Document DMX6aM2JwRQvOjD5zw2wMn1N

British Journal of Industrial Medicine 1985;42: 579-590 Behavioural evaluation of workers exposed to mixtures of organic solvents N A MAIZLISH,' * G D LANGOLF,2 L W WHITEHEAD,3 L J FINE,3 J W ALBERS,4 J GOLDBERG,2 AND P SMITH5 From the Department of Epidemiology,I Centre for Ergonomics,2 and Department of Environmental and Industrial Health,3 University of Michigan, Department of Neurology,4 University of Michigan Medical Center, Ann Arbor, Ml 48109, and Department of Industrial and Systems Engineering,5 Ohio State University, Columbus, OH 43210, USA ABSTRACT Reports from Scandinavia have suggested behavioural impairment among long term workers exposed to solvents below regulatory standards. A cross sectional study of behavioural performance was conducted among printers and spray painters exposed to mixtures of organic solvents to replicate the Scandinavian studies and to examine dose-response relationships. Eligible subjects consisted of 640 hourly workers from four midwestern United States companies. Of these, 269 responded to requests to participate and 240 were selected for study based on restrictions for age, sex, education, and other potentially confounding variables. The subjects tested had been employed on average for six years. Each subject completed an occupational history, underwent a medical examination, and completed a battery of behavioural tests. These included the Fitts law psychomotor task, the Stroop colour-word test, the Stemnberg short term memory scanning test, the short term memory span test, and the continuous recognition memory test. Solvent exposure for each subject was defined as (1) an exposed or non-exposed category based on a plant industrial hygiene walk-through and (2) the concentration of solvents based on an analysis of full shift personal air samples by gas chromatography. The first definition was used to maintain consistency with Scandinavian -studies, but the second was considered to be more accurate. The average full shift solvent concentration was 302 ppm for the printing plant workers and 6-13 ppm for the workers at other plants. Isopropanol and hexane were the major components, compared with toluene in Scandinavian studies. Performance on behavioural tests was analysed using multiple linear regression with solvent concentration as an independent variable. Other relevant demographic variables were also considered for inclusion. No significant (p > 0.05) relation between solvent concentration and impairment on any of the 10 behavioural variables was observed after controlling for confounding variables. Exposed/non-exposed com- parisons showed a significantly poorer digit span among those exposed, but this has not been generally reported in the Scandinavian studies. The medical examination showed no abnormalities of clinical significance. The inability to replicate the findings of the Scandinavian studies could have been due to the shortness of the duration of workers' exposure, the type of solvents in the mixtures, use of different behavioural tests, or to selection factors. The central nervous system is the primary target organ for inhaled organic solvent vapours and an impairment of central nervous system function and neuraesthenic symptoms are commonly reported *Present address: Occupational and Environmental Health Unit, University of California, Davis, California 95616. Received 12 November 1984 Accepted 7 January 1985 after short term, high level exposure. At low levels of exposure, evidence of frank neurotoxicity may be preceded by subtle, subclinical changes in behaviour or psychological function.' 2 Since the 1970s a substantial effort has been devoted to evaluating subclinical behavioural impairment among occupational groups exposed to solvents.3 Sensory perception, motor function, and higher cognitive abilities have been studied using 579 580 Maizlish, Langolf, Whitehead, Fine, Albers, Goldberg, and Smith performance tests adapted from clinical neurology, experimental psychology, intelligence testing, and ergonomics. Laboratory and field studies with short term exposure to relatively pure styrene,4 5 toluene,6 and other solvents5 have shown a slowing of reaction time at concentrations close to threshold limit values (TLVs). Cross sectional epidemiological studies of long term spray painters exposed to mixtures of organic solvents have shown impairments in reaction time, spatial reasoning, visual memory, and psychomotor skills.'8 These studies included exposure levels of less than one third of the TLV for mixtures and raise the possibility that synergism or long term exposure may have adverse effects on a wide range of higher cerebral functions. An important limitation of previous epidemiological studies has been the failure to show that the severity of impairment increases with dose.7-'2 Dose-response is an important criterion for establishing causality'3 and for providing a quantitative basis for risk assessment and standard setting. The objective of the present study was to replicate previous cross sectional epidemiological studies and to examine dose-response effects with quantitative exposure data. Materials and methods SELECTION OF STUDY SITES AND SUBJECTS Subjects were chosen from large midwestern United States companies engaged in manufacturing automotive parts or furniture and printing companies belonging to a regional trade association. Four employers (referred to as plants 1, 2, 3, and 4) agreed to participate (table 1). Investigators were unrestricted in the selection of subjects at plants 2, 3, and 4, but were restricted to self selected volunteers at plant 1. A cross sectional sample of subjects was selected from among hourly workers in a multistep process. Firstly, an industrial hygiene walk-through was conducted at each plant to determine which current job classifications were exposed or non-exposed to solvents (table 1). Secondly, each worker in the study population was assigned a status of "exposed" or "non-exposed," based on the walk-through survey and job classification listed on employee rosters. Thirdly, all, or a sample of exposed, were selected from employee rosters