Subjects and methods

STUDY POPULATION

The production of polyester boats started early 1982. From 1984 to 1987, we carried out several cross sectional industrial hygiene studies in the styrene polyester boat building plant with a workforce that at that time consisted of 185 exposed male employees.16 17In 1989, the usual production activities stopped for economic reasons and only repair and maintenance activities were continued with a reduced workforce (n=31) that could continue to work at the plant. Due to the changed industrial activities, the exposure to styrene dropped considerably to an average of 41 mg/m³. The other workers (n=154) were made redundant on the basis of age (workers more than 50 years old retired) or family income (people who were economically less important to their family—for example, being married to a working spouse and having no children—were preferentially dismissed).

Three years after the event in 1989, we retraced the 185 workers enrolled in our previous studies.16 17 Figure 1 shows how the population of the present retrospective study was recruited and found to be eligible on the basis of initial defined exclusion criteria (craniocerebral trauma with 10 minutes or more unconsciousness, alcoholism (five or more alcoholic drinks/day),18epilepsy, major depression, or other major neurological disease). In total, 117 subjects of the original workforce participated in the present study (participation rate 63%) consisting of 90 formerly exposed workers and 27 workers currently exposed to low concentrations of styrene. A group of 111 control male subjects (electricians and assembly workers) from another plant were invited to cooperate in the study. Age, sex, and schooling level were matched as closely as possible across groups. Twenty seven refused to participate and 20 were excluded according to initial defined exclusion criteria (the same criteria as those applied to the exposed subjects as well as the absence of a history of occupational exposure to solvents). Eventually, 64 control workers (58%) were enrolled in the study. The neurobehavioural testing of these groups occurred from November 1992 to March 1993.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Recruitment of study population. Lost to examination: no address, moved abroad, dead. Exclusion because of confounding diseases (alcoholism, n=6; craniocerebral trauma, n=5; epilepsy, n=1; diabetes, n=1; severe depression, n=3) or current exposure to solvents in the formerly exposed group (n=1).

Unexpectedly, in February 1994 the plant with the workers still employed at low exposure planned to stop activities. A strike period of 1 year followed and eventually the plant shut down in February 1995. These events caused a delay in blood sampling for characterisation of the mEH phenotype. Blood was taken from September 1994 to June 1995—that is, about 6–15 months after the end of exposure in currently exposed workers and 5–6 years after the formerly exposed workers left the plant. In total, 56 workers (48%) from the combined group of currently and formerly exposed workers were willing to provide a blood sample.

CHARACTERISATION OF EXPOSURE TO STYRENE

Despite a well designed ventilation system, 4%–9% of the personal air samples exceeded the previous American Conference of Government Industrial Hygienists (ACGIH) threshold limit value time weighted average (TLV-TWA)19 of 213 mg/m3 and 4% of the peak exposures at the most exposed job sites exceeded the previous TLV-STEL (short time exposure limit)19 of 426 mg/m³ as shown in the earlier industrial hygiene studies from 1984 to 1987.16 17 Detailed individual records were kept for every worker as to the different tasks during his daily working hours (job characteristics, duration, and location of every individual task) allowing calculation of the exact total number of hours every individual worker was effectively exposed to styrene from 1982 to 1993 for currently and from 1982 to 1989 for formerly exposed workers. Also, instant measurements with colorimetric Dräger tubes were carried out daily by the industrial hygienist at locations where exposure to styrene was suspected to be high (comparable with STEL measurements). These data were collected meticulously with reference to the time of the day, the location of the job site, and the work done (in total 4145 measurements). These Dräger tube data (y) were validated by comparison with peak exposure measurements (15 min personal air sampling, n=80) of the industrial hygiene studies (x). A good correlation was found between the two sets of data (linear regression: y=1.066x; R² =0.83, p<0.01). On the basis of the daily Dräger measurements, an overall mean airborn exposure concentration was calculated for each person involved with lamination tasks that caused the bulk of the exposure to styrene. The site of each lamination job was characterised every year by a yearly average median exposure. A cumulative exposure index was then calculated for each worker through addition of the products obtained by multiplying the yearly average exposure at the site of each specific lamination job with the time (hours) yearly spent at that site. The overall mean concentration of exposure to styrene of each individual worker was obtained by dividing his cumulative exposure index by the total number of hours he was involved with lamination tasks from 1982 to 1989.20 Because the workers also did other tasks with much less or nearly no exposure to styrene, a mean time weighted exposure to styrene concentration was also calculated by dividing cumulative exposure by the total number of hours worked in the plant. At the time of the present study, biomonitoring data and full shift personal air sampling confirmed mean (SD, range) low exposure to styrene in currently exposed workers to be 41 (24, 11.5–120) mg/m³. Urinary mandelic and phenylglyoxylic acids showed good relations with styrene in the air suggesting that significant skin absorption of styrene was unlikely to occur (gloves were worn continuously).

