Вы находитесь на странице: 1из 7

Toxicology

Letters
ELSEVIER Toxicology Letters 82/83 (1995) 203-209

Neurobehavioural test batteries: current status, critical


evaluation and new directions

Ann M. Williamson
National Inmtute of Occupational Health and Safety (Worksafe Australra). PO Box 58. Syciney. Austraha. 2001

Abstract

Neurobehavioural techniques are currently making a contribution to knowledge in toxicology which is being used
to set safety limits of exposure to hazards. This, however, puts pressure on researchers in the field to ensure that the
methods used are the best and most informative available. A review of the neurobehavioural tests currently in use
shows that there are a number of issues that need to be addressed to strengthen and increase their utility. Issues
include test selection and the validity, reliability and sensitivity of tests. bias and confounding of measurement in
the testing session and problems in interpreting the results.

Keywords: Neurobehavioral methods; Behavior

1. Introduction 2. Problems of making measures of behaviour

In the assessment of effects of toxic exposure,


2.1. Informed consent
neurobehavioural testing is beginning to come of
age. Increasingly, studies are using neuro-
behavioural techniques to look at the effects of Increasingly, due to ethical considerations,
exposure to potential toxicants for an increasing neurobehavioural toxicity testing requires in-
range of exposures. Despite the very healthy formed consent from study participants. This can
growth, the area is in danger of slowing and bias behaviour measurement through selecting
stagnating because of an apparent tendency to out only volunteers. To avoid such bias, where
avoid facing some critical issues mainly relating possible testing should be mandatory even
to problems of confounding of test measurement. though this often contravenes other ethical prac-
Now that the field is starting to mature, it is tices. Alternatively, we need to establish that
imperative that we tackle some of the issues in a non-participants do not differ from partici-
systematic fashion. This paper focuses on the pants on any important variables. For example,
reasons why we may not be getting accurate only workers who have relatively low exposure
measures of behaviour change in the face of toxic to a hazard may volunteer, so biasing assess-
exposure and why our interpretations of behav- ments of the neurobehavioural function of the
iour change may sometimes be misleading. group.

037%4274/95/$09.50 @ 1995 Elsevier Science Ireland Ltd. All rights reserved


SSDI 0378-4274(95)03556-Z
204 A.M. Williamson I Toxcology Letters 82183 (1995) 203-209