at plants 2, 3, and 4 and from sign-up sheets of volunteers at plant 1. At plants 1, 2, and 3, non-exposed were chosen from employee rosters (or sign-up sheets at plant 1) by matching an exposed subject's plant, sex, age (+ 5 years) and educational level. Subjects at plant 4 were matched only on sex because of the limited number of nonexposed subjects. Subjects were then notified that they had been selected and that they would be paid to take part. They responded by informing their supervisors or the investigators if they wished to do so. Subjects who refused to take part or were otherwise unavailable for testing were replaced with other matched subjects when possible. Of 640 workers in the sample, 269 responded and 240, including 28 replacements, signed consent forms (table 2). A total of 124 were classified as exposed and 116 as non-exposed. The low response rate at plant 1 probably reflected management's requirement that subjects should volunteer rather than be chosen by the investigators. Response rates were higher at other plants where investigators had a direct role in selecting subjects and recruitment. The response rates of the exposed did not differ significantly from those of the non-exposed at any plant or overall (x2, p < 0-05). The respondents were on average two years older than nonrespondents and at plants 1-3, proportionately more women were respondents (37%) than nonrespondents (23%). These comparisons do not show any significant differences between respondents and non-respondents. CHARACTERISTICS OF SUBJECTS Table 3 presents the demographic characteristics of the study group (n = 240). The subjects were young Table 1 Descrption ofstudy sites Plant Product Processes using solvents Job classifications Exposed* Control Date(s) production started 1 Office furniture Spraying paint or glue Sprayer (paint, glue), top Assembler, material handler, 1963 trimmer, paint technician cutter, sewer 2 Office furniture Spraying lacquer or glue Sprayer (glue, lacquer), rubber Assembler, cutter, sewer, 1970s wiper material handler 3 Automotive parts Spraying paint or glue Sprayer (paint, glue), spindle Assembler, material handler, 1973 washer, booth cleaner, paint machine operator mixer 4 Printed matter Offset printing Pressmen, feeder operator Machine operator, material 1963 handler, plate preparer *Exposure category based on industrial hygiene walk-through survey. Behavioural evaluation of workers exposed to mixtures of organic solvents 581 Table 2 Enumeration ofthe study population, sample, respondents, non-respondents, and replacements by plant and exposure status at walk-through. (Number ofexposed in parentheses) Group Plants 1-4 Plant I Plant 2 Plant 3 Plant 4 Study population 1571 (276) Sample*: 640 (227) Respondents 296 (114) Non-respondents 371(113) Replacementst 28 (15) Signed consent forms and tested 240 (124) Response rate (%): 42 Exposed (%)II 50 Non-exposed (%) 38 434 (120) 434 (120) 121t (36) 313(84) 8 (4) 72 (35) 28 30 27 714 (28) 56 (28) 50 (27) 6(1) 1 (0) 51 (27) 89 96 82 345 (85) 72 (36) 52 (25) 20(11) 19 (11) 71 (36) 72 69 75 78 (43) 78 (43) 46 (26) 32(17) 0 (0) 46 (26) 59 60 57 *Sample at plant 1 was self selected volunteers whose names appeared on sign-up sheets. Samples at plants 2-4 were selected from company employee rosters. tOf the 121 names on the sign-up sheets, 72 were selected by the investigators and 64 signed consent forms. tReplacement subjects were drawn from the plant study population. Response rate = (respondents/sample) x 100. IlDifferences between response rates of exposed and non-exposed were not significant (p > 0 05, x2). Table 3 Characteristics ofthe study group Item Plants 1-4 Plant I Plant 2 Plant 3 Plant 4 No* 240 Mean age + SD 35 11 Meanyearseducation+SD 11 2 Meanyearsoncurrent job SD 6 6 Mean alcoholic drinks/day SD 1 2 Men (%) 66 Caucasian (%) 90 Cigarette smoking: Never smokers (%) 32 Current smokers (%) 50 Ex-smokers (%) 18 >1 Alcoholic drink/day(%) 31 Former job as a painter or printer (%) 16 Hobby chemical users (%)t 7 Regular medication takers (%)t18 72 36 10 12 2 76 12 96 96 29 49 23 43 15 6 15 51 30 12 11 2 35 11 61 96 53 31 16 20 4 8 18 71 36 10 11 2 4+3 12 18 79 29 63 8 25 13 0 23 46 39 9 12 1 12 8 22 100 93 22 56 22 41 35 17 15 *Missing data excluded from calculations. No more than five cases missing for any item. tHobby chemicals included solvents, lead, or pesticides. tOf 53 reports of medicine taken in 24 hours before testing, 25% were aspirin or other analgesics, 13% were antihypertensive drugs, and the remainder were uniformly spread among 11 different categories. (mean 35), mostly male (66%) and white (90%), high school educated (mean 11th grade), and had not been employed on their current job and at the plant for long (mean 6, 7 years, respectively). Mean daily alcohol intake was low; 31 % reported consuming more than one alcoholic beverage a day, usually as beer. Sixteen per cent reported a former job as a painter or printer. Plant 3 was mostly female whereas plants 1 and 4 were more than 96% male. TEST PROCEDURES Each subject completed six behavioural tests, a questionnaire covering work and medical history, and a medical evaluation. The medical evaluation included a clinical neurological examination,'4 a blood chemical profile, and analyses of blood and urine for ethanol and drugs. The results of the clinical neurological examination will be reported elsewhere. The subjects were tested at the worksite, usually in batches of four for up to 16 workers a day. Testing was completed at each plant in four or five consecutive days during 1980 and 1981. The matched pairs of exposed and non-exposed were generally tested at the same time of day to minimise the behavioural effects of circadian rhythms. The exposed subjects were tested before their shift, wherever possible, to minimise acute effects of sol- vent exposure. The six tests comprising the battery are described below. Each test took about 15 minutes to complete. The tests were administered by trained and experi- enced personnel who were blind to the exposure status of the subjects. Fiuts law task measured psychomotor speed.'5 16 Subjects tapped between