occupational exposure limits established on the
basis of the "time-weighted average" (TWA) concentrations (Atherley, 1985). In the absence of knowledge about
the dosimetry and mechanism of toxicity of a chemical, timeweighted averaging has also been used as a default approach
in estimating safe levels of ambient airborne chemicals (Jarabek, 1995). When the C X t relationship has been examined,
however, it has usually been found that C influences toxicity
more than t (Atherley, 1985; ten Berge et al, 1986).
The uncertainty associated with risk assessments that are
based upon exposure data and rely on Haber's Rule for
extrapolation across exposure durations generated the need
to examine empirically the relationship between C and t on
a relevant functional endpoint. Thus, the major objective of
this study was to evaluate whether quantitative exposureeffect modeling could improve the estimation of risk from
exposure to a VOC when dosimetric and mechanistic information are lacking. The extrapolation of concern here involved estimation of toxicity for short, high-level exposures
given data from longer, lower-level exposures. This extrapolation is of concern because monitoring data, and hence
estimates of exposure, tend to sample over relatively long
time periods (i.e., 24-hr cycles), and high-level industrial
releases of airborne chemicals tend to be shorter (on the
order of minutes) (EPA, 1995).
This evaluation was approached in two steps. First, the
response to a VOC was assessed at various combinations of
C and t designed to yield constant C X t products. A constant
response obtained at all values of C and t yielding a constant
product would support Haber's Rule and the current strategies used to assess risk. Conversely, if toxicity is not directly
related to the C X t product of an exposure, then the approach
to estimating risk from short-term, high-concentration exposure should be modified. The second step involved calculating the empirical relationship between C and t which produces a constant change in behavior (a "standard" effect).
This relationship was then compared to Haber's Rule, which
was used to provide a frame of reference against which the
observed relationship was compared.
Trichloroethylene (TCE) was chosen as the prototypic
VOC for this work because of its high volume of use, estimated at 42.5 X 10s kg in the United States in 1991 (Chemical Marketing Reporter, 1992) and its consistent appearance
in occupational and ambient atmospheres (Hughes et al,
1994; EPA, 1995). TCE has also been shown to affect neurobehavioral functions in microelectronics workers (Broadwell
et al, 1995) and to elevate auditory thresholds in humans
31
(Szulc-Kuberska et al, 1976) and in animals (Crofton et al,
1993).
Because attentional dysfunctions have been observed in
humans exposed to VOCs (Arlien-S0borg, 1992; Baelum,
1991; Benignus, 1981; Dick et al, 1984), behavioral tests
of attention may be useful for assessing the acute effects of
solvent inhalation. Animal models of some of these processes are being developed for studies in neuroscience
(Bushnell, 1995; Chiba et al, 1995), toxicology (Bushnell
et al, 1994; Bushnell and Crofton, 1997), and pharmacology
(McGaughy and Sarter, 1995; Bushnell et al, submitted for
publication).
Rats were trained to perform a signal detection task that
had been developed for characterizing the effects of VOCs
on sustained attention (Bushnell et al, 1994). Two measures
of performance of the task—sensitivity (SI) and response
time (RT)—were first examined as functions of inhaled TCE
in terms of C, t, and the C X t product. A strict application
of Haber's Rule showed that the C X t product did not
adequately predict changes in either endpoint.
Concentrations of TCE associated with standard effects
on behavior were then calculated for each subject from
functions relating the behavioral endpoints to inhaled vapor concentrations of TCE at three durations of exposure.
The "criterion concentration" for each determination was
defined as the concentration of TCE at which a fitted cubic
polynomial concentration-effect function crossed a given
"effect level." The effect level represents a significant
change from control performance, and represents an important decision point in conducting this kind of analysis
because it implies specification of an adverse effect. Effect
levels may be based upon clinical criteria for impaired
function when normative data are available for the endpoints of concern. Such criteria do not exist for the endpoints used in this study, so effect levels were denned in
the signal detection task as a 0.1-unit decrease in SI and
a 100-msec increase in RT. A 0.1-unit decrease in SI
represents 10% of the scale on this measure, which ranges
from 0 (chance performance) to 1.0 (perfect discrimination). A 100-msec increase in RT is characteristic of deficits observed in rats in a variety of attention tasks (e.g.,
Pang etal, 1993; Bushnell, 1995; Ward and Brown, 1996)
and represents a large and adverse change in studies of
human reaction time (Volz and Sturm, 1995). Both of
these effect levels were previously associated with statistically-significant decrements in signal detection during inhalation of toluene (Bushnell et al, 1994). Here, the criterion concentration for SI will be called the "SI 0 .i," i.e.,
the concentration associated with a 0.1-unit decrease in
SI; the criterion concentration for RT will be called the
" R T 1 0 0 , " i.e., the concentration associated with a 100msec increase in RT.
METHODS
Subjects. Twelve male Long-Evans rats (Charles River, Portage, ME)
were housed individually in suspended plastic cages on heat-treated pine
32
PHILIP J. BUSHNELL
shavings in a housing facility fully accredited by the American Association
for Accreditation of Laboratory Animal Care (AAALAC). As required by
AAALAC, animal care conformed to the guidelines provided by NIH.
Lighting followed a L:D 12 h r 12 hr (0600:1800) photoperiod; all behavioral
testing occurred in the light phase of the cycle. Each animal was maintained
at 350 g body weight by scheduled home cage feeding (Ralston Purina, St.