2.2. Effects of motivation to perform their faking could not be detected. Faking sub-
jects produced poor performance consistent with
Motivational effects can take 2 opposite forms, the nature of instructions on how to fake but
subjects trying to do very well (faking good) and monetary reward by contrast did not produce
trying to do poorly (faking bad). Faking good any additional negative effects on performance.
may be less of a problem in testing for toxic A few techniques have been used to assess the
exposure effects where the potential for com- possible confounding effects of motivation level.
pensation payments is a possible reason for The overall approach is conservative in attempt-
changing motivational state. Even so, in many ing to detect individual subjects who may be
situations some workers will try harder than judged as ‘faking’. Techniques include methods
others, with consequent effects on test perform- such as forced-choice tests [5], examining the
ance. It must be remembered, however, that all pattern of test performance such as variations in
behaviour requires a certain level of motivation the speed-accuracy trade-off [3] and using dis-
or incentive to generate it. This is particularly so criminant function analysis [6]. A method which
for tests involving memory or attention. has not been used in the assessment of neuro-
Research on motivation suggests, however, toxic action involves Signal detection methodolo-
that raising motivational level may not actually gy [7]. This t ec h nique differentiates sensitivity to
produce large changes in performance. For ex- detect the signal from the subject’s bias or
ample, monetary incentives improved informa- criterion which is their expectation about the
tion processing performance to a greater extent likelihood of the correct response occurring.
than did practice on the tests but this effect was Measures of sensitivity are likely to remain
only seen for a composite measure of perform- constant over situations where a subject’s criter-
ance [l]. Incentives did not change the relation- ion setting might be affected by such factors as
ship between performance on information pro- external motivation. Consequently the method
cessing tests and intelligence tests, indicating that would be extremely useful for both detecting the
manipulating the emotional state of the indi- existence and the extent of the motivation ef-
vidual did not result in changes in mental capaci- fects.
ty. For some tests, trying harder will not improve
performance much and can even make perform-
ance worse. Trying harder on a relatively easy 2.3. Classical confounders
task is simply not possible [2]. The effects of
trying hard can sometimes be seen, however, in 2.3.1. Test situation
tasks in which a speed-accuracy trade-off can A large number of influences in the test
occur [3]. situation have been shown to affect performance.
The opposite end of the motivation spectrum, All relate fundamentally to the general state of
faking bad or malingering, is often raised as a the individual. Where the factor makes significant
serious difficulty for the validity of neuro- changes to the individual’s level of arousal,
behavioural testing, because there is often con- resulting in either too little or too much, per-
siderable payoff for occupationally or environ- formance would be expected to suffer [8]. Evi-
mentally exposed individuals to make the most dence is available that factors such as how hot or
of, or exaggerate, any effects of a neurotoxicant. cold it is, the effects of background noise, the
In an investigation of the relationship between lighting level and the time of day can effect
monetary incentives and poor performance [4], performance [8]. While most studies attempt to
subjects were given varying levels of information control for these factors by holding them con-
about how to fake poor performance on a stant or by counterbalancing, test situations such
forced-choice memory test and studied test per- as those based in industry settings may make
formance under circumstances where motivation such actions difficult or impossible. Clearly varia-
to fake was manipulated by monetary reward if tions in the test situation must not be overlooked
A.M. Williamson I Tox~olog~ Letters 82lS.Z (199.f) X-209 205

as they can have relatively large effects on and the extent to which they use alcohol. While
performance. most studies attempt to control for these factors,
they often do so without establishing either the
23.2. Testing procedures extent of the influence on test performance or
Factors due to the test procedure may also be that the factors actually exert unwanted influ-
important. For example, the importance of ences on the tests in question over the age-range
procedural rigour was demonstrated in a study of of the study group. Not all functions nor all
the covariates of peripheral nerve function using aspects of functions are affected by age, unless
vibrotactile and thermal thresholds and nerve the age differences are very great. For example,
conduction velocity and amplitude [9,10]. For all it appears that motor performance slows and
measures, the examiner emerged as a major becomes more variable with age [13], but mem-
covariate, with the difference between examiners ory capacity may not [14]. Factors like practice
being as much as one standard deviation in and physical training have also been shown to
magnitude for some measures. It is not clear why minimise the age-related differences between old
the examiner exerted such a large influence on and young adults [13]. Studies showing differ-
these measures as they differed in the degree of ences in motor performance due to the age-
subjectivity they involve, however these studies related decline compare very great differences in
strongly suggest that measures should be as age, with the older group typically in the 60 plus
objective as possible. Methods such as comput- age group. Most studies of neurobehavioural
erised testing, standardised protocols and interra- effects involve subjects who are considerably
ter reliability need to be brought into play here younger, particularly if focusing on exposures at
to avoid these sorts of problem. the workplace. Consequently, age-related differ-
ences in function may not be a significant prob-
2.3.3. Effects of repeated exposure to the test lem in many studies.
It is well known that repeated exposure to a These issues are also relevant to other as-
test can influence performance through either the sumed confounding or nuisance variables. Taking
effects of practice or fatigue. For many tests the action to remove the effects of these variables
effects of practice or of fatigue are simply not may not simply be a waste of time, particularly if
known. Test performance can be changed sig- post-hoc statistical methods are used. If the
nificantly due to both practice and fatigue. Evi- assumed confounding variable is significantly
dence shows that a single repetition of a test can correlated with exposure to a neurotoxicant, as is
produce a practice effect which is sufficient to often the case with factors like age, education
counter the effects of 20 or more years of ageing level, and alcohol use, attempts to reduce the
[ll]. For all tests. asymptotic levels of perform- effect of the variable can also reduce the power
ance are reached before measures are taken to to detect a difference between exposed and non-
represent stable performance [12], and taking exposed subjects [15]. It is important to analyse
care to avoid long periods of testing. However the relationships between confounding variables
the time taken to benefit from practice and to and both dependent and independent variables
show fatigue-related deterioration in perform- before designing a study.
ance can be used as a useful measure of toxic
effect. The signal detection method described
above would also be useful for establishing the 3. Problems of making particular types of
effects of fatigue on performance. behavioural measures