SCREENING FOR NEUROTOXICITY

Before testing, the participants completed the NSC-60 questionnaire and a self administered questionnaire inquiring about lifestyle and medical history. The exposed and control workers were screened by one and the same interviewer with a structured interview about complaints at work (structured neurological anamnesis of complaints: SNAC). This questionnaire included complaints of headache, personality changes, concentration or memory difficulties, changed sensory functions in hand or foot, pain sensations, increased clumsiness, tremor, sleeping difficulties, diminished strength or endurance, diminished vision, hearing loss, tinnitus, altered smell, intolerance for alcohol, increased tiredness in the morning, increased tiredness in the evening, loss of equilibrium or dizziness, and “other complaints”. The workers were asked if they ever had the complaint during their work as a laminator (exposure to styrene) or as an electrician or assembly worker (control). The answers were scored “0” if the complaint was absent and “1” if present. If the complaint was present, it was asked if it was still present at the time of the interview (1992–3). If the complaint was no longer present the answer was scored “0” (not persistent), if still present it was scored “1” (persistent).

The standardised and validated neurotoxicity symptom checklist-60 (NSC-60)21 was administered for screening complaints at the time of the study. The NSC-60 is a self administered questionnaire which has 10 categories of questions about personality, concentration, equilibrium, sensorimotor functioning, general somatic complaints, sleeping problems, mood, chest symptoms, fatigue, and a general category of neurotoxicity symptoms. The mean score obtained for each category is a value between 1 and 4, and is called the category score. The category scores were dichotomised to “0” (score ⩽33th percentile of the control group) and “1” (score >33th percentile).

A computer assisted neurological test battery, the neurobehavioural evaluation system (NES, version 4.38), was used to administer neurobehavioural performance tests.22 The selected NES tests were hand-eye coordination, simple reaction time, symbol-digit substitution to assess visuomotor performance, associated learning, associated recall (number correct/associated learning×100), digit span forwards, and colour-word vigilance to test concentration and memory. The interviewer was blinded to the intensity and duration of exposure of the workers.

ANALYSIS OF mEH PHENOTYPE ACTIVITY

Buffy coat was collected from two blood samples of 8 ml, each withdrawn by venepuncture with Vacutainer CPTTM cell preparation tubes (Becton-Dickinson USA) containing sodium citrate as anticoagulant. The mononuclear cells were separated and rinsed twice with phosphate buffered saline and then S-TKM buffer (0.25 M sucrose, 80 mM Tris, 25 mM KCl, 5 mM MgCl₂, 0.1 mM EDTA, pH 7.35) was added to obtain a 2 ml suspension which was sonicated on ice followed by microcentrifugation at 9000g at 4°C. Aliquots of the S9 fraction were taken for protein measurement by the BIORAD method,23 and typical protein concentrations were between 1000 and 2000 μg/ml buffer solution. The S9 fraction solutions were stored at −80°C for further use.