two copper plates with a stylus coupled to an electronic timer. After 50 dis- crete tapping motions the task was made more or less difficult by varying the distance between the plates and the size of the target zone on each plate according to pre-established dimensions. The per- 582 Maizlish, Langolf, Whitehead, Fine, Albers, Goldberg, and Smith formance variable was based on the average speed which letters were out of order (item only) were for four blocks of 50 tapping motions and the cor- considered separately from those in which the letters responding plate dimensions. were in the correct order (item and order). Stroop test measured the ability to ignore distract- Contnuous recognition memory measured the ing visual and verbal stimuli.'7 18 Subjects were ability to remember digits after a lapse of up to 40 asked to name the colour of the ink in which letters seconds.22 The subjects were presented with a list of were written. The letters formed either a repeating three digit numbers, displayed singly on a pattern of Os or spelt a word that was the name of a microcomputer screen at the rate of one every five colour-for example blue. The colour in which the seconds. About two thirds of the numbers were word was written always differed from the colour repeated. After each number was displayed, the sub- that the word spelt-for example "blue" in red let- jects indicated whether the number had been previ- ters. Slides of colour words and coloured Os were ously displayed by pushing an "old" or "new" but- cued by a projector controlled by a microcomputer. ton. The response variable was based on the accu- The subjects were electronically timed in 70 trials on racy of responses during 15 minutes of trials. how fast they could name colours with and without interference from colour words. The total time to RELIABILITY AND STATISTICAL POWER OF name colour-words (conflict stimuli), "colour- BEHAVIOURAL TESTS naming time," was the first performance variable. Test-retest reliabilities for the Stroop test, memory The second variable, "interference time," was the scanning, and memory span are greater than time to name colours for distracting colour words 0X852123 (AM Rose, unpublished data). Based on minus the time to name colour Os for the same col- prior research and the number of tested subjects, the our of ink. statistical power of the Fitts test, Stroop colour nam- Mental rotation measured spatial reasoning or vis- ing test, and memory span was sufficient to detect ual intelligence.'9 Visual stimuli, presented by a mic- 5% differences in test scores between exposed and rocomputer, consisted of slides depicting normal non-exposed with a 90% certainty (a = 0.05). The and backwards letters, such as R and A1, which were least significant differences (a = 0.05) in test scores rotated 00, 600, 120, and 1800 in the plane of the detectable with 90% power was 15-20% for mem- slide. Subjects were instructed to mentally "flip up" ory scanning, continuous recognition memory, and the image and call out "normal" or "backwards", Stroop interference. depending on the orientation of the letter. The ver- bal response time was electronically recorded for 80 DESCRIPTION OF WORK PROCESS AND EXPOSURE slides. The performance variable was based on the The materials and work processes were similar at speed of the verbal response. plants 1, 2, and 3 and have been described by Memory scanning measured the rate of short term Whitehead et al.24 Paint vehicles contained alkyd memory scanning.20 Subjects memorised a short list resins and mixtures of aromatic solvents, chlorinated of one digit numbers presented on a flash card. and oxygenated aliphatic solvents, alcohols, and ace- Numbers between 0 and 9 were 'hen displayed tates. Fillers contained various pigments but lead singly on a microcomputer screen. After each pigments were used infrequently. The paints also number was displayed, the subjects indicated contained various additives including biocides, whether the number was included in the memorised stabilisers, and antiskinning agents. Glues contained list by pushing a "yese or "no" button. The time to toluene, hexane, and chlorinated hydrocarbon sol- respond was electronically recorded. The perform- vents. ance variables were based on response time for Metal or plastic parts were placed on a conveyer "yese and "no"' responses as a function of the list line or were positioned at a work station. Parts were length of the memorised list. List lengths of two and sprayed with hand held spray guns that nebulised five were used for a total of 200 trials per subject. the paint, lacquer, or glue with or without a pressur- Memory span measured short term recall for let- ised air stream. Spray booths were ventilated and ters.2' A list of letters was displayed one at a time on equipped with either a water curtain or dry filters. a microcomputer screen at the rate of one a second. Painted or glued parts were taken to areas where The subject wrote (from memory) the letters in these and other components were upholstered, order of presentation on a prepared form. assembled, packed, and shipped. Exposure to sol- This was repeated using 43 lists of various lengths. vent also occurred during paint mixing, cold clean- The performance variables were based on the length ing, wiping, touch up, and disposal of waste paint. of lists that were recalled without error in 50% of At plant 4, a one to four man team operated each trials. Responses with omissions or comissions were of seven offset printing presses that consumed vari- scored as errors, but the memory span for lists in ous inks and solvents. Printing plates were com- Behavioural evaluation of workers exposed to mixtures of organic solvents 583 posed in separate areas using a photoetching or lithographic process. A bindery and material handling department were located in the basement. The original ventilation systems were still in use at each plant at the time of the study. The ventilation system at plant 4 recirculated contaminated pressroom air which led to heavy contamination of all production areas. No workers engaged in solvent operations wore solvent collecting