Louis, MO) after daily test sessions (Ali et al., 1992); tap water was available ad libitum in the home cage. One animal died after Series 2 of causes
unrelated to TCE exposure; its data were excluded from the study. Response
time (RT) data for one additional rat were excluded because it generated
exceptionally long values during Series 1 and 2 and normal values during
Series 3. The long RT of this rat fell almost 10 standard deviations above
the mean RT of the remaining rats and was thus a statistical outlier (p <
0.01;Dixon, 1953). Thus final group size was 11 for measures of sensitivity
and bias, and 10 for RT.
Apparatus.
Four 32.9-L test chambers were constructed of stainless
steel and glass to permit assessment of operant performance of rats breathing
controlled concentrations of solvent vapors (see Bushnell et al., 1994, for
details). The front wall of the rat's section of each test chamber contained
two retractable omnidirectional response levers, a food cup with a hinged,
clear plastic door, a house light, a signal light, and a 5-cm cone loudspeaker.
The house and signal lights were mounted 15 cm above the floor of the
chamber: the signal light was centered in the wall above the food cup,
between the house light and the loudspeaker. The retractable levers (Machine Components, Plainview, NY) were 0.32-cm-diameter stainless-steel
rods, 3.8 cm in length. Each was inserted horizontally 2.5 cm into the
rat's section of the test chamber through a 1.5-cm-diameter opening in the
stimulus-response panel. The two levers were mounted 16 cm apart, on
either side of the food cup, and 3.2 cm above the floor of the chamber.
Access to the chamber was gained by removal of the rear panel, which was
made of clear glass framed in aluminum.
Signal stimuli were generated with the incandescent signal lamp (bulb
No. 3019) by amplifying current from a 256-step digital-to-analog converter
(DAC; Model L65-28, Coulbourn Instruments, Lehigh Valley, PA). Each
lamp was adjusted to 0.30 lux with no attenuation as measured with a
photometer (Model 450, EG&G, Inc., Salem, MA) mounted on the glass
door of the test chamber, 20 cm from the signal lamp. The brightness of
the signal light was varied by adjusting the attenuation of the DAC in 1dB steps yielding signal light levels of 0.300, 0.187, 0.118, 0.072, 0.043,
0.025, and 0.014 lux with the bulb continuously on (values averaged across
the 4 test chambers). Measurements of the light emitted from each bulb
during the 300-msec signal duration used in the study revealed a linear
relationship across signal intensities between steady-state brightness (in lux)
and integrated light output from the 300-msec pulse (in lux-sec). Given
that the integration time for scotopic vision is approximately 100 msec at
absolute threshold (Barlow, 1958), the brightness measure was taken as a
better estimate of stimulus intensity applied to the retina. Each of these 7
intensities was presented during each test session against a background
illumination level from the houselight of 0.1 lux. Signals were generated
using SKED-11 software (State Systems, Kalamazoo, MI) running under
RSX-llM-plus on a PDP11/70 computer (Digital Equipment, Maynard,
MA). The same software controlled vapor concentration monitoring and
behavioral testing (see below).
Background white noise of 65 dB(A) was generated in each test chamber
by Coulbourn interface modules including a diode-based generator (S8102) and solid-state amplifier (S82-24). The noise was produced by the
loudspeaker (El2-01) and measured with a sound level meter (Brtlel &
Kjacr, Marlborough, MA, Model 2235) and a 0.5-in. microphone (Bruel &
Kjacr, Model 4133) centered in the rat's work space and directed at the
stimulus-response panel.
A remote blower pulled conditioned room air (22 ± 2°C, 60 ± 5% RH)
through each chamber at 18 L/min (1.8 min/air change). Air flow rates
were measured and controlled using laminar flow elements and gate valves
in the individual chamber inlet air lines. Internal chamber pressures were
maintained negative 2.8-3.6 cm H2O relative to the test room. Trichloroeth-
ylene (TCE) vapor was generated by metering liquid spectrophotometric
grade TCE (99.5%, Aldrich Chemical Co., Milwaukee, WI) at 1.5 mL/min
into a heated (93 ± 5°C) vertical stainless-steel J-tube (2.4 cm i.d., 71 cm
long) through which dry grade nitrogen gas flowed upward at 3 L/min
(Miller et al, 1980). The resulting TCE vapor (135,000 ppm TCE in N2) was
routed via a manifold into individual chambers using mass flow controllers
(Model FC280, Tylan General, Torrance, CA). Excess TCE vapor was
retrieved from the manifold, condensed, and discarded. The J-tube and
manifold were maintained at 48 ± 2.5 cm H2O pressure; the entire manifold,
mass flow controllers (MFCs), and delivery lines were heated to 63 ± 3°C.
The command voltage to each MFC was set by the computer via Coulbourn
DACs. Each concentration of TCE could be generated in each chamber.
During each exposure session, one (air control) chamber received only
clean air its MFC was set to zero, and a ball valve in its vapor delivery
line was closed. Chamber air oxygen concentrations were >20.1% during
all exposures.