A novel mEH assay has been developed in which a labelled styrene epoxide was used as substrate. Briefly 5 μl of a 10^-4 M styrene glycol solution in isotonic saline was added to 240 μl of the S9 fraction as internal standard. The reaction was initiated by addition of 5 μl [²H₈]styrene-7,8-oxide (10^-4 M in acetone) and incubation for 30 minutes in an oil bath at 37°C. The incubation mixture was extracted with 300 μl ethyl acetate; 100 μl of the ethyl acetate layer was dried under an N₂ steam. The mixture was derivatised for 30 minutes at 110°C with BSTFA/acetonitril (1/10), and analysed by GC-MS. The derivatised styrene glycol samples were injected by means of a solid injector (kept at 253°C) into an HP 5890 series-II gas chromatograph coupled with an HP 5970 quadrupole mass spectrometer. For chromatographic separation of the compounds of interest, a DB-5-ms fused silica capillary column (30 m×0.25 mm, 0.12 μm phase thickness, Alltech Associates, USA) was used with a source pressure of 0.006 Pa, the interface was kept at 270ºC and the oven programmed at 7ºC/minute from 100°C–140°C and at 30°C/minute from 140°C–300°C. The mass spectrometer operated in the electron impact mode with an ionisation energy of 70 eV. By referring to a calibration curve, the monitoring of ion-185 (trimethylsilated deuterated styrene glycol) relative to ion-174 (undeuterated analogue, internal standard) allowed an accurate measurement of the mEH mediated formation of styrene glycol in the incubated samples. The mEH phenotype activity was assessed as the mean of triplicate incubations.

STATISTICAL ANALYSIS

The statistical analysis was performed with the statistical package for social sciences (SPSS) for Windows, version 7.0. When needed log normal transformation was applied (number of SNAC complaints, symbol-digit substitution, simple reaction time). Fisher's exact test (double sided) and odds ratios (ORs) were calculated for dichotomous outcome variables—for example, persistent SNAC complaints and NSC-60 categories. For comparisons between control workers, and currently and formerly exposed workers one way analysis of variance (ANOVA) and Duncan's multiple range test were used for continuous outcome variables (number of complaints and NES test results). General ANOVA models were built allowing for age, alcohol consumption, years of welding, and the number of years at school as covariables for the NES results and additionally smoking and personality score for the questionnaire results. General ANOVA results are presented with ε² values (proportion of variance in the dependent variable explained by the difference across the groups). Multiple linear regression or multiple logistic regression analyses (dichotomous variables) were applied on the combined exposed group (currently and formerly exposed workers) to study dose-effect or response relations. The regression models were run with the forced entry method by introducing one of the exposure indices at the time (duration of exposure, mean concentration of exposure, interaction term between both, or cumulative exposure) together with the dependent variables and covariables. For the SNAC results and the NSC-60 category scores, age, alcohol consumption, the number of years at school, years of welding, smoking, and personality score were the covariables. In multiple regression models of the NES results, age, alcohol consumption, the number of years at school, and years of welding were included as independent covariables. Belonging to the current exposure group or not (1 or 0) was also included as a dummy variable in the regression analyses next to the exposure indices and the covariables to test separately the effect of being currently exposed. For the effect of mEH phenotype activity on the outcome variables, alcohol consumption and age were included as covariables. The level of significance was set at a p value ⩽0.05.