respirators. The breathing zone of 112 "exposed" and 47 " non-exposed" was sampled with a personal breathing zone sampler for roughly a full workshift during the week of behavioural testing. Subjects for whom air samples were not taken were assumed to have the same exposure as a sampled subject within the same plant, job title, and department. Twenty individual solvents were identified using gas chromatography mass spectroscopy and infrared spectrophotometry, and routinely quantitated using gas chromatography. Solvent concentrations, expressed as full shift time weighted averages (TWA), are presented in table 4. The mean total solvent concentrations at the furniture and automotive parts plants (plants 1-3) were low but at the printing plant (plant 4) the mean total solvent concentration, which was dominated by isopropanol, naphtha, hexane, and xylene, exceeded 300 ppm. The mean total concentration among the non-exposed at plant 4 was about eight times greater than that of exposed at the other plants (table 5). This finding indicates that the exposed/non-exposed classification scheme introduces a considerable degree of misclassification with respect to demonstrable solvent exposures. Further analyses rely primarily on the quantitative solvent measurements. To maintain consistency with Scandinavian studies, Table 5 Mean solvent concentration (ppm) by exposure status at walk-through. * (Number ofsubjects in parentheses) Plant Exposed Non-exposed 1 17(35) 137 2 24 (27 9 (24) 3 20 (36 5 (35) 4 385 26 193 (20) 1-4 96(124) 37(116) *Includes assumed values of unsampled subjects. comparisons of "exposed" and "non-exposed" will also be presented. In addition to total solvent concentration, other exposure variables were constructed from measurements of air borne solvents, duration of employment, and a scale that weighted the toxicity of various solvent components relative to toluene and xylene.25 Each of these exposure variables was highly correlated with total solvent concentration and with each other. For analytical purposes, only the total solvent concentration will be presented, although the other exposure variables yielded similar results. STATISTICAL METHODS Based on prior knowledge,' age, sex, education, and alcohol intake were considered to be the most important potential confounders of performance. Potential confounding was examined by correlation analysis (table 6). Because performance on most tests in the battery declines with age26 and possibly with alcohol intake, an impairment due to age or alcohol intake may be misattributed to solvent exposure. Sex affects performance of mental rotation and Stroop colour naming time by 5-10%; men Table 4 Per cent cumulative TLV34 and mean concentration (ppm) ofsolvents in breathing zone air by plant* Solvent Plants 1-4 Mean SD Plant I Mean SD Plant 2 Mean SD Plant 3 Mean SD Plant 4 Mean SD So TLV* Total solvent Isopropanol Methylene chloride Trichloroethylene Acetone 2-Butanone (MEK) Naphtha Toluene Xylene Ethylbenzene Hexane Residual 40 71 6 68 126 9 31 70 Oql 2 10 0 1 22 4 81 0 30 10 22 0 3 62 4 91 1 20 8 17 1 -- 1 9 13 15 9 19 170 65 12 17 18 12 26 302 117 0 0 2 1 6 161 70 4 2 6 4 16 12 31 100 00 2 1 3 2 7 15 9 0000 1 47 5 1 3 0 0 50 20 35924 65 3 1 2 0 1 20 11 1 01 00 53 1 1 2 1 2 39 18 2 3 5 2 4- - *Includes assumed values of unsampled subjects. tMethyl isobutyl ketone (MiBK), propyl benzene, cumene, heptane, butyl acetate, isobutyl acetate, isopropyl acetate, isobutyl butyrate, and butyl cellosolve were present at mean concentrations of less than 1 ppm. Y.Cn/Trvt %TLV = 100 x where C is the concentration of the nth solvent in a mixture and T is the threshold limit value for that solvent. Assumes an average mo ecular weight of 100. = <0-5 ppm; - = not determined at plant 4. 584 Maizlish, Langolf, Whitehead, Fine, Albers, Goldberg, and Smith Table 6 Correlation between demographic variables and solvent exposure (ppm) t Variable Age Sex Education Alcohol intakell Solventd Age Sex Education Alcohol - 004 - -0-29** -0-19** -0-06 -0-24** 0-07 - 009 -0-31** 002 0 16* *p < 0-05; **p < 0-01. tData combined for four plants; four cases missing (n = 236). tSolvent = total solvent concentration (ppm). Sex indicator variable: 0, men; 1, women. IAlcohol intake = mean daily drinks. perform better on tasks needing mental rotation and women perform better on spontaneous verbal tasks.'7 27 The correlation analysis (table 6) indicates that sex may introduce a moderate positive bias on the Stroop test but a moderate negative (masking) bias on the mental rotation test. Education was not considered to be an important confounding factor. Dose-response was investigated with univariate plots of performance versus total solvent concentration, simple linear regression, and stepwise, forward multiple regression models28 in which total solvent concentration (ppm), plant, age, sex, years of education, mean daily alcohol intake, previous job (as a painter or printer), and hobby (entailing potential exposure to lead, solvents, or pesticides) were candidate variables. The significance level for including a term was p < 0*1. In multivariate regressions error residuals were visually inspected and satisfied conditions for normality and independence. The mean performance between those classified as exposed and non-exposed on the walk-through survey was compared using 2 tailed t tests for unpaired data. Fourteen subjects for whom English was a second language and one who admitted taking amphetamine immediately before testing were excluded in the analyses of psychological tests. The medical evaluation did not show a clinically significant physical impairment that warranted any subject's exclusion from the psychological tests. Laboratory studies did not show detectable blood ethanol for any subject. ll IU 3 ._L tn 0 200 400 600 Total solvent concentration (ppm) Fig 1 Fitts law task (speed and difficulty) versus total solvent concentration by plant. Results Performance on selected variables from each behavioural test is plotted against total solvent concentration by plant in figs 1-6. The slope and intercept for a simple linear regression between performance and total solvent concentration is presented in table 7 for each performance variable. The results of stepwise multiple linear regression are summarised in table 8; mean performance by plant is sum- marised in table 9. 