TCE vapor concentrations were monitored using an infrared spectrophotometer (MIRAN 1A: Foxboro Co., East Bridgewater, MA). The spectrophotometer was calibrated before and after each 2-week exposure series
using a static, recirculating closed-loop method. During each exposure, the
TCE vapor concentration in each chamber was sampled sequentially in 5min periods as previously described (Bushnell et al., 1994). Voltage from
the spectrophotometer was digitized with a 256-step analog-to-digital converter (ADC: Coulboum Model SK25-08) and stored for later analysis. The
vapor flow was adjusted by the computer via the MFC to maintain or regain
the target concentration. Three samples were obtained from each chamber
during each test session. The rise time (f^) for the TCE concentration in
each chamber was about 8 min.
Signal detection task. The rats were trained to press either response
lever for 45-mg nutritionally complete food pellets (P. J. Noyes Co., Lancaster, NH) using an autoshaping-operant method in which the levers were
inserted for 15 sec and then retracted; food was delivered when either lever
was pressed or when the levers were retracted (Bushnell, 1988). Next, the
signal light was mounted 2.5 cm above one lever. On half of the trials
(signal trials), the light was illuminated for 2 sec prior to lever insertion
and remained lit until a lever was pressed. On the other half of the trials
(blank trials), the light was not illuminated. Each correct response, i.e., a
press on the lever under the light (signal response) on signal trials or a
press on the other lever (blank response) on blank trials, caused illumination
of the food cup light for 2 sec and delivery of a food pellet. After each
incorrect response, the houselight was turned off for 3 sec. Both levers
were retracted when either lever was pressed. The left lever was designated
as the signal lever in two test chambers and the right lever as signal lever
in the other two chambers. Training sessions were 100 trials in length.
Once the rats mastered this discrimination (accuracy »85% correct), the
signal light was moved to the top center of the stimulus panel (above the
food cup) and the signal duration was shortened in stages from 2 sec to
500 msec. Next, a postsignal interval (between offset of the signal and
insertion of the levers) was added: this interval was initially 0.01 sec and
was extended to 2, 3, or 4 sec (varying randomly across trials) for the final
conditions. The intertrial interval was initially 15 sec; for signal detection
testing, it was shortened to a variable value with a mean of 7 sec, a range
of 0.30 to 24.4 sec, and a constant probability of signal onset over time
(Fleschler and Hoffman, 1962). With the addition of the variable postsignal
interval, the signal could occur during a time interval ranging from 2.3 to
28.4 sec with an average of 9 sec, but was never more than 4 sec, nor less
than 2 sec, prior to insertion of the response levers. These parameters were
chosen to achieve a trial presentation rate of 5 trials/min.
For the final task, the signal duration was set to 300 msec, signals of
variable intensity were delivered (see above), the number of trials was
increased to 300 per session, and the probability of pellet delivery after
each correct response was reduced to 0.8 (with illumination of the food
cup light provided as secondary reinforcement after every correct response).
Each 70-min exposure/test session consisted of a 10-min pretest interval,
during which the TCE vapor concentration rose to its target level, followed
33
SIGNAL DETECTION, TCE, CONCENTRATION, AND TIME
TABLE 1
Schedule of Exposures to TCE Vapor and Concentrations of TCE in Air
Series 1
Series 2
Series 3
Nominal
Actual
Nominal
Actual
Nominal
Actual
Air
0 + 0
Air
0 + 0
Air
0 + 0
400
404
800
815
1600
1596
+
±
±
±
+
±
40
5
80
20
160
31
600
593
1200
1199
2000
2003
± 60
+ 12
+ 120
± 13
± 200
+ 26
1200
1210
2400
2436
2400
2409
± 120
+ 23
± 240
+ 41
± 240
± 28
Note. Each animal was tested twice at each concentration, 1 day per week, with Mondays always an air day. Nominal values refer to the target TCE
concentrations (ppm) and ranges (±10%) for each exposure; actual values show mean ± SD ppm across animals and determinations. Series 1 and 2
were administered consecutively; Series 3 was administered after a 6-week period during which the rats received other TCE exposures in the same
concentration range, but with varying durations of an hour or less. Of the 501 determinations of TCE concentration made, 3 exceeded the target range
(=el3.8%), and one fell below it (-11.8%).
by three 20-min blocks of approximately 100 trials each. Trials were presented in groups of 4 at each of the 7 signal intensities. Each trial group
contained 2 signal trials and 2 blank trials (in random order); trial groups
were also presented in random order in each 100-trial block. To reduce
somnolence during the pretest interval, food pellets were made available
during three 1-min periods, signaled by illumination of the food cup light,
on a progressive ratio schedule of openings of the food cup door.
Trichloroethylene exposure. Each rat was exposed to TCE vapor at 0
(air control), 400, 600, 800, 1200, 1600, 2000, and 2400 ppm (nominal
values; actual values are given in Table 1). Target levels were defined as
the nominal level +10%. Each rat was always tested and exposed in the
same test chamber. TCE exposures were divided into three 2-week series
(Table 1) during which each rat received air and three concentrations of
TCE vapor, one concentration per day, for four consecutive test days in each
week of each 2-week series. Exposures were systematically counterbalanced
across days and test chambers during each series. Except when necessary
to repeat an exposure from the previous week, all animals received air on
Mondays.
Data analysis: Exposure. TCE concentrations were calculated from
digitized voltage readings from the spectrophotometer using standard curves
determined at biweekly intervals. The log of the TCE concentration was
linear with respect to voltage from 200 to 2800 ppm.