2.3.4. Characteristics of individuals Particular types of measures may have par-


Differences between individuals are most often ticular problems or confounders associated with
raised as potential confounding factors. These them. Standard neuropsychological tests are
include the person’s age, their level of education, often used in studies because there is information
206 A.M. Williamson I Toxicology Letters 82183 (1995) 203-209

on their validity and reliability [16,17]. The and intertrial correlations. The measures used in
detection of neurotoxic behavioural effects, how- these batteries have the strength of having
ever, places a number of constraints on the use known, high reliabilities available before the
of both standard and novel tests. measures are used to any great extent. Kantowitz
[20] makes the point, however, that this approach
does not deal with the issue of validity or
3.1. Expectations of a test
uniqueness of measures. He questions the com-
mon practice of classifying measures into per-
In many assessment situations the perceived
ceptual, motor or cognitive domains, in absence
aim or face validity of a test may have an implicit
of any theory to guide test or measure selection.
influence on test performance for some indi-
He argues that functional groupings are some-
viduals. For example, the motivation to perform
what arbitrary, with little other than expert
the test may depend on the subject’s expectations
judgement to support them, and they often
from his/her experience of the test. Factors like
overlook the range of functions required to
the title of the test and the level of complexity of
perform each test and consquently hamper inter-
the test (perceived as too easy or too difficult),
pretation of test results. For example, most tests
may cause subjects to change their performance
in neurobehavioural toxicology rely on visual
in response to their expectations about any
stimulation, a fact that is usually overlooked in
consequences of particular performance on a
interpreting study results. If the toxicant does
test. The extent that these features influence
have effects on the visual system, these could be
performance is not understood. Clearly situation-
fallaciously interpreted as deficits in another
al factors may also play a role, but at the present
functional domain such as motor performance or
time we do not know how big a role. There is a
memory. It has been demonstrated that visual
need for some careful research on the influence
impairments can indeed adversely affect per-
of these implicit features on performance includ-
formance on tests which use vision as a vehicle to
ing the face validity of tests.
deliver test stimuli to other parts of the cortex
WI.
3.2. Test representation and uniqueness A number of authors have put forward an
alternative approach [l&20,22,23] maintaining
Two essential features of good measurement that the field needs tests that relate to an overall
are the extent to which the measures generated model or theoretical framework. Often the
are reliable and valid. Good test reliability is choice of tests is defended on the grounds that
particularly important because errors of mea- they are a pragmatic representation of what
surement can obscure the relationship between functions might be affected based on previous
dependent (test performance) and independent research. Kantowitz [20] in contrast maintains
(exposure) variables. Test developers and users that in selecting human performance tests ‘theory
in this area have used tests regardless of the is the best practical tool’.
reliability of tests and measures. For example,
one study [18] demonstrated that test-retest
reliabilities were only moderate for a large num- 3.3. Sensitivity of tests
ber of measures from the Neurobehavioral
Evaluation System (NES) [19], one of the most Tests need to be sufficiently sensitive to detect
commonly used test batteries. This has implica- neurotoxic effects where they exist. Tests with
tions for the utility of this test system in many high reliability and good validity will help to
settings. In contrast, the approach used in the reduce the error variance so as to reveal effects
development of the automated performance test due to the independent variable, exposure. Tests
system (APTS) [12] selected tests on the basis of from traditional batteries may appear to be
their stability over practice for means, variances useful as they usually have good psychometric
A.M. Williamson I Toxicology Letters 82183 (1995) 203-209 207