Results

CHARACTERISTICS OF THE STUDY GROUPS

The participating groups did not differ greatly for the covariables of interest (table 1). Despite efforts to match the groups according to schooling level, the currently and formerly exposed workers had fewer years of schooling and thus also a lower schooling level than the controls. It is interesting to note that in the three groups the scores for motivation of participation in the study were equally high (> 98%) and that their personality scores were similar. The currently exposed workers who laminated polyester from 1982 to 1992, on average were exposed for 4700 hours, whereas the formerly exposed workers laminated only from 1982 to 1989 (about 1100 hours less). The overall mean concentration of exposure to styrene during their lamination tasks from 1982 to 1989 amounted to 148 and 157 mg/m³ for currently and formerly exposed workers respectively, the combined exposed workers on average had 155 mg/m³ of exposure. Thus, before the 1989 event the exposure to styrene did not differ significantly between currently and formerly exposed workers. This was also shown by the time weighted exposure concentration.

The available characteristics for the workers exposed to styrene who did not participate had a similar age distribution (mean (SD, range) 39.0 (10.6, 26–65)), but their educational level was lower than in participating currently and formerly exposed workers (primary 34%, first grade 58%, second grade or higher 6%). The mean duration of exposure (3230 h) and overall mean exposure concentration (146 mg/m³) were not different from those of participating exposed workers.

QUESTIONNAIRE RESULTS

The results for each single complaint question of the SNAC questionnaire are summarised in table 2. During the moderate to high exposure to styrene, the combined exposed group had a significantly higher frequency of complaints such as irritability, fatigue in the evening, headache, common colds, intolerance for alcohol, diminished endurance, memory, and smell than did the controls. Despite not having been exposed for 3 years at the time of the SNAC questionnaire, 10 of the formerly exposed subjects still complained of a persistently diminished sense of smell (p<0.01). In the currently exposed workers, who experienced lowered exposure to styrene condition after 1989 (mainly due to a new type of boat introduced at the beginning of 1992), some complaints were still overreported—such as fatigue in the evening, common colds, and sleeping difficulties.

By comparison with control workers, of which 29 workers (45%) ever experienced some complaint during work, significantly more formerly exposed workers (79%) experienced some kind of complaint during lamination tasks versus only 52% in currently exposed workers (table3). Although both currently and formerly exposed workers had worked under the same exposure conditions until 1989, there was a significantly higher proportion of currently exposed workers (48%) who never experienced complaints during lamination than formerly exposed workers (21%) (Fisher's exact test, p<0.01; OR 3.5; 95% confidence interval (95% CI) 1.4–8.6). The number of complaints per person during exposure was twice as high in currently exposed and four times as high in formerly exposed workers compared with controls (one way ANOVA, p<0.01). The excess of complaints in the exposed groups remained unchanged after allowing for covariables in the model—such as age, schooling, alcohol consumption, years of welding, smoking, and personality score (ANCOVA: exposure group, ε²=0.18, p<0.01; none of the covariables reached significance). For the period after the 1989 event, low exposure for currently exposed workers and 3 years after the end of exposure for formerly exposed workers, the proportion of formerly exposed workers still reporting complaints markedly decreased from 79% to 31%. The mean number of complaints per person in formerly exposed workers drastically decreased as well and did not differ from that in controls, whereas currently exposed workers still reported more complaints than the two other groups (one way ANOVA, p=0.07). Currently exposed workers still showed an excess number of complaints after allowing for the covariables in the model (ANCOVA: exposure group, ε²=0.03, p=0.07; none of the covariables reached significance).

For the NSC-60 questionnaire, significantly more formerly exposed workers exceeded the cut off scores of several categories: chest (formerly exposed 36 (40%) v control workers 14 (22%), p=0.02); equilibrium (formerly exposed 28 (31%)v control workers 10 (16%), p=0.04); and somatic complaints (formerly exposed 66 (73%)v control workers 34 (53%), p<0.01). In currently exposed workers only the equilibrium category score reached borderline significance (currently exposed nine (33%)v control workers 10 (16%), p=0.09). The proportions of workers that exceeded the cut off scores for the chest and somatic complaints categories were similar in currently and formerly exposed workers, nine (33%) v 36 (40%) and 18 (67%) v 66 (73%) respectively.