Data analysis: Behavior. Nonparametric signal detection analysis was
used to calculate a sensitivity index (SI) and a responsivity index (RI) (Frey
and Colliver, 1973; Sahgal, 1987); values were calculated as described by
Bushnell et al. (1994). SI represents the ability of the subject to discriminate
signal trials from blank trials, and varies typically between 0 (chance accuracy: signals are not discriminable from blanks) and 1.0 (signals are discriminable with complete accuracy). RI provides a measure of response bias,
reflecting the subject's tendency to respond signal or blank: it varies from
— 1.0 (all responses blank) through 0 (equal numbers of signal and blank
responses) to +1.0 (all responses signal). In terms of signal detection analysis, negative bias indicates a strict criterion for responding signal (i.e., few
signal responses are given) and positive bias indicates a lenient criterion
(many signal responses are given). SI and RI were averaged across signal
intensities and calculated separately for each block (third) of each session.
Response time (RT) was measured for each response type as the time
between insertion of the levers into the chamber and the time at which a
lever press was recorded. Low error frequencies at high signal intensities
precluded calculation of latency for incorrect responses, and RTs for
hits and correct rejections were very similar; thus RT was analyzed for
hits only.
Finally, the frequency of response failure (timeout, TO) was recorded
during each test session. If the TO frequency exceeded 50, the session was
terminated. Trials lacking a response were not repeated.
SI, RI, and RT were calculated for each exposure series independently,
and averaged across repetitions of repeated conditions (i.e., all air sessions,
1200 ppm TCE in Series 1 and 2, and 2400 ppm TCE in Series 2 and 3;
see Table 1). The 600 ppm TCE condition was dropped from the analysis
to obtain equal spacing of concentrations. SI, RI, and RT were analyzed
statistically using repeated-measures analyses of variance [SAS General
Linear Model, Version 6.06 (SAS, 1990)] with TCE concentration and test
block as repeated measures. Greenhouse-Geisser degree-of-freedom (df)
corrections were used to minimize the effects of asymmetrical variancecovariance matrices; the multiplicative df correction factor (e) is reported
after each F ratio under Results. After the overall ANOVAs, matched-pair
t tests were performed to compare the effects of exposure at two C X t
products generated with different values of C and t (i.e., 800 ppm-hr obtained from 800 ppm at 1 hr, 1200 ppm at 0.67 hr, and 2400 ppm at 0.33
hr, and 1600 ppm-hr obtained from 1600 ppm at 1 hr and 2400 ppm at
0.67 hr). The criterion for significance in all tests was a = 0.05.
The concentration-effect function for each endpoint (SI and RT) from
each rat at each duration of exposure was fitted with a cubic polynomial
curve by least-squares regression using an iterative procedure (RS/1 V5.2.2:
BBN Software Products, Cambridge, MA). The criterion concentration for
each curve was then calculated as the concentration of TCE (in ppm) at
which the fitted curve crossed the appropriate effect level. For SI, the effect
level was defined as a 0.1-unit decrease in SI from the mean value generated
by that rat in its test sessions in air. For RT, the effect level was defined
as a 100-msec increase above that rat's mean RT in air. As stated in the
introduction, clinical norms have not been established for these endpoints.
Therefore, selection of these values was based upon existing data from this
and other tests of attention in rats (see above), including previous work
with toluene in this test, which demonstrated significant changes in these
measures at these effect levels (Bushnell etal, 1994). The resulting criterion
concentrations (SIo i and RTrOo) were compared across exposure durations
by one-way repeated-measures ANOVAs (SAS, 1990), followed by \-df
contrasts of the value at each shorter duration with the value for that endpoint at the 1.00-hr exposure duration.
RESULTS
Twelve of the 198 exposures to TCE in this experiment
were repeated because of problems cither with the vapor
generator or the data collection system. Of the 501 determinations of TCE concentration comprising the final set of
exposures analyzed, the target range (nominal concentration
± 10%) was exceeded three times and undershot once. Table
1 shows the mean (±SD) concentrations obtained from each
exposure series.
Response failures (TOs) increased from a mean (±SEM)
of 1.1 ± 0.7 per 300-trial session in air to 4.0 ± 1.4 in
34
PHILIP J. BUSHNELL
0.8
co 0.6
i
(0
o
CO 0 . 4
UJ
Air
400 ppm
800 ppm
1200 ppm
1600 ppm
2000 ppm
2400 ppm
S 0.2
0.0
0.33
0.67
1.00
Time (hr)
FIG. 1. (A) Sensitivity (SI) and (B) response time (RT) as functions of TCE concentration (individual lines) and time (abscissae). Each point on the
abscissa represents a 0.33-hr (20-min) period of testing within each session. Values are means (±SEM) across rats. Asterisks indicate points which differ
from air control at corresponding time points (a = 0.05).
2400 ppm TCE [F(6,60) = 3.73, e = 0.28, p = 0.051]. The
maximum number of TOs in any session was 30; thus no
sessions were excluded due to excessive response failure.