qualities. They are, however, designed to assess 4. Problems of the interpretation of behaviour
clinically affected individuals, and are not chal-
lenging enough or do not have a large enough Issues also exist around the interpretation of
range of measurement to pick up deficits before the results of neurobehavioural tests. These
they become either irreversible or too limiting to issues mainly stem from problems due to the
the individual’s function. Alternative test meth- nature of the study design and the consequent
ods for performance testing should be considered analysis of test results.
including more complex tests such as dual tasks
and grammatical reasoning, self-paced testing 4.1. Purpose of the study
[24], adaptive testing [25] and signal detection
theory [ll]. Studies of neurobehavioural effects are con-
The issue of sensitivity in this area is rather ducted with 2 main purposes, to screen or estab-
like walking a tightrope. From the view point of lish whether particular individuals are effected by
interpretation, there is a limit to increasing exposure and to diagnose or determine the
sensitivity of tests. Tests can become too sensi- nature of the effect. It is argued that these 2
tive when they detect differences between in- purposes have implications for the inclusion of
dividuals due to irrelevant factors such that the different types of tests in a battery [29]. Tests for
effects of the independent variable can be over- screening need to be suitable for detecting the
shadowed. The real issue here is the acceptable specific effect of the toxicant. Diagnosis, on the
level of concern. When is the size of the exposure other hand, requires a much more comprehen-
effect large enough to require action? Simply sive battery in order to look at the range of
showing that exposed groups are different from possible effects of the substance. The problem is
non-exposed may not be sufficient in practical that the diagnostic purpose really needs to be
terms if the size of the difference is not taken done first and done well in order to establish the
into account. Choosing a criterion for judgement best tests for screening individuals. The effective-
is not easy. Some authors suggested that the ness of screening depends to a great extent on
effects of aging constitute a good criterion the effectiveness of diagnosing what are the real
[26,27]. Other alternatives include effects on effects of the toxicant. Too often, in this field, the
daily functioning and quality of life [15] and diagnostic work is largely equivocal or does not
methods to assess delayed neurotoxic effects clarify which specific effects are characteristic of
P31. the neurotoxicant. The screening approach is
In a few instances researchers have tried to taken too early. As a consequence it leads to
anchor the size of performance deficits in terms poor screening and is probably one of the
of the effects of an agent which is known and reasons for the lack of resolution about the
accepted to be neurotoxic. For example, 2 groups effects of particular neurotoxicants such as or-
have developed dose-response curves for the ganic solvents.
effects of alcohol on performance in their test
batteries [12,28]. In both cases the aim was to 4.2. Individual differences in basic capacity
create a form of calibration or criterion validity
on which effect sizes of various exposures could This is an issue because studies in neuro-
be judged. It does, however, still beg the ques- behavioural toxicology tend to be cross-sectional
tion of when we should be concerned. Should it in design. Attempts to match exposed and con-
be at the equivalent of 0.05 blood alcohol, or one trol groups on confounding variables may not be
night without sleep? The issue of practical conse- very successful. A variety of methods have been
quences of exposure effects which are picked up used to control for baseline intellectual capacity,
by sensitive tests is an important one. We need to including pre-existing intelligence or vocational
pay attention to it if neurobehavioural testing is test results taken before any toxic exposure had
to be truly useful in toxicology. occurred [30], the use of monozygotic twins [31]
208 A.M. Williamson I Toxico1og.v Letters 82183 (1995) 203-209