200 400 Total solvent concentration (ppm) Fig 2 Stroop interference time versus total solvent concentration by plant. FITTS LAW TASK The plot of Fitts law "speed and difficulty" shows highly variable responses but no association be- Behavioural evaluation of workers exposed to mixtures of organic solvents 1600 8 585 1200 I >113444 44 L ) '800 ~3344 444 44444 12 444 4 4 .3 2 400 0 200 400 60a0 Total solvent concentration (ppm) Fig 3 Mental rotation time versus total solvent concentration by plant. 160 12.o U0, 4 v E 4~4 i120I' 4 34 *- 44 a2.14 x 44 4 ~4 4 4 4 24444 i3 4 23 141 44 C n 2 _s I1 0 200 400 600 Total solvent concentration (ppm) Fig 5 Memory span (item and order) versus total solvent concentration by plant. 0 200 400 600 Total solvent concentration (ppm) Fig 4 Stemnberg memory scanning rate versus total solvent concentration by plant. 0 200 400 600 Total solvent concentration (ppm) Fig 6 Continuous recognition memory (d') versus total solvent concentration by plant. ween performance and solvent level at individual plants or overall (fig 1). In univariate or multivariate regressions (tables 7-8) solvent concentration was not significantly related to performance (p > 0.05). Performance among subjects at plant 3 was slightly poorer (p = 0.03) on average than that of subjects at other plants (table 8). STROOP COLOUR WORD TEST The plot of Stroop interference time does not depict a trend with solvent concentration (fig 2); significant linear dose-response relation were not found in univariate or multiple linear regression. Age and plant 3 were significant (p < 0.005) predictors of performance (table 8). Stroop colour naming time was not significantly associated with solvent level at any plant or overall (tables 7-8). In agreement with previous research colour naming time was significantly slower with increasing age (p < 0.001) and significantly faster among women (p < 0.001) (table 8). In multivariate analysis subjects at plant 2 were slightly slower than other subjects (p = 0.02). MENTAL ROTATION The plot of mental rotation speed shows a highly variable response but no obvious correlation with solvent concentration overall or any plant (fig 3). The simple linear trend among plant 4 subjects was 586 Maizlish, Langolf, Whitehead, Fine, Albers, Goldberg, and Smith Table 7 Simple linear regression ofpsychological performance and total solvent concentration (ppm) by plant Performance testt (psychological functon) Plants 1-4 (nt = 201-217) Plant I (n = 62-70) Plant 2 (n = 41-46) Plant 3 (n = 54-62) Intercept Slope Intercept Slope Intercept Slope Intercept Slope Fitts law task (psychomotor) speed & difficulty (msec/bitt) 124 Stroop test (reedom from distraction) Interference (msec) 230 Colour naming (msec) 1107 Mental rotation (visual intelligence) Speed (msec) 906 Memory scanning (memory) No-scanning (msec/item) 44-7 Yes-scanning (msectitem) 38-4 Memory span (short term memory) Item & order (digits) 4-2 Item only (digits) 45 Continuous recognition memory (intermediate memory) d' 0-88 -0-011 0-016 0-08 -0-10 0-01 -0-02 0-001 0-001 0-001* 124 246 1108 879 50-8 46-1 4-4 4-7 0-90 -0-16 0 70 1-42 0-68 0-37 0 09 -0-005 -0-004 0-001 119 232 1129 901 31-0 32-0 4-2 4-6 0-91 0-072 0 04 -0-16 -0-45 0-17 -0-11 0-002 -0-002 0.001 *Slope significantly different than 0 at p < 0.05. tLower values reflect better performance on all tests except continuous recognition and memory span. WVariable n due to missing values on some tests. All r2s <0-08 except mental rotation at plant 4 (r2 = 0-12). 130 192 1069 943 41-8 34-6 4-0 4-4 0-84 -0-098 0 44 0-25 -0-46 -0.05 -0-11 0-006 0-004 0-003 Plant 4 (n = 40-45) Intercept Slope 109 257 1193 1017 54-1 38-3 3-8 4-5 0-72 0-034 0-07 -0-18 -0-41* -0-03 -0-02 0-001 0-001 0-001 Table 8 Stepwise forward, multiple linear regression ofsix psychometric tests: improvement (+) or impairment (-) ofpsychological performance associated with demographic variables and exposure to mixtures oforganic solvents Test (function) Fitts law task (psychomotor) Stroop colour-word Mental rotation test Sternberg memory Memory span test Continuous test scanning recognition memory (freedom from (visual intelligence) (memory speed) (memory capacity) (intermediate memory, distraction) Variablet Speed & difficulty Interfierence Colour Speed (mmec) (mnseclbit) (snec) naming (msec) No Yes Item Item d' scaning scaning & order only (digits) (digits) Age Sex (0= M, 1 =F) Education Plant 2* Plant 3 - Plant 4 Shift (0 = before 1 = dunng, after) Hour of test Prior employmentt Hobby exposure Total solvent (ppm) r2 0-02 ++ 0 09 0-12 0-18 ++ ++ ++ + ++ ++ + 0-16 0-14 0-12 0-11 0-08 'Plant dummy variables relative to plant 1. tAll (+) and (-) significant at p < 0-01. Alcohol and regular medications did not enter any model at p < 0-01. tFormer printer or painter. significant (p < 0.05); however, performance improved with increasing solvent concentration. Age was not controlled in this simple regression and may have contributed to the negative association through its association with solvent exposure (r(age) = -0-46). No significant trend with solvent concentration was found by combining responses from all four plants (table 7). According to the multiple linear regression model, age (p = 0-001), sex (p < 0-001), hobby chemical use (p = 0.06) were significant predictors of performance. The predicted direction of effect of increasing age, sex, and hobby chemical use was toward poorer performance, as expected. MEMORY SCANNING A plot of the average of "yes" and "no" memory scanning is presented in fig 4; (yes and no scanning times were highly correlated, r = 0.7). No association between performance and total solvent con- centration is apparent at individual plants or overall. Simple linear regression of either yes scanning time or no scanning time and total solvent concentration was non-significant by plant or overall. According to the multiple linear regression model, speed of yes or no scanning