Whereas SI increased slightly across time in air, it de-
creased with increasing concentration of TCE and increasing
duration of exposure (Fig. 1 A). Effects of TCE concentration
[F(6,60) = 23.28, e = 0.49, p < 0.0001], test block [F(2,20)
= 11.65, e = 0.85, p = 0.0004], and concentration by block
0.8
CO,
i
0.6 -
0.4 UJ
Air
400 ppm
SOOppm
1200 ppm
1600 ppm
2000 ppm
2400 ppm
n 0.2 o
0.0
0
400 800 1200 1600 2000 2400
400 800 1200 1600 2000 2400
C x t Product (ppm-tir)
C x t Product (ppm-hr)
FIG. 2. (A) Sensitivity (SI) and (B) response time (RT) as functions of the product of the TCE concentration (C) and time (/). The data are replotted
from Fig. 1. C x t = 0 for all air points because C = 0. Points with matching superscripts differ by matched-pair t tests (a = 0.05) showing that, for
800 and 1600 ppm-hr, neither SI nor RT is determined by the C X t product
SIGNAL DETECTION, TCE, CONCENTRATION, AND TIME
0.33 hr
0.67 hr
35
1.00 hr
0.8
C
B
Air Mean
S 0.4 -|
co
o
W 0.2 J
Air Mean
0.6 0.4 - Effect level
Effect level
0.4 - Effect level
-1821 ppm
0.0
800
1600
• 1373ppm
*
0.2 -
0.2 -
nn
nn
2400
0
800
Air Mean
0.6 -
1600
2400
1
0
1.2
1.2
1.2
1.0 -
1.0 -
1.0 -
800
1600
2400
800
1600
2400
RT,,, • 2317 ppra
w
O 0.6
Effect tevd
Q
0.4 ^
0.2
'Air Mean
800
1600
800
2400
1600
2400
TCE Concentration (ppm)
FIG. 3. Concentration-effect functions for SI (A-C) and RT (D-F) from Fig. 1, plotted separately for each duration of exposure. The mean value
from all air exposures combined is shown as the horizontal dotted line (air mean); the mean effect level—a reduction in SI of 0.1 unit, or an increase
in RT of 100 msec—is shown as the horizontal solid line. The solid curve shows mean of the best-fitting cubic polynomial regression functions fit for
each rat through its empirically determined points. The criterion concentration (SIoi for SI and RT100 for RT) is given in each panel as the mean
concentration of inhaled TCE at which each animal's fitted curve intersected its effect level. Mean SIo , and RT100 values do not always equal the apparent
intersection of the mean function with the mean effect level, due to differences in distribution of the intercepts and the parameters of the curves.
interaction [F(12,120) = 5.60, e = 0.40, p < 0.0001] were
all significant. Analysis of the interaction showed that SI
was significantly reduced below values of about 0.52.
RI was not significantly affected by TCE concentration,
but increased across trial blocks [F(2,20) = 4.53, e = 0.86,
p < 0.024]. The concentration by block interaction was not
significant.
RT increased with increasing concentration of TCE and
with duration of exposure (Fig. IB). Effects of TCE concentration [F(6,54) = 12.55, e = 0.29, p = 0.0003], test block
[F(2,18) = 21.43, e = 0.67, p < 0.0001], and concentration
by block interaction [F(12,108) = 11.57, e = 0.22, p <
0.0001] were all significant. The significant interaction resulted from a greater increase in RT at 2400 ppm TCE than
at any lower concentration.
When plotted as functions of C X t product, both SI (Fig.
2A) and RT (Fig. 2B) showed that the effect of TCE was
not constant at a given C X t product for all values of C
and t. Thus, 800 ppm-hr was delivered after 1 hr at 800 ppm
TCE, after 0.67 hr at 1200 ppm TCE, and after 0.33 hr at
2400 ppm TCE: at this C X t product, SI was reduced
significantly only at 2400 ppm TCE (Fig. 2A), and RT was
increased only after 0.33 hr of TCE at 2400 ppm. Similarly,
1600 ppm-hr TCE was delivered after 1 hr at 1600 ppm
TCE and after 0.67 hr at 2400 ppm TCE: at this C X t
product, SI was decreased more at 2400 ppm TCE for 0.67
hr than at 1600 ppm for 1 hr (Fig. 2A), and RT was increased
significantly only after 0.67 hr at 2400 ppm TCE (Fig. 2B).
Thus concentration appears to drive the effects of TCE more
than does duration of exposure.
Mean concentration-effect functions from each block of
trials are shown in Fig. 3 for SI (top) and RT (bottom).
Table 2 describes the results of the curve-fitting calculations.
Measured at 0.33, 0.67, and 1.00 hr, respectively, the mean
values for SIo., were 2298, 1823, and 1373 ppm; for RTIOO,
the corresponding values were 2317, 2118, and 1983 ppm.
SIoi for the 0.33-hr time point differed significantly from
that of the 1.00-hr time point (F(2,20) = 9.09, e = 0.59, p
< 0.002; 1-J/contrast of 0.33-hr vs 1.00 hrF( 1,10) = 10.67,
p < 0.009; for 0.67-hr vs 1.00-hr F( 1,10) = 4.11, 0.05 <
p < 0.10). Similarly, RT.oo for the 0.33-hr time point differed
significantly from that of the 1.00-hr time point (F(2,18) =
4.77, e = 0.86, p < 0.03; l-df contrast of 0.33-hr vs 1.00
hr F(l,9) = 7.98, p < 0.02; for 0.67-hr vs 1.00-hr F(l,10)
= 2.55, p > 0.10). Thus the concentration of TCE required
to produce a constant effect on either outcome measure de-
36
PHILIP J. BUSHNELL
TABLE 2
Criterion Concentrations (ppm) for Sensitivity (Slai) and Response Time (RTioo) Calculated from Individual Concentration Effect Functions at Each Exposure Duration
Time
Endpoint/parameter
0.67 hr
1.00 hr
1823 ± 153
0.77
(0.32-0.94)
1373 ± 190
0.81
(0.46-0.97)
2118 ± 184
0.79
(0.46-0.97)
1983 ± 124
0.86
(0.54-0.99)
0.33 hr
SI
SIoi
r2 mean
r 2 range
2298 ± 202
0.63
(0.12-0.90)
RT
RT 100
r2 mean
r2 range
2317 ± 144
0.79
(0.46-0.99)
method assesses CNS processes similar to those underlying
sustained attention in humans. The present results thus demonstrate that inhalation of TCE reduces the ability of rats to
maintain attention to visual signals whose detection is required for accurate performance of this task.