NES TEST RESULTS

In general, currently and formerly exposed workers performed worse than control workers (table 4). One way ANOVA and Duncan's multiple range test showed that compared with control workers currently and formerly exposed workers had significantly worse test results for symbol-digit substitution (p=0.01) and digit span forwards (p<0.01). Hand-eye coordination was significantly (p<0.01) worse only in formerly exposed workers compared with control workers. After allowing for covariables (age, alcohol consumption, number of years at school, and years of welding) the group effect for symbol-digit substitution, digit span forwards, and hand-eye coordination remained unchanged (ANCOVA: ε²=0.05, p=0.01; ε²=0.09, p<0.01; ε²=0.05, p<0.01, respectively). Age was a significant covariable for the three neurobehavioural tests (symbol-digit substitution, ε²=0.32; digit span forwards, ε²=0.06; hand-eye coordination, ε²=0.06; all p<0.01) and, also, the years of schooling explained a significant part of the symbol-digit substitution variance (ε²=0.05, p<0.01). Simple reaction time, colour-word vigilance, associated learning, and associated recall did not differ significantly between the groups.

ACTIVITY OF THE mEH PHENOTYPE

The assay for the activity of mEH phenotype had a coefficient of variation of 13% for three repeated measurements in the same person. Enzyme assay blanks with incubation of heat denatured S9-fraction showed a non-enzymatic conversion of 3%–4% of the substrate into [²H₈]styrene glycol. The mEH phenotype activity was normally distributed in the studied population and ranged from 0.54 to 2.68 pmol/mg protein/min (mean (SD) 1.57 (0.49)) which represented a fivefold variation in mEH phenotype activity between people.

DOSE-EFFECT OR DOSE-RESPONSE RELATIONS

In multiple linear regression, the number of complaints per person in currently and formerly exposed workers during the period of high exposure (thus before the 1989 event) was significantly related to the individual overall mean concentration of exposure to styrene (regression coefficient β=0.25, p=0.03), but not to duration of exposure (β=0.09, p=0.34) or cumulative exposure (β=0.08, p=0.44). The interaction term concentration of exposure×duration of exposure did not reach significance (β=0.13, p=0.19). None of the covariables (age, years at school, years of welding, personality score, smoking, alcohol use) was significantly associated with the number of complaints (all p>0.2). Neither exposure indices, nor any of the covariables explained the persistence of complaints at the time of the interview in formerly exposed workers (multiple logistic regression, p⩾0.38). Allowing for the overall mean concentration of exposure and the covariables in the model, the dummy variable “being currently exposed or not (1 or 0)” showed a significant association with the number of complaints during exposure. Being currently exposed gave significantly less complaints than being formerly exposed (β=−0.15, p=0.03). In multiple logistic regression, there was no relation between the NSC-60 category scores and overall mean concentration of exposure to styrene, duration of exposure, interaction term between both, or cumulative exposure (p⩾0.10). Among the covariables only the personality score was related to all NSC-60 category scores (p⩽0.04) and alcohol consumption had a negative influence on the mood, equilibrium, and sensorimotor category scores (p⩽0.04).

In multiple linear regression, symbol-digit substitution in currently and formerly exposed workers was found to slow significantly with increasing duration of exposure (β=0.21, p<0.01, fig 2). Furthermore, there was a trend of diminished digit span forwards (β=−0.16, p=0.08) with increasing numbers of hours of exposure. Although, no group differences were found, increasing duration of exposure was significantly related to worsening of colour-word vigilance (β=0.21, p=0.03) and associated learning (β=−0.18, p=0.05). The overall mean concentration of exposure did not enter the models. Table 4 shows that by comparison with duration of exposure, cumulative exposure or the interaction term, duration of exposure×concentration of exposure, did not produce better p values when entered in the multiple linear regression models. It is, however, interesting to note that symbol-digit substitution and colour-word vigilance (perceptual speed) were quite well associated with the interaction term (β=0.18, p=0.02 and β=0.21, p=0.03, respectively). Among the covariables, a borderline significance (p⩽0.08) was found for age in explaining part of the variance of the neurobehavioural test results, except for associated recall. Similar results were found for the educational level and symbol-digit substitution, digit span forwards, and associated learning. The personal monitoring data at the time of the study did not explain a significant part of the variance of the neurobehavioural results of currently exposed workers (p>0.10). Allowing for the exposure indices and the covariables in the model, the dummy variable “belonging to currently exposed or not (1/0)” did not correlate significantly in any regression analysis.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Linear regression model in the exposed study population (+, currently exposed and ο, formerly exposed) (n=117) between the duration of exposure and symbol-digit substitution test results after adjustment for age, educational level, alcohol consumption, and years of welding (residual log (symbol-digit substitution)).