improved with increasing solvent con- centration (p = 0-09, p < 0-001). Effects related to plant, former employment, and hobbies were also indicated by the model. Subjects who reported using Behavioural evaluation of workers exposed to mixtures of organic solvents 587 ible 9 Mean performance on psychological tests by plant ,formance testt ;ychological fincdon) Plants 1-4 Plants 1-3 Plant 1 Plant2 Plant3 Plant4 (n = 201-217) (n = 159-177) (n = 62-70) (n = 41-46) (n = 54-62) (n = 40-45) ,ts law task (psychomotor) Speed & diffculty SD (msec/bit) 123 24 124 22 123 23 120 24 129 20 119 28 coop test (freedom from distraction) Interference SD (msec)* 231 112 229 111 252 125 238 80 198 107 236 120 Colour naming SD (msec) 1112 185 1105 179 1121 198 1127 156 1073 170 1142 206 ental rotation (visual intelligence) Speed SD (msec) 899 152 901 155 874 137 895 107 938 197 893 140 emory scanning (memory speed) No-scanning SD (msec/item)* 44 28 44 26 54 28 33 22 41 24 45 33 Yes-scanning _SD (mseclitem)** 37 24 38 23 47 28 31 18 33 18 32 24 emory span (short-term memory) Item & order SD (digits) 4-2 07 4-2 0-7 4-3 0-7 4-2 _05 4-1 +07 4-209 Item only SD(digts) 4-6 0-8 4-6 0-8 4-7 0-9 4-5 0-6 4-4 0-8 4-6 07 wntinuous recognition memory (intermediate memory) d' SD 0-93 + 0-5 0.90 0.5 0-91 0.5 0-91 0 5 0.87 + 0.5 1-1 0-6 Jleans between plants significantly different (F-test, ANOVA) p < 0-05. 'p <0-01. ,ower values reflect better performance on all tests except continuous recognition memory and memory span. iariable n due to missing values on some tests. solvents, lead, or pesticides in a hobby were slower than subjects who did not. Subjects reporting former employment as a painter or printer were faster than other subjects. MEMORY SPAN The 50% threshold memory span for letters recalled in correct order was not associated with solvent level (fig 5). In univariate or multivariate analyses, memory span-item and order and span-item only were not significantly related to solvent level. Education was associated with a longer span (p < 0-001), as expected. Subjects tested before their work shift and those tested relatively later in the day (mean hour 1035) performed slightly better (p < 0.06) than other subjects. CONTINUOUS RECOGNITION MEMORY Accuracy of recognition memory is plotted against solvent concentration in fig 6. A significant linear trend, indicating increasing accuracy with increasing solvent concentration, was found overall (p < 0-05) (table 7). This was also found in the multivariate analysis (p < 0.005) (table 8). Subjects reporting previous employment as a painter or printer had slightly better performance (p = 0.06). EXPOSED/NON-EXPOSED COMPARISONS As previously noted, "exposed" and "non-exposed" categories based on the initial walk-through assessment of exposure introduce considerable misclassification of demonstrable exposures. For completeness and consistency with Scandinavian studies, performance between these groups is compared in table 10. Combining over plants, exposed and non-exposed, were similar with respect to mean age, mean educational level, mean daily alcohol intake, and sex distribution. The exposed group at plant 4 was on average six years younger than non-exposed (38 v 44), but the exposed consumed a greater Table 10 Behavioural performance: percentage impairment (-) or improvement (+) ofexposed relatve to non-exposedt Performance testt (psychological function) Plants 1-4 Plant I Fitts law task (psychomotor) Speed & difficulty (msec/bit) -5* Stroop test (freedom from distraction) Interference (msec) -11 Colour naming (msec) -4 Mental rotation (visual intelligence) Speed (msec) -2 Memory scanning (memory) No-scanning (msec/item) -10 Yes-scanning (mseclitem) -4 Memory span (short term memory) Item & order (digits) -5* Item only (digits) -6*** Continuous recognition memory (intermediate memory) d' +2 -2 -34** -10 -3 -16 -2 -11*** -12*** -7 *p < 0-10; **p < 0.05; ***p < 0.01. tlOO x (mean of exposed-mean of non-exposed)/mean of non-exposed. Plant 2 -7 -16 -2 0 +9 +3 -5 -4 +14 Plant 3 -5 -7 -5 -7 -17 -4 -5 -7 -10 Plant 4 -7 +14 +3 +5 0 +2 +2 0 +29 588 Maizlish, Langolf, Whitehead, Fine, Albers, Goldberg, and Smith number of alcoholic beverages a day (2 v 1). These years in other studies concerning solvent mix- differences tend to introduce small confounding tures.7-'2 biases in different directions. Secondly, the main component of solvent expos- On the Fitts task, the exposed performed less well ure in previous studies was toluene, but isopropanol, than non-exposed by approximately 5% (p > 0-05) which is probably less potent, was the dominant at each plant and overall. exposure of this study (table 4). Thus short duration The Stroop interference time was 10% greater (p of exposure and less potent components in the > 0-05) among the exposed than the non-exposed exposure may contribute to the lack of overall. Performance was significantly worse for the performance-exposure associations on behavioural exposed at plant 1, but the exposed at plant 4 were tests. 16% faster than the non-exposed (p > 0.05). A Thirdly, how do the test batteries compare in similar relation was found for colour naming time. scope and sensitivity? Direct comparisons are There were no significant differences in speed of limited to memory span tests for which the version mental rotation between the exposed and the non- in the present study23 has a greater sensitivity than exposed. that based on Wechslers digit span.29 With the The exposed were slightly faster (p > 0-05) than exception of simple reaction time, the range of the non-exposed in no-scanning overall and at abilities appears to be similar in the present and plants 1 and 3. There were small, non-significant Wechsler based Scandinavian batteries. Our power differences in yes-scanning rate between the expos- calculations suggest that the tests comprising the ure groups. present battery can detect small differences in per- Combining over plants, the exposed had a formance of the same magnitude as impairments significantly smaller memory span (item only) than reported by previous studies (5% to 15%). the non-exposed. This was largely accounted