The effects of TCE on SI and RT were closely related to
the vapor concentration (C) of inhaled TCE and the duration
of inhalation (t). However, Fig. 2 shows clearly that neither
effect of TCE could be predicted based upon total TCE
exposure, expressed as the C X / product. Consistent with
measures of other toxicants in other assay systems (Atherley,
1985; ten Berge et al., 1986), C appears to drive the effect
of TCE on signal detection more strongly than t.
In addition, these results provide a means to quantify the
10000
Note. Mean functions and effect levels are shown in Fig. 3. Values are
means (±SEM ppm); bold values differ significantly (p < 0.05) from the
1.0-hr duration. Goodness of fit of each cubic polynomial function fitted to
the data is indicated by the mean and range (across animals) of the proportion of the variance accounted for (r 2 ) by each regression.
creased as the duration of the exposure increased, and was
significantly different at 0.33 hr compared to 1.00 hr.
Finally, the criterion concentrations calculated from the
cubic polynomial functions in Fig. 3 were plotted in logarithmic C X t space as a function of the duration of exposure
for SIo i (Fig. 4A) and RTIOO (Fig. 4B). Based upon the 1.00and 0.33-hr time points, the slope of the observed function
for SIo i was 2.16 (i.e., the line relating log C to log t could
be described by the equation 2.16 log C + log / = 6.77).
The slope of the observed function for RT100 was 7.11 (i.e.,
the line could be described by the equation 7.11 log C +
log t = 23.43).
DISCUSSION
The reduction in sensitivity and increase in response time
during inhalation of TCE (Fig. 1) closely resemble effects
previously reported for inhalation of toluene using auditory
signals in the same task (Bushnell et al., 1994). The changes
in SI and RT have now been observed during inhalation of
solvents of two classes—an alkylated benzene and a chlorinated hydrocarbon. In addition, changes in signal detection
similar to these effects on SI have also been observed after
treatment with benzodiazepine receptor agonists (McGaughy
and Sarter, 1995), nicotinic receptor antagonists (Turchi et
al., 1995), and cholinergic and adrenergic drugs (Bushnell
et al, 1996). These studies and results of parametric and
pharmacological manipulations (McGaughy and Sarter,
1995; Bushnell and Crofton, 1997; Bushnell et al., 1996)
provide strong evidence that reductions in SI observed in
this task reflect a decrement in signal detection and that the
C x t ° 1373 ppm-hr
Calculated Sl o 1 (ppm)
0.33
0.67
1.00
Time (hr)
10000
B
a
a.
C x t " 1983 ppm-hr
o
\
C ° 1983 ppm
O
Calculated RT100 (ppm)
1000
0.33
0.67
1.00
Time (hr)
FIG. 4. Criterion concentrations calculated for (A) SI and (B) RT at
each exposure duration, plotted on log-log coordinates as a function of the
exposure time. The criterion concentrations (points connected by heavy
lines, showing means ± SEM) increased as exposure time was shortened:
for both measures, the value at 0.33 hr differed significantly from the value
at 1.00 hr (p < 0.05). Thus the assumption that time does not influence
toxicity (dashed horizontal line) is not supported by the data. However, the
criterion concentrations do not increase as fast as would be predicted by
Haber's Rule—depicted here as the function with a constant Cxt product
(light solid line)—which assumes that C and t affect toxicity equally.
SIGNAL DETECTION, TCE, CONCENTRATION, AND TIME
relationship between C and t of exposure on the toxicity
of TCE. It is clear from Fig. 4 that the observed criterion
concentrations for SI and RT for the three values of t fall
between two default assumptions regarding the relationship
between C and / of exposure and response. All three curves
are drawn to intersect the criterion concentration observed
after a 1-hr exposure (i.e., 1373 ppm for SI and 1983 ppm
for RT), to indicate a starting point for extrapolation toward
short-duration exposures. The upper (solid) lines reflect the
function based on Haber's Rule, i.e., C X t = 1373 ppm-hr
for SI and C X t = 1983 ppm-hr for RT. The lower (dashed)
lines reflect a more stringent assumption, that time has no
influence on the criterion concentration; thus C = 1373 ppm
for SI and C = 1983 ppm for RT. This latter relationship
has been proposed as a conservative default assumption in
cases where extrapolation from a given set of exposure conditions must be made to shorter exposures at higher concentration (Jarabek, 1995).
The relationship between criterion concentrations and
time shown in Fig. 4 indicates that neither assumption accurately reflects the relationship between C and t in determining the effect of inhaled TCE on either SI or RT. The assumption based upon Haber's Rule overestimates the concentration necessary to produce a given effect, and thus
underestimates the risk of short exposures when extrapolating from longer exposures. Conversely, the assumption that
time does not alter the response to TCE underestimates the
concentration necessary to produce the effect, and thus overestimates the risk of short exposures.