prospective study designs in which baseline non- References


exposed performances are measured and com-
pared to performances at intervals over a period [l] Larson, G.E., Saccuzzo. D.P. and Brown, J. (1994)
of exposure years [32] and through use of mea- Motivation: cause or confound in information process-
sures taken currently but judged to represent ing/intelligence correlations? Acta Psychol. 85. 25-37.
[2] Kahneman. D. (1973) Attention and Effort. Prentice-
basic intellectual capacity such as the vocabulary Hall, Englewood Cliffs, New Jersey.
test from the WAIS [33]. The first 2 methods can [3] Williamson, A.M., Clarke, B. and Edmonds, C.W. (1989)
be difficult to achieve so have not been used very The influence of diving variables on perceptual and
frequently in neurobehavioural toxicology cognitive functions in professional shallow water
(abalone) divers. Environ. Res. 50. 93-102.
studies. The last method is used much more
[4] Martin, R.C., Bolter, J.F.. Todd, M.E.. Gouvier, W.D.
often, but presents some particular problems as it and Niccolls, R. (1993) Effects of sophistication and
is very likely that the measures of basic capacity motivation on the detection of malingered memory
taken after exposure may also be affected by performance using a computerized forced-choice task. J.
exposure [34]. In fact it is possible that any Clin. Exp. Neuropsychol. 15(6), 867-880.
[5] Pankratz, L. (1983) A new technique for the assessment
measure of intellectual capacity may be affected
and modification of feigned memory deficit. Percept.
by the neurotoxicant so making it impossible to Motor Skills 57, 367-372.
use these measures to show prior baseline func- [6] Bernard, L.C.. McGrath, M.J. and Houston. W. (1993)
tioning once exposure has already occurred. Discriminating between simulated malingering and
closed head injury on the Wechsler Memory Scale-Re-
vised. Arch. Clin. Neuropsychol. 8. 539-551.
[7] Green. D.M. and Swets. J.A. (1966) Signal Detection
5. Conclusions Theory and Psychophysics. Wiley, New York.
[8] Hockey, G.R. (1986) Changes in operator efficiency as a
function of environmental stress, fatigue and circadian
Neurobehavioural testing is at an important rhythms. In: K.R. Boff, L. Kaufman and J.P. Thomas
place in its development. It is now clearly estab- (Ed.), Handbook of Perception and Human Perform-
lished and accepted as a methodology which can ance: Cognitive Processes and Performance,Vol. II, John
add to the body of knowledge about how toxic Wiley & Sons. New York, pp. 44-49.
[9] Letz. R. and Gerr. F. (1994) Covariates of human
exposure can affect health and safety. Unfor-
peripheral nerve function. I. Nerve conduction velocity
tunately, however, there are many loose ends to and amplitude. Neurotoxicol. Teratol. 16(l), 95-104.
this development. Many issues are unresolved [lo] Gerr, F. and Letz. R. (1994) Covariates of human
and are increasingly being ignored in favour of a peripheral nerve function. II. Vibrotactile and thermal
pragmatic approach, ‘let’s get measuring’. Now is thresholds. Neurotoxicol. Teratol. 16(l). 105-112.
[ll] Welford, A.T. (1992) Psychological studies of aging:
the time, however, for us to reassess our tools
then origins, development and present challenge. Int. J.
and to attempt to build up the strengths and Aging Hum. Dev. 34(3), 185-197.
work on the weaknesses. Issues associated with [12] Turnage. J.J. and Kennedy, R.S. (1992) The develop-
the choice of tests, reliability, validity, sensitivity ment and use of a computerized human performance
and the role of confounders must be addressed test battery for repeated-measures applications. Hum.
Perf. 5(4), 265-301.
before the field can move on.
[13] Stelmach, G.E. and Nahom, A. (1992) Cognitive-motor
abilities of the elderly driver. Hum. Fact. 34(l), 53-65.
[14] Salthouse, T.A. (1990) Influence of experience on age
differences in cognitive functioning. Hum. Fact. 32(5),
Acknowledgements 551-569.
[15] Gullion. C.M. and Eckerman, D.A. (1986) Field testing.
I am grateful to Deborah Biancotti for her In: Z. Annau (Ed.), Neurobehav. Toxicol., Johns Hop-
assistance in producing this paper. The views kins Press, Baltimore.
[16] Lezak. M.D. (1983) Neuropsychological Assessment.
expressed in this paper are those of the author
2nd Ed., Oxford University Press, New York.
and do not necessarily reflect those of the Na- [17] Wechsler. D. (1981) Manual for the Wechsler Adult
tional Occupational Health and Safety Commis- Intelligence Scale - Revised. The Psychological Corpo-
sion. ration. New York.
A.M. Williamson I Toxicology Letters 8218.3(1995) 20.3-209 NY