for by Moreover, responses correlated with age, sex, and the performance of the exposed at plant 1. Memory education as predicted from previous research.'8 26 27 span (item and order) was slightly smaller (p > 0-05) among the exposed than the non-exposed. METHODOLOGY No significant differences in accuracy between It was assumed that performance was accurately exposure groups was found on the continuous described by linear additive models in which con- recognition memory test. Of note, the exposed at founding was adequately controlled. Goodness of fit plant 4 were 29% more accurate than the non- was examined by comparing model predictions with exposed. means stratified by plant, sex, age, education, and other variables. For the Fitts law, Stroop colour Discussion naming, and memory span (item only) observed ver- sus predicted agreed within 5%. Whereas the good- The dose-response relations observed in this study ness of fit was poorer for other parameters (gener- do not confirm previous reports of behavioural ally, 13-25%), there was agreement in the direction impairments among workers exposed to mixtures of of response. Given the great variability of response, organic solvents below recommended limits.7'- 12 As it is doubtful whether non-linear models would have in previous studies, the comparison of the exposed been more efficient or have led to different conclu- and the non-exposed showed some minor deficien- sions. cies among the exposed. Memory span of the Several potentially confounding relations con- exposed was significantly less than that of the non- cerning age, sex, alcohol, and plant were identified exposed, but this has not been generally reported. (table 6), but only sex in the test of mental rotation Whereas exposed/non-exposed differences are not could have masked solvent related impairments or readily explained, exposure to solvents is probably contributed to the apparent improvement in per- not responsible since impairments were not formance. Other factors, possibly unquantified, may observed in more powerful analyses of dose- have contributed to other associations, suggesting a response. These findings will be discussed below tak- beneficial effect of solvent exposure. Of particular ing into account differences between this and previ- concern are selection biases, subject motivation, ous studies and methodological limitations. skill level of jobs, nature of past exposures, or observer/subject bias. COMPARABILITY OF STUDIES Non-response, job skill, and "healthy worker" There are important differences between the Scan- selection were considered as sources of selection dinavian studies and the present study. Firstly, the biases. More than one third of the eligible subjects average duration of employment (exposure) was at plants 1 and 4 declined to take part in the study. seven years in the present study compared with 14 Because subjects at plant 1 were volunteers, prob- Behavioural evaluation of workers exposed to mixtures of organic solvents 589 ably the more interested or better informed indi- viduals participated. Other than age and sex (which were unrevealing), additional data were not avail- able to investigate whether health or exposure related factors influenced self selection. Twenty of the 30 non-respondants at plant 4 were interviewed; 12 did' not cite a specific reason for nonparticipation and five said that they were "too busy." These data are insufficient to rule out the possibility of a bias due to non-response. Since skilled work may develop cognitive function30 and since people with superior abilities may select skilled trades, job skill has been identified as a potential confounder in other neurobehavioural studies.8 1231 Pressmen at plant 4, who were the exposed group, were among the most highly skilled workers in this study. Nevertheless, multiple linear regression analyses gave essentially the same results whether pressmen were included or excluded. This suggests that some other factor operating among printing plant workers may be associated with improved performance. Similarly, Iregren observed that printers surpassed spray painters and nonexposed controls in several cognitive abilities.9 Cross sectional studies are vulnerable to an underestimation of risk because affected individuals leave the workplace or select out of exposed jobs within the workplace (healthy worker selection). This has been asserted in several investigations of behavioural effects and documented in one study.32 It is not known whether the population in the pres- ent study underwent this type of selection. Many of the exposed jobs in the study were high paying, high seniority jobs. Moreover, the small difference between duration of the current job (6 years) and total employment at the plant (7 years) suggests that migration between jobs was not a significant factor. A second limitation of cross sectional studies is inherent in the study design; subjects were not followed up over time and it is not known whether performance deteriorated over time beyond that attributable to the normal aging process. Current solvent levels are unlikely to have misrepresented current or past exposures. Duration of employment on the current job was relatively short and nearly as long as total duration of employment. Also the plants were relatively new, having been opened in the 1960s and 1970s. Leaded paints were infrequently used; moreover, lead would have contributed to an association of adverse solvent related effects, rather than masked it. Other neurotoxic exposures are unlikely to have been overlooked. Physiological tolerance to short term exposures to solvent has been postulated by Savolainen et al.33 Although tolerance to ethanol is widely recognised, tolerance to isopropanol has not been widely studied but would have tended to mitigate the effects of exposures at plant 4. Through the consent procedure subjects were informed that the focus of the study was to examine the possible effects of solvents. Compared .with the exposed, the non-exposed may have been less motivated, believing that their participation was secondary. The lower response rate of the non-exposed (table 