Calculation of criterion concentrations provides a means
to quantify the relationship between C and t in determining
the effect (£). If one generalizes Haber's Rule as E = C
X tm, one can calculate the values of m and n, the exponents
on C and t. Assuming m = 1, and given the approximately
linear functions relating log(SIo 0 and log(RT100) to log(f) in
Figs. 4A and 4B, one can derive the functions relating C
and t for a given effect level and thus obtain values of m
for Sic, and RT 100 . These calculations result in values of n
of 2.2 for Sic, and 7.1 for RTIOO.
The value of n for SI (2.2) falls within the range of those
reported by ten Berge et al. (1986), who used lethality as
the endpoint of toxicity, and obtained values of n ranging
from 0.8 to 2.7 for 9 systemically-acting compounds. It
is intriguing that the one compound which yielded n < 1
in ten Berge et al.'s analysis was TCE. This discrepancy
indicates a clear difference in mechanism between the
effects of TCE on signal detection behavior and lethality
in rats, with the latter effect being much more time-dependent than the former. In contrast, the value of n for RT
(7.1) is far larger than that obtained for lethality by ten
Berge et al., suggesting that the effects of TCE on response time—as reflecting motor output of the CNS—are
less time-dependent, and more concentration-dependent,
than changes in either SI or lethality.
37
It must be emphasized that these calculations were conducted to evaluate the roles of the concentration and duration
of inhalation exposure to a common VOC in the air. The
experiment was designed to address the C X t extrapolation
issue in risk assessment, based upon the all-too-plausible
assumption that knowledge of the toxicokinetics and dosimetry of the VOC are either unknown or too uncertain to estimate the absorbed dose of the VOC (as concentration at the
target organ). Thus these results do not address the role of
toxicokinetics in determining the outcome. Clearly, knowledge of the dosimetry of a VOC should increase the predictability of its effects. If so, then physiologically based pharmacokinetic (PBPK) models of VOC dosimetry should enable better risk assessment strategies for inhaled VOCs,
based upon estimates of absorbed dose rather than estimates
of exposure. For example, knowledge of the brain concentration of a VOC associated with attentional dysfunction in
rats, coupled with knowledge of the kinetics of that VOC
in humans, should provide a firmer basis for estimating exposure levels likely to produce similar dysfunction in humans.
In summary, this work demonstrates that inhalation of the
volatile solvent TCE impairs two important components of
the ability of animals to sustain attention to brief, unpredictable events: accuracy and speed of response. These results
also confirm previous work suggesting that the toxicity of
airborne chemicals depends upon the combined effects of
concentration and time, and that concentration plays a
stronger role in determining the strength of the response
than does time. Finally, this work quantifies the relative
importance of C and t on the response of the nervous system
to TCE; in so doing, it provides some guidance for assessing
the risk to health of short-term exposures to VOCs, under
typical exposure conditions which vary in both concentration
and time. This guidance will be most applicable when neither
the dosimetry nor the mechanism of action of the compound
is understood; it may also be useful in setting occupational
standards for VOCs with primary effects on the CNS.
ACKNOWLEDGMENTS
DTS. Vernon Benignus, William Boyes, Ila Cote, Kevin Crofton, Robert
MacPhail, Woodrow Setzer, Jane-Ellen Simmons, and John Vandenberg
provided invaluable guidance with the C x t modeling. In addition, James
Allen, E. Baker Bailey, Charles Hamm, Paul Killough, Todd Krantz, Michael McFarland, John McGee, Wendy Oshiro, Beth Padnos, Earl Puckett,
and Kaye Riggsbee cheerfully provided essential technical assistance with
this project Drs. Boyes and Annie Jarabek provided insightful reviews of
me manuscript.
REFERENCES
Ali, J. S., Olszyk, V. B., Dunn, D. D., Lee, K. A., Kendall, S. M., Rhoderick, R. R., and Bushnell, P. J. (1992). A Lotus 1-2-3-based system for
recording and maintaining body weight of laboratory animals. Behav.
Res. Methods Irutrum. Computers 24, 82-87.
Arlien-S0borg, P. (1992). Solvent Neurotoxicity. CRC Press, Boca Raton,
FL.
38
PHILIP J. BUSHNELL
Atherley, G. (1985). A critical review of time-weighted average as an index
of exposure and dose, and of its key elements. Am. lnd Hyg. Assoc. J.
46,481-487.
Baelum, J. (1991). Human solvent exposure: Factors influencing the pharmacokinetics and acute toxicity. Pharmacol. Toxicol. 68(S1), 1-36.
Barlow, H. B. (1958). Temporal and spatial summation in human vision at
different background intensities. J. Physiol. 141, 337-359.
Benignus, V. A. (1981). Health effects of toluene: A review. NeuroToxicology 2, 567-588.
Broadwell, D. K.., Darcey, D. J., Hudnell, H. K., Otto, D. A., and Boyes,
W. K. (1995). Work-site clinical and neurobehavioral assessment of solvent-exposed microelectronics workers. Am. J. lnd. Med. 27, 677-698.
Bushnell, P. J. (1988). Behavioral effects of acute p-xylene inhalation in
rats: Autoshaping, motor activity, and reversal learning. Neurotoxicol.
Teratol. 10, 569-577.