The symbol-digit substitution test in formerly exposed workers showed a delay of 0.19 s/digit after 3600 hours of lamination tasks. Currently exposed workers laminated 1100 hours more, and their symbol-digit substitution time increased by 0.22 s/digit compared with controls. The estimated value for the upper limit of normal for symbol-digit substitution time is 3.3 s/digit (mean of the controls +2SDs), which is similar to normal data reported by Dutch investigators (3.2 s/digit).24 This suggests that under the given exposure conditions, the evolution from a normal mean (2.5 s/digit) to reach pathological values of the symbol-digit substitution time will be reached on average after 15 160 hours of lamination. This is equivalent to a full time exposure of 8.6 years. A similar calculation based on the relation between symbol-digit substitution time and the interaction term, duration of exposure×concentration of exposure, showed that a pathological value for symbol-digit substitution time would be reached at an interaction product of 2 419 220 mg/m³×hour or 1375 mg/m³×year (taking 220 working days a year and 8 hours work a day). This means 6.5 years of full time lamination tasks at the previous styrene TLV-TWA value of 213 mg/m³ or 9.2 full time working years at a mean exposure to styrene concentration of 155 mg/m³. Although the safety measures and the ventilation system in the plant were very well designed and resulted in low to moderate concentrations of exposure to styrene, it should be pointed out that up to 4% of the air measurements were above the previous ACGIH TLV-STEL and up to 9% of the air measurements were above the previous TLV-TWA depending on location and characteristics of the lamination task.16 17

In multiple regression analysis, no relation was found between the number of complaints per person and the mEH phenotype activity (β=0.17, p=0.25), nor with the persistence of complaints (p=0.59). There was a significant relation between the mEH phenotype activity and symbol-digit substitution test results (β=0.34, p=0.01, fig 3) or colour-word vigilance (β=0.28, p=0.05), and a borderline significant relation with digit span forwards (β=−0.24, p=0.08). The test results worsened with increasing activity of the mEH phenotype.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Linear regression model in the mEH phenotype study group (n=56) between the mEH phenotype activity and symbol-digit substitution test results after adjustment for age, alcohol use, educational level, and duration of exposure (residual log (symbol-digit substitution)).

Discussion

As far as we know, only one short term follow up study about occupational exposure to styrene has dealt with the question of reversibility of neurobehavioural effects.2 Although difficult to carry out, the ideal study design in this context is a longitudinal follow up after the end of exposure in a relatively large cohort of workers. Economic and trade union linked instabilities of the late 1980s forced us to change our investigation into a design aiming at collecting retrospectively neurological information in a cohort of workers exposed to styrene in which 83% lost their job while the rest continued to work in the same plant at job sites gradually decreasing in exposure to styrene. In those field conditions, the maximum that could be done to answer the crucial question of whether neurotoxic complaints were persistent in this workforce was to organise a structured neurological anamnesis of complaints in such a way, that reliable responses could be obtained from the participants 3 years after the 1989 event in the plant. We are confident that a potential bias did not operate in the recruitment and selection procedure of the subjects in the present study. Sixty three per cent of the original cohort of workers exposed to styrene participated in the study. For non-participant workers, all basic group and exposure characteristics were available and were similar to the study group, except for a lower educational level. This could have lowered the robustness of the study, as it has been suggested that low intellectual capacity may predispose to the development of neurotoxic effects of organic solvents.8 It can be argued that the division of the original workforce into currently and formerly exposed workers could have introduced a bias due to the procedure by which workers were made redundant. However, no medical or productivity criteria were used in this procedure. The criterion of age (>50 years) is not likely to have introduced a bias as there was no relation between age and duration of exposure in this study population. Older workers were exposed to lower air concentrations of styrene and they had spent less hours on lamination tasks. Similarly, years of welding, years or level of education, and alcohol use were not associated with concentration or duration of exposure.