2) suggests that this may have occurred, and the small systematic differences between the exposed and the non-exposed could reflect differences in motivation. In conclusion, this investigation has not provided evidence of adverse behavioural effects related to exposure to moderately low levels of solvent mixtures for a relatively short duration. Concern over adverse effects of longer term, low level exposure to solvent cannot be completely deferred because of limitations in the study design. Periodic monitoring of workplace solvent exposures and behavioural performance may provide longitudinal data on which to base definitive conclusions. This project was supported by a contract from the National Institute for Occupational Safety & Health, United States Department of Health and Human Services. Requests for reprints to: Dr L J Fine, Room 6007, SPH-II, University of Michigan, Ann Arbor, MI 48109, USA. References 'Johnson BL, Anger K. Behavioural toxicology. In: Rom WN, ed. Environmental and occupational medicine. Boston: Little Brown and Company, 1983. 2 Xintaras C, Johnson BL, deGroot I, eds. Behavioural toxicology. Washington DC: US GPO, 1984. (NIOSH publ No 74-126.) Cherry N, Waldron HA, eds. The neuropsychological effects of solvent exposure. New Lane, England: Colt Foundation, 1983. 4 Cherry N, Waldron HA, Wells GG, Wilkinson RT, Wilson HK. An investigation of the acute behavioural effects of styrene on factory workers. Br J Ind Med 1980;37:234-40. Gamberale F. Behavioural effects of exposure to solvent vapors: experimental and field investigations. In: Horvath M, ed. Adverse effects of environmental chemicals and psychotropic drugs. Amsterdam: Elsevier, 1972. 6 Gamberale F, Hultengren M. Toluene exposure II. Psychophysiologic functions. Work Environ Health 1972;9: 131-9. Elofsson S, Gamberale F, Hindmarsh T, et aL Exposure to organic solvents: a cross-sectional epidemiological investigation of occupationally exposed car and industrial spray painters with special reference to the nervous system. Scand J Work Environ Health 1980;6:239-72. 8 Hanninen H, Eskelinen L, Husman K, Nurminen M. Behavioral effects of long-term exposure to a mixture of organic solvents. Scand J Work Environ Health 1976;4:240-55. Iregren A. Effects on psychological test performance of workers exposed to a single solvent (toluene): a comparison with effects of exposure to a mixture of organic solvents. Neurobehav Toxicol Teratol 1982;4:695-701. 590 Maizlish, Langolf, Whitehead, Fine, Albers, Goldberg, and Smith ' Olson BA. Effects of organic solvents on behavioural performance of workers in the paint industry. Neurobehav Toxicol Teratol 1982;4:703-8. "Hane M, Axelson 0, Blume J, Hogstedt C, Sundell L, Yterborg B. Psychological function changes among house painters. Scand J Work Environ Health 1977;3:91-9. 12 Knave B, Persson HE, Goldberg JM, Westerholm P. Long-term exposure to jet fuel. An investigation on occupationally exposed workers with special reference to the nervous system II. Scand J Work Environ Health 1978;5: 19-44. '3 Hill AB. The environment and disease: association or causation? Proceedings of the Royal Society of Medicine 1965;58:295- 300. 4 Albers JW, Cavender GD, Levine S, Langold GD. Asymptoma- tic sensorimotor polyneuropathy in workers exposed to elemental mercury. Neurology 1982;32: 1168-74. 5 Fitts P. The information capacity of the human motor system in controlling the amplitude of movement. J Exp Psychol 1954;47:381-91. 16 Langolf GD, Chaffin DB, Foulke JA. An investigation of Fitts law using a wide range of movement amplitudes. Journal of Motor Behaviour 1976; 8: 11 3-28. 17 Stroop JR. Studies of interference in serial verbal reactions. J Exp Psychol 1935;18:643-61. 18 Jensen A, Rohmer W. The Stroop color-word test: a review. Acta Psychol 1966;25:36-93. Cooper LA, Sbepard RN. Chronometric studies of the rotation of mental images. In: Chase WG, ed. Visual information processing. New York: Academic Press, 1973. 20 Sternberg S. Memory scanning: new findings and current controversies. Q J Exp Psychol 1975;27: 1-32. 21 Smith PJ, Langolf GD. The use of Sternberges memory scanning paradigm in assessing the effects of chemical exposure. Human 22 Factors 1981;23:701-8. Wickelgren WA. Age and storage dynamics in continuous recog- nition memory. Developmental Psychology 1975;11: 165-9. 23 Smith PJ, Langolf GD, Goldberg J. Effects of occupational exposure to mercury on short term memory. Br J Ind Med 1983;40:413-9. 24Whitehead LW, Ball G, Fine LJ, Langolf GD. Am Ind Hyg Assoc J (in press). 25 Axelson 0, Hane M, Hogstedt C. Current aspects of solvent- related disorders. In: Zenz C, ed. Developments in occupa- tional medicine. Chicago: Year Book Publishers, Inc, 1980. 26 Schumacher J. Effects of aging on four behavioral tests. Ann Arbor, University of Michigan, Center for Ergonomics, 1981. 27 Maccoby EE, Jacklin CN. The psychology of sex differences. Stanford: Stanford University Press, 1974. 28Draper NR, Smith H. Applied regression analysis. New York: John Wiley & Sons, Inc, 1966. 29 Wechsler D. The measurement and appraisal ofadult intelligence. Baltimore: Williams & Wilkins Co, 1958. 30 Kohn ML, Schooler CL. Work and personality. Norwood, NJ: Ablex, 1983. 31 Gregersen P, Stigby B. Reaction time of industrial workers exposed to organic solvents: relationship to degree of exposure and psychologic performance. Am J Ind Med 1981; 2:313-2 1. 32 Husman K. Symptoms of car painters with long term exposure to a mixture of organic solvents. Scand J Work Environ Health 1980;6: 19-32. 33 Savolainen K, Riihimaki V, Linnoila M. Effects of short-term xylene exposure on psychologic functions in man. Arch Occup Environ Health 1979;44:201-1 1. 34 American Conference of Governmental Industrial Hygienists. TLVs: threshold limit values for chemical substances and physi- cal agents in the work environment. Cincinnati: ACGIH, 1982. Destruction of manuscripts From 1 July 1985 articles submitted for publication will not be returned. Authors whose papers are rejected will be advised of the decision and the manuscripts will be kept under security for three months to deal with any inquiries and then destroyed.