Bushnell, P. J. (1995). Overt orienting in the rat: Parametric studies of cued
detection of visual targets. Behav. Neurosci. 109, 1095-1105.
Bushnell, P. J., and Crofton, K. M. (1997). Neurobehavioral toxicology of
organic solvents. In Behavioral Toxicology and Addiction (R. J. M. Niesink and R. Jaspers, Eds.), Open Univ. of The Netherlands, in press.
Bushnell, P. J., Kelly, K. L., and Crofton, K. M. (1994). Effects of toluene
inhalation on detection of auditory signals in rats. Neurotoxicol. Teratol.
16, 149-160.
Bushnell, P. J., Oshiro, W. M., and Padnos, B. K. (1996). Effects of cholinergic and adrenergic drugs on visual signal detection in rats. Soc. Neurosci. Abstr. 22, 437.
Chemical Marketing Reporter (1992). Chemical profile: Trichloroethylene.
Chem. Mark. Rep. February 13.
Chiba, A. A., Bushnell, P. J., Oshiro, W. M., and Gallagher, M. (1995).
Altered selective attention in rats with cholinergic lesions of the substantia
innominata. Neurosci. Abstr. 21, 937.
Crofton, K. M., and Zhao, X. (1993). Mid-frequency hearing loss in rats
following inhalation exposure to trichloroethylene: Evidence from reflex
modification audiometry. Neurotoxicol. Teratol. 15,413-423.
Crump, K. S. (1984). A new method for determining allowable daily intakes. Fundam. Appl. Toxicol. 4, 854-871.
Dick, R. B., Setzer, J. V., Wait, R., Hayden, M. B., Taylor, B. J., Tolos, B.,
and Putz-Anderson, V. (1984). Effects of acute exposure of toluene and
methyl ethyl ketone on psychomotor performance. Int. Arch. Occup.
Environ. Health 54, 91-109.
Dixon, W. J. (1953). Processing data for outliers. Biometrics 9, 74.
EPA (1994). Methods for Derivation of Inhalation Reference Concentrations and Application of Inhalation Dosimetry. EPA Report EPA/600/890/066F. Office of Health and Environmental Assessment, Environmental
Criteria and Assessment Office, Research Triangle Park, NC.
EPA (1995). National Air Pollution Emission Trends, 1900-1994. EPA
Report EPA-454/R-95-011, Office of Air Quality Planning and Standards,
Research Triangle Park, NC.
Fleshier, M., and Hoffman, H. S. (1962). A progression for generating
variable-interval schedules. J. Exp. Anal. Behav. 5, 529-530.
Frey, P. W., and Colliver, J. A. (1973). Sensitivity and responsivity measures for discrimination learning. Learn. Motiv. 4, 327—342.
Hayes, W. J., Jr. (1975). Toxicology of Pesticides, p. 51. Williams & Wilkins, Baltimore, MD.
Hughes, K., Meek, M. E., and Windle, W. (1994). Trichloroethylene: Evaluation of risks to health from environmental exposure in Canada. Environ.
Carcinogen. EcotoxicoL Rev. C12, 527—543.
Jarabek, A. M. (1995). Consideration of temporal toxicity challenges current
default assumptions. Inhal. Toxicol. 7, 927-946.
McGaughy, J., and Sarter, M. (1995). Behavioral signal detection in rats:
Task validation and effects of age, amphetamine, and benzodiazepine
receptor ligands. Psychopharmacology 117, 340—357.
Miller, R. R., Letts, R. L., Potts, W. J., and McKenna, M. J. (1980). Improved methodology for generating controlled test atmospheres. Am. lnd
Hyg. Assoc. J. 41, 844-846.
Office of Technology Assessment (OTA) (1990). Neurotoxicity: Identifying
and Controlling Poisons of the Nervous System. OTA-BA^t36. U.S. Congress, U.S. Government Printing Office, Washington, DC.
Pang, K., Williams, M. J., Egeth, H., and Olton, D. S. (1993). Nucleus
basalis magnocellularis and attention: Effects of muscimol infusions.
Behav. Neurosci. 107, 1031-1038.
Sahgal, A. (1987). Some limitations of indices derived from signal detection
theory: Evaluation of an alternative index for measuring bias in memory
tasks. Psychopharmacology 91, 517-520.
SAS (1990). SAS/Stat, Version 6, 4th ed., SAS Institute, Cary, NC.
Szulc-Kuberska, J., Tronczynska, J., and Latkowski, B. (1976). Otoneurological investigations of chronic trichloroethylene poisoning. Minerva
Otorhinolaryngol. 26, 108-112.
ten Berge, W. F., Zwart, A., and Appleman, L. M. (1986). Concentrationtime mortality response relationship of irritant and systemically acting
vapours and gases. J. Hazard. Mater. 13, 301-309.
Turchi, J., Holley, L. A., and Sarter, M. (1995). Effects of nicotinic acetylcholine receptor ligands on behavioral signal detection in rats. Psychopharmacology 118, 195-205.
Volz, H. P., and Sturm, Y. (1995). Antidepressant drugs and psychomotor
performance: A review. Neuropsychobiology 31, 146—155.
Ward, N. M., and Brown, V. J. (1996). Covert orienting of attention in the
rat and the role of striatal dopamine. J. Neurosci. 16, 3082-3088.
Warren, E. (1900). On the reaction of Daphnia magna (Straus) to certain
changes in its environment. Q. J. Microsc. Sci. 43, 199-224.