In line with complaints described in other studies,1 2 the present study also showed a significantly higher prevalence of complaints (irritability, fatigue, headache, alcohol intolerance, concentration difficulties, imbalance, and diminished sense of smell) when workers were exposed to styrene compared with control workers. The ambient air concentration of styrene during lamination tasks was the most important factor explaining the number of SNAC complaints at work. Interestingly, 3 years after the end of exposure (mean TWA 70 mg/m³ before the 1989 event), formerly exposed workers still reported an excess of atypical SNAC and NSC-60 (chest, equilibrium, somatic) complaints. These included epigastric discomfort, headache, difficulties in swallowing, heart palpitations, dyspnoea, nausea, tinnitus, diminished sense of smell, imbalance, chest oppression, feelings of drunkenness, and increased sensitivity to cold.21 In long term but less exposed workers (mean TWA 36 mg/m³), Edling et al 2 reported that most complaints were reversible after the end of exposure to styrene. In the present study, none of the predictors such as concentration of exposure, duration of exposure, activity of the mEH phenotype, or any other covariable could explain the persistence of complaints at the time of the interview. Unlike those of formerly exposed workers, the NSC-60 and SNAC results of currently exposed workers did not equally reach significance, most probably because of a smaller group size and because of a significantly greater proportion of currently exposed workers that never reported complaints although their exposures did not differ from those formerly exposed. This confirms the view that long term exposed workers may experience less subjective complaints than workers who left exposure earlier. Alternatively, the excess of complaints in formerly exposed workers might be seen as reflecting a revenge attitude of the dismissed workers. Although this could explain part of the variance, it is likely to be less plausible as the number of complaints at the time the workers were exposed was well associated with the concentration of exposure. It should also be pointed out that at the time of the study all the formerly exposed workers had been hired by other companies and were thus no longer unemployed. Also, formerly exposed workers had the same results on the hold questions of the NSC-60 (personality category) as currently exposed and control workers. This is a strong indication that formerly exposed workers had no tendency to tick high scores on the questionnaire items. An interviewer bias for the SNAC results seems unlikely, not only because of the blinding of the interviewer, but also because a comparison between the results of the SNAC questionnaire and the self administered NSC-60 questionnaire proved the consistency of the results.20 Despite lower TWA exposure at the time of the study (40 mg/m³) a slightly higher total number of complaints per person remained in currently exposed workers (diminished sense of smell, increased fatigue in the evening, imbalance, and sleeping difficulties at night) but a dose-effect relation could not be found. This excess of complaints in currently exposed workers could very well be a mixture of effects, due to the current day to day exposure, and chronic effects, due to the integrated dose. This has also been suggested by others.25

The neurobehavioural test findings are in line with most other styrene studies reporting effects on attention demanding tasks, visuomotor performance, learning, and memory.3 4 26 In the present study, hand-eye coordination (visuomotor accuracy), digit span forwards (short term memory, attention), and symbol-digit substitution (visuomotor performance) test results were significantly different between formerly or currently exposed workers and controls. Unlike many who consider reaction time tests to be the most sensitive tests, the lack of effect on simple reaction time in the workers currently exposed to very low concentratios of styrene is consistent with findings of Edling et al 2 and supports the view that slowing of reaction time reflects acute narcotic effects of higher concentrations of exposure.27 28