[18] Arcia. E. and Otto, D.A. (1992) Reliability of selected sures of cogmttve functioning. Neurobehav. Toxmol.
tests from the Neurobehavioral Evaluation System. Teratol. 7, 345350.
Neurotoxrcol. Teratol. 14, 103-l IO. [28] Allen, R.W., Stein, A.C. and Mrller, J.C. (1990) Per-
[lY] Baker. E.L. and Letz, R. (1986) Neurobehavioural formance Testmg as a Determmant of Fitness-For-Duty.
testing in monitoring hazardous workplace exposures. J. Paper presented at the Aerospace Technology Confcr-
Occup. Med. 28( 10). 987-990. ence and Exposition. Long Beach. California
[20] Kdntowitz. B.H. (1992) Selecting measures for human [29] Johnson. B L. (1987) Prevention of Neurotoxrc Illness in
factors research. Hum. Fact. 34, 387-398. Working Populations. Wiley, Chrchester.
[21] Kempen. J.H., Krttchevsky. M. and Feldman. S.T. [30] Hanninen. H.. Eskelmen, L.. Husman, K. and Nur-
(1994) Effect of visual impairment on neuropsychologi- minen. M. (1976) Behavioral effects of long-term expo-
cal test performance. J. Clm. Exp. Neuropsychol. 16(2). sure to a mixture of organic solvents Stand J. Work
223-231. Env. Health 4. 240-255.
[22] Smith. P.J. and Langolf. G.D. (1981) The use of Stern- [31] Hannmen. H.. Antti-Polka. M. Juntunen. J. and Kos-
berg’s memory-scanmng paradigm in assessing the ef- kenvuo, M. (1991) Exposure to orgamc solvents and
fects of chemical exposure. Hum. Fact. 23. 701-708. neuropsychologrcal dysfunction: a study on monozygotrc
[23] Wtlliamson. A.M. (1990) The development of a neuro- twms. Br. J. Ind. Med. 48. 18-25.
behavroral test battery for use m hazard evaluatrons m [32] Williamson, A.M. and Winder. C (1993) A prospective
occupational settings. Neurotoxicol. Teratol. 12, SO9- cohort study of the effects of solvent exposure. Env.
51-t. Rcs. 62. 256-371.
[24] Scerbo. M.W., Greenwald. C.Q. and Sawain, D.A. (1993) [33] Hooisma, J.. Hanninen. H , Emmen. H.H. and Kuhg.
The effects of subject-controlled pacing and task type on B.M. (1993) Behavioural effects of exposure to orgamc
sustamed attention and sublective. J. Gen. Psychol. solvents in Dutch painters. Neurotowicol. Teratol 15,
120(3). 2933307. 397-406.
[2S] Werss, D J. and Vale, C.D. (1987) Adaptrve Testmg. [34] Baker, E.L.. White. R.F., Pothrer. L.J., Berkley. CS..
Appl. Psychol. Int. Rev. 36, 249-262. Dmse. G E , Travers. P.H.. Harley. J.P. and Feldman.
[26] Weiss, B. (1983) Behavioral toxicology and environmen- R.G. (1985) Occupatronal lead ncurotoxtctty: rmprove-
tal health science: opportumty and challenge for psy- ment m behavroural effects after reduction of exposure.
chology. Am Psychol. 38, 1174-1187. Br. J. Ind. Med. 42, 5077516.
[27] Smith, P.J. ( 1985) Neurobehavioural toxicology: mea-

Вам также может понравиться