Symbol-digit substitution time showed significant relations with the duration of exposure and the interaction term, duration of exposure×exposure concentration, whereas digit span forwards showed a trend to worsen as exposure time increased. Although no group differences could be found, reaction times of colour-word vigilance (perceptual speed) and associated learning tests worsened significantly with increasing duration of exposure in the exposed group. None of these variables was related to the overall mean concentration of exposure. The results suggested that less than 10 years of exposure to a mean atmospheric styrene concentration of 155 mg/m³ (STEL measurements) or 70 mg/m³ (TWA measurements) and the concomitant occurrence of up to 9% and up to 4% of these measurements being above the previous TLV-TWA and TLV-STEL values, resulted in the development of dose-dependent effects and the persistence of some complaints and neurobehavioural changes. This would indicate that sporadic peak exposures may play a part in the development of chronic or persistent neurobehavioural effects. This is in line with publications on organic encephalopathies due to exposure to solvents,29 and strengthens the recent lowering of the TLV-TWA and TLV-STEL concentrations by ACGIH from 213 and 416 mg/m³ to 85 and 170 mg/m³, respectively.30

This study also showed a fivefold variation of the activity of the mEH phenotype between people, a variation in line with some of the activity ranges previously reported, ninefold,11fivefold,12 and 1.6-fold.13 There was no association between activity of the mEH phenotype and age, alcohol use, overall mean concentration of exposure, or duration of exposure (all p>0.10). Activity of the mEH phenotype did not seem to influence either the number of complaints during exposure or the persistence of the complaints. By contrast, visuomotor performance (symbol-digit substitution) and perceptual speed (colour-word vigilance) were worse in fast styrene metabolisers, but only if the duration of exposure was included in the multiple regression analysis. In other words, although exposure to styrene is still the primary factor determining the development of neurobehavioural effects, the occurrence of these effects may be modulated by the mEH phenotype. A similar pattern was found for the role of glutathione S-transferase M1 null genotype in the development of organic psychosyndrome, in which the effect became only significant in highly exposed workers.31 The slowing of visuomotor speed (increase of symbol-digit substitution time) with increasing activity of the mEH phenotype (fig 3) was surprising and might have different explanations. Firstly, it could be that styrene-7,8-oxide is not the most toxic metabolite of styrene in humans, but another toxic metabolite further down the metabolic chain could be. Mutti et al 32 33suggested that the ultimate metabolite, phenylglyoxylic acid, could be the active neurotoxicant. A second explanation might be that glutathione S-transferase activity in humans is more important for styrene detoxification in non-hepatic tissues than is commonly thought. On the other hand, the lack of glutathione S-transferase M1 could possibly lead to an aspecific induction of other metabolic pathways, for instance activity of the mEH phenotype,34 which may indirectly contribute to a positive relation between activity of the mEH phenotype and neurotoxic effects.

Conclusions

Some subjective symptoms or complaints and neurobehavioural effects in workers exposed to styrene were most likely persistent, even at TWA exposure concentrations which were thought to be safe in the 1980s. The concentration of styrene in the air during lamination tasks seemed to be the most important factor in the development of complaints, although most of these complaints disappeared after exposure ended. Also, the duration of exposure at lamination tasks seemed to be the best predictor of toxic effects on visuomotor performance and perceptual speed, which could still be detected in workers 3 years after the exposure to styrene had ended. The detoxification metabolism of styrene-7,8-oxide may also influence the development of these dysfunctions, although no clear conclusion about the underlying mechanism can be drawn from the present study. The results suggested that less than 10 years of exposure to a mean concentration of atmospheric styrene of 155 mg/m³ may result in persistent neurotoxic effects.